A Place for Consciousness

Probing the deep structure of the natural world


Table of Contents



Dedication

Acknowledgments

Preface

Abstract

Chapter 1. Liberal Naturalism

1.1 The Sliding Tile Puzzle

1.2 The topics: causation and consciousness

Causation

A recent resurgence

Consciousness

1.3 The structure of the book

1.4 Beginning the journey

Part I. Against Physicalism

Chapter 2. The Form of the Argument

2.1 The dialectic

2.2 The game of Life

A simplified physics

Supervenience, definition, and ill­defined properties

Two kinds of reduction

2.3 Could a Life universe possess phenomenal consciousness?

The challenge

The skeptic's claim

Qualia as observables

The character of the supervenience base in Life

Transferring information between worlds

Adapting anti­physicalist arguments to Life

Chapter 3. A Formulation of Physicalism

3.1 How to get nonphysical facts "for free"

The role of logical supervenience

3.2 Formulating Physicalism

Some things that logical supervenience is not

3.3 Getting back to the real world

Life imitates physics

The intrinsic properties of the physical

Summary

Chapter 4. The Objection from Metaphysical Possibility and Necessity

4.1 The minimal meaning standard

4.2 Are Kripke and Putnam necessities really metaphysical?

A quick summary

Deflating Kripke and Putnam

4.3 The constraints of natures

4.4 Are conceivable worlds not "really" possible?

Chapter 5. Logical Supervenience, Entailment, and Identity

5.1 Why not identity?

5.2 The failure of identity in ordinary cases

An argument against token identities for ordinary objects

Defending premise (3)

Coincidence is not identity

Deflating identity

5.3 The unimportance of token identity within physicalism

The physicalist's world with realization relations

The world with token identity

Logical supervenience is more fundamental than identity

5.4 Why does identity matter so little to physicalism?

Identity is not an entity

The origins of identity

5.5 The incoherence of the primitive identity view

The formal argument

Defending the argument

Hyperintensional differences and identity

The appeals to Kripke and Putnam

Summary

Chapter 6. The Objection from Holism and the Rejection of Analyticity

6.1. Quine's rejection of the analytic/synthetic distinction

A quick overview of Two Dogmas

Preliminary rebuttal

6.2 Meaning holism and the anti­physicalist arguments

The stability of conditionals

The role of systematization

Definitions, ostension, and recognition

Experience and The Background

Summary

6.3 The objection from holism of confirmation

6.4 Kirk's argument for the logical supervenience of consciousness

Part II. Puzzles, Paradoxes, and Tension

Chapter 7. Rethinking Nature: On the Possibility of a Benign Panpsychism

7.1 Nature and the problem of consciousness

7.2 Why we must go beyond the mind

7.3 Complexity

Concepts employed

Simplicity

Empirical questions

7.4 Functionality

Concepts

Simplicity

Empirical questions

7.5 Biology

Concepts employed

Empirical questions

7.6 Is panpsychism palatable?

Chapter 8. Puzzles for Liberal Naturalism

8.1 Where should we go from here?

An obstacle

The strategy

8.2 Category one: The­many­that­are­yet­one

(1) The unity of consciousness

(2) The subjective instant

Summary

8.3 Category two: The paradoxes of epiphenomenalism

(3) The knowledge paradox

(4) The superfluity of consciousness

Summary

8.4 The grain problem

Summary

Chapter 9. The Boundary Problem For Phenomenal Individuals

9.1 Introduction to the problem

9.2 The conceptual foundations of the problem

9.3 The brain and its subsystems

9.4 Moving towards Scylla and Charybdis

9.5 Scylla and Charybdis: The Boundary Problem

9.6 Two objections to the problem

9.7 The teeth of the problem: two examples

9.8 What to do?

Chapter 10. Transitions

10.1 First steps

10.2 The specter of epiphenomenalism

10.3 The space of possible responses

10.4 Problems with Hume's view

10.5 Foundational problems with Humean views

The metaphysical problem: The unity of the world

The epistemic problem: Solipsism of the present moment

10.6 Physical theory and Humean views

10.7 Beyond Hume

Part III. The Two Faces of Causation

Chapter 11. Receptive Connections

11.1 Causal responsibility and causal significance

11.2 Effective properties

11.3 The idea of receptivity

11.4 Receptivity as a connection

11.5 The conceptual grounding for receptive connections

Chapter 12. A Tutorial: Some simplified models

12.1 Overview of the presentation

12.2 How to understand the diagrams

12.3. Suite 12.1: Level one individuals

Diagram (a)

Diagram (b)

Counterfactuals

Two kinds of determinism

Diagram (c)

Diagram (d)

12.4 Suite 12.2: Generalized higher­level individuals

Diagram (a)

Emergent laws

Emergent properties

Diagram (b)

12.5 Suite 12.3: The orderly world of Life

Diagram (a)

Diagram (b)

Chapter 13. Does Receptivity Logically Supervene on the Physical?

13.1 Are the physical properties the receptive properties?

The Humean mosaic, the nomic mosaic, and the causal mesh

13.2 Including receptivity would not enhance the experimental content of physics

13.3 The argument from the possibility of bizarre receptive structures

13.4 Receptivity and spacetime

13.5 An example: positive and negative charge

13.6 From Receptive Connections to Carriers

Chapter 14. The Carrier Theory of Causation

14.1 Circularity in the causal mesh

14.2 Circularity instantiated

Properties of systems and properties of objects

Extrinsic properties within systems

14.3 The circularity of physics

14.4 From natural individuals to phenomenal individuals

Chapter 15. The Consciousness Hypothesis

15.1 Consciousness and high­level individuals

Arriving full circle

The absence of analysis

The knowledge argument

Epistemic asymmetry

The logical possibility of zombies

The inverted spectrum

15.2 The puzzles for Liberal Naturalism

Ubiquity and fundamentalness

The unity of consciousness

The subjective instant

The knowledge paradox

The superfluity of consciousness

The grain problem

The boundary problem for natural individuals

15.3 Complementary strengths and weaknesses

Chapter 16. Space, Time, and the Unity of the World

16.1 The direction of time and the unity of the world

Reflections

16.2 Ingressions, hits, and cascades

The unity of the world

16.3 Space and time

16.4 Spacetime in the Life world

16.5 "Fixed" facts and the puzzle of unidirectional connections

Chapter 17. Conclusion

17.1 The sliding tile puzzle revisited

17.2 Science and The Carrier Theory

Appendix A. David Lewis' Similarity Metric

Appendix B. Prolog program for the Life world's ingressions
















A PLACE FOR CONSCIOUSNESS:

PROBING THE DEEP STRUCTURE OF THE NATURAL WORLD



by

Gregg Howard Rosenberg




© 1997

Gregg Howard Rosenberg

ALL RIGHTS RESERVED




Dedication



This dissertation is dedicated to the memory of David Han.

This was for both of us, buddy. I loved you. Rest in peace.



Acknowledgments

Re­inventing nature is hard work. I could not have done even the little bit of it that I do here without a lot of support from others, both intellectually and emotionally. My intellectual interests and capacities have grown slowly, and my debts are owed to many. My time at Indiana University has been wonderful and invigorating, and the path that took me here was long and winding. I want to begin by acknowledging my debts to those who helped to steer me along that path, so that I finally found my way to philosophy.

I would like to thank Anthony Nemetz for first introducing me to the world of intellectual questioning when I was an undergraduate business major. His demanding eloquence was a revelation to me at that time in my life, as nothing in my background had previously exposed me to intellectual life.

I would like to thank Douglas Hofstadter, whose books "Metamagical Themas" and "Gödel, Escher, and Bach" serendipitously fell into my hands about that same time, steering me towards the philosophy of mind and cognitive science. My interactions with him here at Indiana University have been challenging and provocative.

I owe my deepest debts from my time at the University of Georgia to Donald Nute. Not only did he direct my master's thesis when I was studying Artificial Intelligence there, but he has encouraged and supported me every step of the way since: first in my decision to move into philosophy, and then by encouraging me to come to Indiana University to do my Ph.D..

I would like to thank Ned Block for the helpful conversations we had during my time at MIT in 1991. His insistence that ideas as unusual as mine need to be very strongly motivated has always stuck with me, acting as a burr whenever I have been tempted to cut corners in my writing or thinking.

When I arrived at Indiana, I was already convinced of the explanatory gap between the facts of consciousness and the physical facts, and suspected that there must be a deep link between consciousness and causation. I was extremely fortunate to come here with those interests at the same time that David Chalmers was finishing his dissertation on the very same topic, working through some very similar intuitions. That first year with David was very energizing, as well as exasperating. It was he who convinced me that, given my understanding of the problem, intellectual integrity required me to give up my commitment to physicalism; he also helped me to see that giving up on physicalism was a very different thing from giving up on explaining consciousness. I can scarcely say what this work would look like without his feedback and patience over the last few years, nor could I be sure it would even exist. I am, at heart, an intellectual conservative and a physicalist, and many times I felt discouraged by the distance I was traveling from my intellectual home. Without David's continued assurance that the work was interesting and worthwhile, it would have been very tempting to take the easier path.

The faculty at Indiana have also been very supportive. I would especially like to thank the members of my committee. I would like to thank Mike Dunn for taking the time to read each chapter of this dissertation over the last six months, and for providing his comments. The work would be much poorer without that feedback. Most importantly, I would like to thank him for being liberal enough to let me pursue these ideas.

I would like to give special thanks to Anil Gupta, not only for the helpful discussions we have had over the years, but for providing me with a role model for the way a true philosopher should conduct himself. His probity, patience, gentleness, and integrity have been an inspiration to me.

I would like to thank Tim O'Connor for his enthusiasm, incredible energy, and time at our long lunches. His intellectual support the last few years provided much needed reassurance for me here in Bloomington, reassurance that helped me to keep myself on track.

In the last year Indiana was very fortunate to add Brian Cantwell Smith to its cognitive science program, and I was fortunate to add him to my committee. Like me, Brian is a computer scientist­cum­philosopher, and the perspective that gives is difficult to put into words. I am especially grateful to Brian for the long hours he invested trying to help me improve my writing. Without his help, this already difficult piece of work would be almost unreadable.

As anyone who has completed an advanced degree knows, graduate students learn more outside of class from our peers than we do inside class from professors. Nothing can substitute for heated arguments over beer that last late into the night. I almost have too many of these informal debts to list, mostly to my fellow graduate students in the philosophy department. I would like to single out for special thanks a handful that have provided especially memorable philosophical conversation over the last few years: Tony Chemero, Diarmuid Crowley, Stephen Crowley, Eric Dalton, Craig DeLancey, Jim Hardy, and Adam Kovach. I would also like to thank the Indiana University philosophy faculty for their support over the years, but most especially David McCarty and Paul Spade.

Here in Bloomington, my warmest and most deeply felt thank­you's are owed to Leslie Gabrielle. Not only has she provided me with an important intellectual sounding board, but her friendship and support over the last four years have been priceless on a personal level. Work in graduate school in a highly specialized field can sometimes be isolating, and the connectedness she provided for me was desperately needed. I would not have gotten through some of the rougher times over the last few years without her. Along those same lines, I would like to thank my friends from back home in Atlanta, especially Allen Domenico and Bob Lauth for their support and encouragement. The most precious friendships are the ones that you know will last a lifetime.

Finally, my most speechless thanks are reserved for my parents, Sally and Donald, and for my brother Alan. They have made an investment in my life and identity that is truly staggering to consider. Every word in here reflects their love.



Preface

So, why write about consciousness and causation?

Those two problems are each tough philosophical chestnuts individually, and it's not clear why thumping them together would help us to crack either one of them. The burden of this dissertation is not only to argue that they need to be treated together, but to actually show, in a very concrete way, how they in fact do go together. I argue, in effect, for the radical thesis that conscious experience is an aspect of causation itself, and one that is absolutely vital to it. And I argue that all this is incompatible with physicalism while, nevertheless, being perfectly compatible with physical science. This is a tough ledge to walk in the current intellectual climate. Accordingly, the aims I have for this work exist at several levels, each a bit more ambitious than the previous.

At the least ambitious level, I wish for this dissertation to provoke readers. Within it I defend a bevy of ideas that are at odds with the physicalist orthodoxy within the philosophy of mind. I believe the world picture I begin to flesh out in this work should at least make physicalists uncomfortable, as it brings to the forefront the possibility that a dual aspect or property theory needn't be supernatural, naturalistically untenable, or hopelessly vague. My chief hope for this work is that, after reading it, no physicalist should be able to rest comfortably with an easy assumption that any alternative view must lead to empirical absurdity.

At the next level, I hope that this dissertation challenges readers. Possible challenges come on two fronts. On one front, readers who remain convinced that physicalism is likely true need to find better arguments for it than currently exist. Almost all parties agree that embracing physicalism requires biting some large bullets, but physicalism's strongest support has been the widespread intuition that rejecting it seems to lead to even greater absurdities than does accepting it. Especially, many assert that physicalism is required to guarantee the causal relevance of experience.

One challenge coming out of this dissertation is to show vividly that, if one wishes to hold onto this physicalist shibboleth, one's background theory of causation needs to be fully articulated. For what this dissertation does above all else is present a case that, under at least one possible and substantial view of causation, a view that seems compatible with physical science, the physicalist shibboleth turns into a chimera. Not only does experience turn out to have a place in the causal order even if physicalism is false, I make a case on grounds completely independent of the mind­body problem that it, or something nearly exactly like it, is required for causation to exist.

A second kind of challenge exists for those less sympathetic with physicalism: many open questions remain at the end of this work, and these questions present the possibility for an actual empirical and philosophical research program. The challenge is to see if these open questions lead to fruitful avenues of research, or if, instead, they show that the ideas here lead down a dead end.

At the third and most ambitious level, I hope that this dissertation may actually convince some readers. While I propose some unusual ideas here, I try very hard to accompany my proposals with substantive philosophical arguments. The generic position I suggest, which I call Liberal Naturalism, is currently a minority position, but it at least has contemporary precedents within philosophy, especially in the work of philosophers like David Chalmers and Frank Jackson. My more specific proposal, which I call The Carrier Theory of Causation, involves experience directly in the fundamental causal character of the world. This more specific proposal seems very radical when stated baldly, but I have not pulled a rabbit out of a hat, nor have I motivated it by appeal to any strange aspects of quantum mechanics. I have tried to work with acceptable rigor using only some fairly mundane intuitions about the world, and about consciousness. And I have tried, always, to respect science. While the case I make is not airtight, few philosophical or scientific arguments ever are. It is, nevertheless, highly detailed and rationally motivated. When viewed as a whole, I think it has a pleasant kind of elegance, and promises to be potentially fruitful. I consider those properties its greatest strengths.

Finally, I want to say that this dissertation is intended to be as much philosophical narrative as it is a collection of philosophical arguments, although arguments are produced in abundance. For that reason, the argumentative structure of this dissertation should, in some respects, be viewed as secondary to it. My chief concern when writing it was to convey a possible way of looking at the world. In that sense, I hope the reader comes away with a feeling that there is a cosmological vision that is worth thinking about lying in the vicinity of the ideas here, and that it promises to make the world a rich and interesting place if true. Within philosophy, its closest kin is to be found in the process philosophy descended from Bergson and Whitehead, and I owe Whitehead a debt for inspiration.

The most important thing to know about this dissertation before you begin reading it is that, although it is long, it is possible to take a short tour and still come away with the main thrust of argument. For those interested in the short tour, I recommend reading chapters one through three to understand the set­up of the problem. From there, skip to chapters ten and eleven, and then to chapters fourteen and fifteen. If the short tour has peeked your interest, go back and read the rest. Especially, those who believe in the adequacy of one or all of the standard replies to the anti­physicalist arguments should read Part I. Part II gives more thorough reasons than the short tour does for believing that someone interested in understanding consciousness should look hard at causation itself. Finally, a full and careful reading of Part III is interesting independently of one's views on the mind­body problem, especially the detailed treatment of receptivity in chapter twelve, and the arguments against actualism, and for the reducibility of spacetime, in chapter sixteen.


Chapter 1

Liberal Naturalism

1.1 The Sliding Tile Puzzle

The mystery of consciousness is both profound and exciting. To anyone who has thought hard and long about it, the questions it raises linger and deepen upon reflection. Eventually, they seem to transcend specific questions about consciousness, touching insecurities about our understanding of nature herself. One begins to suspect that solving the puzzle of nature may not be like piecing together a jigsaw puzzle, with pieces nestled stably in their proper place. To solve the puzzle of conscious experience, we may have to view the project as being more like trying to solve a sliding tile puzzle.

What is a sliding tile puzzle? A sliding tile puzzle consists of a rectangular frame with movable tiles within it, each tile decorated with a different part of the puzzle. Initially the tiles are scrambled, and the goal is to unscramble them to retrieve the puzzle's picture. The rectangle contains one empty space, and the puzzle solver must rearrange the tiles by sliding them into and out of this empty space. By doing this, the puzzle solver hopes to undress the confusion and reveal the puzzle's ornamental face.

Sliding tile puzzles contain a trap, a seductive property that lures the unsuspecting. Often the puzzle solver can bring order to almost all of the puzzle, perhaps fitting every piece into its proper slot except the last two tiles. These last two tiles might be transposed, for instance, each in the other's slot. The trap is sprung when the puzzle solver holds stubbornly to the hard won order in the rest of the puzzle, afraid that disturbing it too much will cause it to disappear, never to return. Seduced by the order already in the puzzle, the puzzle solver searches desperately for a minimally disruptive solution, one that places the pieces without disturbing the rest of the puzzle very much.

Unfortunately, we cannot usually solve tile puzzles this way. To fit the final two pieces in place one has to regress first, and then rebuild the old order from a new direction. The trap is that, since the puzzle solver flinches at every challenge to the old order, the ideal of completing the puzzle becomes a hopelessly elusive goal. The irony is that the hard won old order will eventually reappear within a more completely ordered context, but only if it is first challenged and, temporarily, relinquished.

In writing this book, I have approached the problems of consciousness and causation like they are the final two pieces in a sliding tile puzzle. I wish to help put them into their proper places within a naturalist framework, and believe that sound arguments exist that this achievement will carry a cost. This cost will require ruffling, just a bit, the hard won order science has brought to our understanding of nature. The cost is this: we must concede that materialism is an inadequate version of naturalism.

Materialism is an ontological thesis. Within metaphysics, ontology is the study of what exists, with particular emphasis on the different ways of existing possessed by different kinds of things. Traditionally, philosophers studying ontology have been very much concerned with identifying fundamental categories of properties, objects, events, or processes whose existence they could see as grounding the existence of other kinds of things. Materialism (which I will also call physicalism) is the thesis that all other kinds of things derive their existence from the existence of the physical. Among these "other kinds of things" are minds, value, the good, justice, beauty, and meaning.

Materialism has been a dominating influence in this century, and as a working hypothesis it has severely constrained the avenues of explanation open to scientists and philosophers. The constraint it provides has proved very fruitful in the empirical sciences, and a bit less so in philosophy. My own opinion is that its overall effect has been positive. Because of it, we have uncovered a great deal of the startling beauty and subtle order in the puzzle of nature.

I have great intellectual admiration for, and sentimental attachment to, materialism. And yet, in this book I will argue that materialism is false. The book has a positive thesis also. Despite the failure of materialism, I do not believe we should abandon the goal of naturalism, as we should not identify naturalism with materialism. We do best when we think of naturalism as a methodological requirement. It requires that we place human beings in the world without special, ad hoc assumptions that are discontinuous with what else the natural sciences tell us about nature.

Among naturalists, those such as McGinn (1989) occupy one extreme. They are so impressed by the difficulty of the problems associated with explaining consciousness that they have concluded that it is just not soluble. Their opponents on the other extreme have dubbed them, a bit derisively, "The New Mysterians". These opponents sometimes continue to hold out for more traditional science (Churchland & Churchland, 1990; Flanagan 1992), and sometimes simply deny the existence of the phenomenon at issue (Dennett 1988; Wilkes 1988). We can call those on this other extreme, again a bit derisively, "The Gung­Ho Reductionists".

The position I will begin to develop in this dissertation is a Liberal Naturalism. Like materialism, Liberal Naturalism holds that the world is composed from a single set of fundamental properties and entities, related by a single set of fundamental laws. However, unlike materialism, Liberal Naturalism holds that some of these properties and laws are not physical properties and laws. I am a Liberal Naturalist, as is David Chalmers, an earlier version of Frank Jackson, Nagel in some of his moods, and Wilfrid Sellars on some ways of reading his work. The Liberal Naturalists recognize the possibility that what is described by physics, and what subsists in physics, may not circumscribe nature's limits. That allows them comfortably to step outside the standard physicalist ontology, while retaining a naturalist world view. From The Gung­Ho Reductionist and New Mysterian standpoints, this is something of an end­around(1).

The positive project in this book is to identify what these non­physical properties are; to explain why they should exist; and to give reasons for believing that a story exists showing how they fit cohesively within a scientific and naturalistic world view. I will pursue these goals by introducing a substantive view of causation that goes beyond the physical. If successful, this rich picture of causation will provide the bridge that takes us from the physical to consciousness. The view I will develop respects the effective closure of the physical. I attempt to complete our views of causation by adding elements that are complementary to the structure of activity described by physical science, and, for that reason, are every bit as essential to it as the physical.

1.2 The topics: causation and consciousness

Causation ­ The human body is a wondrous system, a marvel whose subtlety and complexity test the word "machine", making it sometimes seem like a recklessly inadequate characterization. Natural science tells us that the body is made ultimately of very tiny and exotic physical entities, and we know that it consists in the motions of, and interactions among, delicately layered physical structures. Our bodies are spatiotemporal organizations of these tiny entities, driven by an enormous number of microphysical interactions between and among them.

According to physical theory, the entire being of these entities consists in the active dispositions that produce their intrinsic dynamics, and their intimate couplings. The mystery of consciousness is the question of why this assembly, this whirlwind of causation, should ever feel. Why cannot this causation go on without feeling, without sensation, without experiencing at all? Viewed in the large, these finely layered patterns are dynamical wonders, but it is hard not to wonder why the dynamics should be conscious. Causation produces motion, but why should a congerie of motion, however complex, ever feel? Many working philosophers of mind consider the task of finding an answer to this fundamental question the greatest obstacle to the task of naturalizing the mind.

A recent resurgence ­ The problem of consciousness is currently a very live topic of interdisciplinary debate. New journals devoted to the topic of consciousness, such as The Journal of Consciousness Studies and Psyche, have appeared in the last few years. In 1987, Ray Jackendoff ventured a computational analysis of consciousness in his book Consciousness and the Computational Mind. Marcel and Bisiach's anthology Consciousness in Contemporary Science (1988) presaged and helped to legitimize the growing scientific interest in the topic, as did Bernard Baar's substantive proposal about the functional and biological organization of consciousness in A Cognitive Theory of Consciousness (1988). More recently, major books on the topic by Nobel laureates like Gerald Edelman (1989) and Francis Crick (1994) have lent scientific prestige to the topic.

Within contemporary analytic philosophy, a slow but steadily building discussion existed in the 1970's and 1980's, provoked largely by contemporary anti­physicalist arguments. The key figures producing this pressure were Saul Kripke (1972), Thomas Nagel (1974), and Frank Jackson (1982). In the late 80's, William Lycan presaged the philosophical resurgence of the topic with his book titled simply Consciousness.

The book that ultimately broke the dike was Daniel Dennett's flamboyantly titled, and popularly successful, Consciousness Explained (1991). Dennett galvanized forces on all sides through his ingenious efforts attempting to explain consciousness by contentiously arguing it out of existence. Though widely read, Dennett's work has garnered more opposition than conversions.

I believe another recent book, David Chalmers' The Conscious Mind (1996), marks a modern watershed of sorts. The Conscious Mind succeeded in systematizing the dispersed and often sketchy versions of anti­physicalist arguments, expressing them within a powerful and perspicuous metaphysical framework. While the logic and metaphysics underlying these arguments stand to be refined even further, the conclusions at stake seem to be these. First, Chalmers argues that we have very good reasons to believe that an in­principle explanatory gap exists between explanations in the physical sciences, and the facts about consciousness. These facts include even the seemingly brute fact that consciousness exists. Second, he argues that the existence of such a gap has ontological repercussions, and cannot be dismissed as a mere epistemic shortcoming. A theory's ontological completeness is bound to its explanatory power by unbreakable chords. In the rest of this book, I use the term anti­physicalists to refer to the group of philosophers who, like me, accept these conclusions.

After defending this negative thesis, Chalmers suggests that we may explain consciousness even if physicalism is false. When reductive explanation fails, he points out, we may turn to non­reductive explanation. The physical facts would play a large and essential role in such an explanation, but they would not be the whole story. In some ways the present book is a sequel to Chalmers' book, taking another few steps down the road of explaining consciousness within a non­reductive framework.

Consciousness ­ Given all this commotion, some readers may wonder what it is that I will be trying to explain. "Consciousness" is a multiply ambiguous term(2), and not all senses of the term pose equally severe problems. This book focuses on what philosophers call phenomenal consciousness, but, unfortunately, no easy way to define phenomenal consciousness seems at hand. An understanding of what consciousness is is as much a matter of acquaintance as definition. At best, one can call attention to it in an increasing level of detail. I will tarry over the task for only a little while here. Along the way, I cite more thorough sources that interested readers are encouraged to turn to for more detail.

The most succinct way to convey the meaning of the term is through a phrase popularized by Thomas Nagel. A creature's phenomenal consciousness is responsible for the facts concerning what it is like to be that creature. For instance, part of what it is like to be a visually normal human being is for purple things to visually appear in a certain way, as having a certain kind of visual quality. Once one becomes aware of these visual qualities as qualities, one may naturally wonder what the colors from a larger color space look like. For example, some birds discriminate color more finely than any human. What is the experience like when these birds see the extra colors presumably available to them? Once we know about their ability, a question about the character of their conscious experience remains. Similarly, just as human visual sensations are different from human auditory sensations, one can wonder what a bat's echolocation is like for it; one may also wonder what the qualities and sensations associated with a manta ray's sensing of electromagnetic currents on the ocean floor are like. At the extreme, one may even wonder, however implausibly, whether there is anything at all that it is like to be these creatures. Perhaps they are unconscious automatons, all "dark inside"?

All of these wonderings seem to involve legitimate questions about matters of fact. Solving the mystery of consciousness requires understanding just what kind of facts these are. Are they ordinary physical facts, explainable in the same basic way that other physical facts are explainable? This book will ask: given all the physical facts ­ the facts about the motion and structure of bodies ­ why are there facts like these at all?

Each of these wonderings addresses an informational deficit in our knowledge of the world. The information about the character of these qualities constitutes phenomenal information for subjects of experience.(3) Additionally, the exact organization of the qualities of experience, and perhaps even their character, seems to be very responsive to conceptualization. One sees this phenomenon when staring at visual illusions like the Necker cube: the qualitative experiences associated with seeing its face as oriented upward or as oriented downward are very distinct. This suggests a location for the world's repository of facts concerning phenomenal consciousness. For a particular creature, the facts concerning what it is like to be that creature arise from (i) its capacities for obtaining phenomenal information in the first person, and (ii) its way of conceptualizing the world.(4)

Moving just slightly beyond Nagel's slogan, one may deliver longer descriptions of the intended target. The following paragraph by Brian Loar (1997) does a good job of concisely expanding the slogan, "What it is like to be...."

On a natural view of ourselves, we introspectively discriminate our own experiences and thereby form conceptions of their qualities, both salient and subtle. These discriminations are of various degrees of generality, from small differences in tactual color experience to broad differences of sensory modality, e.g. those among smell, hearing and pain. What we apparently discern are ways experiences differ and resemble each other in respect of what it is like to have them. Following common usage, I will call those experiential resemblances phenomenal qualities; and the conceptions we have of them, phenomenal concepts. Phenomenal concepts are formed "from one's own case". They are type­demonstratives that derive their reference from a first­person perspective: 'that type of sensation', 'that feature of visual experience'. And so third­person ascriptions of phenomenal qualities are projective ascriptions of what one has grasped in one's own case: 'she has an experience of that type.'

At the next higher level of detail, one can move on from descriptions like this by cataloging varieties of phenomenal experience. By asking the reader to attend to varieties of consciousness available in his or her own case, such efforts hope to direct and refine each reader's awareness of the subject matter. Chalmers engages in this sort of approach in the first chapter of his The Conscious Mind. He calls attention to, and gives short accounts of, the fascinating variety of phenomenal content found in experiences as diverse as: visual experiences, auditory experience, tactile experiences, olfactory experiences, taste experiences, experiences of temperature, pains, other kinematic and proprioceptive sensations, mental imagery, conscious thought, emotions, and the sense of self. Such catalogs are often effective at making the subject matter vivid.

At the most extreme level of detail one can isolate the meaning of 'phenomenal consciousness' by comparing and contrasting it with other senses of the term 'consciousness'. The most detailed contemporary exposition of the meaning of the term seems to be in Charles Siewert's (1994) Understanding Consciousness. Siewert delves into extreme detail, with much care, in an attempt to isolate the sense of the term that picks out the mystery, drawing it out from its hiding place among the other senses of the term. His exposition runs over two hundred pages, and I recommend it to readers who are hungry for further clarification.

1.3 The structure of the book

As the discussion of phenomenal consciousness attempts to progress, new proposals all flounder on the same basic problem: naturalized explanations of consciousness seem forced to postulate unexplainable identities between conscious states and the ontology of the proposed theoretical machinery. Some proposals claim that conscious states are primitively identical with certain kinds of computational states (e.g., Johnson­Laird 1988); others claim that they are primitively identical with biologically based brain states (e.g., Kinsbourne 1988); still others claim that conscious feelings are identical to special kinds of representational contents (e.g., Tye 1995); and so on. Since each of these features is at least arguably present in the one type of consciousness we have access to, our own, and since none of the proposed theories can establish their explanatory power without first appealing to the contentious identities, adjudicating between them is as difficult as reliably generalizing from them.

The sticking point seems to be that what we know about physical causation just does not seem to contain, even implicitly, anything that would ground the possibility of conscious experience. This well­known "explanatory gap" between our physical image of the world and conscious experience forces theorists to treat these identities as primitive, and these, in turn, vitiate the explanatory force of the proposals that use them. Our ontological commitments are embedded in our explanatory frameworks, so failures of explanation signal incompleteness in our ontological commitments. That is, in a nutshell, why physicalism fails.

Within contemporary philosophy of mind, physicalism is something of a church orthodoxy. Anti­physicalists are not completely hostile to the church's principles. Rather, the arguments they produce stand to physicalism in a way that is something like Lutheran attacks on aspects of Catholic orthodoxy and practice. In an effort to fend off these reformist attacks, the Church elders produce a standard set of maneuvers designed to defend the faith.

In part one of this book, Against Physicalism, I argue that a physicalist metaphysics cannot adequately account for consciousness. To establish physicalism's failure, I examine four of the most prominent shields held forth by the architects of this philosophical Restoration, arguing that none of them effectively blunts the blow. Chapter two examines the charge that the reformers are arguing from ignorance, and also offers a working analysis of the physical. Chapter three formulates a version of physicalism that is weak enough so that any plausible version will fail if it fails, and then uses the results from chapter two to argue that it fails. Chapter four examines the prospect that some opaque, metaphysical necessity may restore the orthodoxy, allowing for a non­reductive physicalism. Chapter five examines an analogous appeal to identity, one that proposes the existence of identities that do not follow from any deeper explanatory connection. Chapter six discusses an appeal to alternate conceptions of explanation and ontology stemming from meaning holism.

The last two parts of the book build a defense against a fifth objection, an objection that is perhaps the most simple and powerful available to the Restoration. This fifth objection is simply that physicalism must be true if we are to avoid absurdity. The absurdity arises from an unsavory dilemma that anti­physicalism is supposed to reduce to: consciousness either is epiphenomenal (i.e., causally irrelevant), or interacts in a spooky way with the physical. By entering into a detailed analysis of causation, and thereby producing a third alternative, I will argue that this is a false dilemma.

The case I will build is divided into two parts, one occupying the middle third of the book, and the other presented in the final third. In part two of the book, Puzzles, Paradoxes, and Tensions, I spend time exploring problems and tensions created by the anti­physicalist conclusion that there must exist fundamental nonphysical properties. How can the world have both physical and phenomenal aspects? And why would it? By searching the places where these two aspects seem most incompatible with one another, I try to discover clues about where the incompleteness in our knowledge might lie. Among other conclusions, I argue that the existence of consciousness is evidence for hidden structure within nature. Also, I argue that, at every turn, our search points us toward the need to more fully understand causation itself.

I devote the final part of this book, The Two Faces of Causation, to a direct analysis of causation and the mystery of causal interaction. Perhaps the metaphysics of causation is richer than materialists usually suppose? I will argue for an analysis of causation in which experience shows up as a result of special sorts of causal interactions, interactions in which a set of effective individuals share a common receptive field. As a first point, I build a case that, like the explanation of consciousness, the explanation of causation also requires nature to have two aspects. As a second point, I argue that a realist account of causation needs to posit a structure to the causal nexus. I will argue for a conception of the causal nexus as the connecting point where effective elements of the world become receptive to one another.

An intriguing twist arises from the analysis of the structure needed to understand the causal nexus. On grounds completely independent of the problem of consciousness, I will argue that this structured entity must have properties that precisely parallel the troublesome properties of consciousness. The nexus requires a partially nonphysical nature, a kind of privacy, unity, apparent irrelevance from the perspective of physical explanation, a variety of intrinsic properties, and it contains ties to both individuation and identity.

Given these precise parallels, I will argue that an appealing case exists for postulating that the existence of conscious experience is intimately related to this nonphysical aspect of causation. As such, it turns out that the place of consciousness in the natural world intrinsically connects it to a larger, metaphysical background via its intimacy with causation itself. Under the kind of realist account of causation I will detail, a picture emerges that does not drive a wedge between consciousness and the physical world. Instead, it locates us within a world that is richer than the one previously available. The resulting view provides the foundations for a possible Liberal Naturalism.

1.4 Beginning the journey

If I may be granted the patience to indulge in one final metaphor, in many ways this book will be like a slow boat ride down a long and lazy river. We will travel through an exotic set of landscapes, lingering to take into our sights a sometimes disorienting variety of unusual beasts. Before the trip ends, we will explore questions about the nature of the physical, the meaningfulness of claims involving necessity, the status of high­level token identities within physicalism, the possibility of panpsychism, the boundaries of consciousness, the nature of causation, the primitiveness of space and time, the relation between actuality and possibility, and the very idea of an intrinsic property. With luck, the richness of the landscape will serve to make the journey worth the investment.

I will begin by clarifying why Liberal Naturalism is needed in the first place, and that means understanding why an explanation of consciousness will elude us within a purely physicalistic framework. That is the task taken up in the next five chapters, and it will carry us through the first third of our journey. Pack your bags, it should be fun.

1. An 'end-around' is a misdirection play in American football. The offensive team makes the defensive team believe it is taking the ball in one direction, but then hands it to the split­end, who runs in the other direction around the startled defense. In this case, the Liberal Naturalist rejects any form of physicalism, but goes around traditional forms of defense against dualism by retaining naturalism. The end­around is a high risk/high reward play. If all goes well, it results in a large gain. If it develops poorly, usually a large loss results as the split­end gets tackled behind the line of scrimmage.

2. Baruss (1990) catalogs twenty­nine separate definitions of the term, which he groups into three categories. Chalmers himself distinguishes eight senses of the term in his first chapter.

3. For an extended defense of the existence of phenomenal information see Lycan (1996). Lycan is a physicalist.

4. I do not mean to suggest in any way at all that these are independent capacities.


Chapter 2

The Form of the Argument

2.1 The dialectic

Recent anti­physicalist arguments have relied on a variety of thought experiments, claiming that these point to inevitable limits on the physicalist program for explanation, and, by implication, the metaphysical position of physicalism. The physicalist's response to these maneuvers has wavered between polite incredulity, and a zealous patience with business­as­usual science. This response frustrates the anti­physicalists with its apparent blindness to the deep philosophical issues that seem, from their perspective, to be at stake.

Perhaps the seminal modern paper on the issue is Nagel's What is it like to be a bat? (1974). Nagel argues that the subjectivity of points of view will be left out of any physicalist account of the universe. According to Nagel, our inability to discover what it is like to be creatures very different from ourselves, despite knowing their physiology, makes this failure evident.

Among others, Frank Jackson (1982) and recently David Chalmers (1996) have refined Nagel's guiding intuitions. In Jackson's well ­ known thought experiment, he asks that we consider a super­neuroscientist named Mary. From within a black and white room, through books and observation of a black and white TV, Mary learns everything there is to know about the visual system. Jackson maintains that, nevertheless, Mary learns something the first time she is exposed to color. She learns what the experience of blue is like, for instance. Thus, physicalism must be false because we can know all the physical facts without being able to know, even in principle, all the facts.

Chalmers asks us to conceive a universe physically identical to ours from big bang to big crunch, but with the twist that our counterparts have no conscious mental life. They are phenomenal zombies. Chalmers argues that such a universe is conceivable and, furthermore, logically possible. He argues that this shows the falsity of physicalism by showing that the facts about qualitative consciousness are further facts, not determined in the appropriate way by the physical facts.

If the anti­physicalists are akin to Lutheran reformers, the Churchlands have quite accurately summarized one typical response by defenders of the faith (1990),

The negative arguments here all exploit the very same theme, viz. our inability to imagine how any possible story about the objective nuts and bolts of neurons could ever explain the inarticulable subjective phenomena at issue.

In the quote above, the Churchlands seem to be accusing the anti­physicalists of producing arguments from ignorance. In this chapter I will defend the anti­physicalist arguments against this charge, arguing that this physicalist attempt at restoration fails. I will lay out the form of the argument. On the way to doing this, I will need to suggest a working analysis of the physical, defend the position that knowledge of phenomenal facts comes from observation, and offer a diagnosis of why the phenomenal facts do not logically supervene on the physical. Although the analysis in this chapter is primarily intended as a clarification of current arguments, I believe the end­product also counts as a substantial new argument in its own right.

2.2 The game of Life

A simplified physics ­ In this section, I will try to clarify the anti­physicalist's understanding of the arguments by developing a detailed analogy between the physical and the cellular automaton called Life. The Life world has been used in discussion of the mind­body problem before, most notably in Dennett (1991a). Along the way I will use examples to introduce the philosophical notion of logical supervenience, but only later will I discuss the relevance of this notion to the issues. The immediate purpose is to lock onto intuitive conceptions of what it means to be physical, and to logically supervene upon the physical facts. I do this by drawing out the categorical structure of physical theories, identifying the kinds of information they convey, and exposing the kinds of conditions that make physical properties the kinds of properties they are.

Life is the name of a kind of cellular automaton that evolves on a two­dimensional grid. Cellular automata consist of points, or cells, in an abstract space, all of which can have kinds of "causal" properties. We can start an automaton off by assigning properties to cells at random. The automaton then evolves, changing states according to rules that apply pointwise to the space. Typically, the rules that determine which properties a cell will have at a given time are a function on the properties of neighboring cells at an immediately preceding time.

In Life, we think of each cell on the grid as square, and as having eight neighbors: a neighbor touching it on each side, and a neighbor touching it on each corner. Additionally, a cell can have exactly one of two properties, being "on" or being "off", at any given time step. Figure one pictures a cell and its neighbors. Three simple rules govern the evolution of a Life automaton:

1) If a cell has exactly two on neighbors it maintains its property, on or off, in the next time step.

2) If a cell has exactly three on neighbors it will be on in the next time step.

3) Otherwise the cell will be off in the next time step.

Despite its simple structure, the Life automata can evidence a tremendous variety of patterns. For instance, the mathematician who invented it, John Conway, has proven that a Life grid can act as a universal Turing machine. More remarkably, he has proven that the grid can support extremely complex patterns that are self­replicating in Von Neumann's sense of nontrivial self­replication (Poundstone 1985). These patterns have functional properties similar to DNA and provide the motivation for the name Life. Let us imagine a possible world that is a Life universe consisting of an infinite grid. The two properties possessed by grid cells, "on" and "off", are the basic physical properties in the Life universe. The rules governing the grid's evolution are that universe's laws of physics. When thought of in this way, Life becomes a perfect modeling ground for understanding the notion of logical supervenience that we will discuss later.

Entities called gliders may serve as a simple example of how logical supervenience works in the Life universe. A glider consists of a sequence of patterns, each exactly five contiguous cells, that move across the screen in a characteristic fashion (see figure 2.2). Other cellular automata can also produce gliders, so we cannot define the property of being a glider using Life physics. Life can present sufficient conditions for the existence of gliders, but fails to be necessary for them.

The point that will be most important to the discussions in this chapter, and in the next few, is that the kind of sufficiency Life facts present for other kinds of facts is a very strong kind of sufficiency. It is not the kind of sufficiency that one might assert of causal conditions, as there's not even an air of contingency about it. Given a set of facts about Life cells, the very facts about what it means to be a glider go the rest of the way towards establishing the facts about the existence of gliders in the given situation. In this sense, it is not even conceivable that the facts about gliders could be different, given a set of facts about a Life world. So the kind of sufficiency being pointed to is a kind of logical sufficiency, in a sense of "logical" that determines what philosophers sometimes call "broadly logical" possibilities and necessities.

This is an issue that I will come back to later, but the sense of "logical" at work does not require meanings that are formally definable, nor does it require strict derivation within a formal system. In fact, gliders already give us an example of how reductive definition is inessential. The existence of particular gliders may be logically necessitated by Life facts in an intuitively clear way, and, just as clearly, the type is not definable in terms of structures specifically involving Life's "on" and "off" properties (as pointed out above, other cellular automata, with other kinds of basic properties, may support the existence of gliders).

Understanding these intuitively clear relations in a rigorous way is a project for a meta­theory. No doubt, difficult and subtle issues exist in the meta­theory here. These issues concern what meaning is, how well the behaviour of meanings may be captured by logical syntax, and how to formally understand logical sufficiency. The unsettled character of these philosophical meta­projects should not distract us from the fact that the anti­physicalist arguments are applied arguments. As applied arguments, what is most important to judging their soundness is adequate competence (i.e., skill), and not necessarily an adequately formalized meta­theory regarding what concepts and meanings are. Such competency issues should be addressed at the applied level. On the face of things, finding a completely adequate meta­theory for logic and meaning is no more pertinent to judging the arguments to follow than it is to judging other arguments, inside and outside of science, that require using various interesting concepts.

I will come back to these issues later in this chapter when I discuss just what the physicalist claim is. In the meantime, one of the purposes of this section is to illustrate the sense of logical sufficiency at issue via simple and clear examples. For instance, upon analysis, we see that all it means to be a glider is to have a certain structure and to exhibit a certain evolutionary behavior. The structure produces an orderly and predictable range of successive states that, lacking interference by other phenomena, glide across the grid. Once we have achieved this insight, we may easily see that there might be some circumstances in a Life world that would be logically sufficient to produce the existence of gliders. After all, the physics we are considering will allow for structure to arise, and for the evolution of those structures, and that is not difficult to see. To rule out the possibility that gliders could exist in a Life universe, we would need a specific proof that the physics was insufficient to produce them.

The kind of logical sufficiency being pointed to in the example of gliders is what I will mean by "logical supervenience" in what follows. We may accurately think of it as a kind of necessitation relation: the facts in the supervenience base necessitate the facts in the higher­level ontology by being logically sufficient to produce them. Now, could a complex organization of the properties in Life similarly necessitate the existence of consciousness in that kind of world? The case of gliders suggests a way of approaching this question. We can try to discover the most general classes of properties that can logically supervene on the physics of a Life universe. If these do fall into a few general classes then that fact could be a great guide to us. When trying to decide whether the Life universe could exhibit a property, we could try to give an analysis of the property in terms that one or more of those classes could satisfy. If we find that kind of analysis, then we must hold open the epistemic possibility that the Life universe could realize the property in question(1). Our forbearance would hold until presented with more specific proof, either an existence proof or an impossibility proof. Conversely, if the property in question avoids such analysis, then we must conclude that a pure Life universe could not realize it.

The properties possessed by the atomic cells of Life fall into three general classes: location, their "powers", and their ancestor/descendent relations. The class of causal powers consists of the properties of being "on" or being "off" (for the sake of argument, I will stipulate that being "off" designates a property). Let us call a property a causal role property if it meets two conditions. It must be a property of a basic object (Life cells, in this case), and it must play a counterfactually supportive role in the determination of the causal powers of other basic objects. In the present case, both properties of location and causal powers are causal role properties.

Additionally, a cell's current property may be an antecedent or consequent in a chain of causation to or from other cells. We can define an ancestor and descendent relation using the idea of counterfactual consequence. Let us say the state of a set of cells at a given time is just the conjunction of the states of its member cells. Imagine a set of cells X that is in some state at some given time, and a set of cells Y that was in some state earlier. The state of X at the later time is a counterfactual consequence of the earlier state of Y IFF a change to the state of Y at that earlier time would have led to a change in the state of X at the later time.

One can then define ancestor and descendant relations in the obvious way. Let <x,p,t> denote the state p of cell x at a given time t, and <Y,P,t> denote the state P of a set of cells Y at a given time t. A cell state <x,p,t> is a descendant of some earlier <Y,P,t­n> just in case <x,p,t> is a counterfactual consequence of <Y,P,t­n>. Ancestor relations are the converse of descendant relations. A consequence of these definitions is that they make a cell state at a given time a descendant of any superset of an ancestor, but that does not matter. Here, as with the definitions that follow, being too liberal will not undermine the basic work the analysis will do. If this is bothersome, one can tack on a minimal set condition.

What sorts of properties can exist in a world with just these properties as fundamental? The properties of the combinatorial structures of these basic cells will be supervening properties. For clarity, I will stipulate that structures are a set of contiguous cells in Life. 'Contiguous' means that every cell in the set is reachable from any other cell in the set without having to cross a cell not in the set. The morals that I am going to draw from this analysis will apply to messier notions of structure also.

The ancestral and descendant properties generalize easily to structures. What other interesting properties do structures have? We can see that the world will support other properties analogous to properties that physical things possess. Among these properties, two important classes should be singled out: the evolutionary relations between structures, and their interactive relations. An evolutionary property is instantiated by a structure when it is a member of an evolutionary sequence. For instance, a structure has the property of being a glider if its state is one of the states that occurs in the evolution of a glider. In the general case, an evolutionary sequence is a set of successive states each of which is sufficient to produce the next if no other phenomenon interferes with it. Many interesting phenomena, such as gliders, consist of many different structures linked through time, evolving sequentially from one another in this way.

The class of evolutionary properties is very broad, and all sorts of uninteresting evolutionary properties are specifiable. At every time step the whole grid has at least one evolutionary property, and any random region within a grid will usually have more than one evolutionary property. As before, this liberality should not matter to the later points. We can usefully analyze a whole host of interesting Life properties as evolutionary, from being a glider (which moves across the screen) to being a blinker (which moves back and forth horizontally and vertically) to being a traffic light (which twinkles like a star).

Some objects, like "eaters",(2) are interesting because of evolutionary properties in combination with properties like interactive properties. The idea of an interactive property casts a broad net over the whole array of effects a structure may have. Consider gliders again. As already related, gliders are structures of five cells that move across the grid as they evolve. During its evolution, a glider may meet other structures, and each interaction will produce some disruption or change to both phenomena. The range of possible interactions and results form the interactive properties of gliders. If we wished, we could represent any interactive property using a set of structures and their states, specifying points of contact, and describing the state of the structure that would result in the next time step.

Like with the other properties we have discussed, we can be very liberal about interactive properties. In fact, any given Life object will have an infinity of such properties. It will have one for every possible interaction it might have with other Life objects, and one for each description of the result.

Perhaps a bit surprisingly, we now seem to have sufficient resources for analyzing all the well­defined and really interesting, general classes of properties that can logically supervene on a Life automaton. For example, one can describe the history of any Life object as a sequence of different structures whose evolutions are punctuated by the exemplification of one or more of its interactive properties. Each punctuation results in a new structure, or set of structures. They begin to evolve, exemplify further interactive properties, and so on.

Supervenience, definition, and ill­defined properties ­ Something important about the above treatment is that even the well­defined properties are not reducible, in the sense of being theoretically definable, in terms of Life. These well­defined properties only supervene on potential Life automata, as Life may present sufficient but not necessary conditions for them. Additionally, Life objects that are less well­defined than structures will exist, and these will have less well­defined properties. We have to accommodate this vagueness in a way compatible with the idea that the Life world consists, at bottom, only of well­defined properties.

A proper analysis will proceed by giving accounts of the vague properties involved. Any successful analysis would connect an ill­defined property to the Life world by designating similarity classes of the well­defined properties, allowing members of this class to act as potential realization bases for properties in the ill­defined categories.(3) To see why, note that if one asserts of anything that it is an x or has property y one must be able to justify the assertion based on some aspect of its history, including its present. This implicitly reduces to the assertion that some such combination of well­defined basic properties is logically sufficient for realizing the less well­defined properties. Any adequate analysis of the vague properties will have to show how this can be.

Generally, the introduction of ill­defined properties by using similarity classes just bifurcates each of the basic categories. Each basic category becomes a genus with two species. The category of structural properties breaks into precise structural properties, as defined above, and vague structural properties realizable by a similarity class of well­defined structures. Interactive properties split into a species of precise interactive properties, as defined above, and vague interactive properties realizable by a similarity class. The other properties receive a similar treatment. Always, particular exemplifications of imprecisely defined properties occur through reference to a set of precisely defined properties. Figure three depicts the logical structure of the situation.

Two kinds of reduction ­ Like the well­defined categories, we can naturally construe the vaguer interactive and structural properties to include possibilities that do not involve specific Life facts and physics. So no theoretical reduction, in the sense of defining one category in terms of the other, is at hand, for either the vague or the well­defined properties.

Even so, in another sense of 'reduction', gliders may be reduced to Life facts in virtue of logically supervening on them. This other sense is that the existence of things like particular gliders may be wholly dependent on Life facts and physics, despite the fact that the category outruns definition in those terms. That is, the existence of particular gliders may reduce, even if their categorical being does not. Thus, theoretical reduction fails, but ontological reduction succeeds (pending a deflationary, intentional account of the categorical slack). Furthermore, even though theoretical reduction fails, the kind of reduction that succeeds clearly hinges, partially, on facts about meaning: the circumstances in Life may produce the existence of particular gliders by satisfying the concept that picks out the property of being a glider.

One might think that one could continue to define new sorts of interesting properties in the Life universe by considering systems. A system would be a collection of structures and take into account their relations. However, as we have defined structure, we can view a system as just a larger, more complicated structure. As a structure, a system also has only evolutionary properties, structural properties and interactive properties. Facets of its internal dynamics will be what makes it a system.



2.3 Could a Life universe possess phenomenal consciousness?

The challenge ­ Now that we have done this preliminary work, we can return to the problem of consciousness. Might qualitative consciousness arise in a universe that consisted only of facts satisfying the specifications of Life? This is a serious question. Life can exhibit phenomena of indefinite complexity. Who knows what properties structures of quadrillions of cells might exhibit and what interesting phenomena might arise? Can we rule out the presence of phenomenal consciousness in a pure Life universe?

In such a pure world, any kind of phenomenon that can exist will be, in principle, of a kind that allows Life to exhibit logically sufficient conditions for its existence. Ultimately, this means allowing for entailment by:

(1) Causal role properties

(2) Ancestral properties

(3) Structural properties

(4) Evolutionary properties

(5) Interactive properties

(6) Some overlapping combination of 1­5

Such a world may possess all sorts of "emergent" properties via combinations of (1) ­ (5), along with, perhaps, the addition of vagueness. For instance, we cannot rule out such a universe exhibiting some kind of genuine life. We already know it may contain self­replicating phenomena. That these entities might eventually lead to the existence of animate objects is at least an epistemic possibility. These objects (epistemically) might metabolize elements of their environment, act in a goal­directed manner, evolve to be increasingly complex, and, generally, possess a suite of functional properties sufficient for regarding them as alive. Given that life might exist, ecologies might exist. Given that ecologies might exist, even economies might arise in a Life universe. We can analyze economies into kinds of functional relations between objects within an ecological system, and functional relations are a combination of evolutionary and interactive properties.

So, overall, logical supervenience does not give us grounds to rule out many kinds of phenomena in such a universe. Nevertheless, I will argue that no pure Life world can necessitate the existence of consciousness, or the specific character of its phenomenal qualities. By a "pure" Life world, I mean a world in which the only fundamental facts are those completely describable by the physics of Life. The purity condition is extremely strong, and very important for the discussion to follow. The purity of our hypothetical Life world means that it contains no properties, features, entities, or aspects other than those laid out in the specification of Life's physics. Nothing, absolutely nothing, may be smuggled in. In particular, it cannot be imagined to be "implemented" by anything other than "pure" dispositions corresponding to "on" and "off" properties, whatever might be meant by this (the possibility of such a world is, indirectly, a concession to physicalists, and does not illegitimately help the defense of anti­physicalist arguments I give below).

Arguing for negative conclusions of this type requires producing reasons that justify believing that we cannot properly analyze the property in a certain way. At one level, such assertions will seem brute, simply relying on an adequate grasp of the concept that picks out the property. At other levels, we may approach the proposed analyses from a variety of more concrete perspectives, note the systematic failures, and the general reasons behind the failures. That is the strategy that I will pursue below.

To set the stage for a more vivid articulation of the point, I want to emphasize that we must remain constantly aware that the Life universe is not our universe. We are to imagine an alien dimension, a dimension fully described by Life's physics. No one can decide the question of whether any conscious feeling at all can exist in a pure Life universe by an appeal to first­person evidence, or analogy, or verbal reports. This takes out of play certain ground­level intuitions that affect the discussion about consciousness in our universe. For instance, we cannot claim that we will empirically discover a coordination between different kinds of descriptions (Flanagan, 1992), a coordination that will allow us to attribute an identity or supervenience relation between feeling and the functionality of Life objects.

Establishing such coordination requires us to access facts of both kinds, and the problem is precisely to access the phenomenal facts in a pure Life world, if any. Conversely, if we had access to the phenomenal facts, if any, we would obviously not need any process of "coordination" between them and other kinds of facts. Those other kinds of facts would have been the entailment base from which we obtained our phenomenal information. So the Life universe is alien to us, and only logical supervenience could bring consciousness into existence within it. To satisfy the requirement of logical supervenience we need to look for conceptual connections between the well­defined properties, or life history, of a Life object and the presence of consciousness.

The skeptic's claim ­ I will defend the skeptic's claim, which is merely this: we may consistently acknowledge any kind of structure and functionality for Life objects, and deny the presence of feeling in a Life universe. Thus, positive facts about consciousness cannot logically supervene on the facts about a Life universe. I will begin the defense of the skeptic's claim in this way. At bottom, the basic shortcoming seems to be that Life is only able to present facts about patterns of bare difference. A bare difference between entities is a difference that is merely formal, ungrounded in any further facts about internal structural differences between those entities, or internal relations of difference between a structureless intrinsic content possessed by those entities.

Bare differences are difficult to conceive of, and I believe it is reasonable to doubt the intelligibility of the notion. However, a pure Life world would seem to need such facts, and, as I will argue later, a purely physical world would also need them. So as not to win a cheap victory against the physicalist, then, I am going to grant the intelligibility of these kinds of stipulated bare differences between primitive dispositional properties. I think the best way to conceive of a bare difference between two properties, x and y, is to think of the relation as primary, with the existence of the relata, such as they are, derivative on their participation in an ungrounded relation of difference. In contrast, a contentful notion of difference reverses the order of primacy: the existence of the difference is derivative on further facts of intrinsic difference between the natures of the relata. A pure Life world must instantiate bare differences between the "on" and "off" properties, as a contentful difference requires smuggling in further intrinsic facts about the "on" and "off" properties, facts aside from those stipulated by Life's physics. This bareness is what allows them to exist as pure dispositions.

The skeptic maintains that facts about bare difference are always consistent with the absence of feeling, since the difference structures instantiated by phenomenal consciousness involve differences in particular, qualitative contents. Our taste space, for instance, contains different tastes, and our color space contains different colors. The relevant premise is that these tastes and colors are contents instantiating a structure of difference relations, not structures instantiated merely by difference relations.

No one denies that we can catalog the differences between different colors and different tastes along relevant dimensions. If we do this, we can surely abstract out a content­free difference structure. The skeptic's objection is to the further move of analyzing conscious qualities into the abstract patterns of difference between them. Our acquaintance with the phenomenal qualities yields information about them as contents occupying slots within these difference structures. Reification of the difference structure as basic requires ignoring the content instantiating those structures. That move ignores the grounding of those differences in this specific case.

Qualia as observables ­ For the skeptic about feeling in the Life world, these conclusions about the elements of consciousness are not mere "intuitions". Colors, tastes, sounds, and other sensations have the status of observables, and these "intuitions" have the status of observations. As observations they might be wrong, but it takes very powerful arguments to overturn them.

The claim that qualia are observables is controversial to some. The most common worry is that modeling our knowledge of qualia on perception is misleading, so people are unsure how we can be "observing" them. Some people argue from these kinds of worries to eliminativist conclusions, but, minimally, opponents sympathetic to these worries hold that the knowledge grounding the skeptic's conclusions is highly refined, theoretical and corrigible.

To these worries, the skeptic replies that the objector seems to have an unreasonably narrow concept of observation. By insisting that something can achieve the status of an observable only if we obtain the information about it through perception, the objector is making too strong a claim. The physicalist's objection rules out of court a huge amount of information about consciousness that we have access to, and that a theory of consciousness should have to explain. As examples, here are two statements most would agree express facts that should have the status of observables:

(A) Last night I thought about my childhood.

(B) Sometimes I think about my childhood when no one is around.

By calling these things observables I mean that they meet these four conditions: (1) They belong to a type whose members are potential objects of awareness; (2) We can become aware of them without the aid of special instruments; (3) The dubitability of our belief in facts of the relevant type is almost zero; (4) Our awareness of instances of the type is reliable. These characteristics allow facts like (A) and (B) to attain the status of useful falsifiers for scientific and other theories. For instance, a theory of mind fails to account for some of our evidence about ourselves if it fails to account for how we can sometimes think about our childhoods when no one is around.

The skeptic about the possibility of consciousness in the Life world is claiming only that facts like (A) and (B) are no more problematic than many other facts we count as "observable". The fact that perception does not mediate our awareness of them seems to be a red­herring. Thus, if anything can count as an observable, (A) and (B) both can. Our skeptic firmly insists that a science of consciousness must allow such observables if it wants to be treated as legitimate. Since facts like (A) and (B) turn out to be no more problematic as observables than perceptually mediated facts, a straightforward argument delivers the phenomenal qualities as observables also.

Last night, I lacked behavioral evidence that I was thinking about my childhood. I was not writing about it, nor talking about it, nor acting on it. I was, in fact, scouring my bathtub. How do I know what I was thinking about? What was the evidence of my thoughts? It was the presence of certain kinds of conscious phenomenal imagery, verbal, imagistic, and kinematic. That imagery may have been identical to the thoughts, or it may just be a concomitant of thinking that gives evidence for thoughts.

In either case, my awareness of the conscious phenomenal imagery cannot be considered more dubitable than my awareness of my thoughts. Since the phenomenal imagery is the evidence for such thoughts, it is easy to argue(4): the sentence (A) has the status of an observation claim IFF the phenomenal imagery that is my evidence for it has the status of being an observable. Similarly, I obtain my knowledge of types of thoughts like those referred to in sentence (B) from observables IFF I also obtain my knowledge of types of phenomenal qualities from observables.

Arguments such as this, the skeptic maintains, establish that we obtain knowledge of what the phenomenal qualities (colors, feelings, sounds, imagery, other sensations) are like through conscious awareness of them. For example, I obtain my knowledge of what the shades of blue look like to me by consciously experiencing them. Consequently, phenomenal qualities are observables (which is not to say observation of them is always either easy or incorrigible). Through observation the skeptic obtains phenomenal information that explanations and theories must be held accountable for.

The fact that awareness of these things is not perceptually mediated should not count against them. For centuries we had no idea how perception worked, but that did not stop us from treating information delivered by perceptual awareness as being information about observables. Similarly, the lack of a sound theoretical understanding of non­perceptual awareness should not stop us from recognizing that it also delivers information about observables.

David Lewis (1995) has argued that physicalists can grant the existence of qualia, but only if they deny that we have special, unmediated access to their natures. The claim that qualia are observables does not violate Lewis' injunction. To possess phenomenal information, our skeptic does not need to have a more direct access to qualia than to any other kind of observable. The skeptic is chiefly concerned with the character of the connection between phenomenal qualities, as disclosed through the phenomenal information we do have available, and their hidden natures, if they have hidden natures. Does the phenomenal information we have about qualia logically supervene on the properties of their hidden natures, or is there an opaque, nomic connection? The skeptic merely claims that if the hidden nature of qualia is that of a structure of bare differences, the connection between our phenomenal information and this nature is not one of logical supervenience.

The character of the supervenience base in Life ­ If we can show that the pure Life world can only support bare difference spaces, and patterns of bare difference spaces, the phenomenon of qualitative consciousness would then seem to avoid analysis into the possible categories of explanation available. Any proposed solutions to the problem of consciousness that do not go beyond Life's resources would necessarily be from a different game altogether, and not even eligible to win this one.

Our skeptic can make this concrete case by taking a closer look at the materials available in the Life universe. What does it mean to be an "on" or "off" property? The only two requirements are that (i) they be distinct, and (ii) they should be instantiated in patterns conforming to the counterfactual dependencies described by the laws. In short, the distinction between being "on" or being "off" is a merely formal one. "On" and "off" designate bare, content­free difference.

The Life specification is, at heart, a structural schema for a universe. It specifies certain patterns of contrast between being, patterns that must hold for a universe to count as a Life universe. As we ascend to higher levels of organization in the Life world, we do not escape from the circle of bare difference. Appending schemas to schemas only yields more schemas. A Life structure is a pattern of bare difference, mere contrast.

In Life we have a world potentially consisting of a huge number of simple, bare differences lying side by side, with reliable, regular transitions between them. Does this world present logically sufficient conditions for the existence and specific nature of a higher­level consciousness with qualitative, subjective content? We know from observation that the difference between red and green is not a mere formal difference. The question is therefore equivalent to the question of whether or not qualitative contents like the shades of green are patterns of bare difference.

A pattern of differences between colors can produce another color. For instance, a field of tightly packed yellow and red dots may yield an experience of orange under the right viewing conditions. We can observe that the shade of orange that results is not produced by the mere pattern of difference, though. It has to be a pattern of difference between the appropriate colors, thus providing no explanation of color in terms of mere patterns of difference. If we try to abstract the pattern of difference from their contentful bases, viewing them as mere difference structures, we see that the result is multiply realizable and some of the realizations do not yield orange.

For instance, one can instantiate the same structure of differences between two other colors whose hues lie at the same distance from each other as red and yellow (e.g., yellow and green). A pattern of dots of these colors will yield a different color from orange (e.g., brown). Therefore, we can observe an identical structure of formal difference, but different colors. The fact that, even allowing that we start with colors, one cannot reduce some colors to the mere difference structure among other colors is suggestive. After all, the skeptic is maintaining a much weaker position.

The position the skeptic is defending is that patterns of bare differences do not entail the facts about the phenomenal qualities. Patterns of bare differences are difference structures whose identity obtains because of a mere formal difference, ungrounded by content. The skeptic notes that orange cannot even be reduced to the structure of difference between red and yellow once we abstract from the phenomenal content of red and yellow themselves. We can observe more straighforwardly that red and yellow are not constituted by patterns of mere difference, without any content at all.

As an analysis of phenomenal content, the idea that something like a shade of red is a pattern of bare, merely formal differences is observationally inadequate. To make it work, something must be added. The only tool Life presents for constructing phenomenal content out of patterns of bare difference is its counterfactual content. Unfortunately, known logics of counterfactuals only add things like consistency constraints, or metrics over similar possible worlds, to our toolkit. These are not even the right kind of thing to add to a collection of formally distinct properties to make them add up to properties that are not merely formally distinct. A pattern of bare differences does not become a phenomenal content because another world contains a similar pattern, or because it is consistent with patterns that occur elsewhere in that same world. Yet that is all we have here. If one tells a skeptic that a pattern of bare differences transitions to another pattern of bare differences, the skeptic can consistently deny that either pattern has to support a feeling. Nothing in the logic of counterfactuals requires that the transition should feel like something either. The Life schema thus seems to underdetermine the story about qualitative content. We seem to have good reason for believing that the skeptic's claim is consistent.

Transferring information between worlds ­ Why might someone believe that the skeptic's claim is inconsistent? Sometimes unwarranted assumptions predicated on beliefs about our world might tempt us to smuggle phenomenal facts into the Life universe. Qualia in our world may perform some functions, or correspond to some specific internal structures or processes. Perhaps performance of these functions is necessary to their being objectively classified as pain feelings rather than feelings of another type (phantom pains, maybe). Maybe, in our world, performances of these kinds of functions are always accompanied by the feels.

Even if we grant ourselves the assumption that phenomenal feelings perform some functions in our world, a further question looms. Are we telling the complete story about experiences when we tell the story about their function? We have to avoid automatically assuming that a similarly functioning element of the Life world feels pain. That is to beg the question about the completeness of the functional explanation in our world. More strongly, the Life world presents us with an opportunity to investigate that question. What phenomenology, if any, would similarly functioning objects in the Life world logically require us to attribute to them?

At that point, the alien character of the Life world presents a problem. Given our remove from that universe, we have no first­person knowledge that even one conscious state exists in it. Without even that small bit of information, making grandiose generalizations about the existence of consciousness across these alien worlds, ours and the Life world, becomes tricky business. We need some supporting story about how the facts in this alien world can be sufficient to support facts about consciousness. This supporting story cannot assume the existence of consciousness in that world

We clearly need a supporting story that goes beyond the mere coextensiveness of facts in our world if we want to generalize from our world to a hypothetical Life universe. Our alienation from it seems to block transferring the information so naively. At this point, the existence of an explanatory gap in our world, admitted even by many physicalists, is evidence that mere coextensiveness, or "coordination", is all we really have. If this is so, then the functional information alone cannot be the whole story. It seems to follow that a skeptic is consistent if he admits to any kind of functioning at all in the Life world, and denies that the activity supports consciousness.

Adapting anti­physicalist arguments to Life ­ We can also approach the conclusion through meaning, since the type of logical necessitation we are looking for should have a grounding in meaning. Whatever pattern of events might occur in the Life world, it is very implausible that such patterns satisfy what it means to be conscious. As evidence for this, one may observe how holding that Life physics can support consciousness seems completely ad hoc. Why assert that anything possesses a subjective, qualitative aspect in a Life world? What work would the claim do? It does not seem to explain anything that cries out for explanation within Life's closed world. (this observation corresponds to the sort of argument found in Stapp 1996).

The only temptation one can possibly have is an extension of something one might believe about functioning in our world into another world very different from ours, one where we have no first­hand evidence about the existence of phenomenal consciousness at all. Since any kind of interactive, evolutionary, or structural conditions that a Life world meets are already complete, albeit extrinsic, we are left with an explanatory gap, "Why should this combination of Life properties feel like anything at all?" Such an explanatory gap could not exist if all it meant to have a conscious experience was to instantiate a complicated pattern of formally distinct differences.

The skeptic could also weave a tale about an imaginary super­Life scientist who knew all the facts about possible Life phenomena. He might ask, "should such a scientist conclude that there is a quality of experience for certain Life phenomena?" We know that to justify a consciousness claim all our Life­scientist could point to would be instantiations of properties from the well­defined basic categories that logically supervene. Since the facts about these aspects of consciousness cannot be discovered by knowing the facts about interactive properties, structure, or evolution in the Life universe, the claim goes beyond anything that can logically supervene. The super­Life scientist could never justify a positive claim. Thus, concluding that qualitative consciousness exists in that universe would be illegitimate for the Life­scientist (this argument adapts Jackson's knowledge arguments from his 1982 to our Life world).

The skeptic might ask if we can conceive of a universe obeying all, and only, the Life physics, one which exhibited properties as complex as you please, but which did not have qualitative consciousness at all. The skeptic might help us by pointing out that the extrinsic and formal specification of the Life schema is presumed to completely circumscribe the world. The extant understandings of what makes for truth­conditions on counterfactuals, and of what a formal distinction is, seem to straightforwardly allow for the conceivability of this zombie Life world (this argument adapts Chalmer's Zombie arguments in his 1996).

The skeptic might challenge us by asking,"What is it like to be a (given) Life object?" By asking this, the skeptic intends to call our attention to our irremediable ignorance about the real character of the qualities we might be attributing. If they truly logically supervened, such severe ignorance would not exist (this adapts Nagel's worry in his 1974).

If some people insisted that phenomenology could exist, for instance visual phenomenology, the skeptic might correct them by asking if they could conceive of the visual experiences of Life objects inverted in a variety of ways. Perhaps these critics imagine that the "on" and "off" properties of a realized Life universe would not merely be formally distinct, but would have to be implemented by something with an intrinsic nature (someone might plausibly claim that they cannot make sense of the idea of a mere formal distinction, ungrounded in intrinsic difference)(5). If so, this is a further fact relative to the Life specification. One way to get such critics to consider the possibility of inverted spectra is to point out that the Life specification really does need only the bare distinction between the "on" and "off" properties. As long as they are distinct, and play their roles, the intrinsic natures they might have are irrelevant to the question of whether or not it is a Life universe. Since their intrinsic distinctness should remain even if their job description changes, we should be able to imagine inverting whatever intrinsic natures are supposed to be playing those roles. The Life universe is, in principle, multiply realizable. From this, the possibility of inverted spectrum seems to threaten.

These thought experiments are just vivid ways of trying to make the same general point: qualitative consciousness just does not logically supervene on Life facts and physics. Regardless of how successful one thinks these kinds conceptual tests are when applied to our universe, the Life case seems much more clear cut. We can see the answer directly by considering the kinds of properties that can logically supervene in that universe, and noting that there is no analysis of phenomenal consciousness that allows such properties to entail its existence.

In the end, these arguments for the skeptic's claim do not have the form of arguments from ignorance. They do not have the form, "I cannot imagine how such­and­such could possibly explain consciousness, so such­and­such cannot explain consciousness." They have the form, "We have reasons for thinking that such­and­such can only explain facts of type Y. Since we have observational evidence that the facts of consciousness are not facts of type Y, we have good reason to believe that such­and­such cannot explain the existence of consciousness." Its essential form is an argument from a theory to that theory's failure of prediction. To paraphrase Dennett, a perception of failure is not the same thing as a failure of perception. The successful perception, in this case, is of a failure of prediction.

1. Epistemic possibility is the widest space of possibility, including the space of logical possibilities as a proper subset. For instance, given a mathematical conjecture that has not been proven, we must hold it open as an epistemic possibility that it is true, and also that it is false. Still, only one of those is a logical possibility. I suggest thinking of epistemically possible situations this way: a situation S is epistemically possible for an agent A just in case A cannot discern an inconsistency in the description of S.

2. Eaters are structures that can "absorb" other structures.

3. The questions about how these similarity classes might be specified are interesting. Rosenberg (1997) makes some foundational observations about how circular conceptual systems might pick out such classes.

4. The relevant premise for the argument is (P): If x has the status of being an observable, the evidence for x must also have the status of being observable. For example, if I can observe that it is cold outside based on the evidence that there is snow on the ground, the snow on the ground must also be something that I can observe. Similarly, if I observe a photon in the cloud chamber based on the evidence that a cloud has appeared, then the cloud must be an observable also. The premise (P) gains its plausibility from the principle that the epistemic status of evidence cannot be less secure than the status of that which it is evidence for.

5. In chapter fourteen I will pursue this sort of objection in detail, finally endorsing it.




Chapter 3

A Formulation of Physicalism

3.1 How to get nonphysical facts "for free"

The role of logical supervenience ­ In chapter two I did a lot of work analyzing Life. In section three of that chapter I argued at length that positive facts about phenomenal consciousness do not logically supervene on facts about Life. To understand why doing this is helpful, we need to become clearer about what the physicalists' claims commit them to. We can then home in on the precise nature of the anti­physicalist qualms. I will not go into full detail here since that has been done well in other places (notably Jackson 1993; Chalmers 1996; Kirk 1994 and Poland 1994 do so as physicalist sympathizers), but I wish to at least outline the intuitive idea and the way it must be captured within a principled metaphysical framework. The basic intuition behind physicalism is that the particular physical facts in the world somehow exhaust, or fix in a very strong way, all the particular facts in the world. These intuitions are more fundamental than identity claims because identity claims must fail if these intuitions fail.(1) Every fact fixes itself in the requisite sense. In general, the physicalist must claim that for free connections exist between the particular physical facts in our world, and the particular facts of other ontologies. In David Armstrong's phrasing(2), given all the microphysical facts about our world, all the other facts are an ontological free lunch.

Essentially, its proponents ground the physicalist program on the possibility that the high­level description of any particular fact in economic (or psychological or sociological or . . . .) terms is just a re­description of some physical circumstance using another, more abstract conceptual structure. As a different conception of already existing physical stuff, such facts require no fundamentally new ontological posit. To express this insight, the physicalist has enlisted supervenience. The basic idea underlying supervenience is that if the truth of token facts of one type, called the base facts, in the right way guarantees the truth of token facts of another type, called the supervenient facts, then we require only the base facts in our most fundamental ontology.

The basic idea behind supervenience is simple but, as it stands, will not serve the physicalist's purposes. It has several variations, and some of these variations are compatible with dualism. The nuances we must get comfortable with relate to the modality and the scope of the supervenience claim. Supervenience may come in two relevant modalities, natural and logical(3), that are sometimes not distinguished from each other.

Natural supervenience asserts that the base facts determine the supervenient facts in virtue of some law of nature connecting the two. We can paraphrase the nomic supervenience claim as follows. Let the base facts be the A facts, and the supervenient facts the B facts. Then,

The particular B facts naturally supervene on the particular A facts IFF the laws of nature ensure that whenever the particular A facts hold, the particular B facts will also hold.

A short (and potentially misleading) way of stating the natural supervenience relation is to say that the base facts cause the supervenient facts. Clearly, natural supervenience is not what the physicalist has in mind. Almost any anti­physicalist will freely admit that laws of nature exist to connect the physical facts of the body to the phenomenal facts that are claimed to be nonphysical. The problem here is that laws of nature are contingent relative to the facts they connect, and contingent connections cannot support ontological reduction. As an example, consider the connection between gravity and electromagnetism in physics today. Nature connects them only naturally, and therefore we must consider them quite independent and equally fundamental forces. Not distinguishing natural from logical supervenience sometimes gets philosophers into ontological trouble. Searle's position, for instance, seems to be one of natural supervenience and yet he surprisingly believes he is defending ordinary physicalism (1992).

How can the physicalist get the requisite "for free" connections? We can think of it this way. What, in a physicalist world, allows us to assert statements such as, "There are tables."? The ontological category 'tables' is not part of the fundamental physicalist ontology. Such statements are true just in case the conditions existing in our purely physical world are such that at least one object meets the application conditions of the concept 'table.' The physicalist can then say that the world realizes tableness, exhibits an instance of it. In other words, tables exist in a purely physical world because it presents logically sufficient conditions for predicating 'table.' Whether or not the physical world can support such a predication is a matter both of empirical fact (what the particular physical facts happen to be), and the meaning of the concept (do those particular facts present logically sufficient conditions for satisfying it).

That is the sense in which the physicalist can have tables in the world for free, and dollar bills and inflation as well. So, given a full understanding of the physical facts, the logical room for the high­level facts to vary must disappear. This is the sense in which the particular physical facts necessitate the particular economic facts. This logical room is given partly by the concepts involved, and it involves truths of meaning. Indeed, how else can the physicalist justify the claim that economic facts are nothing new over and above the physical facts since truths of meaning are the only things a physicalist can possibly hope to get for free? We say, therefore, that the physical world makes real (or realizes) the particular high­level fact. So to account for given high­level facts, the physicalist needs to uphold the formula physical facts + conceptual truths yield high­level facts.

An example of logical supervenience at work would be the determination of the shape of a Rubik's cube. Does its (roughly) cubical shape logically supervene on the physical facts about it? The answer seems to be clearly 'yes.' Given all the physical facts about a Rubik's cube, facts that include all the spatial relationships between its component parts, a sufficiently intelligent being would be able to determine that it is (roughly) cube shaped. Its shape is entailed by the physical facts alone with no further ontological suppositions being required. It follows that there is no possible world where all the physical facts about the Rubik's cube are the same, but its shape is different.

One final wrinkle: the supervenient properties of an object may be either local or global in character. If supervenience is local, then the base facts of that object alone are sufficient to set the supervenient facts. Facts about the heat of the sun are of this character. If global supervenience holds, then the supervenient facts about an object also require the base facts involved in its context. Economic facts are globally supervenient. For instance, the fact that a particular piece of paper is money is due to a complex social and historical context that it is embedded in. The same goes for the particular fact that Trey Kirven paid $4.71 for his lunch yesterday at 12:51.

3.2 Formulating Physicalism

Generally, I will say that a fact, S, logically supervenes on a set of facts B just in case there is no logically possible world in which the B facts are true and the S fact is false. Now, this automatically means that all necessary truths, such as mathematical truths, logically supervene on the physical for trivial reasons. Physicalism should not rule out Platonism by fiat, so I will leave open the relationship between the physical facts and truths about abstract entities. With this in mind, we can now state a very weak version of the physicalist position: Physicalism is true IFF all the contingent, particular, positive facts in the world are globally (spatially and temporally), logically supervenient on the particular physical facts including, perhaps, the physical laws.(4)



Some things that logical supervenience is not ­ A typical reaction to the discussion so far is to claim that very little logically supervenes on the physical, as the "logical" criteria is far too strong. From the anti­physicalist side, Chalmers (1996) addresses this reaction by sketching examples of how one might establish logical supervenience in particular cases,(5) as does Jackson (1994). They also offer some general considerations concerning why logical supervenience should hold almost universally. From the physicalist side, Kirk (1994) also defends the claim(6), as do Horgan (1984) and Armstrong (1982). For more detail regarding the positive case, I refer interested readers to these authors. Below, I will address the common basis people's doubts seem to rest upon.

Why do some people believe that logical supervenience fails quite generally? Commonly, the requirement that one set of facts follows "logically" from another is thought to be a very, very strong requirement, too strong to capture the general relationship between the high­level facts and the physical facts. The chief source of this doubt seems to reside in the impression that the notions of "logical", "consistent", and "analytic" must be understood in terms of syntactic derivability within a formal system. For example, some people believe that the supervenience conditionals at issue here could only be "analytically true" if one could produce a syntactically well­formed deduction of the consequent facts (i.e., the supervenient facts) from the antecedent facts (i.e., the base facts). To these people, it seems that the A­facts could logically supervene on the B­facts only if one could produce definitions of the A­facts in B­fact terms, and then use these definitions to syntactically derive the truth of the supervening A­facts from the base of B­facts.

The inevitable failure of most high­level facts to meet this condition seems patently obvious. They should fail to meet the syntactic derivability condition for one (or both) of two reasons: (i) most of the high­level concepts at issue will evade perspicuous definition altogether; or (ii) to the extent that one might be able to define any high­level concept, the definition will not be expandable to purely physical terms, and so will not be in a form appropriate for use in a derivation from a purely physical base of facts. I will not take issue with either of these worries. Instead, my strategy in answering the doubts they raise will be to argue that these worries arise because the successes of modern logic have caused a too strong identification between theory and phenomena.

The relevant observation is that logical consequence, analyticity, and consistency are semantic phenomena. The formal syntactic representations that inform our modern understanding of these notions are merely theories invented to help us explore the phenomena. They are tools for modeling them, nothing more and nothing less. The apparatus of these theories is extraordinarily useful, but also incomplete in important ways. Because they embody theories, both the models built with these tools and the tools themselves must answer to the informal competence we possess (or may develop) with the use of our concepts, with meaning, with consistency, and with logical consequence. Only if they pass these tests may we accept them as completely adequate theoretical tools. Thus, to raise this objection against logical supervenience requires defending some very strong claims about the adequacy of logical syntax to capture informal semantics. These claims are that (i) the syntax of logical definition adequately models the structure and behavior of meanings for non­primitive concepts; and (ii) the logical satisfaction of concepts is adequately analyzed within formal logic.

I believe few people, when the point is put so starkly, will wish to defend these claims. The phenomena that argue against them are ubiquitous, and difficult to deny. In the end, resisting the counter­examples is self­defeating, as it requires abandoning the confidence in our native competencies that we need to justify accepting these modeling tools in the first place. I will argue primarily against claim (i), that the syntax of logical definition adequately models the meaning of non­primitive concepts, because I believe that claim (ii) fails if claim (i) fails.

To argue against claim (i), one may point out that it is in conflict with one important source of the original skepticism about logical supervenience. It conflicts with the claim that most interesting high­level concepts do not have definitions. As an example, consider the concept of "friendship". Perhaps some incisive analysis will yield a crisp, formally modelable definition, but, for the sake of argument, let us assume that no such definition will ultimately be forthcoming. Few would want to claim that the concept of "friendship" is, as a consequence, meaningless, nor, I think, would many want to claim that it is a primitive term. "Friendship" is a meaningful concept whose meaning is neither primitive, nor completely analyzable in terms of a formal definition involving other concepts related to one another by the logical constants. This is trouble for someone needing to defend claim (i) above, as it seems to suggest that meaning outruns the resources of formal definition.

"Friendship" does seem to be just this sort of concept. However, do we need to conclude from this that notions of logical consequence and satisfaction do not ever apply to it? That seems like an extremist reaction. When one considers concrete cases, some conditions certainly present themselves as being entitled to the claim that they logically satisfy the concept of friendship, irrespective of its resistance to formal definition. To see this point, imagine two people, Allen and Gregg, who have known one another for twelve years, genuinely liked one another, shared many experiences, secrets, and adventures, go out of their way to be in one another's company, and who rely on one another for advice and support in times of stress. What would we say to someone who, in full knowledge of these facts, nevertheless claimed that Allen and Gregg were not friends. The most straightforward and sensible answer is to assert that this skeptic does not understand what it means to be a friend.

In any case, the dialectic hinges on an adequate grasp of the concept. Should the skeptic produce contravening evidence, it will be contravening precisely because we may recognize the evidence as inconsistent with what it means to be friends. Should no such evidence exist, we will conclude that the skeptic simply does not have an adequate grasp of the concept, and therefore cannot see what is apparent to everyone else: that Allen and Gregg's history together, and feelings towards one another, imply the fact that they are friends. The capacity to represent the meaning of the concept in a formal logical system plays no role in establishing its ability to be satisfied by sufficient conditions. These conditions are an entailment base consisting in complex circumstances, and yield "broadly logical" conditions for its application.

Reliance on this informal competence, and these informal entailments, are not something that science has been able to transcend. When one finds "definitions" of ordinary concepts in science, they are always acceptable precisely because they are attached to potential elements of theory that intuitively satisfy the informal concept. They are, ultimately, backed by this native competence. For example, early versions of the atomic theory could not explain why objects were solid. Arguments for this failure, like the failure of the pure Life world to support consciousness, involved a pre­theoretical competence with the concept of solidity. Especially, judgments concerning the adequacy of proposed operational definitions, definitions involving criteria such as the ability of a body to retain its shape, depend on this informal competence (otherwise, how would we know that it failed to explain solidity?)

Finally, notice that these "reductive definitions" really use transcendental concepts such as "shape" and "body", and the "definitions" of these concepts in physical terms are thinly veiled conceits. What such "definitions" actually encode is an explicit recognition that some elements of the theoretical framework are logically sufficient for applying the concepts. Not many people seriously believe that the truth of any specific physics is a necessary condition for the existence of shapes or bodies. Indeed, physical space may not even be required for something to have shape if it turns out, as it might, that visual qualia are nonphysical, and non­spatial.

To see the importance of informal competence, let us consider whether the arguments against the logical supervenience of consciousness within Life would be improved by using a definition of qualia. Here is one candidate that is as neutral as one could hope for, and not intuitively repugnant:

Qualia ­ Qualitative contents occupying structural niches in an information space.

Does introducing this definition help us to decide the soundness of the earlier arguments? Obviously, how one understands "qualitative content" and "information space" will make the crucial difference here. In particular, can one further define these concepts in terms that Life facts may satisfy? Life facts can surely instantiate "information spaces", so it is not prejudicial in that way.

The real sticking point will be trying to understand "qualitative content". In an attempt to try to formally derive the existence of qualia in Life, one may propose all sorts of definitions of "qualitative content" in Life terms, but it is obvious that this will be little help. The problem is that the proposed definitions themselves will have to stand up to intuitive scrutiny. We will want to know: Do they really capture what we mean by "qualitative content?"

Together, all the above points make a strong case against the adequacy of formal, logico/syntactic definitions as tools for fully representing meaning. If our current formalisms fail in that, then it follows virtually immediately that formal logic has not adequately represented satisfaction either. From that, it follows that we also do not have an adequate formal representation of logical supervenience or analyticity. These conclusions should not be big surprises. They are just what one should expect given that we do not really know yet what concepts are, how they are structured, or what meaning itself is. These issues lurk in the meta­theory of all sciences, and all argumentation, and they do not present special problems for the arguments used in the science of consciousness.

The moral of all these points is simply this: the relevant notions of satisfaction, logical sufficiency, and analyticity are semantic, not syntactic. We do not yet have an adequate meta­theoretic representation of semantic behavior in terms of syntactic form, and rules on those forms. As we observed with the concept of "friendship", a concept's ability to be pounded into the mold yielded by these incomplete theories is inessential to our invocation of logical supervenience. An appeal to failures of that sort conflates theory with phenomena, projecting the inadequacies of our modeling tools onto the phenomena they are supposed to help us model.

As I have been at pains to emphasize already, what the anti­physicalist arguments depend upon is applied competence at the object level, not misapplied meta­theory. Every theory, argument, and explanation uses concepts of one sort or another. We should not have to possess a fully adequate theory of what concepts are before we are allowed to use them. Unless it turns out that the anti­physicalist arguments are relying on concepts and meaning in a way that is different from scientific explanation and understanding generally, then these open questions within the philosophy of language, although immensely interesting and important, will not be specially germaine to the issues here.

For logical supervenience to hold, the supervenient phenomenon must be analyzable into a set of categories that the physical facts can logically fix. Just on the face of it, just looking at the position statement, one should suspect that physicalism must be open to some kind of conceptual constraint, for it relies so heavily on our concepts to bootstrap us from the ontology of the physical to ontologies of other kinds. At the very least, we can ask just what kind of facts the physical facts can logically fix. What are the kinds of facts that the physical world can present logically sufficient conditions for?

3.3 Getting back to the real world

Life imitates physics ­ Where I am going with all this should be obvious by now. The idea behind cellular automata is quite general, and the considerations about the kinds of properties that can logically supervene on them is not limited to anything special about Life. In fact, since a Life grid can imitate a Universal Turing Machine, we can say that, at a suitable level of abstraction, the Life world can reproduce any general kind of property that can logically supervene on any cellular automaton.

We are now able to see the root of the anti­physicalist qualms with physicalism more clearly. Cellular automata very closely capture our concepts of the physical world and physical properties. In fact, it is not too difficult to imagine that our world might be, in fact, a giant cellular automaton, albeit one with very complicated stochastic causal role properties. By using genetic algorithms to discover evolution rules, researchers at the Santa Fe institute have even discovered automata that produce particle like elements. An automaton can use these particles as information bearing elements useful in solving problems encoded in its initial state (Das, Mitchell, et al, 1994).

Even if the concept of a cellular automaton does not perfectly capture our notion of the physical world, our concept of the physical is sufficiently close to that of cellular automata that it seems as if the same explanatory restrictions apply. They seem to be the same in the relevant respects. The differences lie in such things as the complexity of the causal role properties and laws, the mobility of cells, the number and kind of dimensions the cells exist in, and perhaps non­local causation. In our world, the structure of the basic entities is more exotic. Instead of squares, we have particle­waves and fields, maybe thirty­dimensional strings, or other such exotica. Perhaps causation in the physical world requires infinite calculation, and so a Turing machine cannot simulate it.

These differences add degrees of vagueness to the notions of structure, interactive property, etc. . . . but do not seem to make a fundamental saving difference in kind. The failure in the Life universe does not seem to arise from the fact that the basic objects were squares rather than strings, or that the causal role properties were related simply and locally rather than complexly and non­locally. Rather, the failure was rooted categorically in the stark geometric and bare counterfactual nature of the properties and of the world they made.

So it seems that any phenomena that logically supervenes on the physical must be analyzable into one or more of the basic classes of properties -- properties of location, causal role properties, ancestral properties, structural properties, evolutionary properties, interactive properties -- or some combination of them. Again, the complexity of our universe introduces some vagueness into these general concepts, a vagueness that our specific concepts mirror. Still, that vagueness hardly seems like the kind of thing that will allow us to escape the trap.

The intrinsic properties of the physical ­ One disanalogy that some people find obvious is that physical things have an intrinsic nature, whereas the entities and properties in the Life world are described in a way that is neutral regarding such intrinsic facts. These facts about the intrinsic nature of the physical, many feel, may be responsible for the existence of phenomenal consciousness. If this is so, proponents of this objection argue, consciousness would be physical after all.

I am sympathetic to this sort of view, and defend something like it in chapters fourteen and fifteen. However, I believe it fails to salvage physicalism, yielding, at best, a dual aspect theory in which nature has both extrinsic physical and intrinsic phenomenal aspects. The difficulties this proposal meets in saving physicalism come at three levels.

At the first level of reply, one may point out how difficult it is to show that physical things must have intrinsic properties. All that our best physical theories describe is a network of effective dispositions, each element typed according to its place in a network of relations to other such dispositions. This kind of purely relational world intuitively feels absurd to some, but arguments showing it to be incoherent do not seem at hand. The arguments I will produce later against the purely relational view are only plausibility arguments. They are intended to more fully articulate the intuitive basis for rejecting the proposal that the physical things of our world could be like the physical in a pure Life world. The claim that intrinsic properties must carry the effective dispositions described by physics ultimately must be added to theory as an intuitively justified axiom. As such, it stands out as a primitive further fact relative to our scientific knowledge of the physical

At the second level of reply, one may point out the gap between the observation that the proposed intrinsic properties are properties of the physical, and the conclusion that they are physical properties. After all, not all properties of biological things are biological properties, nor are all properties of economic things economic properties. The best case scenario for the physicalist is that the intrinsic properties of the physical (if they exist) are properties that science indexically designates, via their relational profiles, using physical predicates such as mass, charge, spin, and so forth. Unfortunately, some of the facts about what science designates, such as facts about intrinsic natures, may be different in kind from the facts that science specifies. Since physical theories are only committed to the existence of the facts they specify, the commitments of our physical ontology incompletely catalog the world's properties.

Two arguments support this way of looking at the issue. The first argument comes from considering other cases in the history of science where natures have been designated without being specified. In each case the designation has come through a set of surface properties, and the complete set of facts about the designated nature have included new facts. For instance, prior to discovery of the periodic table, gold was a natural kind whose nature had been designated, but not yet specified. The designation had its origin in historical practice, but was held in place through knowledge of identifying surface characteristics: appearance, malleability, reactivity, and so forth. The eventual chemical specification of its nature revealed new facts, facts about its deep structure. These new facts were not simply extrapolations or articulations of the old facts, but changes to the ontology our world view committed us to.

Furthermore, the surface characteristics used to designate the unspecified natures constitute independent properties in their own right. One may see this by noting that chemical theory may be incorrect, and therefore we may not actually have specified the nature of gold. We nevertheless are familiar with a host of properties that gold manifests (e.g., its shininess). These familiar properties are, in principle, multiply realizable, logically capable of being supported by a variety of more fundamental natures. So it seems that the facts about the designated nature of gold constitute further facts about the deep structure of gold, relative to the surface properties through which we designate its underlying nature. By analogy, the facts about a hidden designated nature of the physical would constitute further facts about the deep structure of the world's properties and particles, relative to the surface physical properties through which we designate them.

The second argument that facts about intrinsic natures should not count as physical facts comes from noting just how utterly irrelevant such facts are to any physical question. To see this, consider the relation between matter and anti­matter. The physics of the two types of matter are exactly symmetric so that, for instance, positrons interact with anti­protons in just the way that electrons interact with protons. For these reasons, the relational profiles of every particle and its anti­particle are just the same, with matter being further differentiated only by the indexical fact that it is the predominant stuff in our environment.

In principle, one may invoke an intrinsic difference to ground a further non­indexical difference in kind, and one might think that matter and anti­matter present the right case. Perhaps their ultimate differences lie exactly in intrinsic characters that are incompatible with one another.(7) If so, this seems to matter not a whit to the physical facts about our universe, as normally understood through our best science. If we woke up tomorrow in a universe where every intrinsic character had transmuted into its anti­matter dual, we could not, even in principle, tell the difference. The laws of physics would seem still the same, and we would seem to still live in a world with properties like mass, charge, spin, and so forth.

As far as physical theory goes, only the indexical facts involved in designating the intrinsic aspect of physical terms would have changed. Indeed, for all we know or could know via physical investigation something like that happens all the time. Perhaps there is a funny law that causes intrinsic characters of the physical to periodically transmute into their anti­matter complements. Perhaps. It just does not seem to matter, though. As far as fixing the physical laws and properties in our universe are concerned, such considerations are simply irrelevant. The facts about such intrinsic properties and laws seem to represent further facts, over and above the physical facts about things.

A third level of reply exists to this kind of attempt to save physicalism. Even if one admits the existence of some intrinsic character to the basic physical entities, and even if one extends one's notion of physical fact to cover these facts, science is still left with a bootstrapping problem. The bootstrapping problem concerns how to get from (i) the claim that electrons, photons, and quantum chromodynamic quarks have an intrinsic character; to (ii) the conclusion that there could exist a human consciousness with further intrinsic character all its own. Why shouldn't intrinsic character be delimited right at the boundaries of the microphysical, with the rest being an abstraction off of the patterns of their interaction? I will return to this problem in chapter eight, and deal with it in detail in part three of the book. For now, I merely want to point out that we have absolutely no reason to believe that we can solve it without appealing to some kind of new fact about the world, over and above those science recognizes as the physical facts.

Summary ­ The specific failure of consciousness to logically supervene in the Life world seems to be a result of the fact that its phenomenal aspect is not a purely functional, structural, or evolutionary one. Given that, the failure should hold in our world for basically the same reasons it holds in Life. Indeed, in many ways the Life example gives us a better proving ground for making the determination since, unlike in our world, we do not start with the knowledge that consciousness exists in that universe. Therefore, the temptation to see entailments where they do not exist is greatly lessened.

1. I will argue for the dependence of identity on logical supervenience in chapter four.

2. During a presentation at Indiana University in the fall of 1993.

3. Also, some people suggest that there is a modality of metaphysical supervenience analogous to an a posteriori metaphysical identity. I shall address this idea in detail in the next chapter. The account I give below owes a great deal to the account given, in much more detail, in Chalmers (1996). Chalmers' treatment deals elegantly with the most serious problems philosophers have raised for supervenience. For the sake of brevity, I pass over most of those problems here.

4. The positive facts are those facts that would remain true even if our world was embedded within a larger world.

5. Appropriately enough, the section is entitled "Almost everything logically supervenes on the physical." pp. 71.

6. Kirk's "Strict Implication Thesis" is equivalent to the logical supervenience requirement.

7. This is not the only possible answer, of course. Feynman (1985) claims that a particle's anti­particle is just a particle of the same type, but traveling in the opposite direction of time. If that is correct, then there would be no intrinsic difference between a particle and its anti­particle. I am only using the suggestion in the text as a way of making a point.


Chapter 4

The Objection from Metaphysical Possibility and Necessity

4.1 The minimal meaning standard

Turning aside from the charge that the anti­physicalist arguments are arguments from ignorance, some attempts at restoration instead focus on the intensional framework active in those arguments. Some physicalists may try to resist these kinds of arguments by appealing to a different relation than logical supervenience. Supervenience, after all, is a necessitation relation, and logical supervenience relations partly depend on meaning.

The suggested defense against reformation is that, in many cases, logical supervenience is really only an epistemic relation, yielding epistemic possibilities and necessities. One alternative sometimes proposed is that the physical facts do necessitate the phenomenal facts in a way that is stronger than mere natural necessitation, despite not logically necessitating them. The grade of necessity that is supposed to be relevant to the ontological questions is called "metaphysical," and it is supposed to be something narrower than logical possibility and necessity, yet wider than mere natural possibility and necessity. So, the defense concludes, while consciousness may not logically supervene on the physical, it does metaphysically supervene.

Philosophical appeals to metaphysical possibility and necessity appear in areas as diverse as the philosophy of mind (Levine 1993), the philosophy of causation (Fales 1990), and discussions of ethical realism (Brink 1991). Given the heavy burden many philosophers are asking it to bear, the idea is worryingly obscure. In this chapter, I will argue that either metaphysical necessity is not strong enough to do the ontological work that the reformation is calling on it to do, or appeals to it seem to fall short of very minimal standards of meaningfulness(1). Given these shortcomings, philosophers who rely heavily on metaphysical possibility to do ontological work for them owe the rest of us a much clearer articulation of what they mean.

To begin the argument, I am going to suggest some very mild constraints that any kind of possibility statement must meet to be meaningful. Consider statements of the form, "X is possible" and "X is not possible." as they occur under normal circumstances. Reflection on ordinary examples strongly suggests that a meaningful possibility statement must meet certain minimal standards. The weakness of these conditions means that any failure to meet them is a severe shortcoming. We can formulate these standards using three criteria:(2)

(1) We always make possibility statements relative to an established or assumed context.

(2) Understanding such assertions tacitly requires holding the truth of the context, or some crucial elements of the context, constant as a constraint on the claim.

(3) An assertion of a possibility involves an assertion that the situation in question can be part of a consistent extension of that context, or of whatever part of the context the speaker(s) is holding constant.

People use possibility statements widely in everyday life, and this large variety of possibility statements all seem to meet the criteria above. Consider a chess player mulling over his options. A friend may suggest a move where, in fact, another piece blocks the path. In deciding that the move is not possible, he clearly is not deciding that the move is contrary to the laws of nature or logic. He is judging the more pertinent question of whether, in a chess game, the move would be consistent with the rules of the game.

Similarly, a worker in a complicated bureaucracy may inform customers of possible avenues they can pursue to have a complaint processed. Why is it that filling out a form for review by the manager is a possible course of action, but not storming the CEO's office? The possibility of the former action, but not the latter, rests in its consistency with the normal processes and rules of the organization.

For a possibility statement to be meaningful, the constraints do not have to be sharp or explicitly understood. In ordinary use we might say that climbing Mount Everest would be something that is impossible for me to accomplish. Someone who makes that claim does not really mean a contradiction is involved in the description, or even that it would violate the laws of nature. Instead, that person is appealing to vague and implicit constraints on what common sense would allow that I could "reasonably" or "ordinarily" accomplish. To claim that something like climbing Mount Everest would be impossible for me is to claim that my accomplishing it would be inconsistent with those implicitly understood constraints.

Even in technical contexts speakers may understand the background constraints only implicitly, as something vague. An evolutionary biologist might claim that birds could have descended from dinosaurs, while simultaneously denying the possibility that the first human could have coagulated from swamp water, like a "Swamp Thing", into a fully formed organism. The difference is not that the existence of a Swamp Thing would violate physical law. It probably would not. The difference is that the biologist is assuming the principles of natural selection as the vehicle for producing organisms. The idea that birds descend from reptiles is consistent with these principles, while the idea behind the Swamp Thing is not.

Similarly, when physicists speak of physically possible situations they mean that the situations are consistent with the known laws of physics. On the other hand, when they speak of a theory being a "possible" theory of everything, they allow that it might be inconsistent with known laws in some respects. In the latter case, they mean that it is consistent with known experimental results (and perhaps, implicitly, with superempirical criteria).

Finally, when we philosophers speak of the logical possibility of a situation, we mean that the description is consistent tout court. The assumed context is one where we have made the intensions on the concepts used in its description suitably definite. The constraint requires holding the meanings of the terms constant in any complete extensions.

I believe we could multiply examples easily. In a sense, we ordinarily speak of as many types of possibility as there are contexts and constraints that can make a possibility statement meaningful. What renders a statement of the type "X is possible" meaningful is at least an implicitly assumed collection of background constraints set by an appropriate context.

As these examples show, our standards about how explicit or definite these constraints have to be are not even very demanding. They can be sharp constraints like the rules of chess. They can also be vague, common­sense constraints involving our understanding of plausibility, a scientist's understanding of a theory, or a philosopher's understanding of the intensions associated with terms. This looseness in the criteria makes the demand that a use of 'possibility' meet them a reasonable demand. Relative to a context C, the intelligibility of "X is possible" seems to require at least that the truth of X would not violate the constraints implicitly taken from C. This seems encodable into something like the following "minimal meaning" standard. For a statement "X is possible" uttered in a context C that contains background constraints BC, the minimal meaning standard is simply:

MM) X is logically consistent with BC.

I want to emphasize that I am not putting forward the minimal meaning standard as anything more than its name suggests. Specifically, it is not intended as an adequate analysis of modality. MM is just a requirement on the meaningfulness of particular modal statements or claims. The complaint about metaphysical necessity that I am about to develop is that the most commonly cited constraints ­ the Putnam­Kripke variety, and the constraints due to the natures of things ­ are not especially metaphysical.(3) If the defenders of physicalism are appealing to one of those constraints, their notion of "metaphysical" necessity is not strong enough to do the work they need it to do. In the search for truly metaphysical constraints it seems that the relevant constraints, if there really are any, are not being even obliquely specified.

Now, we can understand a nomic possibility as being a situation that is consistent with the laws of nature. What is the notion of metaphysical possibility that many current philosophers are appealing to? Should we understand a metaphysical possibility as being a situation that is consistent with the "laws of metaphysics?" If so, just what are these laws? Here I think we have grounds to doubt seriously that we possess any BC­facts to substitute in the MM, and therefore we have grounds to doubt that these appeals to metaphysical possibility are even meaningful.

4.2 Are Kripke and Putnam necessities really metaphysical?

In Naming and Necessity (1972) Kripke argues for a variety of necessity that the philosophers who appeal to metaphysical necessity often cite. Around the same time, Hilary Putnam (1973) offered arguments taken to yield a similar moral. I will argue that these philosophers only identified a type of modality whose constraints, rooted in rigid designation, do not seem to have deep ontological consequences. They are at root linguistic constraints, and Kripke and Putnam's sense of necessity cannot do the kind of ontological work that some philosophers want metaphysical necessity to do. Their sense of possibility focuses our talk about the set of conceivable worlds, but it does not eliminate any of those worlds as real ontological possibilities.(4)

In overview, my argument will be this. Kripke's variety of necessity arises from rigid designation. Rigid designation contains two components: (i) the indexicality of certain of our concepts (emphasized more by Putnam), along with (ii) rules for counterfactual reference that depend on the actual world values of these indices (emphasized more by Kripke). Both of these components take the actual world facts for granted, so they cannot present constraints on the space of worlds that could have been actual. Since a possible world is a world that could have been actual, Kripke and Putnam's discovery makes no ontological difference within the space of possible worlds. Instead, it merely makes a linguistic difference to what terms we can use when describing the worlds within that space.

A quick summary ­ Kripke and Putnam gave arguments that proper names, and certain natural kind terms that act like proper names, refer by picking out essences from the actual world. Which essence a term picks out depends on external facts in the environment of the language community, or person, using the term, and so the facts about reference depend on context sensitive indices. An important moral of their arguments is that these facts about reference are at least partly determined by external, contextual facts rather than the descriptions the speaker believes about the subject matter. Some canonical examples are that 'water' picks out H2O, and 'Aristotle' picks out a particular individual, and they do this just the same regardless of one's beliefs about water or Aristotle. This is due to the indexical component of the concept, which addresses its subject matter according to the physical context and history of the thinker or community using it. Additionally, our language seems to have referential rules dictating that we may correctly apply such terms only to the indexed essences, even when we are describing a situation in a counterfactual world. That makes these terms "rigid designators", meaning that they may properly be used to designate only the essence picked out by the index.

Kripke and Putnam's variety of necessity clearly does meet the minimal meaning standard. The context it assumes is the actual world centered at the point where the language is being used.(5) Linguistic rules for using the index associated with the concept provide the relevant constraints, as they govern how the terms expressing those concepts may be applied to other possible worlds.

Deflating Kripke and Putnam ­ Does the discovery of this indexical rule add or remove any ontological possibilities, revealing that some apparent ontological possibilities are merely epistemic? Several arguments suggest strongly that it does not. At the most intuitive level, we may reflect on how we decide the way a rigid designator may be used. In every case we first imagine a world, and then decide whether that world contains the item in question: Aristotle or Samuel Clemens and so forth for proper names; or, for natural kind terms, gold, water, tigers, or whatever. Even to engage in this process one must first understand the possibility in question, and then decide how to describe it given information about the actual world.

Since the ontological content is in the worlds, not the talk, we need to closely scrutinize any use of Kripke and Putnam's discovery to ground ontological claims. In the end, it is hard to see how to use it to do much work at all. Our understanding of the space of possibilities is evidently prior to, and more fundamental than, application of the rigidly designating term. When we become aware of the standards of rigid designation, what we learn are facts about the way we should use our language to describe possibilities already understood. As a corollary, the fact of rigid designation reflects linguistic rules and not ontological discovery.

Philosophers accustomed to thinking of Kripke and Putnam necessity as "metaphysical" often do not find this point immediately obvious, but reflection on the canonical examples reinforces it. Let us consider the difference between our world, where water has H2O as its essence, and the world where an XYZ that is macroscopically indistinguishable from H2O exists in its place. In the discussion that follows the most important fact to remain aware of is the location of the ontological content in the possibility claims. The way that we indexically designate the content of our world matters less than understanding the non­ indexical specification of the possible contents in the worlds at issue. Notice that 'water' merely indirectly designates an actual world essence via superficial properties and contingent relations, while 'H2O' and (presumably) XYZ specify conceivable essences directly and non­indexically.(6) The ontological content of the possibility claims depends on the content of the two worlds, and that is given to us first and foremost by the non­indexical understanding. Are either of these conceivable essences not "really" possible essences?

That the world should conform to the theories favored by our super­empirical standards is not a necessary truth. The world might be grotesquely complicated, even if we are unwilling to consider that possibility in choosing theories. Thus it is possible, really possible, that our super­empirical standards have led us astray. Perhaps, in fact, we live in the XYZ world. What then? For the sake of argument, suppose that we really do live in the XYZ world, thinking otherwise only because our standards for theory evaluation have led us astray. Would it follow that the world we mistakenly think we live in, the H2O world, is not "really" a possible world? Could it be that the theories of our current science are really necessarily false because they secretly violate some metaphysical constraints on the space of possible worlds?

Philosophers who want to uphold such a position owe us some explanation of what these constraints are. Otherwise, they will fall afoul of the minimal meaning standard. These philosophers need to articulate a set of constraints showing that the H2O world could not have been actual, on the assumption that we live in an XYZ world. Rigid designation itself does not explain why another world could not have been actual since it assumes the relevant facts about the actual world. It treats the actual world as a given when setting the values of the relevant indices, but does not place any constraints on which worlds could have been actual. Without rigid designation to provide an informative constraint, the claim of impossibility does not meet the minimal meaning standard. Neither the appropriate context nor the appropriate constraints have been specified.

Furthermore, the arguments for rigid designation do not provide a reason to think that the H2O world might be something less than "really" possible. The hypothetical (that we live in a XYZ world) makes a relevant difference to the context in which we are using our language, and nothing else. Kripke and Putnam's arguments all rely on these kinds of contextual facts, but a shift in linguistic context fails to make an ontological difference to the space of possibilities. Recasting this as a twin­earth case very vividly brings out the irrelevance. If I have a twin orbiting Betelgeuse, my twin's discovery that he lives on an XYZ world does not affect the possibility of our world. In the twin­earth case, the possibilities that have changed are clearly linguistic, not ontological, as the differences between my twin and myself are in the standards governing our uses of the term 'water'.

As the twin­earth case goes, so goes the Kripkean case involving possible worlds. If we, by chance, live in the XYZ world then our term 'water' happens to refer to XYZ. If that is our linguistic context then, when speaking about other possible worlds, our linguistic conventions forbid us from describing H2O using the predicate 'water'. Like a violation of the rules of chess, that way of speaking becomes impossible because it contradicts certain normative conventions. However, we have removed no ontological possibilities. Physics and chemistry might be wrong, but they at least depict a kind of world that might be.

Contemporary philosophers sometimes take the use of Kripke and Putnam's variety of possibility and necessity beyond the limits of its meaningfulness. Consider this argument given by Sydney Shoemaker,

Mightn't the world have been such that the properties instantiated in it are utterly different from those in the actual world?. . . .I question whether there is any content to this idea.

I want to interrupt Shoemaker's argument here to note that, in the grip of metaphysical possibility, he is unsure whether such a claim even has content. This is puzzling, as such claims certainly meet the minimal meaning standard. Shoemaker's problem here seems to imply that many discussions in modern cosmology might not be meaningful. These days, as a matter of course, educated scientists discuss the possibility that the laws of nature might have been different, say, in the very first moments after the Big Bang, or in different cosmic epochs, or in ways that lead to the existence of universes that could not support the existence of life. Shoemaker seems to want to tie the meaningfulness of such discussions to their truth, so that they might turn out to have been meaningless if they happen to rely on false assumptions about what are the actual laws of nature. What view could prevent someone from seeing the content of these claims? Shoemaker continues,

To be sure, supposing that we had names for all of the properties instantiated in the actual world, there would be no formal contradiction in saying that none of the named properties is ever instantiated and that instead some other set of properties, governed by its own set of laws, is instantiated. But, likewise, there is no formal contradiction in saying "Tully is not Cicero," or "Water is not H2O." Lack of a formal contradiction does not guarantee that a sentence describes a metaphysically possible situation.

Shoemaker argues from standard Kripke and Putnam cases of empirical identity to the conclusion that the consistency of other nomic worlds does not show that they are possible. He appeals to the idea that non­actual properties may lack some metaphysical sense of possibility. Despite his appeals to standard Kripke and Putnam examples here, he clearly cannot mean "possible" and "necessary" in that sense of those terms. By explicitly suggesting that we fix our language appropriately, he has already fixed all the relevant Kripke and Putnam constraints. We will not improperly claim some property or law from our world exists in the imagined alternative worlds. Yet this is not enough to convince Shoemaker that the world imagined is "metaphysically" possible. Whatever notion Shoemaker is appealing to here, he has moved beyond Kripke and Putnam possibility, which he calls metaphysical, to some other sense of possibility that he calls by the same name.

This subtle equivocation takes him out of the range of anything that clearly meets the minimal meaning standard. He has not fixed the sense of this second, subtly non­Kripkean use of metaphysical possibility. As he has described it, the situation is logically possible and Kripke/Putnam­possible. From what context does the further constraint come from, and what is that constraint? Unfortunately, the loaded way he is thinking of Kripke and Putnam possibility seems to have misled him into relying on concepts whose meaning is obscure, at best.

This pattern of begining with Kripke and Putnam possibility and necessity and inflating it into some more substantial kind of "metaphysical" necessity occurs repeatedly in recent philosophy. David Brink attempts to explain Kripkean necessity as follows,

Analyticity and necessity are distinct properties. . . .Necessity is a metaphysical (modal) notion; the claim that something is necessary is the claim that things could not have been otherwise in certain respects. This characterization is epistemologically neutral; in particular, necessary truths are not characterized as a priori.(p. 165-166)

Kripke and Putnam's kind of possibility and necessity are not a priori in the standard sense, and Brink is correct to distinguish it from traditional notions of analyticity. However, they do not escape the a priori because they hinge on some metaphysical constraint, but because they involve facts about what refers to what. Facts about reference are, unsurprisingly, not completely a priori. Even so, Kripke and Putnam's discovery has a perfectly good foundation in the rules of language: both the standards of rationality that guide how we determine reference, and the standards that govern how actual world reference may affect reference across possible worlds, rest importantly on epistemically accessible aspects of meaning. These a priori foundations show themselves in the way Kripke and Putnam argue for their conclusions, which is to use the traditional tool of philosophical thought experiment. All of their arguments are of the form, "If such­and­such were the case, what would we say then?" These clearly are not appeals to metaphysical constraints, but linguistic intuitions. The general constraints that lead to these intuitions are a priori.

Only the decisions about specific cases, which depend on knowing the indexical values our context fixes for us, depend on empirical discovery. Do we live in the XYZ world or the H2O world? We know about the indexical character of the rule of application for the term 'water', and how that dictates its use across counterfactual situations. We simply do not have an a priori answer to the relevant question about our linguistic environment. But why should we?

Meanwhile, we do not need to refrain from drawing conclusions about possibility and necessity. Often, we will want to abstract away from the underlying nature as a mere "realizing" property, focusing instead on aspects of the macroscopic patterns these natures support. For instance, there is (at least) one perfectly good property or entity, call it wateriness, that H2O realizes. Roughly, wateriness consists in the pattern of macroscopic effects and responses exhibited by water. When we admit that our superempirical standards might be misleading us, we are acknowledging that both XYZ and H2O could realize wateriness. This is not a mere "epistemic" fact. Presumably only one of them does realize wateriness in our world, but that gives us no reason to think that either possibility is something less than a "real" possibility. It just gives us reason to think that one of them is something more than a "mere" possibility.

The overall moral is that Kripke and Putnam's possibility and necessity are not really metaphysical, except in some very innocent sense. Facts about them follow from logical possibility, broadly conceived, plus some facts about language. The linguistic facts involve context­sensitive indices whose values govern our talk about counterfactual worlds, given information about our situation in the actual world. Kripke and Putnam's arguments have led many philosophers to think that their brand of possibility and necessity are metaphysical rather than linguistic because they involve a posteriori knowledge. This knowledge, though, just involves the a posteriori aspect of language: reference. When we attempt to use metaphysical possibility to express a kind of possibility that is more mysterious than that, we fall foul of the minimal meaning standard.

The irrelevance of Kripke and Putnam possibility and necessity to the anti­physicalist arguments should be clear now. Before the Kripke and Putnam rules even become relevant, we must first decide on independent grounds what the situation is in the actual world. Does fire come from the release of phlogiston? Are the properties of water best explained by XYZ? Does cognitive neuroscience fully explain phenomenal consciousness? These questions are all on a par, and, most importantly, we settle questions about reference by first settling questions like these. Kripke and Putnam possibility was not relevant in deciding the first two, and is not pertinent to discussion of the third for basically the same reasons.

These kinds of questions are settled by examining the relation between theory and the target of explanation. Before we can justifiably postulate a theoretical necessity, the properties of the explanatory target have to be predictable from the theoretical facts, and they have to be predictable without a circular appeal to the necessity at issue. This is just another way of saying that a relationship of logical supervenience must be able to hold between facts of the type postulated by theory and the target facts. Put another way, pointing out the failure of logical supervenience is just a more philosophical way of pointing out that the physical facts alone fail to predict the facts about phenomenal consciousness.(7)

Even after Kripke and Putnam, conceivability, carefully considered, remains the most ontologically interesting sense of possibility. As philosophers concerned with careful reasoning and exposition, we can thank them for helping point us to a way of clarifying our talk. Nevertheless, as metaphysicians concerned with ontology, their kind of necessity should not be our primary concern.

4.3 The constraints of natures

A second class of philosophers appeals to the fact that things have natures (e.g., Plantinga 1974). The metaphysical possibilities are those that are compatible with the natures of things. As before, an appeal to the nature of a thing must meet the minimal meaning standard. That is easy to do simply by specifying the relevant nature. The context will be the theoretical context within which the nature is specified. The constraints come from the meanings of the terms in that theoretical context. The necessities that result are properly thought of as more than merely linguistic necessities, as they are intended to correspond to a presumed modal character in nature herself. The problem with this proposal is that once a nature is specified, one never seems to have any kind of intermediate metaphysical necessity left over. Instead, one gets transparent conceptual necessities, often in conditional form, that are naturally interpreted in terms of more standard kinds of necessities like natural necessity.

Examples of necessities due to natures are common in the natural sciences and mathematics. To go back to our canonical example, science postulates that water has a nature, H2O. The conceptual content of 'H2O' represents that nature, and its theoretical context sets this content. Within that context the fact that something is H2O entails certain consequences, and is compatible only with a restricted range of facts. These entailments are the necessities due to the nature of water, and these compatibilities are the possibilities for that nature.

Similar kinds of necessity exist in mathematics. For example, Peano's axioms are an attempt to represent something about the nature of the natural numbers (I would claim that even a structuralist position, like that taken by Dedekind, is ultimately a theory about the nature of the natural numbers, rather than a position that they have no nature). The entailments from those axioms are necessary consequences of those natures.

Two important morals exist here. First, the necessities that exist in the mathematical realm are not different from the necessities that exist in nature. They are both necessities due to the natures of things, and are accessible to reason. Second, proposed theories imbue the terms with a conceptual content that represents the natures of those things, with the necessities following as rational consequences of that content.

If necessity is the same for both natural truths and mathematical truths, why does it appear to so many as if it is different in the two cases? I believe the difference lies in the entities, not the necessity. When a conditional necessity issues from the nature of a necessary being, the antecedent and consequent are also necessary. That makes each into a kind of non­conditional necessary truth. When a contingent being necessitates, the antecedent and consequent are contingent. That makes them seem like instances of some kind of "contingent" necessity. To see the difference between the two cases, consider the following two necessities, both grounded in natures,

(1)[][(Ax)(H2O(x) --> (Av)( x has volume v --> (Aw)(Az)(E1y)(A mole of x changes pressure by y when its mean molecular kinetic energy changes from w to z))](8)

and

(2)[][Peano's axioms --> ~(Ex)(x is the largest natural number)]

The first statement asserts a functional relationship between changes in the temperature in a mole of H2O confined to a given volume, and changes in its pressure. This proposition is strictly necessary because the conceptual content of the antecedent, derived from its theoretical context, entails the consequent. The second statement says that Peano's axioms strictly imply that there is no largest natural number, and the conditional in statement two is necessary for the same reason that the conditional in the first statement is necessary.

In both cases, the necessity is represented as a conceptual necessity that issues from the theoretical content of the antecedent. To vitiate the entailments, one would have to change the content of the embedding theories in some way. A change like that represents either a different hypothesis about the nature of the entity, or a different hypothesis about the nature of some other entity that may affect its properties. As conceptual necessities, neither is epistemically opaque in the way metaphysical necessity is supposed to be. In particular, the consistency of a described situation with a hypothesized nature does imply the possibility of that situation. The perceived difference in the necessity of natural truths and mathematical truths arises because the following two necessary existence claims have very different degrees of plausibility.

(3)[][(Ex)H2O(x)]

and

(4)[][(Ex) x = 0]

We are almost all inclined to judge (3) to be false, while many people are inclined to judge (4) true. If (4) is true, then one would think that all other natural numbers necessarily exist for similar reasons. Since natures do not change from world to world (that is what makes them natures), Peano's axioms will be true in every possible world also. In short, the inclination to view the numbers as necessary beings yields an inclination to view mathematical truths as necessary truths tout court. In contrast, the contingent status of natural beings means that natural necessities are, in a certain sense, also contingent.

The only opportunity to introduce a troubling kind of necessity is at the place where one judges the existential statements about the natural numbers to be true. Rather than appealing to a sui generis kind of metaphysical necessity, defenders of (4) should produce arguments showing that the nature of the number zero entails that it will exist in all possible worlds. For instance, on some views numbers are identical to sets or classes of sets (0 = ; 1 = {}; 2 = {0, 1}, etc.). This view is a theory about the nature of numbers, and, according to it, zero will exist in any world where the empty set exists.

One may argue, based on a theory of its nature, that the empty set exists in all possible worlds. Here is one way that argument might go. Assume that a possible world must contain at least one object. What exists in a world is what would exist in that world were it to be actual. Sets will exist in all worlds, as one can construct a set from any group of objects. In any world where a set exists, its power set exists. Since the empty set is a subset of all sets, it is in the power set of every set. So the empty set exists in all worlds. So zero exists in all worlds. The existence of the other numbers then supervenes, as one is the unit set containing the empty set, two is the unit set containing one and zero, etc.

Arguments like this point to a way the mathematical case parallels the natural case in having an air of contingency. Necessities due to natures yield facts about what is necessary if the natures exist. After that, plenty of room remains to argue about whether the natures exist, either those of mathematical entities or of natural entities. A nature may not exist because the entity does not exist, or because the entity exists but does not have that nature, the one postulated by the disputed theory. Perhaps the orthodox view of numbers is wrong?

The possibility of disputing an account of a thing's nature always exists, and does not reflect an epistemically opaque kind of necessity. It is a fact about our epistemic access to the entities, not to the necessity. All statements of necessity will suffer from an air of contingency due to this kind of uncertainty. To compensate, we have to decide questions about natures on a case by case basis, on the strength of rational argument, and empirical investigation. These facts about epistemic access do not yield any kind of intermediate metaphysical necessity. We just have the usual kind of uncertainty that our theories might be wrong.

A failure to appreciate the transparent relationship between natures and what they necessitate often leads philosophers to misidentify ontological gaps as epistemic gaps. For example, Joseph Levine's analysis of the explanatory gap between brain processing and consciousness leads him to conclude that the gap may only be epistemic. Levine recognizes that no explanatory entailment exists between the physical nature of brain processing, and the phenomenal nature of consciousness. To resist drawing an ontological conclusion, Levine appeals to an epistemically inaccessible kind of metaphysical necessity to ground some kind of a posteriori entailment. Because of this metaphysical necessity, Levine says, the explanatory gap is only epistemic. It need have no ontological consequences.

Levine must say something more here for his claim to meet the minimal meaning standard. He must tell us what it is about the nature of brain processing that is incompatible with the absence of consciousness. What is the relevant further constraint? We certainly do have theories about the physical nature of that processing, and those theories entail many facts. Since Levine admits that they cannot entail the facts about consciousness, he must admit that every physical constraint we know about, or could know about, is compatible with the possibility that the physical processing could occur while consciousness is absent.

He maintains, nevertheless, that this is truly impossible. To meet the minimal meaning standard, he must hold that we have inaccurately represented the nature of brain processing. Some constraint must exist of which we are not aware. Levine nowhere suggests that our physical theories are inaccurate or incomplete accounts of the physical nature of things. This implies that the missing information must be about a non­physical aspect of its nature. Once one realizes that theories of natures are just ordinary kinds of theories, theories open to the same standards of criticism, no room exists for an epistemically inaccessible metaphysical necessity that prevents the explanatory gap from having ontological consequences. In the end, Levine's recognition of the explanatory gap is the recognition of a failure of prediction by physical theory. Our theories of physical natures do not predict the facts about consciousness, and need to be supplemented. This is a straightforwardly ontological conclusion drawn in the usual way.

The important point to carry away from this discussion is that natures never yield a necessity resting in an intermediate place metaphysically between natural and logical necessity. Instead, we achieve slightly different varieties of necessary truths depending on the nature of the entity. For natural, contingent beings we come away with conditional natural necessity. For necessary beings such as mathematical entities we come away with stronger kinds of necessities, necessary existential statements. This is a difference due to different kinds of natures, not due to different kinds of necessity.

4.4 Are conceivable worlds not "really" possible?

We are led back to where we started, wondering what metaphysical possibility means. How could a logical possibility not be a "real" or metaphysical possibility? What are the constraints Jones is thinking of when wondering whether something conceivable is also "metaphysically" possible? What does it mean to wonder if there is a metaphysical grade of possibility beyond natural possibility, but short of logical possibility?

Notice how problematic this can be in actual cases. For instance, it is clearly logically possible that there could be a world whose space had a ten­dimensional topology. Let us imagine that Jones objects that such a world is not "metaphysically" possible. We would need to know just what Jones means by "metaphysically" possible. Jones must specify both the metaphysical constraints, and how they make an ontological difference. To avoid falling afoul of any standards of rigid designation, we could easily coin a new term to refer to N­dimensional spaces. Let 'schmace' refer to any entity just like space except in its dimensionality. A ten­dimensional schmace world is possible, and it is really, metaphysically possible in any sense of the term that philosophy has made meaningful.

Along similar lines, physicist friends of mine often conceive of possible worlds that are entirely bizarre, with quirky laws. These worlds are not nomically possible, and they evidence behavior that they probably often describe in ways that violate rigid designation. Despite their rough use of language, I do not think they are being naive about possibility. In fact, I think they understand it at least as surely as my philosopher friends who wonder about metaphysical possibility.

All that really matters to my physicist friends is the logical consistency of the laws. That is all that should matter when wondering about the possibility of a world, as consistency and inconsistency follow from the specification of a relevant nature. Given the nature of a thing, we have good reasons to believe that only some things are compatible with its existence, and perhaps its existence necessitates others. We do not have any good reason to think internally consistent natures, once conceived, are not truly possible. In the end, the kind of possibility these worlds have is something that we truly understand. On the other hand, this other kind of possibility that some of my philosopher friends vex themselves about is not so clear. To the extent that it is given a content that meets the minimal meaning standard, it poses no threat to the anti­physicalist arguments. If it does not meet those standards, it is not at all clear why we should worry about it, and it certainly does not seem legitimate to appeal to it in philosophical argument.

1. A different way to defend the arguments would be to make a case that conceivability is a reliable guide to these metaphysical possibilities. Yablo (1993) gives such an argument, but I take a different tack here since I think there are deeper problems with the attacks on conceivability.

2. Similar criteria are defended and formalized in Dunn (1973). A contextualized notion of possibility is also defended in Barwise (1997).

3. O'Connor (1997) describes a third way to understand the notion of metaphysical possibility, but it is not one that will give any comfort to physicalists. This third view is grounded in philosophical theism. O'Connor appeals to reflection on the seemingly contingent status of the world's existence, and how that leads to the kinds of questions cosmological arguments attempt to solve. He argues that the existence of a necessary being that created the world is the most plausible answer to the existence question. The metaphysically possible worlds, then, may be understood as those that the creator is able to create, given its moral nature. While O'Connor's suggestion will provide no succor for physicalists, it might contain a moral for them. In the end, some theistic view like this might be the only way to make sense of the ways that physicalists use the notion of metaphysical possibility. Where else may we find a grounding for the kinds of "laws of metaphysics" their view implicitly requires? Appeals to mystery may make for strange bedfellows.

4. By an "ontological possibility", I mean a world that could have been actual. Kripke himself alludes to this notion as distinct from his "metaphysical possibility" in footnote seventeen of Identity and Necessity (1971).

5. A centered world is a world with a region of spacetime marked as its center, where facts about the center are available for use in the semantic evaluation of indexical terms.

6. This asymmetry is responsible for the tight relation between conceivability and possibility that I will argue for below. The point I will make undermines Brian Loar's attempted defense of metaphysical possibility in his 1997. There, he tries to argue that it is our direct mode of reference to phenomenal properties that pulls apart conceivability and possibility. If my discussion in the text is correct, Loar seems to have gotten things reversed. When such tight modes of reference exist, modes not mediated by superficial properties and contingent relations, conceivability and possibility walk hand in hand. Loar sometimes seems like he wants to argue that it is our understanding of physical essences that is based on contingent properties (i.e., in some sense our theories do not really or fully specify the essences). If he means this to apply to physical theory, then he is commited to the position that physical theory is aspectual, and his position seems to reduce to Liberal Naturalism. Otherwise, he owes some explanation of why these aspects that elude all physical theory are, nevertheless, physical aspects. What about them makes that the case?

7. Some philosophers believe that the moral of Kripke and Putnam's work is that facts about identity, not necessity, are what elude a priori entailment. In chapter five, I will discuss the relation between logical supervenience, entailment, and identity in more detail. For now, it is enough to be clear that Kripke and Putnam did not discover a mysteriously opaque brand of necessity. Whether or not there is a mysteriously opaque brand of identity is a different issue.

8. Given quantum mechanics, this is really a probabilistic necessitation. The complications of probabilistic necessitation are beside the point here, so I am papering over them. For those skeptical of the compatibility of quantum mechanics with the idea of natural necessity, it should be sufficient to point out that if no natural necessities even exist, the case for a strange sort of metaphysical necessity seems very implausible.

Ceteris paribus conditions, which I am also ignoring, are more relevant to the discussion. Most natural necessities involve interactions, so one must take into account the natures and the contexts they are put into. The schematic form is: [][(Nature & Context) --> Consequence]. The example in the text is only meant to approximate the true situation.


Chapter 5

Logical Supervenience, Entailment, and Identity

5.1 Why not identity?

The arguments against physicalism, including the version that I have spread across chapters two and three, bypass concerns about the identity of conscious states and instead focus on logical supervenience. Some philosophers attempt to fend off the reformation by claiming that it is some hyper­intensional relation of entailment that fails to hold between the physical and phenomenal facts. They claim that logical supervenience may nevertheless obtain because conscious states may be identical with special kinds of physical states (e.g., Lycan 1996), and identities are necessary throughout logical space. For instance, in the last chapter I used Joseph Levine as a foil, challenging him to articulate the missing constraint that makes it impossible for brain states to occur without conscious states. Levine may very well answer, "The constraint is that they are identical, and that is that. If they are identical, then we cannot have one without the other."

The worry philosophers like Levine have usually centers around the fact that anti­physicalists, in arguing for the failure of logical supervenience, invoke meaning in a crucial role. Arguments against identity claims risk committing an intensional fallacy, so many people who hold to the existence of a token identity between conscious states and brain states will be suspicious of these anti­physicalist arguments.

In this chapter I will argue that these fears are misguided. In overview, my argument will be this. Within physicalist frameworks, token identity claims presuppose logical supervenience, and entailment, construed as an a priori relation between facts, is the crucial support for logical supervenience in all the relevant cases(1). As things turn out, we are only in danger of commiting an intensional fallacy if we are in possession of an incomplete set of facts about the subject matter. The danger of commiting an intensional fallacy looms only when some further facts exist from which the identity follows. Since the anti­physicalist arguments involve all the physical facts from the outset, the existence of an identity between brain states and conscious states would imply the existence of some further facts, over and above the physical facts.

Maintaining the identity claim without entailment requires the identity relation between these sets of facts to be "brute" or "primitive" in the sense that even having all of the requisite facts about the "two" things, modulo the fact about identity, would not allow one to derive the indiscernability between them that usually grounds claims about empirical identity.

To answer the primitive identity hypothesis the anti­physicalist must defend two separate, but closely related, theses. The first thesis is that, in the relevant cases, identity holds only if logical supervenience holds. I will call this the supervenience thesis. The second thesis is that, for the relevant cases, if logical supervenience holds, it holds because the facts entail the facts(2). I will call this the entailment thesis.

I will defend the supervenience thesis in parts two through four of this chapter. In part five, I will defend the entailment thesis, using it to argue that the primitive identity hypothesis is, at best, methodologically anomalous, and, likely, conceptually incoherent. Taken as a whole, this chapter points to a tight relation between the identity of an entity, the relations holding between facts about that entity, and entailment. I will argue that, within physicalism, high­level identities provide ontological simplification precisely because they are parasitic on entailment relations. A thing's identity is not something that we primitively refer to, like a shining light inside it, so these facts about identity must supervene on the satisfaction of other kinds of facts.

The moral of my arguments below corresponds to parts of Yablo (1984). Yablo calls conclusions of non­identity "Extremely weak, from the point of view of philosophical materialism." In this chapter I agree with him, turning the point upside down by focusing on the implication that, if loss of these token identities matters so little, the identities themselves must not be very fundamental or important. Like Yablo, I argue that facts about token identity must supervene on other, more fundamental facts. Because of this, we can most plausibly regard the primitive identity hypothesis as a disguised version of the "metaphysical" supervenience position, and it falls prey to the objections to that position developed in the last chapter. The arguments in this chapter are also related to those in Johnston (1992).

5.2 The failure of identity in ordinary cases

Imagine some obviously physical object like a chair, and name that chair Chuckie. We will start the discussion with the question, "Is Chuckie identical to some physical state of the world?" Obviously, Chuckie is not identical to any momentary physical state. It persists through time while these physical states are in flux. Many physicalist proponents of token identity claim that Chuckie is identical to a four­dimensional spacetime worm. Chuckie is vaguely bounded both spatially and temporally, and finding the identity conditions that pick out just which worm Chuckie is supposed to be identical to raises puzzles. For the sake of argument, we can waive those worries since deeper problems exist. It turns out that our most basic, common sense understanding of Chuckie's identity is incompatible with the position that Chuckie is identical to a four­dimensional spacetime worm.

An argument against token identities for ordinary objects ­ Let A designate the worm with whom Chuckie is putatively identical. The entity A consists of segments from the world lines of various microphysical particles composing Chuckie throughout its history, and A is identical to the sum of positions and states of these physical particles. In the actual world I did not bump into Chuckie last Wednesday. If I had bumped into Chuckie last Wednesday, would it have ceased to exist? The common sense answer is certainly no. Yet if I had bumped into Chuckie last Wednesday, A would never have existed at all. I would have jostled some of Chuckie's particles in ways they have never been jostled in the actual world. They would have taken on different states, trajectories, positions, etc. Call this similar but different spacetime worm B. B is not identical to A, yet both are versions of Chuckie. So Chuckie cannot be identical to either of them. Instead, the physicalist should say that either of them could have realized Chuckie. Many other spacetime worms could have realized it also. We can state the argument as a reductio:

(1) A = Chuckie (physicalist identity claim)

(2) [](A = Chuckie) necessity of identity

(3) <>(B = Chuckie) (ordinary counterfactual claim)

(4) ~<>( B = A) (from the identity conditions for spacetime worms).

(5) ~<>( B = Chuckie) substituting Chuckie/A in (4), using (2)

Defending premise (3) ­ Lines (3) and (5) contradict one another. It seems that only the most benign modal intuitions are involved in the premises. Premise three might be mildly controversial, but denying it would lead to bizarre consequences. The intuitions at work in premise three are the same ones that allow us to track the identity of an object through time, as the future is uncertain. Will I bump into Chuckie tomorrow? If I do, will it be the same chair or will it lose its identity? Our judgements about continuation through change are essentially counterfactual. If premise three is false then in our ordinary dealings with the world we have no idea if we are ever encountering the same objects through time.

Reflection on cases of personal identity bolsters these ordinary intuitions. Will you be the same person tomorrow? Even regardless of whether or not you get bumped into? Whatever you answered, ask yourself if you know this. We are in great trouble if we do not. We cannot begin to answer these questions if our basic counterfactual intuitions are suspect, and that is a silly situation. Our epistemic situation regarding identity is not that opaque, and we surely know that a given chair will continue to exist whether we bump into it or not. If premise three is false, then our epistemic situation in the world is absurd.

Coincidence is not identity ­ The argument above requires physicalists to give up identity claims even for simple cases like chairs. The presumption that physicalism needs token identities in the first place, I believe, rests on a misinterpretation of extensional coincidence, which is a strictly weaker variety of identity (as Yablo notes, it is something like "contingent" identity), if it is identity at all. The general physicalist intuition, formulated in extensional terms, is that both names, A and Chuckie, designate the same spatiotemporally extensive object in our world. This intuition masks the fact that names themselves have intensional content. The state of affairs that will count as the object we call A or the object we call Chuckie depends on this content, and a contingent co­extensiveness in the actual world is not sufficient for a true identity. This extensional intuition can be expressed by making the names into predicates, and using a formula such as (E1x)(E1y)(Chuckie(x) & A(y) & x=y). Physicalist's clearly need a stronger claim than this to resist the anti­physicalist arguments, as the claim expressed by that formula is perfectly compatible with either property dualism or a dual aspect view.

The problem is that these individuals differ in their modal properties. Although physicalists commonly overlook these modal properties of individuals, or treat them as if they have a secondary importance, they cannot legitimately do so. The reason physicalists must take modal properties seriously is that we type and individuate the basic physical individuals according to their modal properties also. For example, something would not count as an electron, or as this very electron, unless the proper counterfactual facts about it were true (electrons have certain causal powers, for instance). The actions that I might take, and the adventures that I might have, are at least as important to my personal identity as the actions that an electron might take, and the adventures that it might have, are to its physical identity. A physicalist might choose to regard the modal properties of higher­level individuals as somehow less important, but that position is very implausible. The asymmetry seems very unmotivated.

In short, token identities require achieving a kind of intensional economy across possible worlds, given the facts about coincidence in this world. In practice, though, we rarely achieve this sort of intensional economy. Token identities, like hothouse blooms, are both fragile and rare. This revelation raises a question. Given their fragility and rarity, why do so many physicalists believe that a token identity between brain states and conscious states is important to physicalism?

The chief justification for the primitive identity claim is that an a posteriori identity between them is the best explanation for the empirical coincidences we know exist. It simplifies things. But what such arguments really show is that we have good reasons to believe that the physical facts and the phenomenal facts are coincident facts about some single system. This is not a conclusion that anti­physicalists need to reject. Except for a few Cartesian substance dualists, everyone is agreed that the two sets of facts present us with facts about the same thing. The issue is whether they give us the same facts about that thing, or, failing that, whether the physical facts contain the information available in the phenomenal facts. If neither alternative holds, then the physical facts about that thing are aspectual.

Physicalists often overlook this crucial point: the falsity of physicalism is perfectly compatible with the existence of some kind of a posteriori identity between brain events and conscious events. As proof, by the end of this book I will have a position that some people will certainly be tempted to argue involves token identity. The problem for the physicalist is that brain events may not have wholly physical natures, and the reduction may not be from the phenomenal to the physical, but from the physical to some richer, underlying ontology.

Deflating identity ­ Do the failures of identity in ordinary cases falsify physicalism? I do not think that it can be that large a loss, for the intuitions underlying the physicalist claim should not be so easily defeated. For instance, since all of Chuckie's possible realizations are physical, Chuckie's existence seems consistent with the basic physicalist intuitions about it.

Does the loss of token identity even weaken physicalism? I do not think it does that either. Physicalism is concerned about what things exist, both in kind and in particular. Losing the right to make these token identity claims weakens physicalism only if abandoning them would require physicalists to admit the existence of things whose existence they could otherwise avoid.

5.3 The unimportance of token identity within physicalism

If we think carefully about it, we can see that the token identity claims do not constrict physicalist ontology relative to logical supervenience. For similar reasons, we will see that one cannot call on identity to slip anything new into the ontology either. To begin the discussion, recall the basic physicalist intuition. The intuition driving physicalism is the claim that when God wanted to create our world all he had to worry about was making all the basic physical facts true. Once these were set, everything else (trees, rocks, money, stars, people, etc.), and the facts about them, came along for free.

The physicalist's world with realization relations ­ Imagine a physical world where logical supervenience relations hold. That is, imagine that God created all the physical fundamental entities, and made true all the facts about them; after that, everything else was logically fixed since the physical facts realize everything. Imagine further that, after admitting the existence of these realization relations between the lower­level facts and other facts, we refuse to make any identity claims to supplement them. Remember that realization relations are a kind of necessitation relation. I argued against epistemically inaccessible necessities in the last chapter, and those arguments extend to claims about realization.(3) So the full contents of this world should be accessible to us, in principle.

If we follow the procedure above, we end up with a world picture. What does this picture say the world contains? A world with only realization is ontologically robust. The physicalist has to admit, for instance, the existence of many individual people, trees, rocks, and stars, and these things are not part of their basic ontology. Consequently, the physicalist also must admit the existence of many kinds of things. A world with logical supervenience contains a robust plurality of things and kinds, and not many of these things are, strictly speaking, physical, as they are not identical to any physical thing. So, in one sense, it seems that we have increased the number of tokens in the world.

All this ontology, this cornucopia of things and kinds, comes from logical supervenience (via realization) alone, unsupplemented by any identity claims. Can the physicalist constrict his or her ontological commitments by making identity claims? By arguing for token identities, can the physicalist empty this world of some of its contents? As a way of approaching the question, consider some object close to you right now, say a piece of paper. It's pretty obvious that the existence of pieces of paper logically supervenes, at least globally, on the physical facts: it is strictly and transparently impossible for the physical facts to be what they are, but for the actual world facts about pieces of paper to be different.

As a next step in the thought experiment, modify your world picture by postulating that the object is also token identical with some physical state of affairs. So, in some weird way, we have moved from having two tokens, the piece of paper and its supervenience base, to having one. In what way is the nature of the paper's existence different in the two world pictures? Surely, in both pictures we have a piece of paper, so the "strong" physicalist has not gotten rid of pieces of paper. It still has the same mass, the same color, the same shape, and we can use it in the same ways. Not much has changed.

What about the money in your wallet? What about your body? Consider the entire world previously pictured, a world full of objects logically supervenient on the physical. Now imagine a "strong" physicalist pointing at every contested object, process, event, etc. and proclaiming confidently, "Not only is this thing logically supervenient on the physical, but it is also token identical to a four­dimensional spacetime worm!" How has this dramatic postulate changed the world picture we previously had?

The world with token identity ­ We need to characterize the ontological difference between the world picture that postulates just logical supervenience, and the one in which token identity claims are also recognized. Given that particular rocks will still exist in the second picture, then rocks will still exist. Given that particular stars will still exist, then stars will still exist. Generally, every kind of thing that existed in the first picture will exist in the second picture. Furthermore, since only token identity claims have been added, the kinds will be just the same ontologically as before.

So the space of general existents the physicalist had before the identity claim is just as large as that afterwards. How, then, has the physicalist shrunk the space of particulars? The more one reflects on the difference between the two pictures, the less important the identity claim seems to be from a physicalist perspective. In fact, it begins to look like it makes little ontological difference at all, at least not a difference that makes a difference. The only way it seems to affect things at all is that there is some suspicious sense in which we cannot say we have just the same particulars in the two pictures.

In the two world pictures the "natures" of the tokens are different. In the pure supervenience picture, the particulars the physicalist ascribes to the world extend into counterfactual worlds where their realizing spacetime worms do not exist. Because they have this modal thickness, they are different things than their spacetime worms. In the token identity picture, the particulars are thin in a strange way. The "strong" physicalist chains them more closely to actuality. This difference, and this difference alone, is what makes the one claim different from the other.

Logical supervenience is more fundamental than identity ­Do the identity claims really reflect a difference in physicalist ontology, or merely a difference to their commitments about how they are going to talk about counterfactual worlds? We have a direct way to show that it does not make an ontological difference.

The token identity world contains every spacetime worm that the logical supervenience world contains. Since each worm is, by definition, sufficient to realize the appropriate object from the logical supervenience world, all those objects still exist in the token identity world. They were there all along. The particulars from the logical supervenience world are superposed upon the particulars of the token identity world, sharing a coincident extensive base upon the spacetime worm. So the objects involved in these token identity claims are sufficient, but not necessary, to support the existence of the objects involved in the logical supervenience claims. The idea that the "strong" physicalist had somehow emptied the world of some of its contents by adding token identity claims to the pure supervenience world picture turns out to be a shallow conceit. All that really happens is that the "strong" physicalist refuses to recognize the particulars from the pure supervenience picture.

How can we have this unusual situation involving superposed particulars? What do the differences between the "strong" and the "weak" physicalist points of view really amount to? Within physicalism, these facts about identity across worlds, for high­level objects, reflect a conceptual commitment to carving the world up in a certain way. These superpositions merely represent the different potential commitments. They do not reflect a deeper ontological fact. In short, token identity claims both require and affect commitments about how we are willing to regard the existence of things across counterfactual situations.

As a result, a "weak" physicalist who claims only logical supervenience of the high­level facts upon the physical facts is not really making a weaker ontological claim than one who, in addition, makes a variety of identity claims. They are, in a sense, making different ontological claims about the natures of particulars, but these differences are of an innocent sort. Let's look more deeply into why this is.

5.4 Why does identity matter so little to physicalism?

To find an answer to this question, we must reflect on the kind of relation that identity is. Identity is supposed to be that funny relation that everything holds to itself, and to nothing else. Chuckie the chair is identical to itself, and Sadie the sofa is identical to itself, and, if they are different objects, they are not identical to each other. Furthermore, not being identical to itself is impossible for Chuckie. In short, Chuckie's being identical to itself is an essential property of Chuckie.

How do properties of objects get to be essential properties of those objects within physicalism? Standardly, necessities may hold de re or de dicto. De dicto necessities are propositional, holding between or for pieces of language because of meaning. If the necessity is de re, then the essence is considered part of the object itself. Physicalists can admit de re essences, but reflection on how such essences arise in the physicalist picture makes it implausible that identity could be such an essence. I will use H2O as the de re essence of water as my example.

On the assumption that water is identical to H2O, it must be identical to it in every possible world. Thus, H2O is an essence and, clearly, de re, since H2O is manifestly a non­linguistic, concrete entity. Given that H2O is substantial, a question remains regarding what grounds the fact that it is an essence. How did H2O get to be the essence of water? In the previous chapter, I observed that our concept of water has a context dependent intension. Once we had investigated the context successfully, we could fix the specific value of this intension. From that point, the semantic rules of our language dictated that we should rigidly designate H2O, and nothing else, as water across all possible worlds. For the physicalist this de re essence is innocent as its existence as an essence is merely the consequence of a prior semantic intension. The achievement of "essencehood" is just a kind of coronation. Therefore, the existence of "essencehood" is no further, troubling ontological fact. While the essence of a body of water may be a de re property of it (e.g., being a form of H2O), this further fact about H2O, the fact of its being the essence, is not a de re property. It is grounded in the semantics of our language.

Identity is not an entity ­Like the question of essencehood, the question of identity rests partially on discovered empirical facts, and partly on conceptual criteria for assimilating those facts into a coherent scheme. H2O may be identical to water, but, if it is, the fact that it is identical to water does not designate a distinct ontological fact. 'Identity' does not refer de re, at least when a physicalist applies it to high­level objects.

The quick way to convince oneself of the semantic roots of identity is to try to construct a purely empirical argument that H2O is identical to water. To make it concrete, consider the opposing position that it is simply one of many possible realizations of water (e.g., XYZ). In this imagined disagreement, what empirical fact do you and your opponent disagree on? Clearly, you agree on all the facts of chemistry. You may agree, for instance, on the location and distribution of H2O in our environment, and the properties that such distributions should have.

The dispute quickly peters out, as disputants readily realize they are arguing over semantics. Chemistry is enough to convince us that water's properties logically supervene on the properties of H2O, and the reason chemistry is so convincing is because chemical theory entails that H2O will have properties that make it indiscernible from those we know water has. The entailment grounds the logical supervenience. Whether one concludes from this that H2O is identical to water, or merely a realization base for water, is a decidedly secondary and less interesting question. The proper move is to explain to the proponent of the "realization" view that he is using the term 'water' in an aberrant way, and then leave it to him to decide whether he wishes to continue with that use.

Mark Johnston (1997) has argued that, in fact, water is not identical to H2O. Constitution is the relation he suggests holds instead. To argue his point, he notes that if water = H2O, then so does water vapor, and so does snow. Then, by the transitivity of identity, water vapor = snow, which seems quite wrong. His defense of this argument is very interesting, and I think that at least his negative conclusion goes through: the relation that holds between water and H2O is not identity. Whether his positive conclusion goes through, that the relation is constitution, depends on the logic of constitution, which he leaves obscure. At the very least his positive position requires exploring questions about what it means for one type to constitute another, since the more intuitive idea is that tokens constitute other tokens.

Johnston's arguments are relevant to our discussion because they show how little of scientific interest hinges on this debate. Whether H2O is identical to water, realizes it, constitutes it, or something else, is chiefly a matter of philosophical interest. The exploration is for a relation whose logic matches that of the intuitive relations between H2O and water, where those depend largely on the meanings of the terms. Regardless of the result, science will be unaffected as science has already demonstrated entailments between the properties of H2O and the properties of the various manifest forms of water. These entailment relations, which ground logical supervenience claims, really carry the reductive content of the theory.(4)

So the physicalist cannot claim that identity is a de re property like H2O. Since 'identity' does not refer to some kind of mysterious property or entity that things possess, identity claims cannot be primitive. The truth of an identity has to rest upon some further facts rather than referring primitively to some mysterious essential property. Like facts about "essencehood," facts about identity are partly imposed on the world, as a way of coherently organizing the facts about the things physicalists claim are identical.

Along with the anti­physicalists, the proponents of the primitive identity hypothesis admit the in principle absence of an entailment from the physical to the phenomenal facts. Thus both sides agree that avenue is closed. Accordingly, those antecedently convinced that an identity holds should be looking for some facts about an underlying kind, a kind ontologically deeper than the physical, and which will provide a set of facts from which an identity claim follows. Pursuit of such an underlying kind is exactly the quest the last two parts of this book will embark upon.

As far as I can see, the only way the physicalist may save identity as a de re property that we may primitively refer to is to appeal to haecceities. Under that kind of essentialism, the question of change in identity is a further question even after we become perfectly aware of the change in its physical configuration. No further physical investigation will help us to settle such disagreements. Yet, by hypothesis, no further conceptual or linguistic investigation will help us either. Such things must be mysterious, private and unknowable. Quarks may have haecceities, but configurations of quarks do not have further haecceities all their own. Consequently, such haecceities are not compatible with physicalism.

The origins of identity ­ The identities of ordinary objects, for the physicalist, must result largely from standards of human categorization, grounded importantly in meaning and in the ways we deal with objects around us. The tie between identity and counterfactual judgements is easy to understand once one considers our place in the world. We are creatures wandering in time, trapped amid a world of constant flux. Stabilizing our environment is a challenge of paramount importance to our survival. Our situation makes the task of re­identifying things across differences in time, place, and constitution deeply analogous to the task of identifying things across different counterfactual worlds. So concepts with rich resources for helping us to judge counterfactual truths about identity are the best tools for the re­identification tasks the actual world presents us with.

Our concepts must always hold open a range of possible future states of the world as able to contain largely the same objects, even though the world will realize only one of those possibilities. A concept with identity conditions that are too narrow would be useless. We would rarely be able to ground expectations, reactions, emotional attachment, and other important actions that issue from a conceptualization of something as an enduring object. When investigating the world, we obtain a variety of scattered facts. The roots of an identity claim find purchase precisely because of the kinds of unifying relations that hold between these scattered facts. The needed unification comes from entailment, as entailment is ultimately a kind of containment relation. If the A facts entail the B facts, then, in some sense, they contain them. The B facts cannot then be anything new in the world.

The difference between our "weak" and "strong" physicalist should be clear now. It is most definitely not that one is somehow relying on meaning and conceptual truth while the other avoids such appeals. The "weak"physicalist who only makes a logical supervenience claim is announcing a commitment to continue using our ordinary concepts in describing both this world and counterfactual worlds. The "strong" physicalist who insists on the claims of token identity is basically adopting a bizarre conceptual scheme. The "strong" physicalist urges us to replace the ordinary concepts we use to carve up the world with counterfactually "thin" ones. Therefore, adopting the token identity claims just reflects a strange (and relatively minor) change in our commitment to using certain concepts rather than others when talking about the world. With these lessons in mind, we can show exactly what is wrong with the primitive identity claims.

5.5 The incoherence of the primitive identity view

The strategy in this section is to build an argument reducing the primitive identity view to a contradiction, and then to defend the crucial premise. To acknowledge the role that meaning is playing in the anti­physicalist arguments we must appeal to the entailment thesis. I will assume that if [](A-->B), then the A facts entail the B facts. The modal antecedent, [](A-->B), expresses that the B facts logically supervene on the A facts.

The formal argument ­ I will give the argument first, and defend its use of the entailment thesis afterwards. Let PB be the set of true propositions about the physical nature of someone's brain, and Exp represent the set of true propositions about that person's phenomenal experience. Assume that the propositions in these two sets are about token physical facts, and therefore refer, on the one hand, to particular elementary particles and their states, and, on the other hand, particular phenomenal properties within a phenomenal manifold. The domain of quantification will be the union of those two sets of propositions. I will use a ref(...) function to designate the facts that make the propositions true, stipulating that ref(...) returns the actual world reference in every possible world, and I will also assume our language contains a truth predicate.(5) Finally, I will construe the hypothesis of primitive identity as identity without entailment. We can then derive:

(1) ref(Exp) = ref(PB) Primitive Identity Hypothesis

(2) ~[] (PB--> Exp) Failure of entailment, plus The Entailment Thesis

(3) [] (ref(Exp)=ref(PB)) Necessity of Identity

(4) [](Ax)((ref(x)=ref(x)) --> (true(x) --> true(x))) Logical Truth(6)

(5) [](ref(PB)=ref(PB)) --> (true(PB)--> true(PB))) UI

(6) []((ref(PB)=ref(Exp)) --> (true(PB)--> true(Exp))) Substituting ref(Exp)/ref(PB) in (5), using (3), and also substituting true(Exp)/true(PB)(7)

(7) [] (true(PB)--> true(Exp)) modus ponens, (3) and (6)

(8) [] (PB--> Exp) substitution using the T­Schema

Lines (3) and (8) contradict one another, so a primitive identity theorist must challenge one of the assumptions of the argument. We can now see the connection between logical supervenience and identity formally, as identity plus failure of logical supervenience leads to a contradiction. Although the logic of the argument is standard, it contains two contentious steps. The substitutions involved in the move from step five to step six may be one point of contention, especially the substitution of true(Exp) for true(PB). These substitutions should be harmless on the assumption that identities are necessary, and given that we have stipulated that the propositions involved are about token facts. Since the physical propositions involved are about tokens, they must have the same physical truth­makers in every world. If a physicalist wishes to challenge this substitution, it seems that person must also admit that a possible world exists in which (a) the propositions about those physical tokens from our world are still true, but (b) the propositions about the phenomenal facts those tokens are supposedly identical to are not true. Should (a) and (b) both be true, there's little hope that the physicalist could resist the possibility of the Zombie world. Physicalists might win the battle by challenging that substitution, but they would lose the war.

Defending the argument ­ The truly contentious move is the use of the entailment thesis in step two, as it licenses the move from the PB­facts failing to entail the Exp­facts, to ¬[](PB --> Exp). Some physicalists, for example Lycan (1996) and Tye (1995), maintain that the entailments only fail because of hyperintensional differences between the facts. Such hyperintensional failures, they claim, do not license an invocation of the entailment thesis. If counterexamples to the entailment thesis exist, the anti­physicalist must maintain that it is valid within a limited scope, and that its use within the argument falls within that scope.

That is exactly the position that I will defend. My defense of this use of the entailment thesis is an argument by cases. The primitive identity position is that the material conditional PB-->Exp is true in all possible worlds, despite the lack of entailment from PB to Exp. As we traverse logical space, Exp is not going to be true whenever PB is true just by accident. To meet the minimal meaning standard on necessity statements, the truth of the necessity requires an explanation. Only three "mechanisms" exist that can ensure the truth of a material conditional across all of logical space: ( i) The antecedent facts may entail the consequent facts, so that their truth guarantees the truth of the consequent facts; (ii) the consequent may express a necessary truth; or (iii) the antecedent and consequent may express empirically identical facts despite the hyperintensional difference that blocks entailment.

Hyperintensional differences and identity ­ The primitive identity theorists agree with the anti­physicalist in rejecting case (i). Case (ii) seems to provide some arguable counter­examples, but they are not helpful to the physicalist. For instance, it might seem plausible to some that the fact that Trey is married to Carol fails to entail the fact that 2 + 2 =4 (in the fine grained sense of entail that I am using in this chapter). Nevertheless it is necessary that if Trey is married to Carol, then 2 + 2 = 4, as the first is never true at some world where the second is false.

This kind of counter­example to the entailment thesis does not do any work for the primitive identity theorist. The reason is that the two facts, "Trey is married to Carol." and "2 + 2 = 4", are clearly not identical. If this is truly a counter­example, it is one where the entailment thesis is failing precisely when the two facts are not identical. The primitive identity theorist is postulating an identity between the sets of facts, so pointing out that the entailment thesis fails in cases where the facts are not identical makes no headway against the anti­physicalist's argument.

Note that the counter­example may fail even if the two facts are intensionally equivalent. Consider the two facts, "Everything is self­identical." and "2 + 2 = 4." These facts are both true at precisely the same worlds, arguably one does not entail the other, and yet they provide no harmful counterexample to the anti­physicalist's use of the entailment thesis. The reason is that these facts are not plausibly identical either, despite their intensional equivalence. It does not seem that intensional equivalence is the identity condition for facts, as a hyperintensional difference between facts may individuate them ontologically as well as epistemologically.

The appeals to Kripke and Putnam ­ This takes us to case (iii), the claimed empirical identity between the facts. As we have already discussed, this is a slender thread as identity claims fail very easily. Even ordinary token identity claims for ordinary objects fail. Also, typically, identity claims presuppose a logical supervenience relation that is founded on entailments that help fix the needed facts about co­extensiveness(8) (of course, co­extensiveness is not sufficient for identity, but it is necessary).

The primitive identity theorists, because they deny entailment, appear to be denying the role of logical supervenience in establishing identity. They claim, instead, that the logical supervenience of one kind of fact on the other may be a consequence of the identity rather than a presupposition of it. As the earlier discussion showed, this kind of position may work, but it seems that it may work only if identity refers primitively to some kind of haecceity possessed by these high­level objects, properties, or states. The reason physicalists will be driven to this extreme claim is that, otherwise, the facts about identity will have a supervenience base. Since supervenience is a necessitation relation, and the necessitation in this case cannot be nomological, the facts about the identity will have to logically supervene on other kinds of facts.(9) The primitive identity theorist needs haecceities to stand in as the bearers of a thing's identity, playing the role of a de re essential property (or, perhaps, not quite a property, but a sui generis kind of essence). This is an extraordinary position, especially for a physicalist.

So what does this extraordinary position buy physicalists? By appealing to it, the physicalist avoids the need to postulate epistemically opaque necessities across all possible worlds. By endorsing the identity claim, they enable themselves to make use of the apparatus of rigid designation to make the necessity transparent. Still, they pay a price. The price is that they must posit an opaque identity in the actual world. Is this move legitimate? Is it theoretically legitimate to swap opaque necessities for opaque identities?

The only published justifications rely on Kripke and Putnam­type cases. We have already seen that Putnam and Kripke did not discover an epistemically inaccessible kind of necessity. Did they discover an epistemically inaccessible kind of identity instead? Let us consider the two facts, "Samuel Clemens wrote Tom Sawyer." and "Mark Twain wrote Tom Sawyer." The physicalist needs to maintain that these are identical facts, and that one does not entail the other. If these two things are true, the necessity follows transparently from a fact about identity plus rigid designation. And, the story goes, the identity is opaque.

As a first reply, it is worth pointing out how unclear it is that physicalist's are entitled to trivialize the informational differences that make facts hyperintensionally different. The fact that one is the successor of zero is only hyperintensionally different from the fact that Fermat's Last Theorem is true. That is not plausibly just an "epistemic" difference that hides ontological identity, as it is not plausible that all mathematical facts are identical.

In the case at hand, the hyperintensional difference between the Mark Twain and Samuel Clemens facts is grounded in facts about the differing chains of reference. This is a real ontological difference involving those facts, as chains of reference are as much a part of the furniture of the world as are tables and chairs. A similar point holds about the facts involved in the anti­physicalist arguments. The conditional at issue in the anti­physicalist arguments has the form, "All the physical facts --> The phenomenal facts are like this" where the indexical "this"points to some piece(s) of phenomenal information. Here, the hyperintensional difference between the facts must be grounded in some difference in the world, even if it is just a difference in modes of reference.

Remarkably, this small difference in content is all the anti­physicalists need for their arguments to go through. If this difference holds, then the the truth of the consequent involves information about the world not recoverable from the antecedent. Some factual difference must ground this informational difference, just as the difference in the chains of reference grounds the hyperintensional difference between the Mark Twain and Samuel Clemens facts. These points about hyperintensions will be controversial, though, as these issues involving intensions, hyperintensions, and the identity conditions of facts are difficult to resolve, and involve obscurities.

Fortunately, the anti­physicalists do not have to rest their case on such rarefied points. As a second reply, the anti­physicalists may note that, even granting the de re identity between the Mark Twain and Samuel Clemens facts, that kind of identity is not enough to save physicalism. The relevant point is that these de re considerations only come into play after we have decided the facts about whether or not things are identical. The real work is done in deciding whether 'Mark Twain' and 'Samuel Clemens' refer to the same individual or not. Analogously, the real work the primitive identity theorist has to do comes in making the case that the physical facts about the brain are identical to the phenomenal facts about consciousness. This comes down to making a case for opaque identities of reference.

This point is precisely where the Mark Twain/Samuel Clemens kind of case becomes misleading. The misleading aspect is that, as it turns out, this one Mark Twain fact that we have does not entail the Samuel Clemens fact that we have. As a matter practice, sometimes we endorse empirical identities even though all of the facts that we possess may not be sufficient to entail the identity. Properly understood, this practice is harmless to the anti­physicalist arguments as it gives no support to the claim that identity is a primitive relation that may hold even though all of the other facts, period, do not entail it. In the standard cases we do not only fail to have all the facts, we know that we do not have them all. We only draw conclusions about identities, like the Mark Twain and Samuel Clemens identity, when the facts that we do have suggest that possession of all the facts would make the identity transparent.

What "making the identity transparent" comes to is showing, on independent grounds, an indiscernability between the entities, and inferring the identity as the best explanation for the independently demonstrated indiscernability. For instance, having all the Samuel Clemens facts would mean having facts such as there was a man christened 'Samuel Clemens' at birth who grew up to write novels under the pseudonym 'Mark Twain.' Having all the appropriate facts would include having the fact that they were born in the same place, at the same time, to the same mother. No one can reasonably doubt that possessing all the facts would enable us to decide the identity question in a transparent way, at least in principle.

The other standard examples of empirical identities have this same feature. Having all the facts about the Morning Star includes the fact that it is in the same position of the earth's sky, at the same time, as Venus. Using background assumptions (like the assumption that two things cannot occupy the same place at the same time), these facts are enough to entail the identity. The identity of the Evening Star and Venus is entailed in a similar way. Through the transitivity of identity, the identity of the Morning Star and the Evening Star is entailed. Even the identity of water and H2O, for those who accept that identity, relies on a perceived entailment relation between the properties of the types, although I will defer discussion of that case until the next chapter.

The fact that we never really have all the facts means that we often justify identity claims on grounds less compelling than an entailment that establishes indiscernability. But these justifications are always of the sort that are meant to convince us that there would be such an entailment, if only we had all the facts. In general, if facts A and B are only identical a posteriori, so that there is no entailment directly between them, that is because some further set of facts C exists such that C does not contain the identity statement, but A, B, and C together entail the indiscernability of A and B. The relevant lesson from the Kripke and Putnam cases is this: if there is an identity between facts despite the absence of entailment, then the identity itself should be transparently grounded in some larger set of facts that entail indiscernability, and the entailment does not require a circular appeal to the identity at issue.

Pointing to Kripke and Putnam's cases, it seems, does no work against the anti­physicalist arguments, as the anti­physicalist arguments are designed to show that even all the physical facts fail to entail the facts about consciousness. Since primitive identity theorists agree with this conclusion, their position is dead in the water. To put it bluntly, appeals to mystical identities are no more legitimate than appeals to mystical necessities. If there is an a posteriori identity between brain states and conscious states, it must be in virtue of some set of facts larger than the physical facts alone.

Summary ­ The anti­physicalist holds that the physical facts do not entail the phenomenal facts, and therefore the facts about consciousness do not logically supervene on the physical facts. As a corollary, the anti­physicalist maintains that they cannot be identical to (or constituted by) the physical facts either. Noting that the primitive identity theorists agree with the anti­physicalist that the physical facts do not entail the phenomenal facts, the anti­physicalist provides an argument that maintaining a primitive identity between these facts in the face of this failure yields an incoherent position.

The primitive identity theorists challenge this argument by claiming a misuse of the entailment thesis. That is the thesis that licenses the move from the failure of the A facts to entail the B facts, to ¬[](A --> B). In trying to block this inference, we discover that the only cases where the inference fails are those cases where the facts are not relevantly identical in the first place. The most tempting way of undermining the thesis looks to Kripke and Putnam for apparent counterexamples. However, nothing in the Kripke and Putnam cases provides a harmful counterexample to way the anti­physicalists are using the thesis. So the anti­physicalist's use of the entailment thesis goes through under the primitive identity theorist's assumptions. The primitive identity theorists seem to be advocating a conceptually incoherent position.

1. For the arguments in this chapter, I am construing entailment as a kind of containment relation. I am thinking of it as follows: the A­facts entail the B­facts just in case the A­facts contain information that makes their assertion inconsistent with a denial of the B­facts, given just the concepts involved in establishing the two sets of facts. I am using a very "fine­grained" notion of fact such that facts are as finely grained as propositional senses. In terms of strength, I am regarding material implication < strict implication < entailment. This "fine­grained" notion increases my burden in this chapter since I am arguing that this very fine grained relation is necessary to establish an identity.

2. Strictly speaking, the A­facts will often only entail a set of facts, the B'­facts, indiscernible from the B­facts, but possibly differing from them due to an irreducible indexical component that might exist as part of the identity conditions on the B­facts. In practice, we finesse the indexicality by allowing ourselves to primitively postulate these irreducible indexical facts as part of the entailment base, where this practice is justified by an appeal to the independently established indiscernibility plus inference to the best explanation. The A­facts plus these indexical facts will entail the B­facts proper.

3. Because the arguments in the last chapter cover the realization relation, they undermine the position in Tye (1995). There, Tye formulates a physicalist position by using an opaque realization relation. Tye's chief objection to anti­physicalism is that it implies epiphenomenalism about consciousness. The Liberal Naturalist position I will develop later shows how to save anti­physicalism from that charge without falling into interactionism.

4. Johnston surprisingly misses this point. In a postscript to his paper he claims that the general failure of identity undermines the anti­physicalist position. His text suggests that he thinks the form of the anti­physicalist arguments runs from a failure of identity, to a failure of reduction. His defense of physicalism seems to fail once one sees the important reductive role played by logical supervenience relations that are founded on entailments. From our current perspective, one moral of his own arguments is simply that the ontological importance of identity is derivative in the first place. Furthermore, constitution claims, no less than identity claims, require logical supervenience relations that are founded on entailments.

5. One may do this using the revision theory of Gupta and Belnap (1993). Alternatively, one could treat the argument as being given in a meta­language.

6. The argument for this premise is straightforward, as there will be no world in which the consequent is false.

7. The substitution of Ph for PB is benign given the material conditional, and the presumed identity of their truth­makers. See the discussion in the text.

8. Recall that these same points hold against constitution and realization claims.

9. Because, as I argued in the last chapter, the physicalist may not appeal to a primitive "metaphysical" necessity to do the job.


Chapter 6

The Objection from Holism and the Rejection of Analyticity

6.1. Quine's rejection of the analytic/synthetic distinction

The previous four chapters have defended the anti­physicalist arguments against three common objections: the objection that they are arguments from ignorance; the objection that appeals to metaphysical necessity; and the objection that appeals to primitive identity. In their attempts at restoration, some physicalists appeal to a fourth kind of objection to the anti­physicalist arguments. In following this fourth strategy, these physicalists reject the idea of logical supervenience completely, complaining that it illegitimately relies on the analytic/synthetic distinction. This class of objectors typically embrace some form of holism, either of meaning or confirmation. Their claim is that the anti­physicalist arguments fail within a holist framework, since the conceptual connections the arguments appeal to do not exist.

In this chapter, I will address confirmation holism and meaning holism separately, arguing that neither undermines the anti­physicalist arguments. I will defend the position that the anti­physicalist arguments are more robust than the holist suspects. The entailments they appeal to are no more suspect than any others in use inside or outside of science, and the explanation of consciousness remains problematic even within a holist framework.

A quick overview of Two Dogmas ­ Before responding to the holist's worries, I will first briefly summarize Quine's arguments against the analytic/synthetic distinction in the The Two Dogmas Of Empiricism (1963). The alternative Quine proposes is a kind of radical meaning holism. Although holism itself may come in varieties other than Quine's, I will focus on Quine's version and let my response to it serve as a template for responding to variants.

The philosophical distinction between analytic statements and synthetic ones is this. Analytic statements are those that are true in virtue of the meanings of their terms, but synthetic statements are those whose truth involves matters of contingent fact in the world. The classic example of an analytic statement is, "Bachelors are unmarried." An example of a synthetic statement is, "The Braves beat the Expos today."

In Quine's extraordinarily influential Two Dogmas, he expresses strong skeptical concerns about meaning and analyticity. Quine was especially motivated by reflection on Carnap's work, particularly in the Aufbau (1928), where Carnap attempted and failed to build a phenomenalist construction of our knowledge, and then The Logical Syntax of Language (1937), among later works, in which Carnap tried to move forward.

The historical train of thought Quine was meditating on was Logical Empiricism, and its doctrine that the meaning of empirical terms may be derived from experience. This doctrine was a descendant of conceptual empiricism, the position that every empirical concept derives its meaningfulness from its relations to sense impressions. Within the positivist frameworks prominent when Quine was young, concepts became terms, and the question was transformed into one that asked if terms could be "translated", one by one, between various linguistic frameworks in a constructive manner. In the Aufbau, Carnap treated phenomenal language as primary, but this choice was strictly irrelevant to the larger concerns of the radical empiricists (Friedman 1987). The core idea of translation between different kinds of languages, like phenomenal language and the languages of science, was much more important than questions about which is foundational. The latter question was considered purely "metaphysical". The notion of translation used in Carnap's reductive project relied on things like "meaning postulates" whose plausibility rested on intuitive notions of analyticity.

In Two Dogmas, Quine asked how we are to understand what is meant by 'analytic' in the claim that

(1) No bachelor is married

is analytically true. Following Frege, Quine proposed that it was because

(2) No unmarried man is married

is logically true, and statement (1) is gained from (2) by substitution of synonymous terms. Having diagnosed the first class of these "analytical" statements in terms of logical truth plus synonymy of terms, Quine poses the problem of how to analyze synonymy. According to Quine, synonymy "is in no less need of clarification than analyticity itself." He considers a variety of proposals for understanding analyticity besides synonymy, including verification criteria and semantical rules. For modern readers, though, the force of his arguments rests on the failure to account for synonymy.

He rejects the idea that we can analyze synonymy in terms of definition because the relevant notion of definition presupposes synonymy. Leaving aside the less plausible suggestions concerning verification criteria and semantical rules, Quine's arguments consisted in his failing to find an analysis of what synonymy of terms is, and therefore failing to find an analysis of the analyticity of certain statements, taken in isolation. He concludes that the idea of analyticity is too murky to trust. Those moved by Quine's worries have not found subsequent attempts to articulate the distinction to be satisfying, and skepticism about the distinction remains.

Based on these problems, Quine rejects the notion of analyticity of statements, and therefore the idea that there are statements that are true in virtue of meaning alone. Quine's efforts fall short of showing that no satisfactory account of analyticity exists, but he feels he has shown that we have no good reason to believe such an account exists. Consequently, those who continue to believe in it are being dogmatic.

What moral does Quine actually draw from his exploration? How can we have anything entitled to the name 'empiricism' if we cannot derive the meanings of ordinary language terms from experience, or produce a class of experiences that circumscribe truth­conditions? Quine needed to re­orient empiricism by connecting knowledge exclusively to experience, without thereby presupposing that these connections involve reductions of individual statements, or leave room for statements that are true in virtue of meaning alone. Here are a few quotes from the Two Dogmas where he gives his solution [all emphases added]:

My countersuggestion, issuing essentially from Carnap's doctrine of the physical world in the Aufbau, is that our statements about the external world face the tribunal of experience not individually but only as a corporate body.

My present suggestion is that it is nonsense, and the root of much nonsense, to speak of a linguistic component and a factual component in the truth of any individual statement. Taken collectively, science has its double dependence upon language and experience.

. . . .total science is like a field of force whose boundary conditions are experience. A conflict with experience at the periphery occasions readjustments in the interior of the field. . . .

As Frege had earlier rejected the idea that individual terms were meaningful in isolation from statements, Quine's radical move was to reject the idea that statements are meaningful in isolation from theory. Even theory, Quine suggested, was not meaningful except through the way that it is embedded in the whole of language. Thus, the meaning of every term and every statement implicitly relies on its relations to every other term and statement in the language, and, ultimately, it is language as a whole that is responsive to experience.

Preliminary rebuttal ­ Before beginning my main critique of the meaning holist's objections, I need to address two other possible objections that also come from Quine, and are closely related to his holism. The first objection issues from the part of Quine's program that insists on taking an extensional stance towards meaning. Since the anti­physicalist arguments are explicitly modal, the entailments they rely on require an intensional semantics. Extensionalism, if it could provide an adequate account of meaning, would undermine the arguments. My answer to this objection starts by observing how the plausibility of Quine's extensional stance relies on his metaphysics, and his general rejection of objective modal truths. Consider this passage from "Necessary Truth" (1966),

In principle, therefore, I see no higher or more austere necessity than natural necessity; and in natural necessity, or our attributions of it, I see only Hume's regularities, culminating here and there in what passes for an explanatory trait or the promise of it.

What if Quine's metaphysical commitment to a Humean view of causation is wrong? The inexorably modal character of substantive views of causation means that, if we live in a non­Humean world, we live in a world that contains objective modal facts. A purely extensional account of meaning in a world with objective modal facts woefully mismatches thought with the world it is directed at. In that case, insisting on taking an extensional stance towards meaning becomes unreasonable. The extensional stance becomes otiose. My answer to objections issuing from the extensional stance leverages this dependency, and is therefore contained in chapters ten through thirteen, and explicitly formulated in chapter sixteen.

In chapter ten I critique the Humean view, producing decisive reasons for rejecting it. By chapter thirteen, I will be in position to produce a very strong argument for realism about possibility. Those arguments constitute a justification for rejecting the extensional stance in language, and turning instead to an intensional semantics. I believe that some interesting reasons exist for suspecting that Quine's brand of holism relies on his endorsement of extensionality. If his meaning holism relies essentially on the extensional stance then the two positions fail together. However, for the sake of argument I will assume that they can be separated, and that an intensional holism is viable. In the rest of this chapter, I will critique worries arising from meaning holism in abstraction from any support that position might receive from the extensional stance.

A second objection comes from Quine's quasi­behaviorism, and is directed at the way the anti­physicalist uses the term "experience". To do work in the anti­physicalist arguments it clearly cannot mean what Quine takes it to mean (e.g., Quine 1992): stimulation of the sensory nerve endings. That is not what a theory of consciousness will be about. 'Experience', as is often pointed out, is ambiguous and also refers to phenomenal experience, which involves the first­person experiencing of certain kinds of qualities. We gain knowledge of these qualities by having experiences of Quine's sort, at least.

This presents the holists with a dilemma. Either they insist, with Quine, that "experience" on "the periphery" is univocal, referring behavioristically to "stimulations of sensory nerve endings," or they admit to the legitimacy of experience, phenomenally construed. It is easy to see that taking the first horn of the dilemma does not help the holist. Hung on that horn, the holists are no longer raising an objection to the form of the argument, but rather to one of its premises. Furthermore, this premise is supported by a strong auxilliary argument that qualia are observables, an argument given in section three of chapter two. In the end, horn one is a bald­faced eliminativism and lacks all the plausibility that the position always lacks.

The second case is different. The qualities of experience (in the phenomenal sense) show up as close to the periphery of experience as possible. Our concept of phenomenal experience resides in something like Searle's Background (1983)(1), waiting for science to hook up with it. More strongly, its instances pervade The Background. Various phenomenal concepts are so ubiquitous, our cognitive system seems close to hardwired in its stubborn insistence on throwing the phenomenal properties into the world by latching them onto our intensional objects: that feeling is in my foot; that color is on the wall; the rose smells so sweet. The qualities of phenomenal experience are everywhere, literally fused into our model of the world as intensional support for our understanding of ordinary objects. Any systematic failure of theory, as a whole, to account for the phenomenal qualities of experience constitutes a massive failure of the internal part of Quine's theoretical field (physical science) to account for pervasive elements on the periphery (roughly, The Background). By Quine's own lights, it is the internal elements of the field that must buckle.

When these two elements of Quine's philosophy are removed, the remainder is an intensional holism in which phenomenal consciousness exists as a central target of explanation on the periphery. It is this position that I shall address in the rest of this chapter. We will see that this remainder is quite harmless to the anti­physicalist arguments.

6.2 Meaning holism and the anti­physicalist arguments

Quine's conclusions have been significant within philosophy, especially the philosophies of science and language. Quine's advocacy of meaning holism finds different degrees of expression, but the rejection of the distinction between analytic and synthetic sentences is now at least quasi­orthodoxy in modern Anglo philosophy. From the viewpoint of the holist the problem seems to be this. The anti­physicalist arguments rely on the claim that the facts of consciousness fail to logically supervene on the physical facts. Their arguments against logical supervenience crucially involve meaning: given an analysis of the supervenience base and the target facts about consciousness, the proposed base facts seem unable to entail the facts about consciousness. The appeals to entailments in these arguments look like appeals to proposed analytic truths, so the arguments are suspect.

The stability of conditionals ­ The entailments needed by the anti­physicalist should be interpreted as analytic, in some sense, but they need not be understood in a way that violates any strictures that meaning holism puts on entailment. The anti­physicalist arguments evade sin because they do not rely on judging the truth of single statements independently of theory, nor on accommodating single contrary experiences to theory.

The anti­physicalist arguments are explicitly concerned with systematizing the relation of physical theory as a whole to phenomenal experience as a whole. They do not traffic in isolated statements of theory, comparing them to isolated experiences. Because their arguments have this ambitious form, anti­physicalists can reconstruct them within a holist framework. Within that framework, they have the form of demonstrations that a certain class of theories fail to predict all the facts about experience. Within that reconstruction, we can see the anti­physicalist arguments as special instances of ordinary scientific falsification. On the face of it, nothing in Quine's critique rules out such arguments.

Some people might be uncomfortable even with the idea of entailments from theory as a whole to various sets of facts as a whole. These conditionals from the whole theory to those facts might be seen as analytic themselves. Such worries can be overdone, and are in this case. Nothing in Quine's arguments forbid entailments from theories as a whole to consequences of that theory, so long as the larger embedding network for that theory is held relatively constant.

For example, Euclid's axioms continue to entail the theorems of Euclidean geometry, despite the empirical pre­eminence of non­Euclidean theories of space. These entailments are safe since they follow from the set of axioms as a group, and nothing in the larger web has caused us to reject the rules of inference or interpretations they rely on. We simply reject the idea that they accurately represent the geometry of space, but that kind of empirical failure very rarely, if ever, affects the truth of the purely conditional conclusions at issue. The anti­physicalist arguments only require insulated, and therefore very safe, entailments of this kind.

The role of systematization ­ Some people might feel uncomfortable because the anti­physicalists do not explicitly refer to each and every possible physical theory, or to each and every detail of current theory in drawing their conclusions. The worry might be that one cannot show failure of prediction without appealing to the specifics of a particular theory. Holists need to do much work to make this into a forceful objection, if they can do it at all. The anti­physicalist arguments do not seem to require such detail because they are concerned with systematizing the failure of physical theory, demonstrating the in­principle obstacles. Such in­principle arguments appear in scientific discussion elsewhere, and we do not consider them suspect because of Quine's worries.

Although necessarily less formal, these arguments have a form similar to that used by John Bell in showing that the predictions of quantum mechanics could not be reproduced by any local hidden variable theory. In making his case, Bell did not examine the details of every single possible local theory, demonstrating for each how it failed. Instead, he abstracted out the general conditions that limit any such theory, and showed how no theory meeting those conditions could predict the same results as standard QM. In doing this, Bell showed how we could determine that local theories, in principle, could not do justice to experience on the periphery, conditional on QM being correct in its predictions of experience. Bell showed how, in principle, a certain class of theories fails to entail certain facts from experience.

The anti­physicalist arguments are simply ways of making the same kind of point about the relation of purely physical theories to conscious experience, using the same method Bell used. They point to certain features shared by an extraordinarily broad class of theories, and argue that a theory with just those kinds of features, and no others, will fail to predict some facts about experience. We can possess all the physical facts, without being able to derive the truths about the phenomenal facts. If this is true, it seems that there could be a world where all the postulates of physical theory are satisfied, but in which some facts about experience (the facts concerning its phenomenal character) are absent. The kinds of entailments involved are not of a different type than science usually uses.

If one wishes to manufacture an epitaph for these entailments by calling them "analytic" in the suspect sense of that word, so much the worse for Quine's arguments against analyticity. Nothing about meaning holism invalidates Bell's Theorem. Then, prima facie, nothing about it invalidates the anti ­ physicalist arguments either, in form or content. The practice of extracting predictive content from theories is more secure than any arguments advanced against analyticity. To turn meaning holism into an objection to this form of argument will require a great deal of work, if it can even be done, and will require close discussion of the details. Perhaps the arguments violate Quinean strictures in some subtle way, but meaning holism provides no easy and obvious "in principle" objections to the forms of the arguments per se.

Definitions, ostension, and recognition ­ To violate Quine's critique of the analytic/synthetic distinction, the anti­physicalist would need to implicitly or explicitly rely on a notion of entailment precluded by the breakdown of the distinction. One can see that this is straightforwardly not the case by noting that the notion of analyticity Quine's arguments actually refute is one of statements being true simply in virtue of the meanings of their terms, isolated from theory and experience. Nothing in the rejection of such truths requires rejecting the existence of entailment relations from theories as a whole, interpreted in the context of experience, and against the background of a larger theoretical network.

The anti­physicalist arguments might cross the analytic divide if they required sharp definitions of 'experience' that were said to capture the "meaning" of the term. Fortunately these arguments are perfectly compatible with the absence of such definitions. On the face of it, we no more need such a definition of 'experience' to satisfy the entailment than we needed a sharp definition of 'dissolving' before we could test chemistry, or 'brain' before we could test neural theories. Entailments from theory to experience are produced, confirmed, and refuted all the time with only rough and ready definitions being in view.

A task exists that is far more important than producing a precise definition that captures the "meaning of the term." This task is to leverage our competence with our ordinary concepts about a targeted phenomenon, finding a source of information that is much closer to the periphery than is the theory being tested. Our source of information about the phenomenon must be firmly enough entrenched in the network that it is capable of making significant contributions to the falsification of the theory at issue, which lies further from the periphery. While Quine's arguments make theories resistant to scattered instances of disconfirming evidence, a steady stream of such information, produced at or close to the periphery, can force theory revision on the inner part of the network. This is exactly the tack that the anti­physicalists have taken, as they point to the massive failures in physical theory's attempts to account for the facts of phenomenal consciousness at the periphery. They do not point to isolated failures whose intepretation can be revised away to save the theory.

The reason entailment may exist without definition is that we gain many of the concepts used to test our theories from ostension, and from living in the world with the things that the theory is attempting to explain. All such concepts resist sharp definition. Regardless, we still enjoy a nimble grasp of their use so that we can often recognize when they are and are not being satisfied. Phenomenal experience is an example, par excellence, of something we are familiar with from ostension and through living with it. As an analogy, we did not need a sharp definition of 'gold' before we could test chemistry's predictions about it. We do not need a sharp definition of 'data print out' before we can recognize that one is being produced. Neither do we need a sharp, analytic definition of 'experience.' When testing a theory against the periphery, it is far more important to know what the phenomena is as it ordinarily appears than it is to know that the phenomena is such­and­such. This "knowing what" is the kind of knowledge by acquaintance that guides our skillful competence in gathering the information that theory tries to account for. It is skillful recognition, not definition, that we require.(2)

Experience and The Background ­ If the previous discussion is correct, there is a sense in which the entailments involved in logical supervenience claims are "true by virtue of the meanings of the terms," but it is entirely innocent. It is the exact same sense featured in entailments from physical theory to experience generally. Every successful physical theory instantiates some such entailments. All Quine's worries show is that these entailments do not exist from statements within the theory, taken in isolation. Where they do exist is from the theory as a whole to elements of "the periphery" where we live and experience. These elements of the periphery do not have to be taken as given, incorrigibly, but they are rarely if ever definable, and they do not get their meanings from the science being tested as much as from what Searle calls The Background.

The Background is that vast, unspeakably subtle and intuitively understood network of acquaintance and conceptualization that is built up from living where our concepts are working; concepts in The Background are taken from being in the world. Our grasp of the concepts that exist as part of The Background has never before been so poor that we could not judge the requisite entailments that connect theory to it, and the anti­physicalist can see no special reason to think that with consciousness, sui generis, we should begin doubting.

I have already suggested that such entailments are, in fact, extraordinarily resistant, if not immune, to empirical revision. As mentioned earlier, all the entailments between Euclid's axioms and the theorems of his geometry continue to hold even though the theory that space is Euclidean has been falsified. The entailments between Newtonian mechanics and experience hold (it still gets the perihelion of Mercury wrong), even though Newton's theory has been falsified. The same goes for Ptolemy's astronomy, and phlogistic theory, and. . . .In the usual cases empirical factors yield reasons to adjust the theory occurring in the antecedent of such entailments, or yield reasons to adjust our belief in the existence of the mismatched bit of experience in the consequent. Through it all, the entailments themselves are very robust.

Summary ­ In rejecting materialism, the anti­physicalist is simply following Quine's good advice: science is a continuation of common sense, and it continues the common sense expedient of swelling ontology to simplify theory. While application of his advice in this case might earn Quine's consternation, the fact that he might not like it himself does not undermine its rational foundations. If good arguments exist for the position, as they seem to, we are left with no other choice. The meaning holist does not seem to have an objection that carries through.

Finally, meaning holism often goes hand in hand with the notion of "epistemology naturalized", and the holist often has qualms that the anti­physicalist project is at odds with epistemology naturalized. If it is not clear already, by the end of this work it will be abundantly clear that the question of what nature is like is exactly what is at issue here. One cannot judge whether a theory is compatible with naturalized epistemology without implicitly or explicitly appealing to some view of nature. Since such views are just what are at issue here, these objections beg the most crucial questions and are entirely moot.

6.3 The objection from holism of confirmation

A second objection issues from the idea that confirmation is holisitic. This is how the objection goes. The anti­physicalist maintains that physicalism needs entailments from the physical facts to the facts about consciousness, where these entailments mark something more than mere extensional correlations. In short, the anti­physicalist claims that there are two grades of modality that ground entailments within science. The first is natural necessity and possibility, and it is used in positing contingent laws of nature. These entailments explain mere extensional correlations between distinct kinds of entities in terms of something more substantial. The second is logical possibility and necessity. Logical possibility and necessity involve truths of meaning, and may ground ontological reduction.

The holist about confirmation claims that this distinction does not exist in science as it is actually performed. In particular, the holist claims the only things that ever ground scientific identities and reductions are facts about correlatons between the micro­level story and macro­level observations. If this is true, it is no blotch on the sciences that study consciousness that they do no better.

Holists can produce an argument to back their rejection of the distinction between the different kinds of entailment. Recall my example from chapter four of an entailment yielded by micro­theory.

(1)[] [(Ax)(H2O(x) --> (Av)( x has volume v --> (Aw)(Az)(E1y)(A mole of x changes pressure by y when its mean molecular kinetic energy changes from w to z))]

At the time, I interpreted this as a statement about the functional relation between the volume, pressure and temperature of samples of H2O. Strictly speaking, though, it does not express any such thing. The problem is that the ordinary concept of "temperature" does not occur in the theory. To confirm that water has the property, and thus might be H2O, we need to assume the identity between mean molecular kinetic energy and temperature. However, since confirmation is holistic, to get the identity between MMKE and temperature, we need to confirm the identity between H2 O and water. Consequently, the low­level theory itself never entails any high­level facts. It only does so in conjunction with empirical identities, which are themselves obtained by holistically correlating the two conceptual systems. Therefore, we may conclude that purely physical brain states are identical with conscious states on the same kind of evidentiary basis that we use to conclude any kind of empirical identity: extensional correlation.

The short answer to this holist objection is simply that our microtheories do entail the high­level facts about particular phenomena if we give ourselves all other relevant empirical identities. To establish an identity, we first assume other empirical identities. We then use these other identities to derive that the entity in the theoretical domain has just the same properties as the entity in the target domain, at least as far as we know. The identity claim shows up as the last step, and we use it to explain the independently established indiscernability.

We do not, as this objection seems to suggest, use identity statements to circularly derive indiscernability. For example, on the hypothesis that the other identities hold, we do not need a circular appeal to the fact that H2O = water to derive the fact that H2O has all the properties that water exhibits. With consciousness, however, the situation is completely different. Even if we give ourselves all other empirical identities, the physical facts still do not entail the facts about consciousness. Instead, to avoid failure of prediction we have to include the proposed identity and use it as a "bridge principle" of sorts. Once this is pointed out, we can see immediately that the proposed grounds for asserting brain state identities are sui generis after all.

On a deeper level, I suggest that the global process of coordination between microtheory and macro observations is itself really a process of determining whether an entailment exists. My suggestion is that our concepts of macro phenomena pick them out by their causal roles within the high level network. For a theory to entail that something is light, for example, it must entail that there is something in the world that occupies the causal role that light occupies in our world. Quite obviously, the idea of occupying a role in a causal network requires confirming that the network exists, so the theory must also entail light's contextual facts.

The confirmation procedure for each particular identity will assume other identities, as they each assume that the network is in place. Nevertheless, no such restriction prevents the holistic entailment of the entire macro network by the network of causal relations proposed in the microtheory. The process of coordination in science can be seen as a process of confirming, on a piece by piece basis, that this holistic entailment actually holds. If the network of causal relations in the microtheory mirrors the network of relations by which we pick out the macro phenomena, then it is in fact true that the totality of micro facts entails the totality of macro facts (modulo the need to postulate the irreducible indexical facts). The entailment succeeds without circularity because the facts of explanation are relevantly distinct from the process of confirmation. That is all the anti­physicalist arguments need.

6.4 Kirk's argument for the logical supervenience of consciousness

Some physicalists see that none of the four objections discussed so far succeed against the anit­physicalist arguments. At that point, the final battle line in defending the church is to argue that consciousness does logically supervene on the physical after all. A final objection to the anti­physicalist argument is that consciousness must be functional, and we have every reason to believe that it is the physical stuff of our brains that is performing the relevant functions. This view is very widely held, but has been expressed and defended at greatest length, and with greatest force, by Robert Kirk (1994). Kirk accepts that logical supervenience is necessary for physicalism to be true, and he seems to take consciousness seriously in the sense that he accepts the existence of qualia and subjectivity. He argues explicitly that consciousness, even in the full­blooded sense, may logically supervene on the physical because the functionality of consciousness is essential to it.

Kirk's argument hinges on the idea of Neuron Replacement Therapy, which is by now a common notion in the philosophy of mind. Neuron Replacement Therapy is an imagined scenario in which our biological neurons are replaced by functionally equivalent neurons made of different stuff. The question is, "What would happen to conscious experience in such a circumstance?" In an argument that most will be sympathetic to, he claims that our experiences, if they change at all, will change in ways that we can be aware of. He then argues that it seems incoherent to claim that we are aware of such a change, but cannot in any way act on this awareness: we cannot report it, write about it, nor even act disconcerted. In short, in a normally functioning system awareness of conscious experience cannot be completely disconnected from behavior in this way.

Leveraging this point, Kirk makes the following argument. Assume that conscious experience changes when we undergo Neuron Replacement Therapy. By hypothesis, our neural replacements are functionally equivalent to our original neurons, so there is no room for behavior of any kind acknowledging the change in our experiences. Since awareness is essentially connected to behavior, we are then not aware of the change, and this is absurd. Consequently, functional organization must determine conscious experience, and physical organization determines functional organization.

Kirk's argument is a serious threat to the anti­physicalist position, but I do not think it succeeds in establishing that epiphenomenalism is not coherent. For instance, it is not clear that it establishes logical supervenience on functional organization, as opposed to natural supervenience. It also seems to trade on an equivocation in "awareness" parallel to Chalmers' distinction between functional and phenomenal notions of psychological terms. Still, I will not pursue that avenue of resistance. I bypass that kind of objection because I do not believe it is clear that epiphenomenalism is coherent. At best, the physicalist and anti­physicalist achieve a stand­off in that discussion.

The rest of this work pursues what I think is a more interesting response to Kirk's concerns. I think the central intuitions behind such arguments may be correct, at least in support of the weaker point that consciousness, where it exists, is invariant with respect to the conscious system's functional organization. Still, these intuitions do not yield the conclusion that consciousness is logically supervenient on the physical facts. Surprisingly, we shall see that the anti­physicalists have room to evade the charge that consciousness must be epiphenomenal on their view, and they can do so without falling into an interactionist position.

The trap physicalists fall into is that they overlook certain subtleties involved in deciding whether two systems are functionally identical. Two systems are functionally identical just in case they are causally isomorphic at the appropriate level of organization, meaning that they have the very same causal organization at that level. Hidden in such judgments are some subtle questions about causation, levels of nature, and causal organization at a level.

By the end of this work, I will argue that the problem undermining Kirk's reasoning is this: the facts about causal organization do not logically supervene on the physical either. Causation itself has multiple aspects. The facts about some of these aspects are not strictly implied by the explanations of physical science, or the story about functional organization that we would tell from a purely physical perspective. Whether or not a functional identity exists between Kirk's NRT systems cannot be decided independently of knowing the totality of causal facts, and these facts are not just the physical facts. So it is possible to maintain that the facts about consciousness are fixed by the functional facts about a system, but not by the physical facts about the system.

More strongly, the physical facts necessarily underdetermine the functional facts. It is not a matter of there being some contingent non­physical entity interacting with the physical. The non­physical aspects of causation that I will argue for do not interact with the physical in any way. They complete the metaphysical story about what interactions are in the first place. In short, Kirk's argument presumes that physical explanations of causation must either completely describe causation, or fail because they leave gaps to be filled in by other active entities. Since interactionism does not yield the dualist any plausible avenue of resistance against the functionalist argument, Kirk concludes that dualism is commited to the absurd position that we possess a non­physical and epiphenomenal consciousness. What we will discover in the pages that follow is that physical theory may be incomplete in ways more subtle than this. Kirk's arguments fail by taking causation for granted.

1. Searle describes The Background as a necessary grounding for intentionality that includes (at least) a network of non­representational capacities, practices, and pre­intentional assumptions.

2. Estep (1996b) contains some interesting discussions about the importance of knowledge by acquaintance in the exercise of skills.


Chapter 7

Rethinking Nature: On the Possibility of a Benign Panpsychism

7.1 Nature and the problem of consciousness

If physicalism is false, then experience will not be reducible within a purely physical framework. The irreducible character of experience implies that fundamental natural laws are either governing its existence, or governing the existence of something non­physical that it can be reduced to. These laws will be on the same level as those governing properties such as mass, motion, and gravity. Because these laws are fundamental, superempirical standards are relevant in evaluating candidate formulations of them. The argument in this chapter is an extended, critical reflection on the form those laws would take if we attempt to explain the qualitative aspect of our mental lives in a way that ties it too closely to cognition.(1) Careful reflection yields good reason to believe that our attempt to fit consciousness within an adequate naturalistic framework will force us to go beyond issues specific to psychology and the philosophy of the mind.

Consciousness shows us that our world has another fundamental aspect that we must understand if we are to understand the qualitative character of our mental lives. Now we must do the dirty work of rolling up our sleeves, figuring out just what kind of thing this other aspect is, and how we must revise our understanding of the world to smoothly integrate it.

For now, I am going to call this phenomenon the qualitative field. 'Qualitative' is just meant to capture the close relation to qualia; perhaps it subserves them in some way. I use 'field' to denote a bounded collective. Our phenomenal lives contain many distinct qualia: itches, sounds, smells, emotional tones, and tickles are examples. All these different qualia, however, are merged by nature into one subject of experience, individuated from other qualia not only by their type, but also by the field of experience to which they belong (e.g., mine and not yours). The boundaries of this field individuate subjects of experience by including and excluding feeling. We can think of the unified, bounded collection of qualia that constitutes the experience of an individual as the qualitative field associated with that individual. Liberal Naturalism must face the problem of providing a basis in nature for the existence of such a thing. If successful, it will allow us to identify the kind of individuals that qualitative fields are associated with, and also say something about the structure and content of each individual's field. Below, I show why our best strategy is to first rethink nature, and then fit consciousness within our new, enlarged understanding of the natural world. Chalmers has coined the term 'the hard problem' (1995) for the challenge of explaining phenomenal consciousness. This required rethinking of nature is really what makes the hard problem hard.

7.2 Why we must go beyond the mind

As evidence for how uncomfortable the problem makes us, we can observe how difficult it is to dislodge the physical from its accustomed place of primacy in our world view. Even anti­physicalists like Chalmers continue to pose the problem as one of how "experience arises from physical processes." Such language pushes us to think of the physical as primary, with whatever aspect of nature consciousness is a part of viewed as a mere add­on; it just "arises". Are we to think of experience as an artist's flourish? A more satisfying result would be a deeper view of nature, a view which somehow gets under physics. The fact that we need to get under physics is obscured by this "arising" intuition, which in turn is lent crucial support by the intuition that experience exists only in cognitive systems. Therefore, shedding this blinder is an important move in learning to properly appreciate the scope of the challenge.

The anti­physicalist's have taken a first step by realizing that the problem goes beyond detailing the structure and function of physical systems. This has allowed them to tease apart various problems of consciousness, bringing the hard problem of experience into sharp focus, highlighted against the background of the problems concerning function and structure. Even so, this classification does not cut finely enough. Additionally, we must tease apart the problem of conscious experience from its cognitive entanglements, and learn to see in it the more general problem of finding the basis in nature for qualitative content. That is, even the hard problem of consciousness is really two problems superposed. It is a general problem about finding the basis for these qualitative fields, their place in nature, and the laws governing them. Also, it is a specific problem about how the influence of cognition can give a qualitative field the character of consciousness. I will argue both that in principle and in fact these are separable problems, and that the need to separate them forces us to rethink our view of nature quite generally.

This is a strong conclusion, and many will just as strongly resist it so we should more clearly understand what the natural objection is. The problem poised before us is to find a place in nature for qualitative fields. The suggestion I am defending in this chapter is that such fields must have a basis more fundamental and ubiquitous than cognition. Most of us possess a natural and strong opposing intuition that these qualitative fields are quite special things peculiar to minds. We think that, since systems are only mental when they support cognition, qualitative fields must "arise" within cognitive systems alone. Intuitively, this stipulation is tempting because we want to avoid an unpalatable panpsychism. We do not want experiences popping up in unlikely places.

Things get a little tricky here, since the criteria for a system being cognitive is itself so unclear. To be fair to this common sense intuition, we cannot simply define cognition so loosely or behavioristically that an unacceptably wide net is cast. We do not want to make nearly everything cognitive simply by definition. So we can either allow a narrow definition of cognition, or, for the sake of argument, simply stipulate that the class of cognitive systems supporting experience must be like humans in certain important respects. At least, the systems in question must possess some kind of awareness, memory, perceptual ability, and conceptual capacity in the operational sense of these terms. For the purposes of what follows, we do not have to be too specific about these capacities. But we do need to be clear that they must be relatively high-level cognitive capacities. That way, we can constrict instrumental or behavioristic definitions of 'cognition' enough to save consciousness for some (ill­defined) class of intuitively acceptable systems. From this point on, I will just refer to cognitive and non­cognitive systems. Those attracted to wide definitions of cognition should understand the terms as implicitly specifying high­grade and low­grade cognitive systems, respectively.

These contrary intuitions about qualitative fields arising only in cognitive systems are initially very compelling; I held to them myself for quite awhile. Nevertheless, upon reflection they ultimately must be rejected. In what follows, I will first give reasons for rejecting them. Afterward, I will try to say something to undermine the intuitive pull they possess. Essentially, I will try to make more palatable the idea that our intuitions here are misguided, and they should topple under the weight of the good reasons we have for ignoring them.

So, just what is wrong with the idea that qualitative fields are associated only with cognitive systems? The trouble is that it is very difficult to see how an appropriate set of fundamental laws can meet the constraint. For example, what class of base properties would guarantee the existence of consciousness on such a theory? There seem to be only three possibilities: the level of complexity of the system, the kinds of functionality possessed by the system, or one of those plus biology. I will argue that each of these seems implausible as a feature in a fundamental law of nature. We can, even now, imagine the coarse grained forms such laws must take, and we can use three main criteria to evaluate those laws. We can judge whether the concepts employed are appropriate as features in a fundamental law of nature; we can speculate about the simplicity of the resulting laws; and we can speculate about possible empirical consequences with an eye toward judging the plausibility of such consequences. Below, I am going to apply each of these criteria, where appropriate, to each of the three possibilities.

7.3 Complexity

The first possibility is that some fundamental law of nature guarantees that a qualitative field becomes associated with a system when it reaches a certain level of complexity(2). The coarse form of the fundamental law would have to be, "If a system reaches level of complexity N, then a qualitative field must arise from and co-evolve with it." Of course, to meet the constraint, the level of complexity featuring in the law will have to be sufficiently high to exclude many very complex systems. Additionally, the specific laws would have to say much more: such things as the structure of the field, its character, and the timing of its evolution will have to be systematically related to properties of the physical system. Any theory of consciousness should explain these things, though, so they are not special problems for this option.

Concepts employed -- The antecedent of this law has a very unlikely form for a fundamental law of nature. Minimally, the nature of the complexity would seem to matter. The concepts of 'system' and 'complexity' are both far too vague to govern phenomena as definite as the qualitative fields of our minds. Explicating them will not be easy, at least not in a way respectful of their hypothesized status as characteristics nature is sensitive to on a fundamental level. I am going to leave aside the difficult task of defining 'system,' and concentrate instead on the problems associated with giving a definition of 'complexity.' To begin with, many different kinds of complexity exist: there are the structural complexities inherent in the spatial organization of a system's components, the functional complexities of a system's contribution to a larger system, the computational complexities associated with the range of internal states the system may evolve through, and the relational complexities exhibited between a system and its environment.

Complexity also varies in description relative ways. For instance, is the activity of a cell, described in terms of the molecular and atomic interactions within it, more or less complex than the activity of a brain, described in terms of the interactions of cell­assemblies? Should we include the complexity of the cells in the account of brain functioning, or should cells be abstracted away from and treated as primitive functional units? Also, how these complexities themselves are measured can vary along different dimensions internal to the kind of complexity being considered.

Choosing these appropriate dimensions also presents problems. Consider the problem of quantifying the complexity of the Amazon rain forest. Which dimensions should we pick to measure it? How do we find the "units of measurement" that would quantify it? Then there is an associated problem of finding a way of quantifying complexity that allows comparisons across radically different kinds of systems; economies, brains, weather systems, and ecologies spring to mind as examples. Which is more complex and why: Einstein's brain or the system of global ocean currents?

Simplicity -- The general moral here is that the terms occurring in the antecedents of fundamental laws of nature must be sharp, not vague. Furthermore, articulation of a law relating qualitative fields to complexity will require postulating a favored kind (or kinds) of complexity (computational complexity, perhaps). Nature must be uniquely sensitive to this kind, as it alone gives rise to consciousness. Whatever specification we give will count as a new fundamental feature of nature because nothing like "computational complexity" currently features in any fundamental laws. That means the whole analysis of complexity, added to nature as a fundamental feature, will figure into the complexity of the resulting law. In the end, we will postulate some kind of innate and basic causal sensitivity to "complexity" in nature to provide a basis for our fundamental laws concerning consciousness. This almost guarantees that the fundamental laws governing consciousness will not be simple, and believing that the laws really are fundamental will be difficult.

Empirical questions -- Even if we imagine biting the bullet and accepting this added messiness in nature, can we specify a dimension of complexity that will give intuitively satisfying results? I can see no reason for optimism. The most obvious problem is that an arbitrary cut­off point will exist. At some point, conscious experience will mysteriously disappear (or, alternatively, magically appear). Consciousness then becomes subject to Sorites­type decision: some system will exist, say a typical mouse brain, which has level of complexity N and is conscious, but in which some relatively small change to its internal functioning could decrease this level of complexity and wipe consciousness cleanly away.

Here is why: we know enough about cognitive systems now to understand that their performance degrades gracefully with a decrease in computational or organizational complexity. When conjoined with the graceful degradation of cognition, a theory of consciousness based on complexity will yield paradoxical consequences. They would be similar to, although not quite as troubling as, dancing qualia(3) problems.

Since the law only appeals to complexity, and artificial systems may be complex, the law would govern robots. Now imagine a situation with some artificial system just complex enough to be conscious. Let us imagine outfitting this system with artificial neurons, and attaching a device to the crucial neuron that allows us to enable and disable it with the flick of a switch. We could then get a case of "blinking qualia" where the conscious experience of the system is rapidly appearing and disappearing. Nevertheless, the flickering of the conscious field does not affect the system's behavior or its thoughts, or affects them just barely.

Why should this problem be any more worrisome than other Sorites­type arguments? The problem gets its force from a dilemma created by the need to avoid panpsychism. If consciousness were to change character smoothly with complexity, with no sharp blink­out point, we would not be able to avoid panpsychism. So the concern for panpsychism requires that the law designate a sharp point of complexity where consciousness disappears. Here is the dilemma: either that point is something that can support a non­cognitive system, or some cognitive system can exist without supporting experience. We can reject the first horn, because once we allow consciousness to be associated with systems that are too simple, we might as well throw the constraint away completely. So, by hypothesis, any system that is conscious must be cognitive. Additionally, there must also be some cognitive system that does not support consciousness. That is because the graceful degradation of cognition will make it impossible to match the complexity condition perfectly with cognition. To avoid panpsychism, the law must err on the side of excess, not deficiency. Finally, there must be some cognitive system which is right on that line. Its consciousness, however dim, is still accompanied by complicated perception, some kind of memory, thought, and so forth.

We are to imagine building our robot to be that minimally conscious system, and that same robot with one neuron disabled must be the maximally cognitive non­conscious system. In this system, blinking qualia are possible even when it is exercising its (virtually) full cognitive capacity. An important point here is that the robot's cognitive system is more or less fully functional. Really, we should not expect the loss of one neuron from a sufficiently complex, fully functioning cognitive system to make it incoherent like a drunkard or a sleeper. Nevertheless, the magic complexity line gets crossed, so the creature's conscious experience is gone completely. The only way out is if we allow degradation to occur to a point where the system was supporting experience without being high­grade cognitive, thus opening the door to panpsychism and violating the original purpose of our laws.

We also need to settle the question of an upper bound on the complexity that can support conscious experience. If we do not posit a law that sets an upper bound on complexity then we will not be able to avoid the kind of panpsychism that tying consciousness to minds is supposed to avoid. The reason is that any system in which conscious components are involved likely will count as at least as complex as the conscious components themselves, and so would have to be conscious. Consider two people sitting on a train having a conversation. For the system corresponding to their disciplined interactions to function, it must encompass all the cognitive capacities of both participants. Thus, it will be more complex than either of them alone. According to the no­upper­bound theory of complexity, theirs would be a super­consciousness having experiences distinct from either of them. Our prospects for specifying such an upper bound are extremely dim. We can foresee how a given proposal for an upper bound, motivated by considerations like those above, will always admit cases of single systems (e.g., a creature with a brain twice as complex as a typical human brain) which we would never want to exclude from the consciousness club. Eventually, tremendous problems will arise for those of us who wish to individuate all the consciousnesses that must exist, problems tied in with how we are going to define 'system' as it occurs in the fundamental law.

7.4 Functionality

Mere complexity possesses obvious shortcomings as the basis for consciousness. The natural reaction is to suggest supplementing it with something else, something that would allow us to bypass some of the problems above. The most promising strategy is to delimit some range of capabilities as relevant to the presence of a qualitative field. Perhaps we should explicitly include some paradigmatically "cognitive" capacities such as awareness (functionally construed), the ability to believe, to plan, or to perceive as necessary for the existence of a qualitative field. This is the second option: perhaps qualitative fields are associated only with systems that possess certain kinds of "cognitive functionality" suitably defined. The coarse grained form of the proposed law would be, "If a system evidences paradigmatically cognitive capacities XYZ, then it will have an associated qualitative field co­evolving with it." This position is quite common, although proponents usually put it forward as a version of non­reductive functionalism. Recently Tye (1996) and Lycan (1996) have taken this tack, and a precursor may be found in Van Gulick (1988).

Concepts -- After hearing this kind of suggestion, we should step back to remind ourselves that we are searching after a fundamental law of nature. The kinds of laws we are looking for are on the same level as those governing gravitation, motion, and mass. We must keep in mind that whatever concepts occur in the antecedent of such laws impute direct causal relevance to the things that fall under the concepts, causal relevance not derivable from any constituents. This is very different from the kind of causal relevance typically enjoyed by high­level phenomena. That relevance is typically derivative on the causal natures of constituents. Finding precise criteria for the concepts used by this kind of proposal is a very problematic enterprise, as anyone familiar with twentieth century psychology and philosophy of mind is aware. Including any of them into the laws governing the qualitative field will make those laws very complex, and imputes a character to natural laws not at all in harmony with those we have already discovered.

What makes one kind of system cognitive, and not another? At least at the beginning, defining cognition requires an appeal to norms. Like complexity, this is a vague matter. Presumably a correct account will involve concepts such as "appropriate behavior," "capable of veridical perception," "supports beliefs that can be true or false," "is rational," and so forth. The basic obstacle here is that cognitive systems are the kind of systems that can make mistakes, and the possibility of error requires the specification of a norm (Millikan 1984; Dretske 1986). In contrast, we should expect a fundamental law of nature to appeal only to purely descriptive conditions attaining in the physical system. Again, it would be quite a peculiar law of nature that took effect only when a system was capable of making mistakes. The resulting challenge is to turn our intuitive, norm­filled concept of cognition into some workable, descriptive characterization.

One promising strategy for meeting the challenge is the thesis that cognitive systems are just a subclass of computational systems. If this thesis is correct, then any system implementing any member of a certain set of algorithms would necessarily be a cognitive system. Such implementations would meet the conditions for applying cognitive norms, but they would not, themselves, be described normatively. Unfortunately, this solution just pushes the problem backwards a level. We will still need an explanation of what makes a computational system cognitive so that we can specify the relevant subset. That will depend on our concept of cognition, and not the other way around. Furthermore, our concept of cognition does contain the norms, and it seems to be a vague matter when they can and cannot be applied. This vagueness will infect our decision about which computational systems to cover in our fundamental law, and which to exclude. On the surface it will seem that we have a descriptive account of the conditions necessary for generating a qualitative field. Yet at a deeper, implicit level it will still seem as if nature is mysteriously honoring a set of norms. We will need to invoke norms to explain the lines drawn through the space of possible computational systems. Also, that subspace of systems itself will seem arbitrary along its edges.

Simplicity -- A worse problem arises for how we will formulate the fundamental law. Somehow we must pick out, in our fundamental law, precisely the subset of systems whose implementation produces associated qualitative fields. Obviously, an infinite disjunction will not do. The most coherent alternative is to be able to specify some precise set of rules embodying "criteria for cognition." Again, this leads us back to the problem of defining cognition. Even in the unlikely event that we can specify such rules, they will not be simple. Also, although every specific cognitive system will have some descriptive features that account for why it is cognitive, we cannot assume that they will all share a set of necessary and sufficient descriptive properties. We cannot assume that 'cognitive system' picks out a natural kind. So we cannot assume a general analysis of cognition will be able to eliminate reference to norms. Therefore, if we are to specify the class of functional systems via some precise analysis of what it means to be cognitive, then we do have to worry about including the norms of cognition in a fundamental law of nature.

If a law of nature were to rely on some normative considerations, it would not be at all clear what that implies about our world. How is it that nature could decide when the law should 'kick in' and a qualitative field pops into existence? Imagine the problem in the context of a developing fetus. It would seem to require nature to evaluate the system, as it is functioning in nature, against a Platonic conception of ways such systems should function, and further determine if it is rational (or appropriate) to hold the system to the standard. At that point, it begins to look like we are attributing mentality to nature as a whole at the fundamental level. Therefore, our fundamental law is again expanding the character of our world far beyond just adding a primitive phenomenal component to our psychology's ontology. It is adding, as a fundamental constituent of nature, a power of semantic divination. This is far from the character of the fundamental laws we have thus far discovered. Those laws have a certain mechanical elegance that is ingenious without forcing nature to continuously evidence ingenuity.

Empirical questions -- We may succeed in specifying the class of cognitive systems in some non­normative way. Probably, the best we can hope for is a functional account, either a computational account or something broader. In such a case we likely will still need teleology (Lycan 1987). Teleology is necessary because the functional character of a given state cannot be determined locally in space or time. The functional contribution of a state is relative only to a series of other states, to interconnections between them, and against the background of "normal" environmental conditions. The factual existence of consciousness will always depend, then, on a vast number of counterfactual truths extending widely through space and time. These facts are of the form, "If the system is in state­type X, and were it to encounter input­type Y, then it would transit to state­type Z." Here, the state­types themselves are supposed to be functionally or teleologically defined, and thus their identity-determination requires reference to further counterfactuals, and onwards in a grandly holistic fashion.

We know now that nature does exhibit such counterfactual sensitivities on the quantum scale (Penrose 1994), so it is not something entirely new. Yet the form of the law we are now considering will require such activity routinely and ubiquitously on unheard of scales, and in macro­systems not known to exhibit these quantum effects. A law of that kind cries out for further explanation. It shows the arbitrariness and ad hoc character we normally interpret as signs that something deeper in nature is being exhibited.

7.5 Biology

A third alternative ties consciousness to biology by specifying that only biological systems reaching a certain level of complexity (or capable of certain kinds of functionality) can be conscious. That would rule out (presumably) global weather patterns, economies, and ecologies. Searle (1992) and Block (1980) have recently defended positions something like this. Unfortunately, this proposal has all the same problems plaguing the first two proposals. A system's biological character cannot be taken as a sufficient condition for consciousness. Inevitably, it will need to incorporate one (or both) of the notions of complexity and cognitive functionality. We are asked to believe that there is a fundamental law of nature whose antecedent contains a clause to the effect, " . . . the system is of complexity N and is carbon based . . ." Or, worse, a disjunction of the form, "the system functions cognitively and is carbon or silicon based, suspended in liquid, with a cellular construction . . ." Proponents of the "wetware" position are likely hoping that we will discover something subtle about cellular mechanics or chemical reactivity as relevant to consciousness. Yet it is very hard to see at this point what they have in mind. The coarse grained forms given above are therefore caricatures, admittedly, but they are caricatures that capture the essence of their subject.

Concepts employed -- To avoid panpsychism, the law tying consciousness to biology will incorporate either the complexity or functionality constraints previously discussed, or both. Additionally, it must specify just what about biology (or "wetware") is important for consciousness. Making a satisfactory case for any kind of biological constraint will be difficult. Consider two recent lines of thought that have been popular in the philosophy of mind, the appeal to evolution and the appeal to "wetware".

Millikan (1984) has argued that evolutionary considerations are essential to understanding the intentional properties of systems. Since consciousness seems rife with intentional content, some people might want to suggest that evolutionary considerations are also important for understanding why consciousness arises. Some people with views on representation that are similar to Millikan's views seem to want to do this (e.g., Dretske 1995; Lycan 1996), but the position is extraordinarily implausible. The main concern of indicator semantics is accounting for the normative aspect of content, and we have already seen the implausibility that the kinds of laws we are searching for involve normative considerations. The difference between consciousness and intentionality (as Millikan conceives the latter) is that consciousness is undoubtedly intrinsic, "in the head", and basic. Millikan argues at length that content does not meet any of these conditions. Therefore, the analogy between intentionality and consciousness breaks down.

A much more popular suggestion is that, somehow, the "wetware" of the brain is responsible for producing consciousness. Here, though, it does not seem as if it is biology that is really the issue, but chemistry. What, though, could be so special about chemistry? Why can it produce consciousness if solid state interactions cannot? Finding a plausible answer to this question is difficult. One option is that chemistry might allow for certain kinds of causal relations that solid state physics might not allow for, and these causal relations support consciousness. However, this is just another way of stating the functionality requirement, adding to it the additional bet that only biology could meet the functional conditions (see Edelman, 1989, for suggestive observations along these lines). But then we have not made any advance over the functional criteria. Once we realize that the functional difference chemistry might make is what we are really interested in, then the addition of biology to the requirement of functionality seems superfluous after all.

Empirical questions -- One has to wonder how we are going to fill in the exact association between biological elements and the elements of experience. If the theory is going to have any bite to it, it should say specifically what about biology supports and drives the evolution of our conscious experience. I can see only two possibilities. We are going to tie elements of experience specifically either to biological events, or to biological objects. Furthermore, it seems to me that the first possibility will invariably involve the second.

As our example we can imagine an experience like the light­headed enjoyment of a surprising and funny joke. If the theory ties this feeling to the occurrence of some kind of biological event, the first question that will arise is why it is important that the relevant type of event be a biological event? Holding true to our constraint is going to require tying the raw feeling specifically to biological structures such as cells or cell assemblies, with their peculiar chemical components. That is why the first possibility involves the second.

The biological markers for given experiences likely will not support the events underlying a unique kind of experience. Much more plausibly, at different times the same biological objects will support different kinds of experiences by participating in different kinds of events. Our theory is likely to have the consequence that a cell assembly (for example) can support a little patch of phenomenal blue, and a little patch of pain, and a little patch of angst, and that its participation in different events attaches these patches to it. Yet the biological nature of these assemblies remains relatively constant from one event to another. A neuron's firing as part of a "blue qualia" event should not be significantly, biologically different from its firing as part of a "purple qualia" event. Why can't we factor out the biology by giving a more general characterization of the event types?

The underlying nature of the biological objects is not changing, and the experiences are constantly changing. Clearly, the event types are going to be doing the real explanatory work. They will be differentiable from each other holistically, as playing a certain role in the continuing evolution of the system as a whole, and should be carrying information based on the history of the system. Furthermore, for the purposes of explaining experience these events will be functionally construable since adding "and the neural firings were realized in a system using serotonin, glucose, and . . ." will add nothing new to differentiate the experience being explained from many other experiences. We then have the unpalatable consequence that an explanatorily impotent X­factor, biology, is nevertheless presumed to be a necessary ingredient in conscious experience.

If it turns out that we simply cannot give an account of mental events which abstracts from the biological substrate, then the conclusion above would fail. Should that be the case, however, it will be because the biology is the only substrate capable of supporting the proper functioning of the system; perhaps we can give no description of the functional capacities necessary for being cognitive creatures like us that does not refer to biology. This is a real possibility, but if that is the reason why biology is relevant then the relevance of the biological conditions will be derivative upon the relevance of the functional conditions. In essence, biology will be important only because of its unique capacity for supporting certain kinds of functioning. In this case, the biological requirement becomes redundant since the functional requirement entails it.

7.6 Is panpsychism palatable?

The preceding discussion does not in any sense prove that the fundamental laws do not tie qualitative fields peculiarly to cognitive systems. The point rests, mainly, on convictions concerning the simplicity, clarity, objectivity, and elegance of fundamental laws. They are convincing only to the extent that one possesses this conviction about nature. These kinds of considerations can yield strong reasons for rejecting such a constraint. In rejecting it we should conclude that qualitative fields have a basis much more uniformly spread in nature. It follows that cognition is merely a specific context inside which this basis is being expressed.

I have offered reasons for rejecting our intuitions that qualitative fields are phenomena that simply 'arise' from certain physical systems, but I also promised to try to undermine the intuitive pull of the considerations that lead so many to insist on that constraint. Those intuitions seem to rest mainly on two points: (1) we have no evidence for the existence of qualitative fields outside of cognitive contexts, and (2) the mere supposition is incoherent since it requires experiences without experiencers.

We can answer (1) simply by noting that every theory goes beyond the direct evidence that we have, because we have direct evidence only in our own cases. From my own perspective, any theory that attributes consciousness to people other than myself is going beyond my evidence. More generally, we can note that what counts as evidence for attributing consciousness beyond one's own case depends on the theory one might have of it. Therefore, in the general case, the concept of 'going beyond the evidence' is not well­defined. For instance, if I believe consciousness is dependent on language ability then verbal reports of conscious states are the only "evidence" that such a belief will allow. Under the influence of such a theory, I would deny consciousness to animals because we only have evidence for consciousness, via verbal reports, in people. Alternatively, if I believed that consciousness depends primarily upon biology then I would extend attributions of consciousness down the phylogenetic chain to other animals. The basis of the extension will be a claim that the common biology we share with them is evidence that they also are conscious. Similarly, if I adopted the stance that only cognitive functioning is relevant to the presence of consciousness then I would, based on the evidence, suppose competently functioning silicon robots to be conscious.

The moral is that what we will count as evidence for consciousness, and what our theory of consciousness is, are heavily intertwined. Thus no pre­theoretical bias about the evidential base that we should not "go beyond" can carry an overriding veto power on the form of the final theory. Surely, the theory must include certain systems as conscious. Any theory that had the result that only Gregg Rosenberg was conscious should give us pause. In general, our theory should include people, and expecting it to include other mammals, fish and birds is reasonable also. Things get a bit fuzzy when we begin to consider insects, perhaps, and also artificial systems, but that is alright.

What should be important to us initially is that our evidence makes its strongest demands concerning certain kinds of organisms our theory must include, and can make only much weaker demands concerning which kinds of systems our theory must exclude. It follows that we should concentrate primarily on finding the simplest, most coherent set of laws that include the systems which intuitively should be included. If an otherwise exceptional theory has the consequence of also associating qualitative fields with some surprising class of systems, then we should accept that consequence as a discovery about nature.

The second intuitive basis for wishing to tie qualitative fields to minds alone is the charge of incoherence for any attribution outside of a cognitive context. For most people, this is probably the main reason for the constraint. Appropriately, this is also the most difficult hurdle to clear on the way to shedding the prejudices that can keep us from finding a simple and satisfying theory. I am not sure one can ever fully shake this intuition, but there are ways a person can come to lose confidence in it. The thesis I suggest to undermine this intuition is that only certain kinds of qualia force themselves on us as essentially the experience of a mind, and other kinds, with very different characters from those we know, might subsist outside of minds.

One observation that can loosen the hold of this objection appeals to the open­ended character of the concept of phenomenal consciousness. The privacy of consciousness forces us to build in a kind of tolerance for alien experiences and feelings. We have to allow that a manta ray sensing the electromagnetic structures on the ocean floor may experience qualities we could never imagine. We also have to allow that simpler and simpler organisms may have experiences of simpler and simpler kinds, as well as alien kinds. It could be that this open­endedness puts us on the slippery slope to an incoherence, but we do not know that it does. The slope could just as easily turn out to be harmless, rather than vicious. If so, the idea of non­cognitive experiencing subjects would be perfectly coherent. At the very least, the open­endedness of the concept means that we have no definitive reason to reject the notion. That is all this defense of the idea requires. Given this, we can perhaps look for ways to bolster ourselves against the skepticism that accompanies the suggestion.

A first step toward undermining our confidence that such a thing is inconceivable is simply to notice that, for most qualia, learning to see them as objects of awareness in and of themselves is possible. That is how speculation and puzzlement about consciousness gets off of the ground. For instance, after some effort I can consider a blue region of my visual field as an object in front of my mind, being experienced, and not just an experience itself. The change in attitude is not so much reflected in my visual field, as in the background feelings which my visual field contributes to my overall experience. What I mean by this is something analagous to how a transforming idea system, or a moving experience, can change one's perceptions of the world or oneself. In these situations there is a definite sense in which one's perceptual fields, say the visual field, stay largely the same. Nevertheless, the contributions they make to one's overall experience are distinctly different in ways that are clearly responsive to the overall conceptualization of the world currently active. In a difficult to articulate way, one's experiences are both the same and still very different than they were before. The sameness seems reflected at a lower, relatively concrete organizational level of perception, and the differences at a much more abstract level of conceptualization which is clearly tightly connected to one's perception, but not completely coupled to it.

Here is what I try to do. I force myself to regard the blue region as an object of awareness; here, I regard it as being experienced like a dog, or a rock, or a flower petal (it is much like putting the secondary qualities out into the world). This contrasts with regarding it as being an experience, and, as such, constituting part of my representational awareness of the world (it is much like putting the secondary qualities back into the mind). The feeling flips back and forth, and in that respect is a bit like when one is staring at a Necker cube. At a moment, it seems that the blue region is part of my awareness, and partly constitutes my experience of the world. Then, at the next moment, it is an object of my awareness, an object of something that constitutes my experience of myself. If dwelt upon, it comes to seem a very strange capability. That I can do it, however, is very clear to me. Most other qualia can be reconceived similarly; sounds in themselves, and even pleasures and pains. It is a very peculiar fact that my experiences, under the appropriate conceptualization, can come to seem objects of experience discovered by my awareness. The only feeling that escapes this illusion is the feeling associated with awareness(4) itself. The feeling of awareness seems conceptually, irreducibly cognitive.

As our awareness divides the elements of consciousness into separable elements, these individual elements seem conceivable as pure experiences of a simpler sort. It is a short step from conceiving of them as pure experiences of a simpler sort to conceiving of the systems generating them as being pure experiencers of a simpler sort. I do not wish to make too much of this. The fact that I can regard some parts of my inner self in that funny way does not show that my experiences are, or could be, objects out there, existing independently of my mind. Of course my experiences are essentially experiences of my mind.

Also, I want to head off a potential point of confusion lurking here. The phenomenal individual corresponding to some subsystem that produces an auditory element in a normal human's experience does not need to be experiencing sound. Instead, such a system might experience qualia uniquely appropriate to its own level of reality, and responsive to its own finer grained causal organization. Just as our experiences seem to reflect our causal organization in some way, the experiencings of a subsystem could reflect its causal organization in some way.

If this is true, it is a mistake to think that its experiencings must be of simple, structureless elements within our own consciousness. Someone might believe that it could be that way, but we do not have good reason to believe it has to be that way. To conceive of the alternative, imagine that all phenomenal individuals experience in a way that corresponds to the causal contributions their physical components make to their physical state. A higher­level subject, such as ourselves, may have a subsystem that contributes an atomic element to its experience. That subsystem, meanwhile, might be a subject of complex experience itself. Its unique experiencings correspond to the contribution that components in its rich, internal causal structure make to its overall state. Why couldn't our world be this way?

These considerations seem to undermine my confidence that such a hypothesis is incoherent. We can assign some sense to the proposition that the producer of a feeling of pain, or tickle, could exist as an experiencer in itself, as something that enters awareness but is not dependent upon it. More tellingly, it reinforces some further considerations and, together with them, lends much more credibility to the underlying coherence of the position.

First among these other considerations is the simple fact that when we speak of the qualitative field of some other, non­cognitive system we are obviously not attributing to it the qualities of our own experiences. We are not attributing little pangs of pain, or experiences of tiny blue dots to non­cognitive systems. Whatever we are attributing, it is not any kind of experience that we can empathize into. Instead, we are attributing to it a qualitative field that has a character in some very abstract sense like that of our experiences, but specifically unimaginable to us and unlike our own qualia. The best attitude toward such attributions is to maintain that the properties in question must fill a place holder for the solution to an analogy problem, for example: "X is to a thermostat as conscious experience is to the human mind."(5) where we know X must have a solution in nature, but we do not really know what that solution is. It is an existential claim whose instantiation is something that we cannot be acquainted with, and hence should not pretend to understand deeply.

Finally, even if some sort of panpsychism is true, we should not naively assume that every perceptual or conceptual individual, such as a thermostat or a rock, will be a kind that possesses a qualitative field. It may be that large scale, enduring, coherent experiencers are extremely rare. The panpsychism we end up with may be as benign as would occur if mainly the interactions between very simple atoms or molecules produced flashes of extraordinary simple and brief feeling, like fireflies quietly flickering in the night. For these reasons, referring to the qualitative fields of non­cognitive systems as proto­experiential rather than experiential is probably best. The term suggests both an hypothesized kinship between the qualitative fields of such systems and our own qualitative fields, and also the alienation from its richness, variety, and, most importantly, its cognitive awareness.

The answers to the second intuitive objection are then (1) We seem able to reconceive our own qualia, in many cases, as objects of experience as well as experiences themselves. (2) The qualitative fields we might attribute to non­cognitive systems do not contain "little pains" or "little specks of blue", but instead have some kind of qualitative character very alien to us. (3) The best way to conceive of those fields is via a mental place holder for the solution to the analogy problem, "Y is to system X as experience is to the human mind." which sets up Y as a qualitative field that we know must exist, but which we cannot concretely imagine. (4) The best term for the alien character of these fields is proto­experiential, a term meant to suggest that they contain qualitative objects that are not, strictly speaking, being experienced by a mind. These proto­experiential states need not have semantic content, and certainly no cognition will occur within the manifold.

Although no experiencing mind is present, there may be a perfectly good sense in which each is the essential part of a subject of experience. By saying this, I am just pointing out the panpsychist conclusion that not all subjects of experience are cognitive (and hence mental) systems. Considerations 1­4 show us how to regard coherently the qualitative fields of non­cognitive systems, even if they do not give us concrete ways of conceiving them. Collectively, what 1­4 suggest is that the difficulty of imagining qualitative fields that are not associated with minds comes from a shortcoming in our empathy, and not from a fundamental conceptual incoherence.

These conclusions also suggest that both parts of the etymology of "panpsychism" are misleading, if we choose to call the resulting position "panpsychist". The "pan" is misleading because it will not be the case that everything is conscious. By assumption, only some feature(s) of the world correspond to experiencing systems. Although we have good reasons to expect that some non­cognitive systems may possess this feature, the theory will still constrain which kinds of things will be experiencers, and the ultimate position may end up being milder than is often feared. As I remarked above, rocks need not be conscious, nor thermostats, even if some variety of panpsychism is true.

The "psychism" is misleading because one need not associate experiencing systems exclusively with cognitive activity, and hence not exclusively with minds. Therefore, even if some variety of panpsychism is true it does not follow, without further assumptions, that minds are everywhere. In the end, it may be that the kind of high­level, conceptualized experience we find in ourselves is a rare variety of experience. Appealing again to the metaphor I used above, perhaps most experiencing entities are much closer to the ground level of reality, little fireflies in the night supporting brief flashes of sensation as they interact. This relatively unthreatening state of affairs could be the state of affairs in our world, and I believe it is counterproductive to prejudge the possibilities and, thereby, be tempted to overreact.

Taken on balance, these considerations show that we have good reason to believe the problem of qualitative content is a general problem for our view of nature, and not just a problem for psychology. We can argue for the conclusion on positive grounds by considering the form of the fundamental laws that we would have to posit to tie qualitative fields to conscious minds; also, we can argue that the intuitions that lead us to want to do so are not so sturdy as they seem at first glance. Especially, considerations about the arbitrary seeming forms the fundamental laws would have to take under the proposed constraints should make us suspect that there is something very deep about the structure and character of nature that is missing from our current, purely physicalistic world view. If this is right, then we must look for clues to tell us how our view of nature must be revised.

1. I am using 'cognition' to refer to the operational equivalent of mentation. My use of the term from this point forward is to identify the functional components involved in having a mind. This includes basically everything studied within cognitive science. Where I use other psychological terms such as 'memory', 'conceptualization', or 'perception' in association with 'cognition', I intend the operational sense these terms have within cognitive science.

2. People rarely endorse the position so baldly in print, but it is very common to hear it appealed to in an off­hand way in conversation and informal correspondence.

3. 'Dancing qualia' is Chalmers' name for qualia whose character shift without change in the functioning of the system. See his 1996 for more detail.

4. Awareness is here being construed as a functional notion related to the ability to use information in control of inference and behavior. Associated with it, however, is a feeling of being aware.

5. This analogy problem can rightly be regarded as a promissory note. The note is partially paid in chapter fourteen, where the kinds of similarities and differences at issue are eventually specified in more detail.


Chapter 8

Puzzles for Liberal Naturalism

8.1 Where should we go from here?

The anti­physicalist conclusion seems unavoidable. In addition, initial reflections on the failure of physicalism have disclosed good reasons for believing that we must cut deeply into its foundations to find an alternative. The most rational conclusion, it seems, is that our understanding of nature is incomplete in a way that is both fundamental, and of widespread importance.

An obstacle ­ These conclusions yield a warrant for exploring a new way to understand nature. An immediate problem is that we simply have too much freedom, far too much, to search effectively. Trying to discover something both "fundamental and widespread" is a futile exercise, as those constraints do not really provide any guidance taken in isolation. We need a better idea about the kinds of puzzles that a fundamental theory of consciousness must provide a solution to. Really, a proper set of puzzles should fill two needs. It should help us to construct an adequacy condition for proposed explanations, and hopefully offer clues about where to find the fundamental incompleteness in our knowledge.

Where should we look for the kinds of puzzles that a foundational theory will be most concerned with solving? On this point, I will depart from the recent standard a bit, as I believe that puzzles posed by the most overtly cognitive aspects of consciousness may not, in the end, be the most helpful kind of puzzles to focus on. By 'cognitive puzzles', I have in mind the sorts of questions raised by facts like: conscious states tend to be reliably reportable; conscious states are representational; conscious states contain information that is globally available in the control of behavior; and the fact that the structure of consciousness mirrors the structure of cognitive processing. All of those facts will be very important eventually, and any theory must allow us to understand why consciousness has those features. In the present context of a foundational search, however, they are not likely the best pointers to follow.

I fear that those kinds features, features so obviously involving cognition, will not be generic enough to point us toward something fundamental and ubiquitous. By focusing on overtly cognitive features of consciousness, we run the danger of confusing the inessential with the essential, and overlooking promising paths in our search. For example, the links between conscious experience and functional awareness lead to very interesting puzzles when considering multiple personality cases (Braude 1991), or commissurotomy patients (Martin 1981). These puzzle cases can be very seductive, philosophically, but they are likely much more intriguing than fundamental.

I suggest that, instead, we should try to identify features of consciousness potentially more generic than these overtly cognitive features. We should look for features of our experiencing that we can plausibly expect any subject of experience, cognitive or not, to share with us. Ideally, the features we focus on will strain our physical image of nature, as the most helpful features will yield a paradoxical view of the world when combined with our physical image of it. The task of removing the paradox driven tension will provide constraints for our search, as each will act as an explanatory target for a Liberal Naturalist view of nature. They may also provide guidance, since, with luck, at least one of those features will provide a clue about the location of our missing knowledge.

The strategy ­ So the strategy I will follow is this. Consciousness seems to have some very generic features that are violently and obviously in conflict with our physical image of the world. With this in mind, I will probe the places of greatest tension. These areas of tension are more likely to lead us to the gaps in our knowledge because they are where the paradoxes of explanation occur for a theory of consciousness. The history of discovery should lead us to expect the deepest insights to come from reflection on the places of paradox.

This chapter and the next articulate a set of six issues for the Liberal Naturalist to consider. Some of these have the character of paradoxes, while others are puzzles. Chapter ten is a transition chapter. It begins with further reflections on one class of paradox, and then moves us from these reflections toward a positive proposal for developing Liberal Naturalism.

Before beginning the exposition of these problems, I want to point out that, for now, I am remaining noncommittal about where any errors might lie. My purpose is to expose the intuitions so that we might later diagnose the problems from within a Liberal Naturalist framework. I am self­consciously withholding judgment about just where any possible naiveté lies.

I would like to ask the reader to follow me in this act of tolerance, at least for the time being. Specifically, I would like to ask the reader to start from scratch, so to speak, and take each of these problems seriously. This request is important because the consciousness studies community represents a diverse variety of intellectual traditions. Some readers will easily grant that each problem involves some relatively subtle error about nature, or the nature of consciousness. Others will react to at least some of the problems (e.g., the problem about the subjective instant) with a much more deflationary or dismissive attitude. These readers might find some of the problems below to be relatively frivolous, arising from gullibility concerning phenomenology, or the deliverance of introspection. To this second group of readers I want to say, "Be patient."

What reasons do we have to be patient at this juncture? On the one hand, we are no longer under any pressure from physicalism to embrace deflationary claims about consciousness. Since the arguments against physicalism used only minimally controversial observations about the nature of conscious experience, quickly dismissing more controversial claims about consciousness would be incautious. Were we to indulge in a quick dismissal on the grounds that these features of consciousness seem incompatible with the physical facts, we would be especially unjustified.

On the other hand, we are in search of a new perspective on nature. We may be able to see, from a new perspective, that our situation does not warrant a deflationary attitude. At the very least, not all of the paradoxes rest on obvious errors or naiveté. By the end of this work, I will be in position to argue that we have been overlooking some possibilities, and these possibilities will provide a license to treat each puzzle with proper seriousness. After we enrich our view of nature, we will be able to resolve the paradoxes without convicting ourselves of hopeless phenomenological confusion or naiveté.

8.2 Category one: The­many­that­are­yet­one

(1) The unity of consciousness ­ My visual field right now is teeming with information. It represents depth, color, shape, motion, and, at a more conceptual level, saliency. From an external perspective, one might think that each piece of information could be present in a separate phenomenal modality, if present in a phenomenal modality at all. In reality, the coherence of presentation, in some sense, transcends the separation of content. These pieces of information have coalesced into what seems a unified field of perception, each piece superposed in an orderly way with the others. Even the separations between different sensory modalities seem superposed, in a subtler way, with one another in a common field. The remarkable character of this coalescence almost forces an inchoate belief in something we feel inclined to call the unity of consciousness. In a recent paper (1996), Thomas Metzinger attempts to describe it this way,

I think that there is a highest­order phenomenal property corresponding to this classical concept of 'indivisibility': The property of wholeness. The wholeness of our reality (and of ourselves in it) can easily be discovered by all of us from our own experience. This wholeness is much more than a simple unity in the sense of the concept of numerical identity mentioned above: I am not able voluntarily to split or dissolve my global experiential space ­ this reality ­ or my own experienced identity ­ myself.

The unity of consciousness presents both a challenge and a paradox. Directly and intuitively stated, the paradox of unity is this: we have unity of experience despite being composed of many diverse parts. The challenge of unity, exemplified by Metzinger's appeal to what "can easily be discovered" from experience, in lieu of description, is this: what do we mean when we refer to the "unity of experience," and why exactly is it problematic? Among the puzzles and paradoxes of consciousness, the driving intuition behind the idea that there exists a subjective unity of consciousness is easily the most elusive.

To say that its exact articulation is elusive is not to say that no definite intuition is there, or that it is inarticulable. For instance, some people point to the non­locality of quantum coherence, claiming it is evidence for consciousness in nature (e.g., Kafatos and Nadeau, 1990). At first glance, seeing why quantum non­locality should make the existence of consciousness in the world seem more intelligible to some people is hard to do. If local causation gives no evidence of consciousness in the world, why should non­local causation yield such evidence? It seems that the inseparability of the components in these entangled states is resonating with certain prior intuitions people carry about the unity of consciousness.

On a more mundane level, I believe the same sort of resonance with prior intuitions underlies the current search for a solution to the "binding problem" for percepts. The binding problem is a puzzle about how the different information presented within a percept can achieve the unified phenomenal character that we find in experience. For instance, as I noted in the opening paragraph of this section, the visual information in the separate processing channels for color, shape, depth, and movement all seem superposed into a single visual representation of the world. Translated into cognitive terms, the binding problem requires understanding how the information in these separate channels functionally "comes together" in this fashion.

No external evidence makes the binding problem compelling. From such a perspective, one could make a good case that it is a pseudo­problem. To make this case, one could point to the absence of any detailed understanding of the architectural constraints needed to produce human performance. From this perspective of ignorance, we are not justified in believing that the brain must somehow globally "bind" separate information streams to achieve its results. Where, then, do the roots of people's belief in the binding problem really lie? It seems the phenomenology of perception is the chief source of nourishment. The feeling that a binding problem exists is feeding off of the conviction that something in the story of perceptual processing, when finally fully developed, should resonate with the phenomenology of inseparable, coherently superposed perceptual elements. Somehow, the idea that coherent neural oscillations explain binding resonates better with our intuitions about the unity of experience than do stories of independent, or semi­independent processes.

Something about causal inseparability and functional unity reminds us of the unity of consciousness. I suggest that beliefs about this inseparability of part from whole underlie intuitions concerning the unity of experience. Conscious states are clearly complex, involving a tremendous variety of qualia coexisting within the experience of the subject. Obviously, when we avow the unity of consciousness, we are not disavowing this complexity.

What, then, are we doing? I suggest the unity intuition springs from the observation that, although the phenomenal manifold is clearly complex, it is not clearly composite. By a "composite" system, I mean a complex system whose existence derives from independently existing elements, its components. The relation between a composite and its components is the standard one in which the existence of the components, and their relations, constitutes the existence of the composite. The components do not presuppose the existence of the whole in turn.

To many, the relationship between composites and their components seems to be the wrong model for understanding the relationship between our complex phenomenology and its elements. For instance, while a clear distinction exists between the feelings involved in orgasms and headaches, it is not plausible that these feelings can exist independently of the whole experiencing of the subject. Similarly, visual experiences are not plausibly constructed from tiny colored dots that exist independently of the experience as a whole, getting drafted into service in its construction. The intrinsic distinctions between these elements of experience do mark them as different, making for a complex and richly structured phenomenology. Their holistic dependence means that, despite being distinct from one another, they are not components of a composite whole. Instead, the elements of a given experiencing come into existence together, each dependent on the existence of the whole, and strongly inseparable from it.

Our physical image of the world could not be more different from this, at least at the classical level. The brain is not only a complex system, but a composite system with components. It consists ultimately of fundamental particles, and these particles support many layers of higher organization. Each of these higher layers of organization has individuals that are components in even higher levels. Eventually we hit the biological level of organic molecules, nerve cells, neurons, neural assemblies, super­assemblies, the brain, and the central nervous system in its entirety. The individuals produced at each stage seem strictly separable from the systems in which they are components (for an accessible discussion of the levels of nature, see Scott 1995).

With these sketchy remarks in mind, we can articulate both the paradox and the challenge of the unity of consciousness. The paradox of unity is this: how can a single system seem to be both composite and noncomposite? The challenge of unity is this: a good theory should help us to articulate what the unity of consciousness is better than this. It should give voice to our unarticulated intuitions in such a way that we recognize them.

The paradox is surmountable, and I believe that current philosophy of mind even has a head start on it. Notice that a functional system, qua functional system, has no component parts although it is complex. Functional systems have functional entities as elements, and functional entities are only the kinds of things they are relative to the role they play within the larger system. In short, causal role typing is holistic, and any characterization of a type of causal role contains an essential reference to the whole context within which the role operates.

For a physicalist, this is no help for the problem. From a physicalist perspective, the kind of causal role typing at issue must be considered a mere conceptual exercise. The true reality is completely derivative on the underlying component system. However, reductionist demands do not constrain Liberal Naturalism. Liberal Naturalists are free to consider the ontology of functional entities in themselves. Perhaps they may be "screened off" from the physical being of an entity in some way.

As we have seen, functional explanation cannot take us all the way to an explanation of consciousness. Nevertheless, it may still play an essential role in solving certain problems. The idea of a functional role is that of a causal role within a specified, canonical context. A natural research program that grows out of reflection on the ontology of functional entities is one that seeks to understand causation: what is it to have a causal role, and what is it to be a canonical context? Perhaps, as Liberal Naturalists, we should try to understand these issues more completely.

(2) The subjective instant ­ The second paradox arises from the apparent simultaneity of experiences within consciousness. Critics might call this, with William James, the specious subjective instant. The paradox is this: we have simultaneity of experience although temporally asynchronous brain events seem to correspond to those experiences. In some sense, we can understand the simultaneity of consciousness as the temporal analog of the unity of consciousness. Subjectively, the elements of our conscious experiencing ­ sight, smell, sound, proprioceptive monitoring, autonomic monitoring, etc. ­ all occur concurrently, at a single moment. Remarkably, a single subjective instant seems to exist.

However, it also seems that brain events cause, or correlate with, the stream of conscious experiences. The brain is a complex physical system extended in space, and the laws of special relativity apply to it. For a set of brain events, no absolute time exists at which we can say they simultaneously occur, no matter how finely we localize the physical correlate of the experience. Thus no objectively specifiable set of brain events exists that are all occurring at the "same time." These asynchronous events seem to be responsible for our subjective experience despite experience containing synchronous elements.

In short, the problem is this. When it comes to the brain, no privileged reference frame exists. When it comes to conscious experiencing, one does. Thus it seems that something exists to which relativistic space­time both does and does not apply.

The usual response to this paradox is to call into question the phenomenology of the subjective instant. For example, it is tempting to see Daniel Dennett's Consciousness Explained as a book length response to the paradox, using it to explain consciousness away. One unsatisfying aspect of this approach is that it seems to beg the question. It presumes that the physical account of time is complete. By applying that account to what we know about brain processing, one can show trivially that there can be no simultaneity of experience. Dennett offers an account in which the simultaneity of the experience is the result of a narrative rationalization by our reporting mechanisms. A probable mechanism is one whose job would be constructing a representation that represents them ­ falsely ­ as occurring simultaneously.

The Liberal Naturalist is not under the same constraint as Dennett, and is free to consider the possibility that our physical understanding of time is incomplete in some subtle way. Some presumptive evidence exists that this is the case. That evidence rests in the well­known problems physics has in explaining the direction of time. Perhaps we may eventually explain the direction of time using only resources within physical theory, but, until that day, the Liberal Naturalist should be willing to consider the possibility that something outside physical theory fixes the facts about the direction of its flow. Along these lines, it seems plausible that the direction of causation fixes the direction of time, and it is not clear at all that physics tells the complete story about the process of causation. If this is true, a promising place for the Liberal Naturalist to begin searching for a resolution to the paradox is in a theory of causation.

Summary ­ Together, these two problems constitute what Nagel (1986) calls "the combination problem." Nagel feels that the combination problem, along with panpsychism, constitute the two biggest hurdles for a dual aspect theory of the mind. We have already seen that current orthodoxy miscasts panpsychism as a problem. If a theory avoids panpsychism, that's great. If it does not, that's great too. Nagel may also be overstating the case about the centrality of the combination problems, but they do point out that some common sense assumptions we have about either the physical, the phenomenal, or both, are deeply wrong.



8.3 Category two: The paradoxes of epiphenomenalism

(3) The knowledge paradox ­ Sidney Shoemaker introduced the first paradox involving epiphenomenalism succinctly in his 1975 article, Functionalism and Qualia. He wrote,

To hold that it is logically possible (or, worse, nomologically possible) that a state lacking qualitative character should be functionally identical to a state having qualitative character is to make qualitative character irrelevant both to what we can take ourselves to know in knowing about the mental states of others, and also to what we can take ourselves to know in knowing about our own mental states.

I have discussed and defended some compelling arguments that qualia do not logically supervene on the physical, and that they are not physical for this reason. Barring the unlikely discovery of physical/non­physical interaction, it seems that the qualia we have could not contribute in any way to the brain states we have. That's the crux of the problem. Shoemaker is worried that, if functionalism is false ( and certainly if physicalism is false), the relations between brain states and conscious states will be "accidental" in this sense: the qualia involved in consciousness make no contribution to the brain states. Because our brain states drive our behavior, including our knowledge claims, it seems that qualia would be irrelevant to what we could or could not know.

The following argument produces the paradox. We accept the supposition that the world is causally closed under physics. This means that every event in the world that has a sufficient cause has a physical sufficient cause, and any probabilistic causes that are active are also physical. Observe first that our brain states are central in producing our knowledge claims, and that it seems very implausible that the objects of knowledge can be completely irrelevant to our knowledge claims. Given this need to connect our objects of knowledge to our knowledge claims, it seems plausible that whatever makes something a piece of knowledge involves possessing certain brain states as essential intermediaries between our claims and the objects of our knowledge.

We must, it seems, connect our knowledge claims to our objects of knowledge in a way that accrues justification. The brain states producing knowledge claims are solidly physical, and it seems like all appropriate relations involving them must somehow involve a causal difference to them. Since we are assuming the causal closure of the physical, nothing non­physical can make a causal difference to our brain states either directly or indirectly. Consequently, any plausibly justificatory relations between knowledge claims and other elements of the world (via brain states) must obtain because of the difference those elements make to physical states of affairs. Given the hypothesis that qualia are epiphenomenal, it seems that we could not know we have qualia, and yet we do. Hence, Liberal Naturalism is caught in a paradox. The third paradox facing Liberal Naturalism stems from the fact that we know about our conscious experiences even though they seem to be causally irrelevant to our knowledge claims.

In chapter ten I will return to the basic argument for epiphenomenalism, giving it a more careful formulation. There I will begin to address it in more detail, eventually producing a refutation by challenging a hidden premise. For now, remember, I am just stating the intuitive problem. The Liberal Naturalist seems committed to conceding that consciousness makes no contribution to the fact that we make the claims about it that we do. It makes no contribution to any of the ways we currently behave, and that is highly puzzling. Since any accuracy in our claims about it would seemingly be based on fortuitous coincidence, it seems impossible, then, that we could know about it.

I think this is an almost unbearably subtle problem. The two most obvious moves do not obviously succeed. One temptation is just to deny the first premise, concluding that the world is not causally closed under physics. Most modern philosophers and scientists reject this move, being justifiably wary of the interactionist dualism implied by it. It might work if causal gaps exist.(1) However, we do not have any evidence that such gaps exist, and the reply is not convincing without empirical support. Given our current state of knowledge, it seems like a bet on a long shot.

A second common move involves positing a nomological correlation between conscious states and brain states, and asserting that this is sufficient for justification even though the conscious states are not causally involved in producing our claims about them. Perhaps, according to this view, a law exists that connects conscious states to brain states? This law assures that we make correct reports of our experiences through some appropriate parallelism between it and an associated brain state. Perhaps the associated brain state is the same one that eventually leads to claims about that experience, and the law assures that this associated brain state reliably produces a truth­maker for the claims it produces. So, usually, when I say that I am having an experience of phenomenal red I actually am having an experience of phenomenal red. My brain is producing one via this law.

This move is problematic even if we put aside its obvious ad hoc character. It establishes a nomological correlation between conscious states and brain states, but this seems inadequate to impart justification. Consider the parallel case of Trey, who is nomologically connected to Java, a volcano on Mars. Imagine a very strange law, one assuring that whenever Trey thinks about Mars, Java erupts. Therefore, Trey's brain states are reliably correlated with Java's eruptions. Now imagine that Trey comes to believe that a volcano on Mars erupts whenever he thinks about Mars. He therefore continually claims, "Right now, a volcano is erupting on Mars." as he thinks of the planet. Are those claims justified?

It seems that Trey may come to believe in the connection for completely unjustified reasons, and therefore those claims would not be justified. For example, perhaps he was reading a long philosophical work on consciousness. He reached a point in the work that discussed this bizarre idea, and convinced himself that it must be true of him. If he were making his claims for those reasons, they would not be justified. Their truth would be luck, despite the reliable nomic connection between them and the volcanic eruptions.

Trey is really only justified in his claims if he is justified in believing in the connection in the first place. In short, while Trey's brain states are reliably, nomologically correlated with volcanic eruptions on Mars, he is not justified in believing in the eruptions because they are not reliably correlated in the right way. Justification is more than nomological correlation.

So the two most obvious methods for shoring up the paradox seem unpromising. We cannot deny the causal closure of physics, nor attribute knowledge of consciousness to reliable correlations between brain states and conscious states. Given the presumption that conscious states are epiphenomenal, it looks like its qualitative character should be completely hidden from our cognitive psychology. How can we explain the peculiar familiarity we have with our own feelings?

What we need is a way of explaining the intimacy between the phenomenal subject and its physical states. The ultimate expression of this paradox is the conclusion that the Liberal Naturalist is committed to epiphenomenalism or interactionism. That seems to rule out anything but the most implausible external connections between consciousness and behavior. However, room exists to doubt whether this is a genuine dilemma. To fight the paradox, the Liberal Naturalist really needs a deeper understanding of causation itself. Perhaps a full analysis of causation will yield a place for consciousness that is neither epiphenomenal, nor interactive.

(4) The superfluity of consciousness ­ Assume that we reject interactionism. This leaves us with an apparent radical epiphenomenalism that produces a fourth problem. To generate the problem, we note that experience is central to our mental lives. Unfortunately, since it is purely epiphenomenal, it does not do anything. This kind of revelation about nature is very problematic, as it means that nature is not parsimonious. The discovery of one such phenomenon conflicts with the convictions needed by a thoughtful realist.

Realists must take Occam's razor to be a metaphysical tool, rather than just an epistemological shortcut. Realists should take it to be a statement of faith in the parsimony of nature. Because of realism's closeness to the razor's keen edge, our confidence in the approximate truth of our scientific theories is always in danger of fatally cutting itself. However, the existence of consciousness, if it is truly a superfluous epiphenomenon, shows that at least one wholly superfluous set of properties exists in nature. It is a clear counter­example to the realist's faith. Consequently, we should be suspicious that nature may absolutely abound with superfluous properties.

Worse yet, if consciousness is superfluous, we could not find grounds to resist the idea that nature may be full of countless efficacious entities that are, from a theoretical standpoint, formally superfluous. After all, we would have one central and important piece of evidence that nature is an absurd sort of place. Can the theories that we use to describe the world be radically wrong precisely because they are the simplest theories? Perhaps God is politically correct and tries to maximize diversity. The superfluity of consciousness would make Occam's razor seem more like a pragmatic principle, and less like a metaphysical tool. As such, it would undermine our confidence in both philosophical and scientific explanation to get at anything like ontological truth.

Summary ­ The knowledge paradox is frequently cited, but the paradox of superfluity is not frequently acknowledged in the literature. Penrose (1989) and Hodgson (1991) take good shots at it, suggesting that consciousness gives the mind some non­algorithmic power. However, even if some proposal like this is correct, it will remain unclear why performance of the proposed functions is linked in any way to the existence of qualitative feels. As I have argued, it seems logically possible that any account of what function consciousness might perform could be made true without experience. Such accounts are just the wrong place to look. To address these problems in the depth they deserve, we will have to return to the question of epiphenomenalism in chapter ten. I will begin to attack the problem directly at that time.

8.4 The grain problem

(5) ­ The fifth puzzle is the "grain problem", as introduced by Wilfrid Sellars (1963) and recently taken up again by William Lycan (1987)(2). What is the grain problem? It comes from noting that the physical character of brain processing involves structure not possessed by phenomenal qualities. For instance, the structure of an expanse of phenomenally experienced color does not divide into finer and finer substructures corresponding to the microphysical structure of the brain, or brain events. Occurrent phenomenal colors, like blue, are structurally homogenous despite their physical correlates having a highly variegated structure. The problem posed is to understand how a homogenous quality can be identical with, or constituted by, a richly structured physical entity.

As physicalists, Sellars and Lycan propose that our physical understanding of the world must be incomplete. Sellars challenges us to find the "non­particulate" foundations underlying our particle­filled understanding of the world. Other physicalists suggest phenomenal qualities may have structure hidden from awareness after all (Van Gulick, 1993). This latter suggestion seems untestable, as any such test would have to assume the materialist answer it is supposed to be investigating.

Interestingly, by again appealing to the functional aspect of an entity, in isolation from its physical aspects, the Liberal Naturalist may make some headway on this problem. The grain problem relies on the assumption that the physical character of the system is the ultimate basis of consciousness. For a physicalist maintaining that something physical constitutes, or is token identical to, every mental event, this truly does raise a problem. Physicalists are not free to ignore the presumed physical basis of a mental entity's existence.

Liberal Naturalists are free of the assumption that conscious states ultimately reduce to physical states at the token level. They may ask, "What follows for the structure of conscious experience if we assume that it is the functional character of the system, and only the functional character, that is implicated in determining the character of consciousness?"

Attempts to answer this kind of question could prove to be fruitful. Consider any Finite State Automaton that consists entirely of a network of states connected by transition rules. These states are the kinds of states they are because of their connection to other states within the network, and not because of any internal structure they might have. For the purposes of the abstraction that defines their type, each state is essentially atomic. Imagining logically possible worlds with structureless particles implementing given FSA's is easy. In general, no internal structure is essential for a given FSA to have the type of states it has, and none needs to be specified.

This way of thinking about computational entities is familiar to computer scientists and engineers, as they also individuate real world functional states at a coarse grain, a grain abstracted from lower­level details. In design situations, specifying a functional entity requires picking out a system, and a canonical set of states for the system. The target system and its target states provide a contextual point of reference relative to which we can define the existence of functional entities within that system.

As things turn out, a variety of ways usually exist for realizing the causal contribution that these components make to the functioning of the larger system. This old point about multiple realization requires those components to be defined in a way that divides through the fine­grained details. Therefore, a functional entity's type will always be categorically independent of the fine grained structure possessed by any particular realization.

How does this help with the grain problem? To sum up: functional objects are defined relative to a canonical reference system by the causal role they play within that system. These specifications of their type need not involve lower­level structural details. Instead, they are individuated relative to some level­encapsulated pattern of interactions. These interactions chiefly involve other objects at the same level, and implicitly invoke the referent systems they partially constitute. Implicit in the identification of a functional being is a choice of level of nature at which the functionally characterized system exists and is interacting. Ontologically, such entities subsist in the pattern of high­level differences their states can make to other systems at the same and higher levels. This means that objects and states are, qua functional beings, level­encapsulated. As such, functional Being has no finer grain.

Summary ­ For a Liberal Naturalist, the prospect of solving the grain problem by appealing to the coarse grained nature of functional Being is intriguing. As before, an appeal like that cannot take us the whole way, but may play an essential part in a final explanation. Also, some detailed analysis of the terms involved must accompany it, terms such as "interaction," "level of nature," and "canonical reference system." In short, it requires investigating causation, and the way causation individuates, and stratifies, entities in the world. This brings us to the sixth problem, the problem that has been most ignored in the literature, yet seems to be the most crucial clue of all in our search. It is a problem about the very existence of phenomenal individuals with boundaries at mid­levels of the natural world. This Boundary Problem for Phenomenal Individuals is interesting and subtle, and I will devote the entire next chapter to it.

1. I am making the usual assumption that the randomness in QM is either a discovery that causation is probabilistic, or a stopgap to be filled by hidden physical variables. I am purposely not treating it as a discovery that there are causal gaps in the physical world.

2. In his more recent book, Lycan (1996) takes a more deflationary attitude towards the problem.


Chapter 9

The Boundary Problem For Phenomenal Individuals

9.1 Introduction to the problem

Bertrand Russell once said that the aim of philosophy is to start with something so obvious as to not be worth mentioning, and to end up with something so absurd that no one will believe it. The boundary problem for phenomenal individuals puts a new twist on Russell's view. The problem starts with some obvious observations, but ends with an unusual enough puzzle that the most problematical part might be bringing ourselves to accept it as a puzzle. As with many issues surrounding phenomenal consciousness, perhaps pointing to the problem is not the greatest achievement. A greater achievement may be understanding it as presenting a challenge, something to be explained rather than explained away, or hidden from.

What is the boundary problem for phenomenal individuals? It starts with the observation that the consciousness of an individual has inherent boudaries. Only some experiences are part of my consciousness; most experiences in the world are not. Arguably, these boundaries are what individuate me as a phenomenal individual in the world. I will argue that this poses a problem any theory of consciousness must answer. How can consciousness have boundaries? What element of the natural world dictates the way that these boundaries are drawn? On the face of it, the physical world does not have appropriate boundary­makers at the midlevel where consciousness exists, but consciousness has boundaries all the same. This is the boundary problem for phenomenal individuals.

Unfortunately, the problem is not immediately obvious, so we will have to do a bit of hard work to see it. Still, the challenge that comes out of it is clear. We must find something in nature that can ground the natural possibility of a phenomenal individual bounded in just the way our consciousness is.

9.2 The conceptual foundations of the problem

The problem begins with some obvious observations, observations that help define for us what it is to be a phenomenal individual. To begin, reflect on the fact that phenomenal individuals come in discrete tokens. Without too much strain, we can think of each subject of experience as being a kind of quantum. I am one such quantum, and so is Bill Clinton, and so are you(1). These quanta, the individuated phenomenal fields of experiencing subjects, contain co­evolving elements. In some vague and notorious sense of "unified", these co­evolving elements are naturally unified into a subject of experience.

Second, the phenomenal field also has boundaries. Not every feeling is part of my phenomenal field since I do not feel the pains produced by damage to your body. The unity and boundedness of the phenomenal field stand together at the core of the concept of a phenomenal individual. The driving intuition is that phenomenal individuals are inherent individuals in a sense of "inherent" we must try to make clear.

If these boundaries could not exist, then nothing like human consciousness would be possible. The problem comes from appreciating that these boundaries are primitive relative to our physical image of the world. It is only understood once we appreciate that the possibility that these boundaries could exist is not inevitable, that it seems the universe could have been quite different. By doing this, we can begin to become aware that nature must have intrinsic structure we do not yet fully appreciate.

Third, our human consciousness is only a species of phenomenal individuality. A phenomenal individual is a manifold of qualitative entities. We only roughly name these entities in our own case with words like feeling, sensation, and appearance. Other kinds of phenomenal individuality may exist in other kinds of beings.

Fourth, in the human case, the phenomenal individual is associated with a human body and its cognitive processing. Humans, and the activity of human cognitive systems, are individuated at a midlevel of the physical world. Typically, our individuation of objects at this midlevel of nature is fluid, context sensitive, and interest relative.

This kind of individuation is highly conceptual, and hinges on facts about the abstract organization and causal cohesion of lower­level activity. As a consequence, events or objects may form parts of many individuals simultaneously, depending on how one wishes to conceive of the organization of the world, and draw the individuating boundaries. For example, a cell may be an individual; also, at the same time, it may be part of an organ; at the same time, it may be part of an individuated bodily system like the reproductive system; at the same time, it may be part of the organism as a whole, and of that organism's society. For each of these different individuals, a different kind of causal and abstract organization exists in the world. Finally, levels of abstract organization and causal cohesion exist between microphysics and human cognition, and between human cognition and the universe as a whole.

All those things should be more or less obvious. What is the unobvious part that reveals the boundary problem? The unobvious part requires realizing that the phenomenal individuals could have been different individuals than they are, and different in ways that would prohibit human consciousness. Therefore, we need to explain why phenomenal individuals are associated with the specific, midlevel patterns of interaction and organization that they happen, in fact, to be associated with.

Here is one way to think of it, just to orient the imagination before I begin in earnest. Internally, these midlevel patterns consist of other patterns of interaction and organization. The phenomenal field of the subject stretches across the boundaries of these interactions. Yet our cognitive systems are also part of even larger patterns than ours. Somehow, the boundaries of the phenomenal individual do not continue to overflow and enlarge into these patterns.

Why do the boundaries exist here rather than there? To head off misunderstanding, I want to emphasize that this is not a question about why my (or your) consciousness is not larger or smaller than it is. It is a question about why there is a phenomenal individual that can be me, or you, rather than one which is both of us, or many that are parts of us. After accepting the boundary conditions that human consciousness has in fact, we still need to ask: what is special about those patterns that allow them to provide boundaries for phenomenal individuals?

The very obviousness of our own existence as midlevel phenomenal individuals is an obstacle to appreciating the question. Because it is something which is always with us, it may not be easy to realize how remarkable it is that things even can be this way. I now want to bring out the unobviousness of the brute fact that phenomenal individuals like us could even exist, individuals localized at a midlevel, with midlevel boundaries to what we feel. The main points I want to bring out are this: (a) if it were not possible to draw these boundaries to the phenomenal field, humanlike phenomenal individuals could not exist; (b) the fact that such boundaries exist where they do is surprising, and the grounds for their possibility are not obvious. To bring out the problem more vividly, I will use several thought experiments designed to loosen our intuitive grip on the inevitability of the boundaries that actually exist.

9.3 The brain and its subsystems

Abnormal forms of consciousness, such as multiple personality disorder, open the door to the possibility that, in some circumstances, multiple phenomenal individuals may co­exist within a single brain. Braude (1991) describes cases of MPD in which different personalities may be co­present, each claiming to be a distinct center of awareness. Among many peculiarities, these centers of awareness (which Braude describes as apperceptive centers) make claims to sharing a variety of relations among their experiences. Sometimes they claim distinct experiences altogether. In these cases, the experiences of each personality are "screened off" from the others, so the different personalities achieve, apparently, privacy of experience. In other situations, their experiences partially overlap, some belonging to multiple centers of awareness, and others only to one. In still other cases, experience may be completely shared although particular experiences may sometimes claim to be owned only by one or another center of awareness. In such cases, John and Mary may both claim to have an experience, but only John claims it as his experience. These cases raise very puzzling issues about the facts of the matter. What are the number and boundaries of the phenomenal individuals that exist in these cases? Is there really one for each personality, or are the claims issuing from confabulation? It seems to me that each hypothesis is at least coherent.

Realizing the coherence of each of those hypotheses is a first step in beginning to see the possibilities that bring on the boundary problem. Most likely, the boundaries of phenomenal consciousness correspond to the boundaries of certain kinds of activity in our brains. Some evidence suggests that specially synchronized activity in and around the cortex constitutes the boundary maker for human phenomenal individuals. For example, Crick (1995) hypothesizes that each level of visual processing is coordinated by its own region of the thalamus. A good point to begin the project of making the boundary problem clear is to ask whether any of the subsystems oscillating within the whole might also constitute a phenomenal individual.

Consider the patterns of synchronized activity that carry and organize auditory information from our ears. Is there a phenomenal individual associated with this activity alone? If there were, we would possess a picture something like this: within ourselves as fully human phenomenal individuals, experiencing subjects, there would exist other phenomenal individuals, themselves perfectly complete subjects of experience, although much simpler. Like Russian dolls, there would be individuals within individuals within individuals, all of them phenomenal subjects.

An answer to this question should exist. The very definiteness of what it means to be a phenomenal individual seems to require it. Given that we understood the obvious observations about phenomenal individuals, extension by analogy should allow us to make sense of this question. Bring to mind the physical image of the world: we have a multitude of interacting microphysical entities at places and times, this multitude congealing into a macroscopic whirlwind of finely layered patterns of organization. Simply imagine looking at the patterns of physical activity in the world from the perspective of a third person observer. Note the coherence of causal and abstract organization at the many levels, and the many ways it exists. We know that a set of these patterns supports boundaries that allow for the existence of us, where we are one kind of phenomenal individual.

The question here is merely if nature supports other kinds of feeling subjects, other kinds of phenomenal individuals. Specifically, we are merely asking if, analogous to us as phenomenal individuals, there might exist simpler, experiencing subjects whose boundaries are given by subsets of the activity that determines our experience as a whole. After all, the subsets we are considering are like the more complete set in many ways. They share common biology with the larger set of events; they carry information, and are processing it; they process it in a very similar way; within themselves, they are internally synchronized and coherent. Is any one of these subsets, itself, a phenomenal individual? Remember, as I discussed in chapter seven (section 7.6), the phenomenal individual corresponding to an auditory element does not need to be experiencing sound. Instead, such a system might experience qualia uniquely appropriate to its own level of reality, and responsive to its own finer grained causal organization. Is this the way our world is?

Nature surely has an answer to this question. Since the question only arises after one is aware of the physical facts, and is in fact suggested by them, the answer seems to be indeterminate relative to those facts. After reflecting upon the physical situation, both yes and no seem possible.

This is merely a conceptual point, a point about how our intuitive understanding of what it is to be a phenomenal individual leaves the question open. Why couldn't there be phenomenal individuals at many levels of processing, associated with subsets of the cognitive activity that corresponds to our own phenomenal individuality? On the other hand, why would there be?

Since such brain activity, taken as a whole, can correspond to the existence of phenomenal individuals, us, the point seems to generalize to a relatively mild claim. Our intuitive concept of what it is to be a phenomenal individual allows for the possibility of simpler phenomenal individuals, individuals whose manifold of experience consists of much less rich and less cognitive experience. Given this, the physical activity in fact corresponding to the existence of a phenomenal individual might also support other, simpler phenomenal individuals via the simpler patterns of organization it contains. Of course, it might not.

In the first case phenomenal individuals would be molecular in a certain sense. Where they exist, the same activity that determines their existence contains simpler, local eddies of organization supporting the existence of simpler, self­contained individuals. In the second case phenomenal individuals would be, in an analogous sense, atomic.

Traditional panpsychism takes advantage of this looseness in the concept to make its position intelligible. Process philosophies, such as Whitehead's, typically take advantage of it by making feeling and phenomenal individuality pre­cognitive primitives. I now want to take this looseness in the concept and, in stages, parlay it into the full­blown boundary problem for phenomenal individuals.

9.4 Moving towards Scylla and Charybdis

We are going to approach the boundary problem in stages. The first scenario we need to consider is a variant on Ned Block's well­known fiction of the Chinese nation simulating the functionality of a human brain (1980). To make it a little less fantastic, we can imagine the simulation of some other, simpler kind of organism's brain, maybe a fish. Very likely, a fish is a phenomenal individual.

Imagine building a robot fish. Imagine also that we have designed its nervous system to be functionally isomorphic to the nervous system of a naturally occurring fish. The processing has been made remote in the usual way, with inputs and outputs to the fish's central nervous system employing relays. These relays send signals to remote stations manned by human beings. The humans monitor the signals as they come in, and relay an output to other destinations. Some signals are sent between the remote stations, and some are relayed back to the fish as motor outputs. In this way, we imagine that the relay system is functionally isomorphic to an actual fish's brain.

The question we are supposed to ponder is whether or not this system is conscious. Block uses the example in an attempt to show that our concept of phenomenal consciousness does not necessitate a positive answer, and therefore analytic functionalism seems endangered. Such considerations support the view that consciousness does not logically supervene upon functional organization. However the argument fails to show that the system will not support the existence of consciousness, and hence of a phenomenal individual, in fact. As physicalists sometimes point out, it seems just as surprising that our brains would support consciousness, but we nevertheless know first­hand that they do.

The fact that the system has parts that are phenomenal homunculi should not make a difference. In the previous section I argued that the idea of phenomenal individuals whose physical organization supported the existence of other phenomenal individuals seems consistent. Since the relay system is functionally like a fish's brain, it is certainly conceivable that this system actually supports a phenomenal individual. In fact, both of these seem to be possibilities:

1) Each homunculus is a phenomenal individual, but the whole system is not.

2) Each homunculus is a phenomenal individual, and so is the whole system.

These possibilities are not so innocent. We can redirect the principles that make them plausible back to a local system for the fish. The homunculi system is functionally isomorphic to the fish's cognitive system. Each homunculus maps onto some important part of the organizational structure of a naturally occurring fish. Imagine the mapping being made with one of the homunculi, call her Edna, mapped onto some functional part of the fish, call it the E­system. No principled reason exists for restricting possibility (1) to the homunculi fish alone. By analogy, it would seem to ground the possibility that the fish's E­system, the part corresponding to Edna, could be a phenomenal individual even though the fish as a whole would not be.

How does the analogy go? By admitting possibility (1), we are admitting the coherence of the idea that the homunculi system as a whole may not be a phenomenal individual. In doing so, we are admitting the coherence of a world where (a) a system may contain phenomenal individuals, and (b) that system may be functionally isomorphic to the fish's system, and yet (c) that system is not a phenomenal individual. In the previous section I gave reasons why it seemed coherent that ordinary cognitive subsystems could be phenomenal individuals themselves. To imagine the E­system as a phenomenal individual, but not the fish, all we have to do is combine the two points.

To combine them, we need to conceptually shift the boundaries that make phenomenal individuals. Shift one's view of nature so that the phenomenal boundaries stretch through the E­system, encompassing all the activity within it, but do not overflow the boundaries of the E­system. The larger individual is abolished. In its place is a collection of simpler phenomenal individuals in interaction and cooperation. Of course, the experience of the E­system would be vastly different from the experiences of Edna. This difference is minor, and it does not stand in the way of our being able to change the intuitively assigned phenomenal boundaries. By doing that, we rob the fish as a whole of its status as a phenomenal individual. Lacking clear criteria for natural boundaries, we can conceptually rearrange the boundaries, forcing the individuality down to the E­system level.

Once we have seen the essential analogy between Edna and the E­system, we can begin to engage in other conceptual shifts. Obviously the E­system might not be a phenomenal individual. It seems perfectly coherent that the only phenomenal individual associated with the fish's brain is the one existing at the global fish level. The coherence of the idea that natural fish do not have phenomenal E­systems seems to support a third possibility:

(3) The homunculi system would be a phenomenal individual, but none of the homunculi would be.

Everyone is confident that in fact Edna would be a phenomenal individual. That should not lead us to overlook the failure of consciousness to logically supervene on the physical. This failure raises the logical possibility of phenomenal Zombies. Phenomenal Zombies are physical systems organizationally just like human beings, but without consciousness. With the spectre of Zombies looming, we need to explain why (3) is not true of our world. For instance, I noted in chapter seven that people often think that phenomenal individuals emerge at a certain level of complexity. I argued there that the proposal is very implausible, but it does seem like an empirical possibility. If it is possible, then imagining (3) merely requires imagining such a world, dictating an appropriate kind of complexity, and then moving the starting point upwards, past Edna. The complexity point where phenomenal individuals emerge would be higher than that possessed by the homunculi, but not the homunculi fish. The result is that Edna could be a zombie, and a component in a system supporting a phenomenal individual. What keeps phenomenal individuality right there, between Edna's ears?

This reconception of boundaries is just the flip side of the earlier suggestion. Earlier the reconception was a movement of the phenomenal boundaries to lower levels of organization, robbing wholes of their phenomenal individuality. Here, the reconception is to a higher level of organization, robbing parts of their phenomenal individuality. Such a world would be one where the eddies of coherent causation that are human beings would not support phenomenal individuality. No human phenomenal consciousness would exist. Instead, the phenomenal individuals exist at a higher level. As human bodies move and swirl and interact with other eddies, the phenomenal individual arises only for the super­system. The possibility is analogous to the way we (or many of us) normally imagine that our cognitive subsystems contribute to our conscious lives without themselves being phenomenal individuals.

These science fiction tales are simply meant to make vivid that our intuitive understanding of phenomenal individuals does support a great deal of possible variation in their boundaries. Once the basic point has been appreciated, we do not need to use philosopher's thought experiments to make it. Even actual systems like economies or political systems or nation states bring it out.

An economy is an extremely complex system of interacting components. In its own way, it processes and reacts to information and self­organizes. As a physical system, it stores information and seems to have a kind of memory, synchrony between its parts, massive parallel communication, and so forth. The scale at which this all takes place is much larger than in an individual brain, and much different in detail, but most of the basic kinds of activity exist. Suggesting that an economy might represent some kind of group mind is nothing new, and for good reason. And it is not a philosopher's question, really. It is a legitimate question of fact about something that actually exists in the real world.

What is the main reason for rejecting the idea? Mostly, it is the bizarreness of believing that the U.S. economy supports a phenomenal individual, and most people consider possession of phenomenal consciousness to be essential to having a mind. Now, I do not know if the economy in fact supports the existence of a phenomenal individual, and actually tend to doubt it myself. Still, it seems every bit as coherent to me that an economy would than it does that my brain would. And I am quite sure my brain does. The economy would have to possess a very different kind of phenomenology than we humans, but there does not seem to be good reason for thinking human­type phenomenal individuals are the only kinds that could exist. And, anyway, it certainly shares many of our mind's most salient characteristics, just stretched out vastly in scale over space and time.

Even the scale differences do not amount to much once one considers that our phenomenal individuality arises from collections of neurons or molecules. If bunches of neurons (or molecules) stand to us as we do to the economy, and their organization supports us as a phenomenal individual, why couldn't we support the economy similarly? Once we see the possibility that both the economy and our bodies might support phenomenal individuals, we are only a short step away from seeing another possibility. It might have been that we are not phenomenally conscious, but the economy nevertheless would be. After all, it seems coherent that our neurons are not phenomenal individuals even though we are. The boundaries of phenomenal individuality, once loosened, can begin to shift radically. Again, why couldn't our bodies just be local, non­phenomenal causal eddies within a larger phenomenal individual? What grounds the bruteness of these boundaries?

9.5 Scylla and Charybdis: The Boundary Problem

The boundary shifting that occurs in these thought experiments is enabled by the fact that information about pattern and organization alone does not fix the boundaries. Most objects at higher levels of organization in the physical world are individuated via an act of abstraction. This involves extracting some significant pattern from the flux of microphysical interaction. Consistent with a given pattern of microphysical causation, innumerable ways exist of conceiving and reconceiving the abstract organizations that supervene.

Just adding these facts about pattern, or abstract organization, to the causation between the microphysical entities does not seem to go far enough in determining the proper sense of "inherent" in the idea that a phenomenal individual enjoys a kind of inherent individuality. One can coherently hypothesize almost as many ways of determining boundaries for phenomenal individuals as there are of abstractly organizing and re­organizing the patterns of microphysical interaction in the world. The resulting scenarios are intuitively bizarre, but bizarreness is not inconsistency. The fact that the boundaries are drawn as they are, are drawn in a way that yields human consciousness, stands out as a brute fact.

We have reached this point: we seem to be faced with the need to understand more deeply what it is to be an inherent individual in the natural world. We need a natural criterion for individuation, one that illuminates the specialness of some patterns over others as supporters of phenomenal individuality. One good set of candidates are the fields of the most primitive particles (or strings or whatnot). Each of these has a natural dynamic unity, one that seems inherent. A phenomenal individual might be associated with each of these.

This suggestion threatens human consciousness. If the fields of the primitive individuals of physics are the only natural individuals, the rest of us are mere abstractions off the pattern of their interaction. Each primitive physical individual may be a simple phenomenal individual, supporting brief firefly flicks of feeling buzzing at the lowest levels of spacetime, but above them the world is dark. This world is the panpsychist's world painted surrealistically: feeling, feeling everywhere, but none of it can think. It does not include a mechanism to allow the creation of higher­level individuals. No mechanism can bootstrap us to human consciousness.

Perhaps by flowing through the lines of interaction the phenomenal individuals could outrun the boundaries of the primitive individuals of physics. Here the trap concerns stopping the flow of interaction. It can seem that the flow of interaction in the universe is inherently unbounded, and no merely abstract pattern presents a natural condition for containing it. Those patterns merely direct it from one area to another, orchestrating it, moving it along through the continuity of the universe. According to this view, phenomenal individuality must follow the boundaries to their limits along these lines of interaction. This makes for the possibility of a universal phenomenal individual, perhaps some kind of cosmic consciousness. Unfortunately, no room exists for the more mundane, midlevel boundaries necessary for human consciousness to exist. Like the first view, this view banishes midlevel individuals from existence.

These two views are a Scylla and Charybdis for theories of consciousness. One view pushes us inward, past the point of midlevel individuation, and into the realm of the subatomic. There, and only there, do we find our natural, inherent individuals. Another pushes us outward, past the boundaries of the subatomic individuals, ever outward along the lines of interaction between them, racing past the midlevel to the continuous unfolding of the cosmos. Only there, at the level of the universe, do we find our inherent individual. Neither view allows for human beings. To navigate the middle ground, we must find a principle that allows us to push those boundaries outward, but only just so. We must be able to go only so far past the microphysical level, but no further. That is the boundary problem for phenomenal individuals.

9.6 Two objections to the problem

The boundary problem, upon first being presented, can make one a little dizzy. Two things about it seem to be responsible. First, certain steps in the argument rely on a disassociation of phenomenal individuality from cognition. The second is that it seems to be introducing an intractable problem, but it is unclear that it is contributing anything constructive. Against both these intuitions, I want to argue that the boundary problem should be accepted as a real problem for our understanding of nature. Furthermore, if we accept it as something to be explained, rather than explained away, it turns out to be a helpful problem. It stands out as a vital clue that may be helpful in construction of a viable and detailed Liberal Naturalism.

The first wave of resistance comes from the idea that all these alternative cases fail because they seem to require phenomenal individuals that do not have the appropriate cognitive structure. As such, they cannot possibly be conscious, for consciousness has a definite functional aspect. Because of this, it requires a very definite range of function to exist.

I think, even granting this objection, it is not clear the problem goes away. One might produce arguments for both the thesis that subsystems of the brain are cognitive in the appropriate sense, and also that the U.S. economy is. A modified version of the problem then could be presented based solely around those cases. However, this grants the objector too much, and would cause us to overlook the most basic place where this objection goes wrong.

The most basic reply to the objection is to point out that, if it were correct, panpsychism would seem to be ruled out a priori. As I argued in chapter seven, no compelling a priori reasons exist for thinking that panpsychism is false, and some good reasons exist for thinking it will turn out to be true. If panpsychism is false, that is an empirical fact and does not follow solely from our understanding of what it is to be a phenomenal individual. In my exposition of the problem, I was careful to put it in terms of phenomenal individuals, but not conscious individuals, or phenomenal consciousness. The basic stratagem, as before, is to decompose the notion of phenomenal consciousness into the dual notion of phenomenal individuality, plus suitable cognition. The idea of phenomenal individuality then stands out as more basic than consciousness, as it does not presuppose specific kinds of cognition, or any cognition at all.

This distinction is essentially what stands behind panpsychist suggestions. The world may consist of many simple kinds of phenomenal individuals, bounded subjects of experience or qualia, alongside the fully conscious individuals such as ourselves. These individuals would not reason, or reminisce, or have conceptual capacity, but they would be nexi of feeling and, perhaps, significance. Again, these are the ideas that show up in panpsychic philosophies such as Whitehead's. Those philosophies may (or may not) be wrong, but someone who wants to convict them of incoherence needs to produce some compelling argument. I do not think such an argument exists.

Once phenomenal consciousness has been decomposed into the concept of cognition plus phenomenal individuality, we see that the existence of human consciousness presupposes some special answer to the boundary problem for phenomenal individuals. Many, many different kinds of phenomenal individuals might exist, with those corresponding to cognitive systems being just one type. Furthermore, the existence of our special type is predicated on something in nature that enables the boundaries of phenomenal individuals to be drawn just so, corresponding precisely to the cognitive activity of certain systems. Therefore, a solution to the boundary problem for phenomenal individuals is a prerequisite for a complete theory of phenomenal consciousness.

Admittedly, we can make important progress toward a theory of consciousness by simply taking the fact of cognitive boundaries for granted. Nevertheless, this will not be a truly fundamental theory, one imparting deep understanding, unless it actually solves the problem. A solution would allow us to derive the boundaries that we know exist in our own case from the non­cognitive facts about nature. We would also be able to address the issue of panpsychism, which I simply take to be a question about whether or not there are any noncognitive phenomenal individuals. This certainly seems to be an empirical question, requiring some knowledge about nature that we currently do not have.

Even if we set aside qualms that the boundary problem requires us to disassociate phenomenal individuality from cognition, we may still be left with the feeling that a problem has been introduced that is counter­productive. What is the point of wasting resources thinking about this Boundary Problem when so many meaty, empirical issues surrounding consciousness in cognitive science and neuroscience still remain? Shouldn't we keep our feet closer to the ground? In reply, I think that reflecting on the problem is worthwhile for Liberal Naturalists, as we are interested in constructing a dual aspect or dual property view of phenomenal consciousness.

On a Liberal Naturalist view of phenomenal consciousness, the natural world has both physical and phenomenal aspects. Liberal Naturalists think of these two things as each being an aspect of a deeper kind, and we think of the natural world as being fundamentally structured by this deeper kind. A viable, detailed Liberal Naturalism will have to identify the deeper kind, and rethink nature in terms of it. This rethinking will allow us to make intelligible the place of the physical facts, the phenomenal facts, and the relations between them in a larger, natural scheme of things.

The absence of explanatory targets (over and above the mere existence of phenomenal consciousness) is a crucial hurdle this project must overcome. Fruitful targets will sit at a suitably deep level, a level fundamental enough to allow them to point beyond themselves at possible research programs, or even solutions. A target of that nature will have to be divorceable from cognition, for instance, since cognition is explicable in high­level, physical terms. So, it is one thing to recognize that nature must be rethought, and quite another to find any clues or guideposts to help in that rethinking.

I suggest that the boundary problem for phenomenal individuals meets all of these criteria, and should be viewed in a positive light, as a crucial clue. The inherent individuation of the phenomenal individuals at the midlevel of human consciousness is a clue to hidden structure in nature itself. One of the explanatory targets for a viable Liberal Naturalism, then, should be discovery of a fundamental kind whose nature gives a compelling solution to the boundary problem. The boundary problem is clear, unlike the the paradox of unity; it is undeniable, unlike the problem of the subjective instant; it avoids entanglements with epistemology, unlike the problems with epiphenomenalism; and it is reasonably clear, unlike the grain problem. Just as importantly, it is unequivocal evidence for hidden structure, and so it promises to point to something both concrete and ubiquitous while the other problems do not.

9.7 The teeth of the problem: two examples

By considering two examples of dual aspect theories that falter on this problem, we can get a better sense of its importance. The two proposals I will briefly critique are the materialism of Michael Lockwood (1989), and the information theory of David Chalmers (1996). I do not believe that either successfully navigates the way between Scylla and Charybdis.

Michael Lockwood's materialism is a resurrection of Bertrand Russell's neutral monism in the context of quantum mechanics. In an argument similar to my argument from Life in chapter three, Lockwood suggests that phenomenal consciousness fails to logically supervene on the physical because physical concepts are content­neutral. They merely specify the structure of the causal flux. Phenomenal qualities and consciousness, on the other hand, are defined precisely by their content. Lockwood suggests that a nice solution to the problem is to simply draft phenomenal properties into duty as the content of the causal flux whose structure is described by physics. The result is a kind of dual aspect theory. Physical concepts are about the structural aspects of the causal flux, and our phenomenal concepts are about the intrinsic content that is in flux.

This is an interesting proposal, but something needs to be added before it can hope to account for the individuation of human consciousness at the midlevel. After all, if we are to believe physics, the individuals who are the natural candidates for this basic phenomenal content are the fundamental fields. Lockwood's theory needs to take us from this simple phenomenal content of simple individuals to the complex, midlevel phenomenal individual necessary for human consciousness. According to one horn of the dilemma, he is stuck at the microphysical level.

Lockwood, instead, appeals to the other horn for help, postulating that phenomenology flows along the lines of interaction in the world. Unfortunately, he has no principle to allow him to resist being hung on this horn, as it urges that the boundaries be pushed further and further outward. Once the bootstrapping process has begun, Lockwood's theory gives us no way to stop it. Actually, the problem is a little worse for Lockwood since he is sympathetic to the Everett interpretation of quantum mechanics. Interaction, although structured, is seamless in Schroedinger's world, and Charybdis demands a reason for stopping it here, where there are human cognitive systems in one eigenstate. There does not seem to be a compelling reason to think Lockwood's proposal would result in anything less than a many­worlds­sized individual.

David Chalmers proposes that phenomenal properties and physical properties might be two aspects of information spaces. If we take his suggestion as being unrestricted, we immediately are confronted with the problem of individuating information spaces. On Shannon's view, which Chalmers appeals to, information is a difference that makes a difference along some causal pathway. But a difference that makes a difference to what along the pathway? Falling on the first horn, we can recognize informational differences to the basic individuals, but that banishes human consciousness.

Falling on the second horn, we can recognize the universe as a whole as an information space. Its structured state changes as interactions occur within it, but this also banishes human consciousness. On Chalmers proposal we should be able to save midlevel individuals by allowing for all co­varying subportions of spacetime to be information spaces. But then we are left with panpsychism run wild. Even within one brain we will have astronomically large numbers of phenomenal individuals, separately experiencing, each corresponding to different ways of carving up the activity of the brain and its causal pathways. An explanation that promiscuous is not very illuminating.

9.8 What to do?

The Liberal Naturalist should take the boundary problem seriously, and think hard about what might be missing from our current view of individuation in the world. The suggestion that we allow inherent individuality to flow along the paths of interaction between individuals sounds promising. After all, we are looking for something more than abstract organization to ground judgments of natural individuality, and causation seems to be an inherent, natural connection par excellence. Also, an interaction divides the world by its very nature, partitioning it into different spaces that mutually condition one another. The second horn of the dilemma gains its conclusion by taking advantage of a naive view of interaction, one that capitalizes on the rough understanding of causation we currently have.

One good strategy to follow would be to think harder and more carefully about interactions in the world. We should think in more detail about the way they might condition nature into individual, mutually influencing regions, and do so at many levels simultaneously. We very well might discover that interactions have certain important aspects we can use to mark off candidates for natural individuation. These individuals would then be candidates for supporting phenomenal individuals. The job would then be to look for a physical reflection of this special feature of interactions.

Along this line, I can suggest two possibilities for the physical reflection of individuality. The first is quantum coherence. This is hardly a novel suggestion, but it would be the simplest solution if it were true. The large scale coherence a quantum individual shows strongly suggests some innate, system­level connection among the parts. The boundaries of the coherence could perhaps explain the boundaries of an associated phenomenal individual. My own feelings are that this would be the simplest solution, but that it is not promising. Perhaps researchers like Penrose and Hameroff (1996) will discover large scale coherence in the brain. If they do, that will be wonderful. Nevertheless, from our vantage point today I do not think we have good reason to be optimistic.

If the brain does not support large scale coherence, we will have to look somewhere else to understand the hidden structure. One place we might look is at decoherence between the brain and its environment. The process by which certain systems decouple from all of the quantum possibilities but one may very well be a process by which nature determines that system as a whole. One possibility that we must consider is that inherent individuation in nature may exist at many levels, and it may very well be that certain patterns of interaction must decohere as wholes, if they are to decohere at all. In doing so, the shared causal fate of its constituents may bound the experience of the system.

These are just two suggestions, and neither may be right. What I do think is very likely right is that we must understand the causal structure of our world better. Causal connections seem the best candidates for helping to understand more deeply the naturally individuated, midlevel structure exhibited in our phenomenal existence. It seems that wherever we turn in trying to understand consciousness, we end up spun around and facing questions about causation. Let us now turn our attention to that issue, reconsidering the argument from the failure of logical supervenience to epiphenomenalism. That argument will supply our final transition to the topic of causation.

1. I am using "quanta" in its classical sense. I do not mean to suggest quantum mechanical ties.


Chapter 10

Transitions

10.1 First steps

Do our physical explanations present the full story about the process of causation, and the causal structure of the world? Beginning in this chapter, I am going to continue motivating, and begin to develop, a view that involves nonphysical aspects of causation. These next few chapters meet several burdens. First, they further motivate the need for a substantial ontological analysis of causation; second, they argue for the counterintuitive thesis that the predictive adequacy of physical theories fails to imply their causal completeness; finally, they begin to fill the gaps in our understanding of causation suggested by earlier arguments, showing more concretely just how physical theory is aspectual.

I will finish with a foundation for a version of Liberal Naturalism. The threads of causation and consciousness come back together later when I introduce the Carrier Theory of Causation. There, I give reasons for believing that the phenomenal field of consciousness may be an aspect of causation.

10.2 The specter of epiphenomenalism

We have already seen that the problem of consciousness presents reasons to look more closely at causation, especially causal interactions. One set of arguments linking the two problems stemmed from the worry that consciousness, if it does not logically supervene on the physical, must be epiphenomenal. The claim that consciousness is epiphenomenal, in its broadest construal, implies that consciousness plays no role in the rest of the world's dynamical evolution.

Epiphenomenalist worries concerning consciousness are easy to come by. We have seen that the physical story about the dynamics of the nervous system does not entail the existence and nature of consciousness. Postulating that a system is conscious does not seem to do any explanatory work, and some feel a great temptation simply to eliminate the concept since the phenomenon it represents would apparently be detached in this radical way from physical processing.

Eliminativists are rightfully skeptical about the existence of something as causally irrelevant as phenomenal consciousness seems to be. If they are right, we must be operating under some kind of conceptual confusion. The eliminativist move is unconvincing primarily because our epistemic status with regard to consciousness does seem quite special. We are acquainted with consciousness, and do not postulate its existence to explain other things. Consequently, its failing to explain something, or our capacity to explain something (even our behavioral capacities) without invoking it, does not give adequate reason to eliminate it.

Our evidential situation does not allow us to eliminate the notion of phenomenal consciousness, so we are presented with a problem. The physical basis of things determines their dynamic and structural properties, and we have good reason to believe that an adequate account exists of the world's physical basis. In the absence of any positive evidence for causal gaps, these points seem to count strongly against forms of interactionist dualism. As a corollary, it is natural to assume that the physical explanation of our behavior is the complete causal explanation of our behavior. Here, "complete" has the sense that every entity relevant to a causal explanation either is physical, or derives its causal relevance via its instantiation in a physical basis.

Once we reject interactionist varieties of dualism, the most attractive remaining option is a dual aspect theory. Perhaps we can best understand consciousness and its physical correlates as two aspects of a single underlying kind? Once one moves to the dual aspect view, the predictive adequacy of the physical story raises the specter of epiphenomenalism. After all, the dynamics of behavior are completely derivable from just one aspect of the world, the physical aspect. We naturally infer from this that the physical aspect is doing all the causal work, with the subjective aspect being the notoriously frivolous 'nomological dangler'. This is immediately counterintuitive, as few of us typically believe that our subjective experiences are completely irrelevant to why we behave the way that we do.

Upon reflection, the situation becomes paradoxical. The position seems to entail that we would say everything we say about consciousness even if we were not conscious. This includes all the pronouncements made by people like me to the effect that consciousness is unexplained and nonphysical. Why should a physical system ever accurately talk about its noneffective, nonphysical aspects? Why should the physical processing reliably represent anything about an epiphenomenal, nomological dangler?

10.3 The space of possible responses

These worries are very serious, and they should inspire deep reflection about the premises that lead to them. We can summarize the basic considerations they follow from as follows:

(1) The physical facts alone do not entail the facts about conscious experience.

(2) We can conclude, from (1), that

(2') experience is a nonphysical aspect of the world.

(3) Our physical theories are, in principle, descriptively adequate characterizations of the dynamical evolution of the physical world.

(4) We can conclude, from (3), that

(4') our physical explanations are complete explanations of the causation involved in producing behavior.

(5) We can conclude, from (2') and (4'), that

(5') consciousness lies outside the causal structure of the world, i.e., it is an irrelevant epiphenomenon.

We can place the extant views about consciousness against the background of responses to this argument. None of them are very happy options in my opinion. Each has an unwelcome bite­the­bullet character. Denial of (1) ultimately leads to an equation of consciousness with some objective feature that lies at an unbridgeable conceptual distance from our prior notion. At heart, this is a kind of eliminativism. It is usually accompanied by attempts to show that our historic view of consciousness, the one taken from our everyday existence as conscious subjects, issues from some kind of deep conceptual confusion.

In defense of this position, proponents make appeals to the idea that a self­monitoring robot might say similar things to what we say. We would know that the robot was confused. This response faces the problem that it avoids addressing our most basic evidence about ourselves. Our problem in explaining consciousness was never to give a physical account of what produces our utterances, as our utterances are not the grounds for our belief in consciousness. Consciousness itself is the grounding.

Rather than approaching the issue in good faith, the arguments for confusion in the concept always seem to need to incorporate the truth of physicalism as a premise. Otherwise, the idea that an explanation of consciousness should consist of uncovering a design specification that might reproduce our talk about consciousness would not get off the ground. The problem of consciousness is, prima facie, a different problem than this.

Denial of (2) ultimately leads to a kind of very radical non­reductive physicalism, one that incorporates either brute identities or necessities. The position boils down to one where the existence of some facts can float free of an intelligible basis in the physical, but are, nevertheless, somehow physical. This kind of physicalism typically relies on "identity" as a magic pill that can overcome the structural problems in the explanatory story. I argued earlier that brute identity claims have serious internal problems. We have grounds to doubt strongly that brute identity claims can do the ontological work identity claims are normally supposed to do. Accepting this kind of claim seems to change the meaning of "identity" in a way that robs it of its ontological significance. Appeals to "metaphysical" necessities are no help either, as the sense of necessity needed has never been specified well enough to make such appeals simultaneously effective and meaningful.

Denial of (3) requires either an appeal to interactionist dualism, or an appeal to brute emergence along with downward causation. It faces the problem of being without empirical support. The spirit of the view is also apparently in conflict with everything else we know about how nature works. Considering these problems, a bet on one of these views is a bet on a longshot.

Finally, some accept all the steps along with the conclusion (5'). This move leads to a kind of parallelism view, maybe supplemented with one way causation from the physical to the phenomenal. It suffers from extreme counterintuiveness, and the air of paradox discussed earlier.

The one strategy that philosophers have not explored is a denial of (4), where (4) is the inference from the adequacy of physical theory to conclusions about the causal completeness of physical explanation. That is the strategy the next three chapters will pursue. Making a plausible denial of (4) requires undertaking a detailed naturalistic analysis of just what causation is, and the relation our physical theories bare to causation.

Probably, this is a major reason premise (4) has been ignored. Until very recently many philosophers have been irrealists or reductionists about causation. These philosophers accept a Humean or quasi­Humean view in which causal connections are simply interpretive projections of the human mind upon certain kinds of regularities. Scientists, on the other hand, have considered probing questions about causation to belong to metaphysics. Such questions are, in many ways, of a different kind than lead to fruitful empirical research, research of the type science is most successful at carrying out.

Although philosophy has seen a strong movement away from Humean views in the last decade or so, the alternative explorations of causation have often been ontologically timid. These explorations have focused mainly on certain logical points about the concept of causation, its relation to evidence, modality, probability, and so forth. Some philosophers have been bolder. D. H. Mellor (1995) postulates that objective 'chances' regarding other facts are primitive properties of some facts, guaranteed by the existence of universals with propensities to produce those chances. Mellor then identifies causes with facts that increase the objective chances of their effects, given some background situation. The desire to fully articulate this basic idea motivates several other recent accounts of causation, such as Cartwright (1983) and Tooley (1993). Evan Fales (1990) allows causation to be the instantiation of a universal relation, suggesting that it involves particular spacetime points in the relata as well as properties. Harré and Madden (1975) postulate enduring things with causal powers.

While these contributions are all thoughtful and intriguing, they do not give us enough information about the structure or character of nature to help much with our present concerns. In particular, they say little or nothing about the causal nexus, its structure, or physical theory's relation to causation. Whitehead's Process and Reality (1929) approaches most closely to what we need here. Unfortunately, the obscurity of his language, coupled with the complexity of his concerns, masks the crucial questions that we will focus on.

Among scientists, concerns with causation have involved a mainly ill­defined and primitive notion of information transfer. Few have undertaken the metaphysical project. We must take a step back and ask what the existence of causation implies about the natural and non­natural structure of our world, and whether we fully represent this causal structure in physical theory.

10.4 Problems with Hume's view

The kind of substantial analysis I have in mind will require postulating natural ontology on philosophical grounds, and will still broach metaphysics at many points. For many people, the idea that understanding consciousness or causation should carry such a heavy price will be unwelcome news. As a rule, such global speculation and reorganization are the kinds of activities we can reasonably engage in only as a response to equally global, categorical failures of explanation. The anti­physicalist arguments defended earlier, and the attendant set of paradoxes accompanying the anti­physicalist conclusion, already establish this kind of categorical failure regarding consciousness. Before also engaging in this kind of speculation regarding causation, I wish to make the causal gaps in physical theories easier to see. In doing this, I hope to make the scope of the theoretical problems more urgently felt.

The kind of view of causation I want to consider is best contrasted with the regularity view of causation put forward by David Hume. The reason Hume makes a good foil is that his view leaves out exactly the element I most want to explore, the causal connections between distinct individuals and events.

Hume's view requires the evolution of the universe to be unconstrained. We are left with an unusual view of the regularity of nature, and of the orderliness of the universe's dynamics. Humeans hold that our causal stories are interpretive projections of the mind, a kind of psychological habit. In the world no connections of dependency, constraint or production hold between individuals or events. This story has appealed to the anti­metaphysical preferences of empiricists for a long time because the kinds of "connections" that we will consider have seemed too obscure and "metaphysical."

After a long period of prominence, Hume's regularity view has begun to fall into disrepute, even among empiricists. Defenders of the view have never found a truly satisfactory account of what distinguishes certain regularities as the causal regularities. Also, several insightful critiques have emerged. These critiques argue that the view has many other substantial shortcomings.

Many of these critiques center on issues of explanation and induction. For instance, Armstrong (1983) raises a host of serious problems for the regularity view of laws. Although the problem of accounting for natural laws is different from the problem of causation, problems that the regularity view faces in accounting for laws cast long shadows on regularity accounts of causation. First, because motivating a regularity view about one but not the other would be difficult. Secondly, because we naturally expect that at least some instantiations of laws will be causal. Those causal facts would not be mere regularities.

On the topic of explanation, Armstrong points out that, if the regularity view is correct, laws cannot be explanatory of the regularities they describe. Those regularities constitute the law, so citing the law in an explanation presupposes just what we are supposed to explain. Therefore, the regularity view cannot account for the explanatory role laws play in our practices.

On the issue of confirmation, Armstrong points out that the regularity view is susceptible to paradox. Since the basic regularity view is expressed using material implications of the form (x) (F(x) G(x)), it is confirmed by instances in which F(x) is false and G(x) is true. For example, if it is a law that rising prices decrease demand, this law is confirmed by instances in which prices do not rise and demand falls anyway (perhaps due to changes in quality, or demographics, or simply random chance).

Armstrong also points to ontological problems with the regularity view. Neither probabilistic laws nor laws expressed by functions of varying magnitudes logically supervene on the regularities they are supposed to describe. The problem is that any sequence of regularities will be logically compatible with more than one probability distribution, or more than one function. These ontological problems require bringing in factors involving human standards of explanation. Thus, the Humean view must move very firmly away from being a reduction of laws to regularities, and towards anti­realism or irrealism about causation.

Others have challenged the basic Humean point that causation cannot involve necessity since we can coherently imagine any arbitrary relations of cause and effect holding between things and events. Harré and Madden point out that many referring descriptions incorporate a notion of causal production into their meaning. As an example, Harré and Madden claim that Joe could not be John's father unless he played a certain causal role in producing John. This role is not merely having a place in a sequence. Although no contradiction is involved in imagining that Joe might not have had this causal role in producing John, nevertheless it is contradictory to imagine that someone could be John's father if he did not causally produce John. The conceptual scheme that gives meaning to 'father' incorporates the idea as a truth­condition.

They also argue that we can properly apply certain natural kind terms only to things that have specified causal natures. Discovery of a substance that seemed to be copper by many tests, but that still produced some different effects, would present us with important conceptual challenges. In the end, one of two outcomes would have to occur. Either we would decide it was not copper after all, or we would have to engage in an extensive revision to our conceptual scheme. The revision would have to allow us to account for the unusual effects.

Both possible results seem to reflect a conceptual necessity meant to mirror a necessity in the nature of the thing. We self­consciously involve the pattern of effects copper produces as part of the meaning of the term 'copper'. The discovery of antiparticles is an example of the kind of classification response at issue. A positron, which is the antiparticle of the electron, is indiscernible from the electron except that it has certain effects and responses that only a positively charged particle should have. Within the conceptual scheme of modern quantum theory that is enough to disqualify it from being an electron, and the response was to create a new class of particle.

This strongly suggests that the tacit understanding of what it is to be an electron is to produce the appropriate kinds of effects and responses. Someone who asserts that a particle that behaves differently from an electron is, nevertheless, an electron, is either being incoherent, or expressing an aberrant meaning with the term. Delimiting exactly what kinds of deviation force reclassification is a complicated issue, but the point remains that it is a partly conceptual issue about how we should rationally apply the concept. The conceptual issue centers on questions about what causal powers should enter into the meanings of the terms.

Harré and Madden urge that these observations demonstrate the unsoundness of Hume's argument that, in principle, we can coherently imagine anything having any effect whatever. The possibility that we can coherently, arbitrarily vary the effects of a thing is description relative at best. What we can do is associate arbitrary effects with surface appearances, or imagine one thing suddenly replaced by another with an entirely different nature. In the general case, our conceptual systems often do delimit the applicability of many terms by postulating causal natures and roles. We represent these natures and roles by a conceptual necessity that is not merely formal, and that renders certain kinds of situations incoherent.

I think all these criticisms, along with many others, present serious problems for the Humean view. However, many regularity theorists have suggested more sophisticated formulations of the theory in attempts to meet these kinds of criticisms. The development of the regularity view has paralleled Ptolemy's astronomy, adding epicycle after epicycle to a poorly designed theory.

We could short circuit this pattern by producing criticisms that transcended the details of particular Humean or quasi­Humean views. What we need are criticisms that will apply to any theory incorporating the basic Humean premise. We need criticisms that will apply to any theory that denies the existence of connections of constraint between what exists or occurs in the world.

10.5 Foundational problems with Humean views

Instead of becoming embroiled in a discussion of the epicycles that Humean theorists have produced, I am going to leave in depth discussion of them aside. I want to consider less well­discussed problems that I believe are broader and deeper. To create these problems, one needs only the bare Humean assertion that events do not constrain one another, and so they seem to apply equally well to any possible formulations of the view. The first of these problems is metaphysical, and the second is epistemological.

The metaphysical problem: The unity of the world ­Taking the metaphysical problem first, the Humean view faces the prima facie problem that it seems to rupture the unity of the world. Here is how the problem arises. The regularity theorist must postulate atomic events. Each atomic event is itself a complete entity insulated from influence or constraint by the occurrence of any other events. Because this first step splinters the world into a multitude of separate and internally complete shards, the regularity theorist is faced with the problem of putting the world back together.

One might think that the whole collection of insulated events obviously constitutes a single world, but this is not at all obvious. The Humean cannot rest with a logical conjunction of events, letting their world be hauntingly reminiscent of the world in Wittgenstein's Tractatus. The problem is that it seems possible for a collection of independent events to fail to be a world, or at least to fail to be the kind of world that can serve the regularity theory.

Consider a collection of causally separated dimensions, something like a set of parallel universes, like in a science fiction novel. We can coherently conceive of each separated world as possessing its own, internal time dimension. In this kind of multiverse, each world's time dimension would sequence the events within it. Nevertheless, there does not need to be an overarching, trans­world time that would sequence events across worlds. Thus there would be no answer to questions about whether event X in world A occurred before or after event Y in world B.

If complex worlds like that are possible, simpler worlds should also be possible. In the limiting case, it seems possible that each world could contain only a single event. Each parallel world has an internal time that gives duration to the single event that constitutes it. A Humean 'world', with its insulated events, could very well be such a multiverse of small worlds: each event instantiates an internal time dimension that gives it duration, but there does not need to be a common temporal framework within which they all exist. Nothing would order them relative to one another, so they need not form one world rather than many separated, single­event worlds.

The problem for the regularity view is that it holds that causal facts correspond to temporal regularities within a world. If the events fail to belong to the same world, or the appropriate temporal relations fail to hold between events in a world, no causal relations can hold. The usual closure condition on worldhood is causal closure. On pain of circularity, the regularity theorist cannot appeal to this condition. The only promising alternative is the suggestion that the insulated events are in the same world if they are instantiated in the same spacetime structure.

The suggestion is tempting. For it to work, the regularity theorist must give some account of the facts about space and time that is independent of the facts about causation. They face two large hurdles in doing this. The first hurdle is that they cannot construct a single world if slices of space are as independent as the events they house. They must argue that slices of space are themselves ordered through time.

Here, the regularity theorist may appeal to the view that spacetime is a single, four­dimensional structure with a primitive, sui generis kind of unity. Space and time together form a kind of seamless sculpture that houses events. This four­dimensional spacetime view may work as an identity condition on a world, but it is a strange kind of entity for the regularity theorist to take as basic. After all, a desire to avoid postulating strange, metaphysical 'connections' between events motivates the regularity theorist. Here, however, the transpatial, transtemporal connections that give spacetime its fundamental unity are at least as mysterious as the causal connections the regularity theorist is trying to avoid. To that extent, the regularity theorist has already paid a large cost.

It also leads right to the second hurdle for the regularity theorist. The regularity theorist needs to account for the direction of causal facts. To account for causal facts, the regularity theorist needs a special kind of world, one that has facts about events preceding other events. Unfortunately for the regularity theorist, the temporal dimension of the four­dimensional structure has no direction that exists independently of a specification of causality!

One can see this most clearly by considering the light cone associated with an event in relativity. Think of each event as occurring at a point. This point provides a frame of reference through which we can view time in the rest of the universe. Exactly what temporal facts hold are relative to each frame of reference, and they may differ in different frames of reference. Associated with each frame of reference is a light cone that delimits the temporal facts for it. We can picture the light cone as a cylinder squeezed in the middle to a point, that midpoint being the frame of reference itself. The two conic halves represent the past and future.

Which half is the past, and which half is the future? The answer in the standard theory is that the future cone is the one in which the event in the frame of reference may have causal effects; the past cone is the one that contains events that may causally effect the event in the frame of reference. Apart from this specification of how causal influence propagates, nothing in the theory itself determines which half is the past and which is the future.

Regularity theorists obviously cannot accept this account. To make their view succeed, they must add a direction of time as a primitive feature of the model. As such, it shows up as logically independent of everything else in the model. Furthermore, its inclusion as a primitive is for ad hoc reasons. The only motivation for taking it to be logically prior to causality is to save the regularity view.

The final view is very ugly. These insulated atomic events all exist within one world due to their coinstantiation within a four­dimensional spacetime structure. This structure has one dimension, and only one, with an inherent direction. This inherent difference between the time dimension and the other dimensions is completely unexplained by anything inside or outside of the theory.

We should now take inventory of how far we have come. Regularity theorists start by wanting to avoid making unintelligible metaphysical postulates. To do that, they need to give a criterion for world identity and temporal precedence that is independent of causation. At the first step, then, the view becomes embroiled in metaphysics. The only plausible way to meet the challenge is to reify spacetime, postulating the existence of a four dimensional spacetime structure to house events. This structure must be basic, for if it were reducible to the causal relations between the events it could not serve its Humean purpose. Its basicness requires it to have a kind of unity and internal connectedness that seems every bit as unintelligible as the causal connections the Humean was originally trying to avoid.

Furthermore, the contours of this structure seem to have no inherent direction. Therefore, the structure itself does not support facts of succession or precedence that can ground causal facts. In response, regularity theorists must simply assert that one dimension of this structure, and only one, does have an inherent direction. This fact is an extraneous adhesion to the model. Why does the temporal dimension have this property when the others do not? The regularity theorist has no answer. The lack of explanation means that the structure itself lacks any difference to ground this very peculiar and important difference between the dimensions. At best, the view starkly violates the principle of sufficient reason, and so the postulated difference appears as magic. It may be that the total view is incoherent, as that kind of difference cannot exist as basic. In either case, we are now very far from Hume's original spirit. Appendix A discusses a popular class of response to this worry.

The epistemic problem: Solipsism of the present moment ­ The regularity theorists seem cornered into postulating a seamless spacetime structure, and an unexplainable direction to one of its dimensions. In doing so, they are saddled with metaphysical claims at least as problematic as those they were trying to avoid in the first place. That is the metaphysical problem that confronts all Humean views. The quandary might tempt regularity theorists to resist the charge that their temporal postulate is ad hoc. They might say we know from experience that our world has direction in the time dimension.

Unfortunately, that move leads them directly into the epistemic predicament. While the metaphysical problem is serious, the epistemic situation is devastating. The regularity view simply cannot be shorn free and saved from global skeptical consequences. It destroys empirical knowledge and reduces to solipsistic skepticism.

The epistemic problems exist because the Humean view reduces to something like a version of Leibniz' monad view. Leibniz' monads were absolutely unitary beings whose evolutions were completely independent and insulated from influence by other monads. Leibniz' coined the term 'windowless monad' to refer to this independence, the image being that every monad was absolutely closed to information about things other than itself. If the regularity view is correct, the basic individuals are separated events, each occurring independently of any other event. The insulated events play the part of monads.

What happens to our conscious minds in this picture, with their collection of occurrent mental events? Applied to our minds, the Humean position implies that the mental events occurring in a subject are unconstrained by anything else in spacetime. This includes events such as perceptual events, occurrent thoughts, occurrent memories, and occurrent beliefs. It seems that we must treat our conscious minds as monads, or monad complexes.

Now consider the situation from the perspective of a given subject of these experiences. At a given moment, what gives the occurrent perceptions, and other mental events, force enough to justify a belief that anything systematically relates their occurrence to any other events that might be occurring outside the region of spacetime they occupy (if there are any such events)? If the Humean view of causation is correct, any regularities that might exist among events seem metaphysically inconsequential, and therefore epistemically inconsequential. These regularities are irrelevant to what occurs within any given mind on any given occasion. We can sum the grounds for skepticism by saying that mental events in the Humean world will not carry any information about the rest of the world.

Notice that none of the events within a mind would be responses to what happens in the rest of the world. Even to give the appearance that the Humean world contains information states or events, the regularity theorist must give truth­conditions for the counterfactuals supported by information bearing states and events. The problem is that the kind of counterfactual force that epistemically useful ideas of information require will be missing.

The regularity theorist cannot appeal to powers, causal natures, or natural constraints. As an example of the kinds of alternatives they have available, we will consider the appeal to a semantics of possible worlds. Within this framework, they must immediately confront the problem of cross­world identity. What do we mean when we say that this event would not have occurred had those events not occurred? This leads us directly to David Lewis' preferred device of a similarity metric on possible worlds. The statement will be true just in case there is a world where no events similar to the ones we are concerned with occurred, and where this world is closer to actuality than any world where the mental events occurred, but the extramental events did not occur.

This relationship between events in the actual world and those in other possible worlds imparts too weak a content to 'information' for it to ground knowledge claims. It is no advance on the original conundrum. Our original problem was how the occurrence of a mental event in a Humean world could give the subject reason to believe something about events elsewhere in that world. An appeal to similarity relations between the subject'sworld and other possible worlds leaves that problem untouched. When all is said and done, we still need some good reason for believing that our world looks like such­and­such, and therefore is similar to possible world A rather than possible world B. And that requires, first, that we have knowledge about our world. Since the similarity relations between worlds are themselves contingent on how our world looks, access to them presupposes access to facts about our world. The problem of how we gain access to facts about our world remains. This problem looks hopeless as long as we must suppose that the mental events occurring within our minds are unconstrained by anything else occurring in spacetime.

The epistemic and metaphysical problems tie into one another. We started the discussion of the epistemic problem because our imagined regularity theorists had proposed that experience reveals a primitive direction on time. Thus, they deny the charge that postulating an inherent and unexplained direction to the temporal dimension of spacetime is ad hoc. The epistemic problem shows that we do not have information about the direction of events in our world. Specifically, if the direction on time is primitive and ungrounded, then we should be able to postulate coherently that it varies between different regions within the structure. How do we know we do not live in a world where time is directed this way at some regions, and that way at others? Actually, the problem is worse. The skepticism brought on by the problem affects memory also. We would have no more reason to trust the deliverances of an occurrent memory than an occurrent perception. Subjectivity in the Humean world reduces to solipsism of the present moment. In a Humean world we would just have no idea what was going on in spacetime.

10.6 Physical theory and Humean views

I have been building a case for the pressing importance of gaining a fuller understanding of causation. We have seen that a better understanding of causal interaction promises to throw light on the boundary problem for phenomenal individuals. As Liberal Naturalists, we have seen that we need a better understanding of causation to avoid the dilemma of epiphenomenalism and interactionism. We have seen that a regularity view of causation leads to a highly contrived, and perhaps incoherent, metaphysical view of the unity of the world. We have seen that the Humean view of causation entails that we are in a fatally absurd epistemic situation. The problems with the Humean view provide motivation to explore the nature of the causal connections whose existence Hume denied, trying to make something knowable out of what he claimed is unknowable.

On the road to understanding causation, the best place to start is with the aspect of causation that we best understand. In this section I will argue that a realist interpretation of our physical theories is compatible with a Humean interpretation of the nature of our world. The easy availability of this Humean interpretation helps to expose the danger in assuming that our theories are telling us the whole story about causation. Once we realize that physical theories might be aspectual, we will be in a better position to find what is missing.

Currently, the idea of coevolving fields forms the centerpiece of our most basic physical theories. These fields periodically contract, for reasons still unknown, to something like classical, localized particles. They then begin again to spread in spacetime. The basic dynamical laws tell us two things. First, they tell us how any given field will evolve given its state at some time in the past. Second, they tell us how the evolutions of different fields become correlated.

The current theory does have a gap in its dynamic story as it cannot precisely specify the evolution of a field. It must appeal to the ill­defined notion of a measurement to specify when the contractions of the fields occur. This gap in the theory should not matter to the discussion that follows as it is leading to a more general point about the character of physical theories. To make this general point, I do not need to become embroiled in specific issues about the interpretation of quantum mechanics, the measurement problem, or the nature of the superposition of states that constitute the evolution of these fields.

The essence of my argument does not concern these specifics. Instead, I am focusing on the general forms used in physical explanation to make a point about their representational character. What I will claim is missing in the simplified caricature of quantum mechanics I am using as a prop is also missing from more complicated expressions of the theory. Some good reasons exist for believing they will be missing from any purely physical theory about the fundamental furniture of the universe.

The evolution of a field is represented by a dynamical equation. These equations plot states of the system against points in time. Given an initial state, the mathematical rules they express describe a temporal trajectory through the field's space of possible states. The relevant feature of such dynamical equations is that their successful use only requires us to assume regularity in the succession of states. They merely associate, or correlate, field states with points in time. Association is a very weak relation.

Specifically, the theoretical apparatus is neutral with regard to how these regularities arise. Nowhere does it mention or need the idea of causal production or dependency between states of the system at different times. The only explicit associations in the function are between states of the system and points in time. The associations between properties of the system are implicit in the function, and are expressed explicitly in the form of the dynamical equations describing the function. It is these explicit and implicit associations that contain the 'causal' content of the theory.

These equations are rules that describe the functions. Each function is lawlike in that it contains physical magnitudes that vary in regular ways across the associative structure. The dynamical rules refer to these specific physical magnitudes, using them in descriptions that compact the temporal regularities in each function's associative structure. The equations will be true even if each state of the system is an independently determined event, and these events coincidentally follow the trajectories given by the equation. For these reasons, a given dynamical equation can be true of a Humean, acausal world. To paraphrase Laplace, we have no need of the hypothesis that one state of the system might causally depend on or be connected to another. If we choose to interpret it causally anyway, this interpretation is adducing something into the theory not represented or required by it.

The second component of our theories describes how these fields 'interact.' I put 'interact' in scare­quotes because this part of the theory is also compatible with a Humean view of nature. What the laws governing interaction express are correlations between the evolutions of different fields. Correlation, like association, is a weak relation and compatible with the absence of any real connection between the fields. It is true that physical forces are supposed to mediate these interactions, but virtual particles carry these forces. We can always interpret virtual particles, as such, as further field elements entering the correlation story. We can, when push comes to shove, explain away the talk of 'information' carried by them as a useful fiction.

In the end, all the theory forces on us is recognition of the highly regular correlation between the evolutions of different fields. Connections of interaction and exchange of "information" is, like talk of connections of causal dependency, adduced into the theory. We do this because the world the theory tells us about would be impossible to believe in without such connections, and not because the theoretical apparatus we are using to describe nature compels us to think that way. Particularly, the theory does not represent these causal connections. If we choose to interpret physical theory in a non­Humean way, we must take it as assuming causal connections implicitly while explicitly describing some aspect of their outcomes. As we proceed in this chapter and the next, I will try to make it clearer why this is.

Admittedly, I have not said anything about the hypothesized quantum collapse of the wave function, or alternatives to quantum mechanics like hidden variable theories. None of these things make a difference to the general point. This point rests on an observation about what our physical theories actually require from us to deliver their results. To do their job, they only require us to possess certain, minimal information. The needed information must be about the values of physical properties involved in some other events that have occurred elsewhere in spacetime. Our theories do not make the further demand that we wonder why or how occurrences elsewhere in spacetime matter in a given case.

One can think of this theoretical apparatus as a kind of probabilistic roadmap. It helps us navigate the four­dimensional surface of spacetime using landmarks to help fix our expectations. To be a successful map, it needs to make only modest demands on nature. It does not require anything more of nature beyond the regularity of relations between the landmarks.

That metaphor tells us how we can be both realists and Humeans about physical theory. Corresponding to every physical property in the theory, we postulate something in nature. We can think of 'mass', 'charge', 'spin' and so forth as each denoting a universal that is instantiated in the appropriate magnitude at the appropriate places in spacetime. These universals only need to be distinct and capable of the specified quantitative variations. They act as the landmarks on our maps. That makes us realists.

The theory can be true, and true in a realist sense, even if we do not postulate further things such as connections between the landmarks. The landmarks simply have to vary in the regular ways that the theory describes. Spacetime must have the appropriate layout. We do not need to suppose that some landmarks produce others, or constrain the production of others. Since we do not have to postulate these things to make the theory true, and true in a realist sense, we will not postulate them. That makes us Humeans.

Given the form of representation in physical theory, we should not be surprised that a Humean interpretation comes so easily. The closest thing we have to a physical theory with built in connections is the pilot wave in Bohm's hidden variables theory. Yet, under the usual interpretation of this theory we can realistically interpret the whole system as describing mere regular fluctuation. Usually, we interpret the pilot wave as pushing around its particle. This interpretation is not forced on us though; a diehard Humean can be happy to note that all the theory describes are very complicated, regular correlations.

We can easily see how to be realists and Humeans about physical theories, but that does not mean we have to be Humeans about them. The most common and compelling interpretations are causal. I am suggesting that the ease and directness with which we can construct a Humean interpretation should serve as a warning. It may be that we cannot make the move to a fully causal interpretation for free. We have to be careful about smuggling background assumptions into our world view. To make sense of unnoticed background assumptions, we may require ontology that physical theory does not explicitly represent. Perhaps we will have to take physical theories to be explicitly representing some aspect (or aspects) of causation, while allowing others to live implicitly in the background. If we interpret physical theories causally, we must check to make sure that they really are complete specifications of the causality.

10.7 Beyond Hume

The Humean view falters on many fronts. It fails at explanation; it fails on confirmation; it fails at avoiding metaphysics; and it falls prey to skepticism of the worst kind. The arguments against Hume show that causation, like consciousness, presents severe explanatory difficulties. It is not enough simply to say that events occur, or that some events constrain what else may occur in spacetime. We need a general and enlightening view of what kinds of properties a world with those kinds of facts, facts about connection and constraint, would require to exist. What kinds of properties and structures must a world with real causation possess? What does a world with a network of causal nexi require for its existence?


Chapter 11

Receptive Connections

11.1 Causal responsibility and causal significance When approaching the topic of causation, many philosophers tacitly assume that the first choice point is between a Humean(1) approach or some form of non­Humean approach. Under the Humean branch there seems to be a further choice between extensional and counterfactual accounts, and under the non­Humean branch there seems to be a choice between singularist and law governed views of causation. The resulting decision tree looks like that in figure 11.1.

The root node of the tree is labeled causal responsibility, and the intended notion is that of a productive cause. We should interpret the branches as choice points along the way to developing a theory of causal responsibility. The search for a theory of causal responsibility is the search for general conditions under which a specified something ­ event, agent, fact or process ­ can be credited with being the cause, partial or total, of some specified event(s), its effect(s). Many assume that these kinds of facts, fundamentally, are what a theory of causation must explain. The approach I will develop below differs from the standard approaches at this very first choice point. It eschews even the root node by rejecting the notion of causal responsibility as the fundamental explanandum for a theory of causation.

The reasons for my divergence from the norm rest in suspicions about the objectivity of the notion, and caution about incorporating some of the assumptions it involves into the start of the investigation. Voicing concerns about its objectivity first, several features of our ordinary notion of causal responsibility raise suspicions that it has both strong intentional and interest relative components. First, negative facts show up routinely as causes in both ordinary and scientific explanation. For example, when an animal starves to death we judge that the cause of death was the lack of food. Along the same lines, we often say that a person's disappointment in himself or another was caused by some failure, where failures are understood as things that were not achieved. The prominence of negative facts in stories appealing to causal responsibility is strong evidence that it is an intentional, rather than a natural, relation.

Second, judgments about causal responsibility seem to essentially involve a kind of conceptual analog to perceptual figure/ground relations. Imagine a typical morning when Trey goes to work. Before getting on the road he puts the car key in its slot, turns it, and starts the engine. Our common notion of causal responsibility will credit Trey's turning of the key as being the cause of the engine's starting. Notice that the counterfactuals involved underdetermine this kind of judgement. It is true that the starting of the car would not have occurred had the key not been turned, but this same counterfactual holds of many other facts: had the alarm not gone off, Trey would still be sleeping and thus the starting of the car would not have occurred; had the spark plugs not fired, the car would not have started; had the earth stopped turning, the car would not have started; and so forth. The counterfactual seems to be an important condition, but the truth of such counterfactuals is not sufficient to yield facts about causal responsibility. Trouble arises in the move to sufficiency, as it seems to bring in interest relative factors, and other idiosyncracies in human judgment (such as how we might judge the similarity relations between two possible worlds). For these reasons, facts about causal responsibility are unlikely to be like facts about rocks, things that we simply trip over while investigating the world's objective causal structure. This point has been especially emphasized by R.C. Collingwood (1940). More recently, D.H. Mellor (1995) has emphasized the tight relation between the notion of cause, and being a means to an end. Most likely facts about causal responsibility arise from a mixed notion, one that contains an objective core on which the more intentional and interest relative concept rests.

In addition to these intentional and interest relative features, the concept of causal responsibility comes loaded with default assumptions about the character of causal relations. Among these are the ideas that causal relations exist only forward in time, only locally in space, and are dyadic. These assumptions may (or may not) be correct, but they strike me as ornaments on a more fundamental notion, and I think they serve to distract attention from it. This more fundamental notion is one that I shall call causal significance.

The causal significance of an event is the constraint its occurrence adds to the space of possible ways the world could be. If a non­Humean view of causation is correct, then the occurrence of an event has significance beyond itself, a significance that ripples widely through the causal mesh. The rest of the world must, in some sense, be compatible with its occurrence (and therefore consistent with the existence of the facts its occurrence instantiates). A non­Humean theory of causation will give an account of what causal significance, in this sense, is. A successful theory of causal significance should lay bare an objective base of facts upon which less objective facts about causal responsibility might rest.

How should we think about causal significance? It seems promising to think of causal significance as resting in the nomic content possessed by the world's individuals. Nomic content itself can be understood as the set of causally relevant properties an individual possesses. So, I will attempt to tackle this task by trying to understand the metaphysics of nomic content.(2)

11.2 Effective properties

To avoid the problems with Humean causation, we must interpret physical theory in a non­Humean way. The interpretive question before us is this: what place do the physical properties have in the causal story about a non­Humean world? To answer this question I will develop a theory of the causal nexus, focusing on the most general kinds of properties a non­Humean world would possess. As I do this, I will ask what physical theory has to say about each aspect of the nexus.

Here is how I have structured the discussion that follows. In this chapter we will introduce and discuss an aspect of the nexus I call receptivity, which I will model as a connection between the effective states of individuals. An individual's effective state is a function of one or more effective properties the individual possesses, where I will understand effective properties as categorical properties that present constraints on the possible states of individuals that are bound within a causal nexus. In the next chapter I will illustrate the ideas from this chapter through a series of simple models. In the chapter after that, I will argue that the existence of receptive connections is only implicit in causal interpretations of physical theory. Finally, I will complete the analysis by introducing the idea of a carrier, and by developing The Carrier Theory of Causation in chapter fourteen.

Let me start the inquiry by reflecting on the process through which we create physical theories. We are creatures embedded fully within the natural world, and physical theories are our attempts to understand something about our place in the causal order of that world. The character of the information we gather, and then hold physical theories accountable for, becomes more apparent when we self­consciously consider the position we occupy.

Nature places human beings, like all organisms, within an effective loop. We must understand how the world may affect us, and the effects we may have on it. Fortunately, perception provides information to help with this challenge. Through perception we become systematically sensitive to enviromental influence, and can treat some of its effects on us as providing information. Quite literally, perception selectively processes effects that the environment may have on us, using them for the informational fuel that we burn and store in forming our interpretations.

We have strongly tuned the methodology of physics to the effective nature of the world. The genius of the experimentalist is in solving the following challenge: assuming the entities we postulate are present, how can we isolate them and identify their states? The basic measuring devices they begin with are those of our biological endowment: eyes, ears, hands, tongue, and touch. The experimentalist must find ways for perceptible and non­perceptible entities to make a distinctive difference to us via our biological endowment.

For non­perceptible entities, the experimental physicist must first find something else that the ultimate object of investigation can affect; then, the experimental setup must magnify this effective difference through a chain. Near the end of this chain is something ­ perhaps a pointer, a colored flame, or a computer printout ­ that can affect our senses without the further aid of special instruments. At this last step, the effective properties of our instruments act on our biological endowment, completing the chain. In short, when we measure, we find effects of the hypothesized entities that we can magnify to a level of reality that we can perceive directly. The character of the entire process forces the effective properties of things into our theoretical fold. It is always a chain of effects, from hypothesized entities to us, whose explanation we require.

These observations should not be controversial. Basic physical theory attempts to specify all the basic effective properties of things. The fundamental physics of our universe will be the science that at least discloses to us the effective properties of the fundamental individuals(3) of our universe, assuming such individuals exist. If this is true, however, it does not automatically become clear that physics will yield a complete account of the world's causal structure.

The reason for doubt is that effective properties require the existence of other kinds of properties. The two questions that will be the focus of the rest of this chapter, as well as the next three chapters, are (1) What other aspects of causation exist?, and (2) Are these other aspects physical? In the end, we shall see that causation has two further aspects, and neither is plausibly physical.

11.3 The idea of receptivity

The effective properties of an individual cannot be effective unless some individual can be receptive to their instantiation. This seems to be a conceptual truth: a property of an individual may be effective only if some individual is receptive to the property's presence. The two notions, effectiveness and receptivity, are logically intertwined. Since they are complementary aspects of causation, the world cannot have one without the other. Thinkers in the history of philosophy have often recognized this duality, usually only briefly and obliquely. For example, in Plato's Sophist the character of the Stranger speaks for the materialists of antiquity, saying,

I suggest that anything has real being that is so constituted as to possess any sort of power either to affect anything else or to be affected, in however small a degree, by the most insignificant agent, though it be only once. (247e, Hamilton and Cairns, 1961)

Receptivity is something like this "power to be affected" Plato briefly points to, as does Locke in chapter XXI of An Essay Concerning Human Understanding,

Power thus considered is two­fold, viz. as able to make, or able to receive, any change. The one may be called active, and the other passive power. Whether matter be not wholly destitute of active power, as its author, God, is truly above all passive power; and whether the intermediate state of created spirits be not that alone which is capable of both active and passive power, may be worth consideration.

This old distinction between active and passive power has fallen to the periphery of modern thought. Likely, part of the reason is the previously discussed empiricist deflation of causation begun by Hume. Another part of the reason is the unfortunately oxymoronic name, "passive power." Despite the empiricist neglect, the idea remains an important part of process philosophy, where process philosophers recognize the logical need for something that does its work (e.g., Griffin 1996).

At times, the conceptual distinctness of receptivity and effectiveness has led us to postulate special kinds of individuals possessing only one of these aspects. For instance, the medieval conception of God as a purely active force (mentioned above by Locke), or unmoved mover, is an isolation of effective properties within a non­receptive individual. On the other hand, dualist proposals about consciousness are sometimes epiphenomenal. They postulate that phenomenal consciousness is determined by the physical properties of the brain, but is nevertheless causally inert. This is the postulation of an individual with properties that are receptive but not effective.

One can intuitively triangulate in on the distinction by considering each case, and then identifying the complementary kind of property as what is missing in that case. What would an unmoved mover be missing so that it, alone among all beings, would be unresponsive? Equivalently, what is it that other beings have that it does not? Answer: it is missing a receptive aspect. What would an epiphenomenal consciousness be missing that would make it, alone among all beings, epiphenomenal? Equivalently, what is it that other beings have that it does not? Answer: it is missing an effective aspect.

Because of obvious problems in gaining knowledge about the presence of a purely receptive being, we would not expect any established science to have accepted the existence of one (modulo, controversially, consciousness itself). Has science ever found it intelligible to propose purely effective beings analogous to unmoved movers? Surprisingly, at least one example exists and, maybe, another. The clearest example of a purely effective entity is Newtonian space. Its Euclidean geometry constrained the movement of objects within it, although it was entirely unresponsive to its occupants. From the perspective we are now discussing, the causal difference between Newtonian space and Einstenian space is twofold. First, the introduction of a different geometry represents a change in its effective nature. Second, Einstein introduced a responsiveness to the distribution of mass within it. This second change is an entirely different kind of addition, ontologically, and the more revolutionary. Einstein added receptivity to space.

Ironically, although Einstein robbed Newtonian mechanics of its only unmoved mover, he may have introduced another kind of his own: singularities. As entities with infinite density, singularities seem to have great effect on the rest of the universe. For instance, they create black holes. On the other hand, it is not clear that anything can, even in principle, affect them in return. Singularities may lack receptivity.

Collectively, these examples show the conceptual and empirical distinctness of effectiveness and receptivity. This distinctness marks an important point: they are not identical aspects of causation. These two aspects of the causal process do different jobs, and they need distinct accounts. A proper account will detail how each aspect helps to ground the very possibility of causal activity. Importantly, the conceptual interdependence between these two aspects of causation, the effective and receptive, has a circular structure. Each aspect presupposes the other's existence in a way that will become clearer as I step through the examples in the next chapter. They are thus interdependent and equally fundamental aspects of the causal nexus, despite being distinct.

On a terminological note, maybe the best way to express the idea is this. Every causal world posseses individuals with these two aspects, an effective and a receptive. The two aspects are irreducibly distinct although logically intertwined. When speaking of elements of these aspects in an isolated way, I will call them properties. When speaking of them against the background of the individual or world to whom they belong, I will call them aspects. This is how I will try to use the terms.

11.4 Receptivity as a connection

The model of receptivity that I will develop may conflict in certain ways with the intuitive image that leaps to mind for some people. In particular, I am treating the receptive aspect of an individual as a connection between a collection of effective states rather than as a monadic property of an individual. Figure 11.2 visually contrasts the two alternative pictures.

I have several reasons for preferring to model receptivity as a connection, eschewing the monadic alternative. One reason is that, if one postulates a picture like the monadic view of receptivity depicted at the top of figure 11.2, it seems that one is still left with the problem of "activating" an individual's receptivity relative to the effective states of other individuals. That is, it is not enough that an individual be receptive simpliciter; it must be receptive to the effective state of some other individual(s). To complete the account, we would have to specify some conditions for selectively determining which individuals a given individual will be receptive to. This further condition, whatever it might be, is a complication to the model that does not arise if one models receptivity as a connection to begin with.

Aside from the inelegance this extra step introduces, it may also limit the account in unnecessary and unfortunate ways. For instance, consider what happens if one introduces spatial or temporal contiguity as the further condition. Such a move rules out non­local causal connection by definition, which seems undesirable. It also brings spacetime into the picture in a fundamental role, precluding the otherwise attractive possibility of reducing it to more fundamental facts about causal connection.

A second reason for preferring the connection view is that, as we shall see, it allows a much more elegant modeling of levels of nature, and of the emergence of higher­level individuals that incorporate lower­level individuals. What the connection view offers is an opportunity to specify the conditions of a substantial internal unity that may ground a notion of natural individuals, and natural individuation.

A third reason for preferring the connection view is that it produces a causal mesh with a topological structure of its own. This topology provides hope for grounding a reduction of the facts about space and time themselves, increasing the explanatory power of the theory.

The idea of the receptivity of individuals is tantalizing, yet can still feel very elusive. This is especially true when first considering the distinction. To help gain a clearer understanding, I want to step through some models of idealized causal situations. I have individually designed each toy situation (presented in the next chapter) to highlight salient features of receptive connections. Their role in causation should become clearer as I step through the examples in the next chapter. Because this work might seem tedious without proper motivation, I want to give a short preview of where the discussion is heading. I hope that the significance of what follows will be more clearly visible from the vantage point of the whole.

To this point I have argued that (i) we should interpret physical theory in a non­Humean way; (ii) that the ideal physics will include all the effective properties of our world's fundamental individuals; and (iii) for the conceptual and empirical distinctness of effective and receptive aspects of causation. For now I will assume something that I will argue for later, that (iv) physics exhibits only the chain of regularity between instantiations of the effective properties. If (i) ­ (iv) are true, it follows that causation in our world has at least two equally fundamental aspects, and that one of them is non­physical. It also follows that physical theory has not explicitly represented the non­physical aspect. Receptivity is an explanatory luxury for physical science, but is nevertheless metaphysically relevant to the causal structure and evolution of the world. Finally, we should note that these two aspects share interdependent categorical natures. Even after giving an account of them, we must discharge these circular remnants. That task is taken up in chapter fourteen.

The overall ontological picture is very interesting. Receptive properties are necessarily related to the physical in that something's being an effective property presupposes something's being receptive (and vice versa). We have a necessary co­instantiation of logically distinct essences. I will argue that, if we interpret physical theory in a non­Humean way, it becomes logically impossible for there to be a purely physical world. The causal individuals of any causal world will possess non­physical, receptive aspects. Nevertheless, the logical connection between these aspects is not one of supervenience (because it is mutual), and the necessity connecting them is not merely nomic (because it is not logically contingent). It is a natural dualism of necessarily connected dualities, but not one that involves a merely nomic, external connection.

Before proceeding to the next chapter, the reader should be aware that the it is essentially a tutorial. It introduces most of the basic concepts this view of causation will use. In some ways, the chapters following the next chapter contain the richer, more interesting material. They are certainly more accessible. Since the next chapter is technical and difficult, readers might prefer to skim over the discussions of its diagrams to try to get the main ideas, and then head straight for chapters thirteen through sixteen. After reading the discussion there, readers should be in a better position to decide if it is worthwhile to do the hard work involved in following the details of the next chapter.

11.5 The conceptual grounding for receptive connections

The metaphysical system elaborated in the rest of this chapter, and the next, is an articulation of three reasonably intuitive ideas.

(1) The world contains effective properties.

(2) A causal nexus consists of at least one effective and at least one receptive individual.

(3) Unmediated interactions are law governed relations between individuals within a nexus.

In the diagrammatic models that follow I try to apply these basic ideas in a way that captures the core of causation in its maximal generality. For the reasons cited at the beginning of this chapter, this goal requires abandoning the idea of causal responsibility as the centerpiece of causation because assignments of causal responsibility involve both objective and subjective factors. The objective factor is the idea of causal significance, built around the idea of symmetric and asymmetric causal constraint. The subjective part of causal responsibility is a kind of figure/ground relation, in which part of the world is intentionally held constant so that some special state of affairs may be singled out as being the cause, or a cause, of some event. Because I am placing causal significance at the heart of my account, some central traditional disputes about the nature of causation fall to the background. For instance, I do not discuss whether causation can occur backwards in time, or whether there can be simultaneous causation.

The idea of receptivity that I am going to work with is a generalization of that found in the stereotypical cause/effect relationship. Here is an example to help illustrate the basis for the generalization. As a kind of fiction, imagine that a burning piece of paper and the flame it produces are basic, natural individuals (whatever this might mean). As part of this fiction, imagine that the effective properties of the piece of paper are causally affecting the flame, whose exact properties (heat, shape, color, etc.) are receptively constrained by the paper's effective state. I am going to conceive of a situation like this as a causal nexus with the paper and flame as component individuals. The nexus is analyzable into the two individuals (paper and flame), a set of effective properties possessed by the paper, and a one way receptive connection binding the two individuals together.

A generalization of this stereotype forms the conceptual foundations of the approach to causation below. This generalization consists of the following abstractions from the basic stereotype (I introduce the generalizations here to help prime the reader, but I am afraid that the full meanings of the terms can only be made clear through examples. I give appropriate examples in the tutorial).

First, I generalize the notion of a natural individual in an inductive way. By definition, the instantiation of a basic effective property will count as a basic natural individual. I inductively define other natural individuals to be any set of natural individuals sharing a common receptivity.

Second, I generalize the idea of receptivity. The generalized notion of receptivity allows the receptive connections to bind individuals in mutually receptive ways, and allows for receptive connections of arbitrarily large arity. In abstract isolation, we can think of an instance of receptivity as an unsaturated entity containing slots. Occupation of these slots by individuals binds the individuals to one another in constraining relationships. I define an instance of receptivity as an N­place connection with at least one slot that opens its occupant to constraint by the effective states of the occupants of other slots.

Third, I will define an effective property. A property is effective just in case it constrains the states of other individuals that share a receptive connection with it. As the base case, the effective state of a basic effective property is just that property itself. The effective state of a more complex individual is a function of the effective states of the individuals constituting it.

I also will use some derivative terminology that I understand by reference to these generalizations. When a set of individuals share a common receptivity I will say that they are bound within a common receptive field. That means that they interact as a natural unit within higher­level individuals. The tutorial expands on the meaning of "natural unit." For brevity, I sometimes refer to a natural individual just as an individual. Also, I sometimes refer to a natural individual with both receptive and effective aspects as a causal nexus. I use causal nexi as a pluralization of causal nexus.

Finally, I want to point out that the natural individuals of our world, the causal nexi, are not intended to correspond in any direct way to the perceptual and conceptual individuals we speak of in daily life. I even take it to be a substantial empirical question whether the individuals within a successful scientific theory are natural individuals in the sense that I am concerned with. For example, societies may appear as individuals within sociology, and galaxies may appear as individuals within astronomy, but it does not follow that they are natural individuals. The natural individuals, the causal nexi, are individuals in virtue of the fact that they play a special, unitary role in causation, and causal interaction. They are ontologically unitary in a sense that common sense and scientific individuals may or may not be.

In the next chapter I shall try to help readers develop a sense of the kinds of constraint structures that may exist given just the core ideas I am begining with. This should give a wider picture of all possible causal worlds. What these models show, if successful, is that the original set of simple, common sense ideas, suitably generalized, is very powerful.

1. In my discussion, I am using the name 'Humean' to describe a class of views broader than simply the regularity view and its variants.What I do mean by the term should become clear as the discussion progresses. The class of views I call Humean includes all those that do not posit a basic connection of constraint or production in the world, so it includes some views that issue from explicit dissatisfaction with Hume, such as Mackie's views on INUS conditions (1974).

2. Which notion, causal responsibility or causal significance, deserves the name "causation?" I think the ordinary language use of "causation" names causal responsibility, but I am going to extend the term for the rest of this book. In most places, when I use the term "causation" I will be talking about causal significance. In the few places where I use "causation" to mean causal responsibility, I hope that the context makes the switch clear. With luck, no harmful confusion will result from these slight equivocations.

3. I am following Strawson (1959) in taking a liberal attitude towards the meaning of "individual." An individual is simply an entity that bears properties. The reason for being so liberal is to avoid heavy commitments at this early stage to what the ultimate causal ontology will be like. To quote Strawson in full,

So anything whatever can appear as a logical subject, an individual. If we define 'being an individual' as 'being able to appear as an individual', then anything whatever is an individual. So we have an endless variety of categories of individual other than particulars ­ categories indicated by by such words as 'quality', 'property', 'characteristic', 'relation', 'class', 'kind', 'sort', 'species', 'number', 'proposition','fact', 'type', etc.(p. 227)


Chapter 12

A Tutorial: Some simplified models

12.1 Overview of the presentation

The diagrams below divide into three suites. The four diagrams in suite one show how one might use the basic ideas to give an account of causal processes, strong and weak determinisms, indeterminism, causal counterfactuals, and both mediate and immediate interaction.

The four diagrams in suite two generalize the ideas in figure one, applying them to possible worlds with multiple layers of individuals. That part of the chapter shows how to understand the possibility of brutely emergent laws, higher­level processes and individuals, epiphenomenal individuals, and emergent effective properties.

Finally, the third suite introduces two diagrams that give possible causal structures for the Life world discussed in chapter two. It uses the Life world to show how the framework may produce orderly worlds. In the process, it illustrates the type­identity conditions on effective properties, drawing out the consequences of their circular dependence on receptivity.

12.2 How to understand the diagrams

This chapter uses diagrams to explore some models of the causal nexus. Although the models themselves are just toys used to explicate the basic notions, I am proposing that the general kinds of elements featuring in the models have analogous elements in the actual world. The arguments presented so far justify introducing a model of these general features. From this point on, it will be the explanatory power of the ideas that must bear the burden of justifying this species of models.



The essential information readers need to understand the diagrammatic "language" I use in the rest of this chapter, and in chapter sixteen, is contained in box 12.1. Some sample diagrams illustrating the grammatical elements in the diagrams are presented in figures 12.1­12.5





In the text below, the phrase "the nomic content of an individual" refers to a set of individual effective states bound within a common receptive field. This means that the instantiation of a lone level zero individual, a lone effective property without a receptive binding, does not have nomic content. In the idiom I am introducing, an individual must be at least a level one individual to have nomic content. Level one individuals are defined to be one or more primitive effective properties sharing a receptive connection. It is this shared receptivity that binds them within a common receptive field. In the general case, level N individuals consist of one or more level N­1 individuals sharing a receptive connection.

The diagrams represent two kinds of receptive connections between the members of a nexus, polydirectional connections and unidirectional connections (shown in figure 12.5). I do not think of one kind of connection as more fundamental than the other. Unidirectional and polydirectional connections support different counterfactuals, as the examples below will illustrate.

The diagrams represent unidirectional connections by using lines whose beads are attached to the individuals that have causal significance in the connection (i.e., the individuals that are doing the constraining). An individual A is unidirectionally connected to B just in case A is constrained by the effective state of B, but not vice versa.

One may intuitively state the idea behind a unidirectional connection in this way. If A is unidirectionally receptive to B, then the effective state of B constrains the state of A, and yet B remains independent of A. The notation ik[i1,i2,...,ij]im will represent unidirectional connections between a nexus of elements i1,...,ij constrained by an individual ik. When present, the subscript im represents the name of the higher­level individual constituted by the nexus of elements. To save clutter, I will suppress im when clarity does not require its use. The brackets represent im's constituent structure, and I use the arrow '' to symbolize the asymmetric constraint between ik and the members of im. I will discuss unidirectional connections further in chapter sixteen, where they feature prominently in the explanation of the direction of time.

The diagrams represent polydirectional connections using lines with beads on both ends, and perhaps in the middle depending on the number of bound individuals. I try to place the beads in a way that indicates which individuals the connection binds, and the number of beads represents the arity of the connection.The simplest case of a polydirectional connection is a bidirectional connection. Two individuals, A and B, are bidirectionally connected just in case a two­place receptive connection exists through which A is receptive to B, and B is also receptive to A. I will then say that A and B share a common receptivity that binds them within a common receptive field.

To say that A is receptive to B is to say that the effective state of A within the nexus must be compatible with the effective state of B within the nexus, as determined by the laws governing effective properties of the type that A and B may instantiate. To say that A and B are bound within a common receptive field is to say that they can collectively constrain and be constrained within further nexi. The discussion of the diagrams in suite 12.2 will further explain the importance of being bound within a common receptive field.

In the general case, a set of individuals a,b,c,...of size N is polydirectionally connected just in case they share an N­place receptive connection through which each element of the set is made receptive to the effective state of every other element of the set. An N­place polydirectional connection forms an N­member nexus of mutually constrained and constraining individuals, individuals whose effective states are instantiated through reference to the whole. The examples we discuss will make the meaning of this condition clearer. The notation I will use to represent a set of individuals bound within a common receptive field is [i1,i2,i3,...,ij]ik, where i1 through ij name the bound individuals, and ik (usually suppressed) names the individual they are bound within.

12.3. Suite 12.1: Level one individuals

I will be examining a conception of the causal nexus as a collection of effective individuals sharing a common receptivity. All the situations depicted in suites 12.1 and 12.2 contain two effective properties, + and o, governed by a single law. The effective law for suite 12.1 is:

This way of stating the law has equivalent formulations that will help to show the connections between the terminology. Some of these alternative formulations are:





Diagram (a) ­ The boxes around the + and o properties indicate that they are each bound within a level one individual, and the scope of the receptive field is just the single level zero individual that can have a value of + or o. In notation, each individual is represented by a formula such as[ij]ik, where the name ij may be replaced by the effective state of a level zero individual, and ik is replaced by a name for the higher­level individual. These replacements yield representations such as [+]i1 and [o]i2. In cases where the name of the higher­level individual is not important, I will suppress the subscript.

The causal situation in diagram (a) trivially constrains each effective property to be compatible only with itself, and so we may see it as presenting the limiting case of nomic content. The question marks within the boxes at times two and three represent the fact that those individuals can instantiate either effective property, independently of the state of any other individual. Despite this world's operational equivalence to a Humean world, it is not a world without causation. This world has true causal structure.

Trying to imagine a very similar world with a different law vividly brings this out. In this other world, the law governing the effective properties would prohibit odd numbers of property types from being instantiated within a nexus, rather than prohibiting an even number of properties. In such a world, degenerate individuals like those in diagram (a) could not exist because any lone occurrence of a property within a nexus would violate the law. In the current context, diagram (a) represents a possible situation only because the existence of individuals like those in (a) is consistent with the law. They meet the real constraints on the instantiation of the effective properties. So even in an anarchic world like this, it is a substantive causal matter whether or not there can be trivial individuals of this type.

Diagram (b) ­ Box 12.2 names the individuals in diagram (b). Diagram (b) remedies the degeneracy from diagram (a) by giving the receptive fields temporal depth. In the notation for this diagram, I have organized the names so that the left­right order of the names within the brackets mirrors the order from earlier times to later times in the diagram.

Looking at the diagram, one may see that the lefthand process exhibits two overlapping unidirectional connections across time. In notation this pair of level one individuals is represented as +[o] from time one to time two, and o[+] from time two to time three. Additionally, the value of the + effective property in time one is "fixed" for the rest of the process: it is given, and also immune from influence through any receptive connection.

Like the lefthand process, the righthand process also exhibits two overlapping connections across time. Unlike the lefthand process, its two level one individuals contain bidirectional receptive connections, making the bound individuals mutually receptive. These individuals are represented as [o,+]i9 stretching across times one and two, and [+,o]i10 stretching across times two and three. The overlaps here illustrate an important general feature of the model: one individual may share indefinitely many receptive connections, creating overlapping receptive fields. Finally, to reiterate the meanings of the terms I will be using, each receptive field with bound individuals constitutes an individual with nomic content, and each individual with nomic content is a causal nexus.

The diagram also illustrates, in a simple way, the chief theme of this approach to understanding causation. I believe that the most effective way to think of the problem of causation is as being about how nature distributes the effective properties within a world. To be a "problem" it must have constraints, and these constraints have a three­tiered character.

On the first tier, we have the general laws governing the interactions of effective states when individuals are receptive to one another. These laws present universal constraints on the possible states of individuals within possible, more particular problem spaces. We can think of them as being the parameters that delimit the natural possibilities for the effective states of the universe. In the models of this suite, only one law is active, the law prohibiting even numbers of type instantiations, and we have only two potential effective properties that may be bound within individuals, either + or o.

On the second tier, we have a network of receptive connections. These receptive connections overlap, binding individuals together in a meshlike way. We can see that the world's receptive structure, abstracted from any particular instantiations of effective properties, instantiates a kind of skeleton that determines a generic constraint structure within which application of the laws becomes important. This second tier is very important, since different topologies for the receptive network may exclude different possibilities for the effective states of the world.

On the third tier, any effective states whose values are "fixed" relative to a nexus present constraints directly to that nexus, and indirectly to other nexi through the overlaps between receptive fields. The scheme only allows solutions including them at the places that instantiate them. With these three tiers of constraint in mind, we can think of causation in terms of solving a multiple constraint satisfaction problem. The constraints operate on a space of possibility, and receptive connections themselves emerge as operators on this space of possibilities.

The two processes in diagram (b) illustrate two simple constraint structures. The set of constraints associated with the process on the left consists of the one general law active in these diagrams, and the partial set of causal facts represented by +[i3] i3[i5]. The constraint structure contains overlapping receptive connections, and one fixed effective property. It has only one solution, a sequence that oscillates back and forth between the two effective properties, beginning with the fixed + at time one.

The process on the right, because it possesses bidirectional connections, represents a different constraint structure. The formula [i2,i4] [i4,i6] represents that structure of constraints. For this process, either of two oscillating sequences satisfies it, so either would be acceptable.

Counterfactuals ­ The difference of connection type makes a subtle difference to the counterfactuals that hold in the two cases. On the left the connections are asymmetric, and the value instantiated at time one is fixed relative to the rest of the model. Consider the following counterfactual, "If the value of i3 had been +, then the value of i1 would have been o." The causal structure presents two complications that we need to consider closely in evaluating the counterfactual. First, the antecedent is impossible, and, second, the asymmetric connection from time one to time two makes it irrelevant to the consequent.

Why is the antecedent impossible? Its causal structure makes the process on the left, as a whole, very strongly deterministic. This strong determinism shows up in the fact that the constraints of the situation (given by the formula +[i3] i3[i5], plus the general law) are consistent with only one solution in the space of possibilities. The constraints on the situation, particularly the presence of the fixed + property at time one, necessitate the values at times two and three. Additionally, they make i1 immune from influence. These facts, together, raise problems for evaluating the counterfactual.

The puzzle comes out more clearly by contrasting the lefthand process with its near twin on the right. The receptive structure on the right differs from the one on the left only by containing bidirectional connections across each time slice, and by not having the value of an effective property fixed relative to the whole process. The general law, and the formula [i2,i4] [i4,i6], represent the causal constraints on the process. The space of possibilities contains two solutions to these constraints, (i) [+,o]i9 [o,+]i10 and (ii) [o,+]i9 [+,o]i10.

The righthand process is therefore indeterministic, as the causal constraints do not serve to constrain the space of possibilities to a unique set of facts. The counterfactual, "If the value of i4 had been o, then the value of i2 would have been +.", which is analogous to the previous counterfactual, seems to be clearly true. The antecedent corresponds to solution (i) above, and the consequent is satisfied by that solution. There is no analogous truth­maker for the first counterfactual we considered, the one about the lefthand process.

So how should we evaluate the counterfactual for the lefthand process? We should resist one natural reading that tempts us to evaluate it as true. Under that reading, we look at the situation and unconsciously let the distribution of receptive constraints shift. We are tempted to only keep the general law constant, while letting the unidirectional connections slip over into bidirectional connections. By allowing ourselves to change the particular causal constraints on the left, we can come to see its counterfactual as true.

Although it is easy to lose sight of the causal connections as part of the relevant background constraints, they are part of the relevant background as I am concerned with it here. The lefthand process is not the same kind of entity as the righthand process, and should not be shape­shifted into it. The subjunctive conditional for the lefthand side has an antecedent with two problematic features. First, it could not possibly be true in the circumstances, so it has no truth­maker. Second, the state of the individual in the consequent could not possibly be influenced by the state of the individual in the antecedent, due to the unidirectional connection.

If only the former held true, I think the best response would be to refrain from giving a truth value to the counterfactual. The reason is that, in addition to having no truth­maker, it has no falsity­maker either. A falsity­maker would be a possible solution in which the antecedent is true, and the consequent false. Since the antecedent is impossible, there is no such solution. The difference between counterfactuals with falsity­makers and those without them should be respected. I believe the best way to respect this difference is to say that some counterfactuals like the one we are considering are neither true nor false.

However, the most important feature of the situation is the absence of a direction of influence from i3 to i1. The problem is that i1 is fixed relative to i3. Even if we allowed i1's value to vary as the result of some other feature in the model, it would not be variable as a function of the value of i3. I suggest that, intuitively, when we consider causal counterfactuals we typically care more about the flow of effect than we care about natural possibility. Therefore, the fact of irrelevance overrides the impossibility. Consequently, I believe the counterfactual for the lefthand side is false, despite the truth of its analog on the right.

Two kinds of determinism ­ Although the righthand process is metaphysically indeterministic, in a weaker sense it is also deterministic. All the forward looking counterfactuals such as, "Had i2 had the value of +, then i4 would have had the value of o." are true. Additionally, all backwards looking hypotheticals such as, "If i4 has the value of o, then i2 has the value of +." are also true. Together, these yield a variety of informational or epistemic determinism shared by both processes in the diagram. Given background knowledge of the other causal constraints, the information about the effective state of any cell yields the information about the effective states of cells at all other time slices in the process. In fact, one would not even need the correct background knowledge. As long as one had a "law" describing the oscillation, the information would become available regardless of whether one knew the particular receptive constraints.

The two processes in diagram (b) highlight two different varieties of determinism, a metaphysical determinism and an epistemic one. The left­hand process is metaphysically determined by the causal constraints and laws in the sense that they necessitate its effective character. It is also informationally deterministic in the sense defined above. In contrast, the process on the right is metaphysically indeterministic since the effective character of that process is underdetermined by the causal laws and receptive constraints. Nevertheless, it is epistemically determined since the solution that is instantiated can be determined entirely from an examination of any part of it, plus knowledge of those causal laws and receptive constraints. Any metaphysically deterministic process will be epistemically deterministic, but not vice versa. Thus, they seem to be a weak and strong variety of determinism.



Diagram (c) ­ Box 12.3 names the individuals in diagram (c). The receptive connections in diagram (c) are temporally shallow, but have spatial breadth. Although each time slice in the diagram is independent of the others, the world is not Humean. The receptive constraints in the diagram cut the space of possibilities in half by excluding the instantiation of either two + properites or two o properties in a time slice. Individuals i7 and i8 possess instantiated values to represent the two possibilities that remain, while the question marks in the receptive field of i9 emphasize its independence from the individuals at the previous times.

The states of the level one individuals in diagram (c) are metaphysically indeterministic, but epistemically deterministic. Each receptive field represents a constraint of the form [ij,ik], which has two solutions in the space of possibilities: [+,o] and [o,+]. Thus each field represents an individual whose total state is indeterministically constrained. Nevertheless, if one were to make a measurement of either the right or the left side of the receptive field, one could infer the effective state of the individual on the other side. Note that the model does not use the assumption that individuals bound within a common receptive field must be temporal or spatial neighbors. We do not have good reason to restrict receptive connections locally in that way.

Diagram (d) ­ Box 12.4 names the individuals in diagram (d), which represents a slightly more complicated network of receptive fields. It has one field with spatial breadth but no temporal depth, one field with only bidirectional temporal depth, and one field with only unidirectional temporal depth. Once again, the shaded areas represent overlapping receptive fields.

This diagram chiefly illustrates the difference between mediate and immediate causal connection. Consider the spatial individual i6 instantiated at time one. I6 is just like the spatial individuals from diagram (c). Like them, i6 must contain one + property and one o property, but the places of their instantiation are left open. The temporal process on the left, i5, is like the bidirectional temporal process from diagram (b).

Notice that the counterfactuals "If i2 were o, then i3 would be o." and "If i2 were o, then i1 would be +" are both true, but true for different reasons. The counterfactual involving i1 and i2 is true because they are both bound within i6, the level one spatial individual. The presence of i6 constrains them to "opposite" values, and the constraint it represents is direct in the sense that other effective properties do not mediate it. The only two relevant facts are: (i) they are bound to one another within a nexus, and (ii) their effective states must conform to the laws governing effective facts within that type of nexus.

On the other hand, the counterfactual involving i2 and i3 is made true by the overlap of the level one individuals, i5 and i6, on the level zero individual i1. In this case, i1 mediates the relationship between i2 and i3. As represented, i3 is responsive to i2 through i1, but it is not bound to i2 directly within any higher­level individual. The proper phrasing is to say that i3 is responsive to the value at i2, but not receptive to it. The difference between responsive and receptive relations corresponds to the difference between mediate and immediate causal interaction.

The practical differences between being responsive to an individual at the same level, and being receptively bound to it, will be illustrated in suite 12.2. There I extend the model to accommodate individuals of indefinitely high levels. I will discuss how the existence of level one individuals such as i6 opens the possibility of stratified layers of individuals: level two individuals, level three individuals, and so forth. In these nexi, lower level individuals act as wholes to constrain other individuals. Individuals connected in a merely responsive way are separable: they do not need to share significance within the same higher­level individuals.

The network in diagram (d) supports other responsive relationships, including one from i3 backwards in time, then forward in time again, and finally settling across space at i4. For example, if i3 had been o then i4 would have been +. Because i4 shares a unidirectional connection to i2, the converse does not hold. In this diagram, i4 is idle.

12.4 Suite 12.2: Generalized higher­level individuals

The four diagrams in this suite illustrate aspects of higher­level individuals, including individuals at level two. The diagrams in suite 12.2 will use the most straightforward extension of the general law from the diagrams in suite 12.1. Let v+(x)

and vo(x) be two functions that accept any individual as an argument, and that return the number of + or o properties, respectively, that are instantiated within that individual. The general law for a higher­level individual is that each of those functions, when summed over all of the higher­level individual's bound individuals, must yield either zero or an odd number.

Diagram (a) ­ Box 12.5 names the individuals in diagram (a). For convenience, I adopt a notation for higher­level individuals that marks them for quick identification of their level. A level one individual will be represented by [....]ik, a level two individual by ^[....]ik, a level three individual by ^^[....]ik, and so forth. The carat character signifies the level of the bound individuals within the brackets. Looking at box 12.5, notice that the level two individual, i10, is a bonding of the level one individuals; this is represented by ^[i7,i8,i9]. The existence of i10 makes i7, i8, and i9 ­ the level one individuals bound within it ­ directly receptive to each other. It is here that a difference between being receptive to an individual, rather than merely being responsive to it, shows up. As we have already seen, if two lower­level individuals are receptive to one another, their mutual receptivity instantiates a higher­level individual that directly assimilates both of them. When individuals of yet higher levels directly bind these new individuals, they must also indirectly incorporate all the constituents at even lower levels. In contrast, a higher­level individual may "carve off" two individuals that are merely responsive, incorporating one but not the other. One may be part of it, while the other one is not.

Looking at diagram (a), how do we determine the possible states for i7, i8, and i9 in the depicted situation? Obviously, the internal constraints on the states of i7, i8, and i9 are still active. These possible individual states constitute a set of prior potentialities for each individual, and the combinations of those solutions constitute the possibility space that i10 further constrains. To illustrate by example, in diagram (a) the presence of i10, the level two individual, shrinks a prior possibility space made up of the potential states of its constituents. The prior possibility space presented to i10 is:

i7 i8 i9

Possibility one ­ [+,+,+]i7 [o,+]i8 [o,+]i9

Possibility two ­ [o,o,o]i7 [+,o]i8 [+,o]i9

Possibility three ­ [+,+,+]i7 [+,o]i8 [+,o]i9

Possibility four ­ [o,o,o]i7 [o,+]i8 [o,+]i9

Read each row above as constituting an ordered triple, and interpret each triple as a member of the prior possibility space given to i10. The existence of i10 shrinks this possibility space in half by excluding possibilities three and four. The problem with those two possibilities is that they would instantiate an even number of +'s and o's within i10, thus violating the general law (remember that i4 is the first element of both i8 and i9, so do not double count its value). As a side effect, the presence of i10 forces a correlation between the state of i7, and the combined state of i8 and i9.

The existence of higher­level individuals also gives us a more general way to view effective properties. For instance, smaller nexi like i8 and i9 have two possible states, [+,o]ik and [o,+]ik. Each state adds one + and one o to a nexus, so within a higher­level individual these two states present the same constraint for the other members (notice that things are not quite this simple in the specific case of i10. The overlapping bond i8 and i9 share makes them similar to a three member nexus with two possible states: [+,o,+]ik, and [o,+,o]ik).

To identify effective properties in a consistent way across­levels, we should always identify them with the constraints they present to one another within higher­level nexi. The constituent differences noted above are not differences that make a difference, since each represents exactly the same constraint within the higher­level individuals that might bind them. Thus, they present the same effective state within that individual, the distinction between them collapsing as they move through the higher­level filters.

Emergent laws ­ Once we have achieved this insight, the policy of adopting higher­level laws that are simple continuations of the laws governing lower­levels pops out as a convenience. We can easily imagine semi­independent laws governing effective states of individuals at different levels. These emergent laws would have to be consistent with the laws governing the effective properties at lower levels, but would not have to be entailed by them. For instance, laws governing the level­two individuals easily could require that only prime numbers of property types instantiate within them.

A consistency requirement is easy to enforce since the system builds consistency in from the ground up. The lower­level individuals present the domain of prior possibilities to the higher­level individuals. Since the domain they present is already constricted, the restrictions of the lower levels will always be already present at the higher­levels. Therefore, the scheme necessarily produces interlevel consistency. Semi­independent, higher­level laws only require the lower­levels to sometimes yield indeterministic results. That way, there is work for them to do. Despite making a difference to the behavior of systems, the operation of emergent laws of this sort would not be detectable through a violation of the laws governing the behavior of lower­level individuals. They could be present, but invisible (or nearly invisible).

Emergent properties ­Whether or not a world possesses emergent laws of this sort, it can possess emergent effective properties. A formal way to think of these emergent effective properties is this. We will say that possible effective states of A possess a prior difference to one another if they instantiate a different constituent structure for A. In this sense, [+,o]ik and [o,+]ik instantiate effective states with a prior difference even if subsequent levels may filter this difference out.

How can these prior differences support the emergence of effective properties? Each receptive field associated with a possible individual ^[i1,i2,...,ij] instantiates a function. This function maps the prior possibility space for <i1, i2,...,ij> into a subset of that space. Each vector of this prior possibility space contains a possible combination of states for i1 through ij, like in the example we discussed earlier. In the simplest case, the prior possibility space will just consist of the rows in the Cartesian product i1×i2× . . . ×ij, where the values of i1 through ij are their independently determined prior potentialities.

Lower­level individuals linked together as a process, such as i8 and i9 in diagram (a), present a minor complication. The overlapping instances of receptivity provide for a shared constituent that acts as a mediate, responsive link between the individuals in the process. Through this link, the set of possibilities for their joint instantiation may already be constrained. This is what happened with i8 and i9 in the example above. Instead of the eight prior possibilities that would be available for the joint instantiation of two independent individuals like i8 and i9, only four prior possibilities actually exist for i10. The shrinkage in the prior possibility space occurs because, for the purposes of determining it, they need to be treated "as if" they formed a single three member individual with a unique kind of constraint. Of course, they only form an "as if" individual, as the two individuals may, in fact, be carved off from one another for exclusive binding within a higher­level individual.

We have to accommodate the presence of such processes by eliminating the impossible instantiations from the Cartesian product. Thus, linked processes may sometimes count as de facto individuals for higher­levels, although they are not naturally individual. Once this adjustment is made, the function representing the higher­level receptivitity maps allowed vector states onto themselves, and disallowed vector states onto the empty set. What is allowed or disallowed depends on the general laws pertinent to nexi of its sort. The allowed states that remain are all the vector states from the original domain that are still possible, given the presence of the higher­level individual. Unless the higher­level individual is epiphenomenal, this remainder will be a proper subset of the original domain.

Using this conception, we can individuate the possible effective states that an individual A may realize within a higher­level individual B. This will capture the posterior differences in the effective states of B's constituents. Let s1 and s2 be two state types for A that possess a prior difference, and let tuples such as <...s1...> and <...s2...> represent their occurrence(s) in the prior possibility space presented to B. We will say that <...s1...> and <...s2...> are counterparts of each other just in case they are exactly the same length, and s1 and s2 occur in exactly the same places in their respective tuples. We will say that <...s1...> and <...s2...> are exact counterparts just in case they are counterparts, and they differ in no way except these occurrences of s1 and s2. S1 and s2 instantiate the same posterior effective state within B just in case s1 and s2 are interchangeable within B. This means:

(i) Every <...s1...> has an <...s2...> counterpart in the prior possibility space, and vice versa.

(ii) Whenever the function that represents B's receptivity maps a tuple <...s1...> from the prior possibility space onto the empty set, it also maps an <...s2...> onto the empty set where that <...s1...> and <...s2...> are exact counterparts.

(iii) Whenever the function that represents B's receptivity maps a tuple <...s1...> from the prior possibility space onto <...s1...>, it also maps an <...s2...> onto <...s2...> where that <...s1...> and <...s2...> are exact counterparts.

A little more concretely, s1 and s2 instantiate the same effective property within B just in case it is always possible to exchange one state for the other without changing the set of possible states available to any of the other individuals within B. In such a case, the prior differences in A's states, s1 and s2, are differences that do not make a difference within B. They are informationally irrelevant within B, as they represent the same constraint on the other individuals. Given the notion of an effective state, these informationally relevant states of A are A's true effective states within B, and they need not be identical with any member from the class of states sharing a prior difference. States such as s1 and s2 are the possible realization bases for the higher­level effective property that has emerged. One important consequence of this is that the receptive connections of high­level individuals may actually bring irreducible effective properties into existence. Finally, s1 and s2 are the same effective state tout court just in case they are the same effective state within all possible B's. Applied to our example, these definitions yield that state types such as [+,o]ik and [o,+]ik, although they share a prior difference, are actually the same effective state tout court.

Diagram (b) ­ Box 12.6 names the individuals in diagram (b). This diagram illustrates how changes at lower levels may prevent the continued existence of a higher­level receptive field. The level one individuals i11 and i12 can support the existence of the level two individual, i17, because their prior possibility space contains a solution to the constraint it presents. In the formal sense outlined above, this means that the function representing the receptivity of i17 maps the prior possibility space for i11 and i12 onto a non­empty set.

The relations between lower­level individuals may change through time, and it is easy to imagine that subsequent arrangements may not support the kinds of higher­level individuals supported by earlier arrangements. These kinds of changes may prevent a higher­level process from continuing, and diagram (b) depicts this kind of circumstance. A new level zero individual appears in time three, breaking the triple bonding from time two into two double bondings. If an analog to the individual i17 could be instantiated to bind i12, i13, and i15 (e.g., ^[i12,i13,i15]), it would be reasonable to see this as a continuation of the high­level process begun as i12. Unfortunately, no solution exists for ^[i12,i13,i15], and so no continuation is possible. Formally, the receptive connection of the proposed individual, i18, maps the proposed prior possibility space onto the empty set. In the sense that processes are de facto individuals, a high­level individual ceases to exist.

Diagrams (c) and (d) simply represent the two opposite ends of the spectrum of

significance for a higher­level individual. In diagram (c), the presence the level two individual forces a unique solution for the instantiations of the basic effective properties. The solution is easy to figure out, and I leave it open so the reader may enjoy discovering it. Diagram (c) represents a case where the existence of a higher­level individual makes an indeterministic world into a strongly deterministic one. Diagram (d) represents an "epiphenomenal" higher­level individual. Its presence does not constrain the prior possibility space given by its constituents at all.

12.5 Suite 12.3: The orderly world of Life

The examples in suites 12.1 and 12.2 were chosen to highlight how the basic features of this model of causation may be put to work. They were designed especially to show the differences that patterns of receptive connection make, and do not represent interesting worlds, or situations within interesting worlds. An important question looms: can this view of causation account for the existence of an interesting, orderly world? For instance, could the Life world analyzed in chapter two be a causal world in this sense? Figure 12.14 shows two possible causal structures that would yield Life worlds if conjoined with suitable laws.



Diagram (a) ­ Box 12.7 names the individuals in diagram (a). The structure of this Life world contains a polydirectional receptive field that binds the neighbors of every cell, and a unidirectional receptive field that binds each cell to its immediate ancestor. These level one individuals, i10 and i11 in the diagram, are bound together into a level two individual, i12. If we reproduce this array of structures for each cell in a world, laws that straightforwardly mirror Life's rules of evolution will produce a Life world. First, we allow all combinations, without restriction, to instantiate in the structures (like i10) that contain a cell's neighbors, and the structures (like i11) that represent the cell's temporal depth. The Life world arises from the laws that govern how these level one individuals interact within level two individuals like i12. To express the laws, we will use a function v(x) that accepts any individual as an argument, and returns the number of "on" properties it contains. I will let i5, i10 and i11 stand in for all individuals of their type. The three rules of Life translate into the following three laws:

(1) If v(i10) = 2, then v(i11) = 2 ×v(i5)

(2) If v(i10) = 3, then v(i11) = v(i5) + 1

(3) Otherwise, v(i11) = v(i5)

Law one corresponds directly to the two neighbor rule, law two corresponds to the three neighbor rule, and law three corresponds to the "otherwise" condition in Life. Apart from producing regular patterns of effective property instantiations, this example also illustrates how receptive connections in a realistically orderly world will also instantiate in a regular, lawlike way.

Diagram (b) ­ Box 12.8 names the individuals in diagram (b). This diagram represents a different causal structure that can also produce a Life world. The important differences here are at levels one and two, where the receptive structure of this world is slightly different from that in the previous example. The laws required to make this universe produce a Life world are more awkward. I will not list them all, but they have forms such as,

If v(i11) + v(i12) + v(i13) + v(i14) = 3, then v(i15) = v(i10) + 1

We need this slightly more complicated form to reflect the true causal structure of the world. It is a trivial matter to produce more radically different laws. In an extreme case, it takes eighty­one laws to specify the behavior of a world where i1­i9 (i.e., including the center cell) are collectively bound within a level one individual.

The most important consequence of the different receptive structure in diagram (b) is that it has different laws than diagram (a), and therefore different individuals with different effective states. This follows from the fact that two effective states are the same only if they are interchangeable, in the way defined earlier. Unfortunately, the constituent structures of the level two individuals in diagrams (a) and (b) are different. Consequently, the world of diagram (a) simply has no laws to handle the individuals of diagram (b), and vice versa. There's no fact of the matter about what constraints each world's individuals would present in the other world, and so no fact of the matter about what kind of effective states those individuals would have in the other world. The two worlds literally instantiate distinct effective properties, as well as receptive facts, despite both resulting in Life worlds.

One might object that these two worlds are the same world, appealing to the equivalence of their respective physics. The most direct answer to this worry is to observe that the two diagrams contain, quite literally, different receptive connections and so different kinds of individuals. Therefore the causal laws describing the causal structure of those two worlds quantify over different individuals. To ignore this difference in the structure of individuals, and in their connections to one another, is to slip back into irrealism about the world's causal structure. The important point these examples emphasize is that the causal structure of the world is not identical to the nomic mosaic it instantiates, where we may understand its nomic mosaic as the simplest description of the lawful ways in which its basic effective properties instantiate through spacetime. If we are being realists about causation, the causal structure of the world will include the patterns of receptive connection between individuals, and different causal ontologies will result from different patterns of receptivity. Even so, these different ontologies may produce an identical nomic mosaic as long as their causal laws vary as well (as they inevitably will).

The chief moral here is that a world's causal facts and laws do not logically supervene on the set of facts including just (i) instantiations of lower­level effective properties, specified as merely distinct (and, one could add, varying in quantity), and (ii) the laws that describe the regularities and correlations between their instantiations. In the next chapter, I will use this conclusion as an important part of an argument that the receptive properties of our world are, at best, only implicit in physical theory.

A secondary moral is that the circularity between receptivity and effectiveness is categorically important. The answer to the question, "What is it to be an effective property of type P?" always involves the kind of receptivity in the world. Conversely, the answer to the question, "What is it to be a receptive property of type P?" always involves the effective properties in the world.


Chapter 13

Does Receptivity Logically Supervene on the Physical?

13.1 Are the physical properties the receptive properties?

The structure of receptive connections proposed and explored in the previous two chapters is a relatively new thing in our world view, or at least considering it so realistically is new. The arguments in chapter eleven made it clear that the effective properties are physical properties. Are the receptive properties also physical? Chapter twelve's discussion of the Life world seems to rule against that conclusion. In this chapter I will give arguments that the facts about receptivity do not logically supervene on the physical facts.

The Humean mosaic, the nomic mosaic, and the causal mesh ­ How may one understand the discussion in this chapter against the background of recent discussions of causation? Realists about causation typically concern themselves with two levels of structure in the world. The "shallowest" level is the Humean mosaic, which is the historical, merely extensional pattern of instantiations of the physical properties through spacetime. The other recognized level is the nomic mosaic, which is the pattern of instantiations of the physical properties brought under reliable, counterfactual supporting laws. After making this distinction, causal realists argue that the facts about the nomic mosaic do not logically supervene on the facts about the Humean mosaic, as a variety of physical laws are compatible with the facts about the Humean mosaic. The implication of these arguments is that we must go beyond Humean views of causation if we are to be realist about the laws of nature.

The three arguments below defend an even stronger conclusion. Aside from the Humean mosaic and the nomic mosaic, the world must have an intrinsic causal structure that includes receptive connections between individuals. If successful, the arguments below show that the world's causal structure does not logically supervene on the nomic mosaic. In fact, it is the other way around. The implication of the arguments is that one must go beyond a mere description of the nomic mosaic if one wishes to complete a realist theory about causation.

The arguments in this chapter have force against physicalism because they give reasons for believing that physical theory is concerned with describing the nomic mosaic only. By limiting its concerns to a description of the nomic mosaic in the simplest possible terms, physical theory leaves the world's receptive structure, its connective tissue, so to speak, as something we understand only implicitly through the way we use physical theory. In particular, because the world's receptive structure is only implicit in physical theories, a variety of structures for the underlying causal mesh are compatible with those theories. It seems, then, that physical theory is explicitly about only an aspect of the causal mesh, its effective side.

Below, I put forward three arguments for this conclusion. The first argument is that adding the information about receptivity to a theory of the nomic mosaic complicates the theory without increasing its experimental content. Given the methodological restrictions active in the scientific culture, it seems unlikely that physical theory would explicitly go beyond a theory of the nomic mosaic. The second argument uses the example of an equation describing the movement of a pendulum to argue that the functional character of physical theory leaves room for bizarre receptive structures. If we postulate increasingly baroque causal laws governing causal nexi, then we may counterbalance any increasingly bizarre receptive structures, reproducing the nomic mosaic that physic's functional laws describe. The third argument raises problems concerning the spatiotemporal location of receptivity. Physical theory seems to make predictions about circumstances in which physical properties will have definite locations, but in which one could only implausibly identify the location of the receptive properties with the location of the physical properties. Indeed, placing the receptive properties in spacetime seems to be problematic in general. That strongly suggests that the physical properties are not the receptive properties.

13.2 Including receptivity would not enhance the experimental content of physics

My first argument in support of the conclusion that the facts about receptive connection do not logically supervene on the physical facts uses these premises:

(1) Physics at least designates the low­level effective properties (from chapter eleven, section one).

(2) When proposing the physical laws, we at least postulate regularities in the instantiation of effective properties. (from chapter ten, and section one of chapter eleven).

(3) Physical theories represent the least set of properties that we need to explain experimental results. (simplicity constraint).

(4) The receptive structure of the world does not logically supervene on the facts about the pattern of instantiations of the low­level effective properties, designated as merely distinct (from chapter twelve, suite 12.3).

To make the argument against the logical supervenience of the world's receptive structure on the physical facts, let us imagine that we have a theory P that meets the requirements of (2) above. This means that it describes the lawlike and regular ways that effective properties are instantiated, including the correlations between instantiations. Now, let P* designate an extension of that theory that includes the laws concerning the receptive structure of the world. By (1) above, we know that P's named properties will at least be part of physical theory. By (4) above, we know that P* must be a proper extension of P. The simplicity condition, (3) above, means P* is also a part of physical theory only if the experimental content of P* is greater than that of P.

To argue that physical theory does not include the receptive properties, I must argue that the experimental content of P is equal to that of P*. Let E be an experimental design used to measure some physical property, p. The experimental design E will be adequate just in case we can predict how our measuring devices will vary with the presence of p. Recall that the important aspects of our measuring devices are those that make an effective difference to perception. The scientist's problem is to use the structure of the experimental situation E to derive conditionals of the form, "If property p is present (ceteris paribus), then the measuring device will be in effective state s."

We need a theory that, given E, will allow us to infer the patterns of causation involved in magnifying the effects of the property p. The problem reduces to one of establishing the requisite sub­conditionals governing significant links in the magnifying chain. By hypothesis, P alone will be sufficient to derive the character of the sub­conditionals, since it describes the covariations between effective states at every stage. It follows that P is sufficient to derive the experimental content of our physical theories. As a corollary, physical theory should not include the receptive properties since they will complicate the theory while being explanatorily superfluous to its guiding theoretical concerns. Physical theory should fail to reveal the world's receptive structure for basically the same reasons Life rules fail to reveal a receptive structure for Life.

13.3 The argument from the possibility of bizarre receptive structures

Another way to argue that the receptive facts do not logically supervene on the physical facts is to consider a concrete example of a dynamical system, and see how its physical description underdetermines an answer to questions about the receptive structure of the individual process whose dynamics it describes. Consider the simple case of an isolated, oscillating system like that depicted in figure 13.1. Physical theory's dynamical equations tell us the shape of the system's trajectory through its state space, represented by the cosine­like curve in diagrams (a)­(d). If we assume, as physical theory does, that spacetime is a primitive entity, then diagrams (a) through (d) should also represent a variety of receptive structures consistent with what physical theory tells us.

In diagrams (a) and (b) the receptively bound states are those that intersect with the two different boxes as they fall through the XY plane. The two sets of intersections yield two different receptive structures. In diagram (c), all the states in the half­circle enveloped by the box, and shaded, are bound within a common receptive field, as are the two boxes themselves. In diagram (d), each receptive field has temporal depth, represented by the overlapping circles covering system states. These receptive fields overlap through time, one following upon the other. The reason each possible field is compatible with the truth of the physical function describing the system is that, given a particular structure of causal nexi, we must invoke causal laws governing the states of those nexi. These laws may take forms that vary arbitrarily far from the physical laws describing the evolution of the system, and baroque laws should exist that reproduce a nomic mosaic conforming to the physical function even when the receptive structures are bizarre.

Diagram (d) clearly represents the most intuitive structure, although even it is compatible with multiple interpretations. A structure like the one in (d) may consist of unidirectional constraint forward in time, unidirectional constraint backward in time, and polydirectional constraint. Finally, nothing in physical theory can decide between the hypothesis that the system is metaphysically deterministic and the hypothesis that it is merely epistemically deterministic, where those terms have the senses discussed in chapter twelve.

Each diagram represents a distinct receptive structure, and nothing in physical theory gives conclusive reason to assert the existence of one structure instead of one of the others. Just as pertinently, from a physical perspective the issue is one of indifference as the function itself does all the work in producing the experimental content of the theory. It seems to follow that physical theory is describing the nomic mosaic, and does not contain the theory about the receptive structure of the world.

13.4 Receptivity and spacetime

A third argument concerns the localizability of receptive properties. At least during measurement, the physical properties of an individual are localized in the sense that one can designate a bounded region of spacetime in which the properties are located. It is not clear that receptive properties are strictly located in spacetime in this same way. Some good reasons exist for thinking that they are not.

For instance, in the famous EPR experiments we find that the physical properties of spacelike separated particles are correlated upon measurement. The present framework might naturally explain these correlations by appealing to the presence of a spatially distributed higher­level individual that binds the particles. In these experiments, the physical properties of each particle are located at their places of measurement. Where is their shared receptivity located? Claiming that it is wholly at either of those places seems strange, as does the claim that it is solely at both of them, or even that it permeates all of the space between them. The location of instances of receptivity in spacetime is a puzzling matter, in a way that the location of physical properties in spacetime is not.

In chapter sixteen I will explore a view that might help to explain this awkwardness. There, I will make a case that spacetime is a construct projected from the structure of the causal mesh, where that mesh has both effective and receptive aspects. One can plausibly argue that spacetime is useful chiefly for structuring the effective aspects of individuals by projecting them into a simpler geometry. From this perspective, it should not surprise us that receptivity eludes easy placement in spacetime. A structure of receptive connections supports spacetime, not vice versa, and, since they are a primitive part of a more fundamental reality than spacetime (namely, the mesh of causal nexi), the receptive properties need not project straightforwardly into spacetime. The relation of the receptive properties to spacetime needs a separate treatment from the relation of the effective properties to spacetime, and this strongly suggests that the physical properties are not the receptive properties.

13.5 An example: positive and negative charge

All of the arguments above for the conclusion that the receptive properties are not physical are at a general level. To double check the reasoning, I want to consider a concrete example. One apparent objection centers on properties like the positive and negative charges on particles. Oppositely charged particles attract each other. The fact that a positively charged proton attracts a negatively charged electron seems to be an effective fact about the charge. Yet, the electron seems to receive the attraction due to its negative charge, and we can say the same for the proton's positive charge receiving the attraction of the electron. Is this a case of physical properties being both receptive and effective?

I can produce two kinds of responses to this example, and they will serve as prototypes that one may extend to other examples. The simplest answer is to maintain that the objection misinterprets the situation: the negative charge does not receive the action of the proton. One would not properly predicate the receptiveness to these properties, but to the individuals themselves. The electron receives the action as a whole individual, and so it is the physical object, not the physical property, that is receptive. The reception of the attraction includes the particle's negative charge, but it is not received through its negative charge. A more plausible alternative is that the negative charge enters the story through the explanation of the particle's response.

An electron's response is a reaction, not a reception. That an electron should receive the action of a positive charge, and then have its reaction partially determined by the presence of its own charge, is consistent with the claim that charges are merely effective properties. The account given here requires individuals to respond in such ways, since a nexus is holistic. Even the etymology (re­action) suggests that charge is acting effectively on both ends.

Here is an analogy, very crude but hopefully a little helpful. Imagine two baseball players playing toss over a wall that blinds them to each other's position. Each receives the ball with their gloved hand, and throws it with the other arm. Think of the gloved hand as their receptive aspect, and the throwing hand as their effective aspect. They use the trajectory and velocity of the ball to gain information about the whereabouts of the other player. The two players as wholes are playing toss, and they could not do it without their receiving gloves (imagine the ball is too hard to catch barehanded, or thrown with too much velocity). Each player is trying to return the ball accurately to the other. Ultimately, the combination of his own throwing arm (his active capacity) and the information he has about the other player (the result of the other player's active capacity) determines the positioning of a player, and the character of a player's throws. This is what it is like for the negative and positive charge to both be effective properties.

I prefer the response above because it makes the overall picture simpler to express and explore. An opponent may press the issue at this point, insisting all the same that we should not introduce a new set of properties to account for the receptivity involved. I think such a line of further resistance by an opponent will degenerate to a merely semantic quibble about the word 'physical' for the following reasons. Even if one was to acquiesce to the opponent's demand, an equivalent, if messier, picture emerges since effectiveness and receptivity are still distinct: we have not collapsed the conceptual distinction, and the failure of logical supervenience stands, even if the same, physical property is playing both roles.

If a single physical property is playing both roles, then the conceptual distinction requires explaining how it is that the single property plays both roles. In short, each role raises distinct explanatory demands, and the performance of both roles by a property would require a double story involving that property. In the end, each story will correspond to a different aspect of the physical property. Our opponent has simply insisted that physical terms designate properties with two aspects. The opponent's demand, even if complied with, does not establish that physical theory adequately details both of these two aspects.

Unfortunately, nothing in the representation of the physical property by physical theory marks the conceptual distinction that leads to the explanatory problem. In short, physical theory becomes artificially unitary in its representation of physical properties since it fails to separately mark irreducible aspects of these properties. We are back in the situation that we started with, the only difference being that we have saved the name 'physical'.

Why should we insist that theory mark this distinction? The situation is parallel to what would have occurred had mathematics failed to mark the distinction between ordinal and cardinal numbers. It would have distinct aspects of the concept of number conflated, and would prevent us from appreciating their differences properly. In the end, once we mark the distinction, continuing to insist that only one kind of number exists seems silly. Similarly, the in­name­only unification of properties as "physical" comes to seem silly. Under the pressure to make the physical properties two­sided in this way, we save the physical completeness of the world, but only in name. It remains the case that physical theory is not sufficiently expressive, and it papers over distinct (now "physical") aspects of the world.

As far as I can tell, all apparent counter­examples to the thesis that the physical properties are just the effective properties fails in one of these two ways: (i) the objector does not properly appreciate that receptivity is a property of whole individuals, usually accompanied by a failure to appreciate the role an individual's own effective properties can play in determining that individual's reactions; or (ii) the objector's position implies that physical theory is conflating these distinct aspects by giving unitary representations to physical properties that, in reality, are inherently dual. In the first case, everything I have said goes through. In the second, we get to save the word "physical", but everything important about what I have said goes through: we must make "physical properties" things that have dual aspects, and tell a different, more comprehensive story about them. This latter move seems silly, as representing the duality of the properties would be an explanatorily irrelevant complication from the perspective of physical science.

13.6 From Receptive Connections to Carriers

The last three chapters have explained the role receptivity will play in a theory of causation. At this point I have created a picture of physics as describing the nomic mosaic, which is the description of the lawlike regularities in the instantiation of the basic effective properties through spacetime. I have painted a picture of the world as a mesh of individual causal nexi, each possessing nomic content constituted by receptive and effective aspects within the nexus. The differences between what physics describes and what the world contains leads to Liberal Naturalism. While the account of the world could surely be more complete, it may serve as an adequate sketch except for one detail left to discharge: the circularity in the notions of effective and receptive properties. In the next chapter I will discharge this circularity by introducing the Carrier Theory of Causation, a theory that will bring consciousness back into the picture.


Chapter 14

The Carrier Theory of Causation

14.1 Circularity in the causal mesh

In chapters two and three I examined the anti­physicalist arguments by drawing an analogy to an artificial world, the cellular automaton called Life. I examined the nature of the properties needed to define what it is to be a Life world, and convicted them of being unable to support consciousness. The root of the problem, I claimed, is that the categories essential to Life are schematic. Consider the question, "What is it to be an "on" property within Life?" To answer this question adequately, I pointed out that we only needed to cite the conditions that it be distinct from the "off" property, and that its patterns of instantiation exhibit the lawlike regularity prescribed by the rules. This schematic account entirely encompasses the categorical being of an "on" property within Life.

Over the last few chapters I have significantly enriched the minimal view of causal content explicit in the bare physics of Life. I have argued that causation has two aspects, the effective and receptive, and I have given a positive account of the relations between these aspects, detailing the distinct contributions they make to the causal character of a world. Despite this progress, our new picture remains limited by the fundamental shortcoming of the Life world. It seems we still have a merely schematic understanding of what it is to be a world with causal content.

Why do I say this? To describe a system with effective properties we only need to take three steps: (i) make names for the types; (ii) stipulate that the names must designate distinct entities; and (iii) describe how the presence of each effective property within a shared receptive connection contributes to the constraints on the overall state of the nexus. In such descriptions, the identity conditions on each kind of effective property involves the existence of the other kinds of effective properties. Implicitly, we understand each property through its stipulated distinctness from the others, and its ability to enter into constraining relationships with them. The kind of circularity involved is an instance of what I will call contrastive circularity. When a contrastive circularity exists, a set of distinct elements is stipulatively established, and then that stipulated distinctness is bootstrapped into a relational story that yields the rest of each element's identity conditions. For any kind of element X, we may answer the question "What is it to be an X?" by citing its distinctness from the other elements of the system, and the relational story involving it within that system.

Circularity also infests the concept of what it is to be a receptive or effective property more generally. When explicating the concept, I defined a receptive connection as a connection with at least one slot that opens its occupant to constraint by the effective states of occupants in other slots. An individual's effective state was explicated as a function of that individual's effective properties, where an effective property was defined as a property that, when instantiated by an individual sharing a receptive connection with other individuals, may constrain the possible states of those other individuals. So receptivity was not defined independently of the idea of an effective property, and effective properties are not defined independently of the idea of a receptive connection. It looks like receptivity presupposes effectiveness, and vice versa. The kind of circularity here is a compositional circularity. Each element is partially defined by the way that it presupposes the other as a positive component in its own nature.

Circular identity conditions, especially contrastive circularity, are hallmarks of schematic thinking. Reflection on such relationships naturally raises questions about how such circularly defined categories come to exist in a world. After all, if A is defined as that which is distinct from B, B is defined as that which is distinct from A, and no other categorical facts are true of them, it seems that the existence of each presupposes the existence of the other. The circularity has the logical feel of a vicious regress, and questions about how such properties get their footholds on existence in the first place forcefully present themselves. Understanding the ontological ground that these kinds of circularly interdependent properties need to exist is the chief task of this chapter.

14.2 Circularity instantiated

Circularity is puzzling, but not deadly. The world is full of systems whose elements instantiate circularly interlocking categories. The Life world itself provides an example, as people implement finite Life worlds in a variety of ways. The original Life world was implemented on a checker board using checker pieces, and the implementations most familiar to people today are computational. I think we can gain some insight by examining how a checkerboard can become a Life world.

How is the circularity in its relational definition satisfied? The secret is that the existence of Life is parasitic on the presence of properties whose categorical natures(1) are not wholly defined by the Life schema. These properties carry the distinctions needed to instantiate Life properties. For instance, red and black checkers play the role of the "on" and "off" properties in a checkerboard implementation of Life. What it is to be a red or a black checker is extrinsic, alien, to Life's physics. The color distinction between them, a distinction that is not intrinsic to the conceptual schema demarcating Life, carries the needed distinctness between the "on" and "off" properties. The checkerboard carries the geometry of Life's two-dimensional space, so we may understand the identities of the places on the checkerboard against their place in the larger geometry of the world. Finally, the intentions of the human beings manipulating the board guarantee the nomically reliable pattern of instantiation between the properties (to the extent that human intentions are nomically reliable). Once again, these human intentions have a nature that outruns the terms of the Life schema.

The form of the solution seems to be this. The circular properties of Life may exist if one can find something to carry the relations that each element of the Life schema must satisfy. Carriers are objects or properties whose natures outrun the categories of the schema, but which can enter into the appropriate relations with one another.

The same general solution transfers to the computational implementations of Life. In those cases, the physical distinctness of the machine states carries the distinctness between the "on" and "off" properties. The major difference between the computational case and the checkerboard case is this. The functional character of the tokens in the computational case allows them to directly carry the nomic reliability of the "on" and "off" properties, bypassing reliance on human intentions.

In each case, one naturally sees the carriers as having natures that are at least partially extrinsic to the Life world that they are carrying. For instance, one naturally views the specific distinction between the redness and blackness of the checkers as extrinsic relative to the generic distinctness required to be an "on" or an "off" property. The specific character of the distinctness is an additional fact about the world, beyond the facts about its being a Life world. This moral applies to more than just the facts about their distinctness. The causal contents of the elements of the system seem to transcend the causal powers attributable to them as "on" or "off" properties. This suggests a useful definition of what it is to be an extrinsic property, relative to a system the property occurs within.

P is an extrinsic property within a system S IFF P is instantiated within an instance of S, and P has a nature that is not exhausted by its relations to other elements as they are defined within S.

The idea of being an extrinsic property within the system S may be better understood by being contrasted with two other, related ideas. The first idea is that of being a property that is intrinsic to the system S. The second idea is that of a property that is intrinsic or extrinsic to an object o.

Properties of systems and properties of objects ­ First consider the idea of a property that is intrinsic to a system. One can articulate the idea in a positive or a negative form. Stating the idea positively: the properties intrinsic to a system are those whose identity conditions are grounded in relations between an object (or objects) in the system, and other elements of the system. Thus, properties intrinsic to systems presuppose the existence of the systems they exist within. Stating the idea negatively: the properties intrinsic to a system have no nature entering into their categorical being except those appearing within a system of the relevant type. In this sense, the categories of their systematic contexts exhaust their categorical natures.

The statements above may sound very abstract, but the following discussion should be able to show that the intuition they are trying to capture is not so elusive. To begin, the properties of being "on" and "off" are properties intrinsic to the system of concepts defining Life, as such properties presuppose the existence of a Life world, and vice versa. In contrast, being a red checker is an extrinsic property within some implementations of the Life world. Redness, while it exists within some implementations of such worlds, is neither dependent upon, nor definable by, the system of concepts constitutive of Life. The categorical being of redness, unlike the being of "offness", is independent of the conceptual scheme demarcating Life.

Second, the fact that a property is extrinsic within a system does not entail that it is extrinsic tout court, as it may be intrinsic to some other system. This is the case with computational states that may implement Life, as those states, qua computational states, are extrinsic within Life but intrinsic to the computational system they exist within.

How do these system relative notions of intrinsicness and extrinsicness relate to more standard notions dealing with properties intrinsic and extrinsic to objects? Jaegwon Kim (1982) has proposed that a property is an intrinsic property of an object just in case the object could possess it regardless of the properties possessed by other objects. Partly in response to criticisms by David Lewis (1983), Michael Dunn (1990) has proposed a refinement of Kim's suggestion by replacing the coarse modal condition by a finer grained notion involving relevant predication. Roughly, one may say that a property is intrinsic to an object if a change in the property requires a modification of the object.

Might properties intrinsic to systems also be intrinsic to objects, in Dunn's sense of intrinsic? Not if the property associated with being that object is also a property intrinsic to the system. For instance, the properties of being "on" or "off" are intrinsic to Life as a system, and also to Life cells, considered as objects within that system. Life cells are the objects in Life, but what is the status of the property: being a Life cell? It seems that it is a property intrinsic to the Life system, but extrinsic to its carrier, as an object may cease to be a Life cell without being modified. One need only change its context to make it the case that the carrier is no longer a Life cell. If that happened, it would then automatically cease to have an "off" or "on" property.

Does this leave the "on" and "off" properties as intrinsic to the objects, Life cells, or extrinsic? As an example, consider the checkerboard implementation of Life. A red checker on a square may represent the fact that the cell instantiated by the square is "off". If the checkerboard around it ceases to be used as a Life implementation, that square may cease to be "off" without having been modified. That seems to make being "off" an extrinsic property of that square, by Dunn's criteria. I believe we are forced to see the property of being "off" as intrinsic to the object, qua Life cell, but extrinsic to it, qua checkerboard square. Some may see this ambiguity as problematic. A question that would give voice to this qualm would be "What is intrinsic to the object itself?" I sympathize with this discomfort, but do not believe that it arises from problematic features of the notions of intrinsic and extrinsic properties. It springs, instead, from problematic facts about the notion of object. 'Objecthood' itself may not name an intrinsic property . Which object is the object "itself?" Despite their interest, I will leave such questions aside.

Extrinsic properties within systems ­ The examples canvassed earlier suggest that some kinds of (system relative) extrinsic properties may bring a system into existence. In doing this, one feature of the examples that seems particularly important is this. These extrinsic properties implement the stipulative contrasts within the system by having a relation that rests on an internal contrast between them. What do I mean by an "internal contrast"? A relation arises due to an internal contrast between A and B just in case the relation is such that necessarily, if A exists and B exists, then R(A,B), where R designates some comparitive relation.

Internal contrasts subsume the earlier stipulative contrasts that we observed exist within schemas. The most relevant difference between stipulative and internal contrasts is that the existence of an internal contrast between A and B may be a consequence of their natures, while a stipulative contrast is designated as a constituent in the contrasting element's natures. Therefore, entities whose type involves a stipulative contrast may not exist independently of one another, since the existence of either presupposes that the difference is instantiated. Entities whose type involves a non­stipulative internal constrast do not presuppose the instantiation of the difference relation, and therefore enjoy an independence from one another. Additionally, given an internal contrast that is not stipulative, as long as both A and B are present within a world, the relation R that holds between them may carry a stipulative contrast. Finally, all stipulative contrasts must be internal, but not all internal contrasts need to be stipulative.

For example, the redness and blackness of two checkers possesses an internal contrast that can carry the stipulative contrast between the "on" and "off" properties of a Life world. The contrast is internal because the distinctness between the colors necessarily holds as long as both exist. Also, the constrast between the colors is not stipulative itself because, as I argued in chapter two, we cannot reduce the natures of the phenomenal colors to a structure of pure difference relations holding between them. Finally, it is worth anticipating that R may be a more complex kind of relation than mere distinctness, although we have not surveyed any such examples yet.

Reflection on the variety of circular systems that actually exist in the world strongly suggests that each exists in virtue of carriers that are extrinsic to the system. For example, a chess game consists of a circularly interdependent set of types: pawns, rooks, kings, queens, etc. Each type is defined by the set of allowable moves it may make within the game as a whole. Without the context of the game, no particular type could exist. The circularity between their categorical natures makes it look as if the existence of each part of the game presupposes the existence of the game as a whole, which, in turn, presupposes the existence of the parts (Sellars 1963; Haugeland 1993). Why isn't this circularity deadly?

The reason chess games can actually exist is that each implementation takes advantage of external properties to introduce, piecemeal, the distinctions and dependencies that are defined whole cloth within the conceptual system. For instance, pieces that are recognizably physically distinct are used to stand in for tokens of distinct types. Board positions are defined relative to a physical space within which it has geometric relations to human players. These extrinsic distinctions and relations allow players to form intentions to move the pieces only in accord with the rules governing the types they will be tokens of. When each of these extrinsic factors slips into place in a way that carries the circular relations of the system, a chess game exists.

Notice how the existence of a chess game seems essentially dependent on these extrinsic properties and relations: were it not for extrinsic factors that have internal relations able to carry the circular identity conditions, the game simply could not exist. If the pieces were not physically distinct, the players would soon lose track of which piece was standing in for each type, and could not form the proper intentions to play them according to chess rules. If the board did not exist in physical spatial relations to the players, the pieces could not be set up in accord with the rules, and the players could not decide the legality of moves. In other implementations, like computer implementations, some other set of extrinsic properties always performs the carrier role.

Computer programs are similarly circular, as the nature of computational elements subsists in their functional relations to other elements within the system. As before, computer programs may exist by being carried by physical states that have natures extrinsic within the computational system. As we widen our view to other conceptual systems ­ economics, biology, and psychology are examples ­ we see the same pattern repeated.

In economics, what things count as goods and services? To a first approximation, goods and services are those things that consumers and producers barter. Who are the consumers and producers? Consumers and producers, in their turn, are people occupying distinct positions in the system of bartering for the goods and services.

In biology, organisms pass heritable characteristics through their genes. A heritable characteristic is one that parents pass from their generation to later generations. A parent, in its turn, is an organism that passes along its genes, or a significant portion of its genes, to the young.

In psychology, beliefs, desires, and perceptions are at least partially definable in terms of their functional role within the cognitive economy. An entity's functional role is its disposition to interact with other entities in the system.

In each case, a closed or semi­closed system of theoretical concepts, many of which are directly or indirectly circularly dependent on one another, exist. This circularity is hardly fatal, or even objectionable. The reason the circularity is innocuous is that, in each case, extrinsic properties exist within the systems, and these properties have internal contrasts between them. These contrasts help carry the circular relations. In economics, we can appeal to the desires, needs, and opportunities of individuals in the wider social system the economy is part of. In biology, we can appeal to the mechanics of molecular biochemistry. In psychology, we can appeal to computational or dynamical properties of neural systems, and the way these properties help the organism survive in an ecological niche.

14.3 The circularity of physics

Reflections on those kinds of examples lead one inevitably to wider and wider systems of concepts. In the case of the natural sciences, this expanding arena of circularly looping systems traces the same path as intuitive expectations of reduction. When we look at a circular system of concepts, we find that its instances are carried by objects with properties extrinsic within that system, but intrinsic to some other system. Inevitably, these other systems are themselves circular, partially or completely, and thus we find them carried by yet another set of objects with properties extrinsic within them. From economics we look to social relations of a broader sort, then from those to psychology, ecology and biology, then to chemistry, and finally to physics.

If this way of looking at these things is correct, these higher­level domains are not just in fact realized by the entities of some other domain, the domain of the carriers, but they have to be. The existence of carriers seems like an essential ingredient, a metaphysical presupposition, for the satisfaction of circularly interdependent kinds of being. This raises a serious problem when we hit physics.

When we reach physics, we find the same kind of circularity. The circularity in physics is easy to see by asking questions about the identity conditions on the basic physical entities, as these conditions are broadly functional. What it is to be a photon, for instance, is to play the functional role in our environment that photons play in physics. What it is to be charge, mass, or spin is to be distinct from the other physical properties, and to nomically instantiate the pattern of regularities prescribed by the laws (again, in our environment). What it is to be gravity is to play the role gravity plays, and similarly for the other basic physical properties. As a result, physics incorporates the same kind of circularity all functional systems of concepts incorporate.

When making this claim, I do not wish to deny the importance of indexicality (i.e., designation) in fixing reference. Likely, these concepts contain indexical components, as "electron" may express a rigid designator. Electrons are arguably just the natures that play the electron role in our world. The more important point is this. Even if some specific nature is picked out indexically by these concepts, the indexical place functions much like a variable in the conceptual structure. Even if the value of the index anchors the categorical system to natures, and even if it does so in a way that depends on the deictic orientation of the concept user within its physical context, it is still functional roles that do the most essential work in fixing the physical category applied to that nature. The indexically designated nature is a nature that is otherwise extrinsic to these entities, relative to the system of physical concepts we employ in science. That is, these natures, if they exist, are extrinsic within the system of physical properties they are carrying.

Also, I want to be clear that I am not claiming that any of these roles are rigidly defined. Concepts have fluid natures, and they always possess an elasticity that outruns specific formulations of the concept's meaning. I am only pointing out the fact that the role, while elastic, is not indefinitely elastic. And it is occupancy of this functional role that makes something the electron in our world, rather than the proton or the muon. In general, basic physical concepts are functional concepts.

So the problem arises: what is extrinsic within physics? What carries the contrasts and relations needed for physics to exist?(2) Perhaps the physicalist can take a hard line here, insisting that nothing carries the physics. The claim that a circular system may be instantiated without any extrinsic carrier is unprecedented and problematic. It is unprecedented in that the extrinsic properties in other systems are not spandrels. They are essential to the instantiation of those systems. It is problematic in that the resulting metaphysics seems unintelligible, if looked at too closely. The metaphysics requires a system of contrast, and relations between the contrasts, in which these contrasts are not carried by any internal contrast between natures extrinsic to the stipulation. It requires a notion of pure contrast, contrasts that seem not to be contrasts between anything. The champion of such a metaphysic takes on a large burden in trying to explain it. Unfortunately, the largeness of the circle in physics makes it easy to overlook the problem, or feel it less vividly, and current philosophy tends not to press the issue. I think the Life world makes the strangeness clearer (it is, after all, just a toy physics). Enlarging the circle makes the problem easier to ignore, but does not make it go away.

The puzzle is to find a set of properties extrinsic within physics. It will not help to appeal to some wider system of properties, or to circle back around in a constructivist way to society or to human psychology. Those maneuvers just enlarge the circle, presenting the same problem again. A proper solution will be one that short­circuits the puzzle, not one that moves it to a new arena. What the world needs are properties whose being is extrinsic within every such system, and yet still have the requisite internal relations to one another.

Perhaps surprisingly, the properties answering to this description are best thought of as properties that are intrinsic tout court. A property whose categorical nature is extrinsic within every system of properties is simply one whose being is intrinsic to itself, rather than to its contextual relationships. That is, it is a property that we cannot understand in purely systematic terms without leaving something out. The least strained way of understanding the physics of the world is to suppose that some kind of intrinsic property carries each effective property, where we understand intrinsic as intrinsic tout court, rather than intrinsic to a system.(3) Having come to this notion, perhaps it will help to summarize the kinds of properties I have discussed so far:

(1) Property intrinsic to objects ­ A property that one cannot modify without modifying the object itself.

(2) Property intrinsic to a system ­ A property whose identity conditions are given entirely by relations to other entities within some system.

(3) Property that is intrinsic tout court ­ A property that is not intrinsic to any system.

(4) Property extrinsic within a system ­ A property that is instantiated within an instance of a system, and that has a nature not exhausted by its relations to other elements as they are defined within that system.

14.4 From natural individuals to phenomenal individuals

One shortcoming of physics, I have argued, is that physics suppresses the receptivity of the world in its theorizing, and thereby leaves out its receptive structure. The addition of receptivity to physics, by itself, does not evade the circularity problem. The kind of causal mesh I have postulated in the last few chapters adds a kind of compositional circularity to the causal character of the world. By doing so, it magnifies the problems that arise from the mere contrastive circularity present in the specification of the effective properties alone. The effective and receptive aspects of natural individuals, together, yield their nomic content. The question now is, "What carries the nomic content of a natural individual?" The answer I am about to propose is the central thesis of the carrier theory of causation.

The Central Thesis ­ Every phenomenal individual in our world carries the nomic content of a natural individual, and every natural individual is a phenomenal individual.

I will defend The Central Thesis shortly, but first a few preliminary remarks. The ontology implicit in the Central Thesis is a neutral monism. The fundamental kind is the causal nexus itself, and the nexus has multiple aspects. It follows immediately from The Central Thesis that phenomenal individuals are not entirely physical, nor epiphenomenal, nor do they interact with the physical. A variety of panpsychism, as discussed in chapter seven, also holds if The Central Thesis is true.

Some strong reasons exist for adopting The Central Thesis. Whatever carries the nomic content of a natural individual must be something with the following character. It must possess properties that are (i) intrinsic properties that are intrinsic tout court; (ii) these properties must have internal contrasts with one another that can carry the requisite necessary relations; and (iii) each of these intrinsic properties must have a nature that is dependent on the nature of something distinct from it in the compositionally circular way that effective properties are dependent on receptivity. Condition (i) must hold because the carriers of physics must be ultimate carriers, carriers that do not enlarge the circle. Condition (ii) must hold because the carriers of physics must carry the relations specified by physical theory, and this means they will carry a set of stipulative contrasts, variations in magnitude, and various kinds of compatibility and incompatibility relations. Finally, condition (iii) must hold because the carriers of physics must carry non­Humean effective properties, and so these properties must exist by being bound to a shared receptivity within every causal nexus where they exist at all.

How many carrier candidates can there be? The phenomenal qualities of phenomenal consciousness meet the first two conditions fairly plausibly. They meet the first condition since one cannot understand what it is to be yellow, for instance, in terms of a system of relations. That is one of the lessons of the anti­physicalist arguments. This intrinsicness is plausibly what makes qualia the funny things that they are, and what makes full knowledge of them attainable only by acquaintance with them. Their natures seem intrinsic in the sense that they are not categorically constituted by the structure of relations they enter into.(4)

Phenomenal qualities also plausibly meet the second condition. They seem to have the needed kinds of internal contrasts, ones that can support the needed necessary relations. For example, distinct sounds exist such that, if each exists, then they are necessarily distinct types of sound. It seems like a small point, but it is very important. The difference between phenomenal qualities is grounded in the difference in their intrinsic natures so that, necessarily, if they exist then the differences obtain. However, as before, they are not definable in terms of this difference structure between them, so one cannot abstract it out as an account of their natures. The very intrinsicality of their nature means they outrun this description in terms of the difference structures they can carry. Their transcendence makes them perfect for the role of carriers.

Secondly, and this is also extremely important, phenomenal properties may enter into much more complex internal relations than mere difference or distinctness. In particular, it seems plausible that they can possess internal scalar relations. Scalar comparisons within (but not necessarily between) phenomenal groups like colors, sounds, tastes and so forth come naturally to us. For example, some sounds are louder than others, some colors brighter than others, and so forth. The most natural way to think of these groups is in terms of phenomenal spaces that they instantiate, with natural orderings of various types between the elements of these spaces. These scalar relationships at least suggest that some kinds of phenomenal properties, when they exist, can carry the kinds of quantitive variations required by physics. The idea is this. An internal contrast between phenomenal quality A and phenomenal quality B exists such that, when they both exist, necessarily A > B is true along some natural metric. If that is so, then A and B (and presumably other members of the phenomenal group they belong to) may carry the more complex kinds of contrastive relations required by physics.

Thirdly, one may plausibly assume that a variety of compatibility and incompatibility relationships hold between phenomenal properties, and possible phenomenal fields. The most straightforward kind of case like this is the postulated red/green incompatibility in our color space(5). One may also reasonably suspect that much more subtle and sophisticated kinds of compatibility restrictions hold for experience, restrictions that apply to whole fields or sub­fields of a phenomenal manifold. For instance, it is not clear that one could simultaneously experience the Necker cube as having face up and face down in the same visual manifold. If this restriction holds, then it is a very interesting sort of exclusion relation, one that incorporates the semantics of the conceptualization right inside the formation conditions on the qualitative experience.

On the other side, one might plausibly postulate that shapeless instantiations of color are impossible, so that the existence of a color property necessitates the existence of a shape. Or, on an even finer grained level, one might postulate that the existence of a hue necessitates the existence of a brightness value (no hue without brightness). In the gap between simple red/green incompatibility and the very subtle Necker cube face up/face down incompatibility, there might exist a whole host of subtle and interesting relations of exclusion, allowance, and necessitation between the possible experiencings of higher­level phenomenal individuals, or the possible proto­experiencings of lower level phenomenal individuals. In a sense, these relationships will be what "drive" causation, as it is these relationships that will carry the natural possibilities, impossibilities, and necessities of physics.

I want to be clear about the claim here. The claim is not that the phenomenal qualities of human consciousness exist at the microphysical level, carrying the effective dispositions of microphysical entities. Rather, the claim is like that canvassed in chapter seven. Alien phenomenal (or maybe more appropriately: proto­phenomenal) intrinsic properties exist, properties in a very abstract sense like the qualia of our own consciousness, that carry the effective dispositions of the world's basic natural individuals. The abstract sense they are like the qualities of our consciousness comes to precisely this: they are intrinsic tout court (in the sense defined above), and they are essentially experiential (for reasons discussed below). The ways they are different from the qualities of our consciousness are this: their specific characters are presumably entirely different from those of our own qualia, and the internal constrasts that hold between them organize them into very different kinds of phenomenal spaces.(6)

So the proposal is that phenomenal qualities carry the effective dispositions in our world. Unfortunately, we still have a problem. It is not enough to merely have some properties that are intrinsic tout court instantiated in the world, just lying next to one another in a Humean way. These properties must be effective, and that means they presuppose receptive connections as positive components in their own being. The further problem here is that effective properties do not merely contrast with receptivity. The relationship between the effective and receptive aspects of an individual must be more intimate, as an effective property may exist only if it incorporates at least one instance of receptivity as positive component in its own being. Here one has a kind of de re logical relationship between aspects of causation, as receptivity penetrates the being of effective properties, occurring as a presupposition in the very notion that they are effective.

The logical intimacy between effective properties and receptivity is not an extravagant luxury, as it plays an important metaphysical role. It is only through this intimacy that, via a shared receptivity in their very constitution, the effective states of different individuals may penetrate one another's being and present their constraints immediately. In a sense, having a shared receptivity provides a principle of substantial unity that activates the relations of compatibility and incompatibility between effective properties, making these internal constraints between them relevant in specific ways to specific cases.

The upshot of these observations is this. We must suppose that, if something is carrying the effective properties in our world, the effective carriers must be appropriately related to something that carries the role of receptivity in our world. How is this compositional circularity between effectiveness and receptivity carried? What is it that every phenomenal property presupposes as a component in its own nature?

I suggest that the ontological relation between phenomenal qualities and their participation in the experiencings of subjects matches the logical structure of the relationship between effective properties and their shared receptivity. Just as causation divides into the duality of effective constraints, and the reception of those constraints, phenomenal experience splits into the duality of phenomenal quality, and the experiencing of that quality. The concepts of phenomenal individual and nomic content share a logical structure in that (i) each divides into a duality; (ii) within each duality the two halves are logically fused, each necessarily presupposing the positive participation of the other half of the duality in its nature, so that (iii) one exists only by being positively dependent on the other. If this suggested isomorphism holds, phenomenal individuals qualify as carriers of nomic content. The Central Thesis encodes the empirical postulate that phenomenal qualities carry the effective constraints in our world, and the experiencing of those qualities carries the receptivity to those constraints.

The postulated mapping goes like this. The being of phenomenal qualities is to the nexus of experience as the being of effective properties is to the nexus of receptivity. Focusing first on the phenomenological side, it seems (at least to me, and I hope this is an intuition that is widely shared), that something's being a phenomenal quality implies that it may enter into the experiences of a subject. I find it implausible that kinds of pain exist that could not be experienced. If this is right, its possible role in experiencings is essential to pain. Similarly, to claim that something is an experiencing subject implies that it experiences phenomenal qualities. That is, its capacity to host and experience phenomenal being is essential to it. I would go as far as proposing that phenomenal qualities could not exist unless some subject was experiencing them, and experiences could not exist unless they were experiences of phenomenal qualities. Despite this mutual participation in one another's natures, they are distinct essences. A phenomenal quality is an object of experience, and should not be identified with the experiencing of it.

This intermingling of being between phenomenal qualities and experiencing exactly mirrors the compositional circularity between the being of effective properties and receptivity. Just as its possible role in a subject's act of experiencing is essential to a quality, so too is the possible role in constraining a nexus essential to effective properties. Just as being an experiencing subject implies that it experiences qualities, so too does being a receptive individual imply that it is receptive to effective properties. Just as phenomenal qualities are only potential unless actually being experienced, so too are effective properties merely potential unless actually receptively bound within a nexus. Just as experience is only potential unless it is experiencing the character of a phenomenal quality, so too is the receptivity of an individual only potential unless it is receiving the constraint of an effective property. And yet, despite this mutual participation in one another's nature, effective and receptive properties are distinct essences.

In addition to all this, as I argued above, it is not a stretch to speculate that relations of inclusion and exclusion, compatibility and incompatibility, exist between phenomenal qualities within a nexus of experience. Some supposition like this is obviously required to complete the account, as the idea that phenomenal properties are carrying the effective dispositions in our world requires such relations. However, all the requirement really comes down to is the postulation that there exist restrictions on the possible forms of experience. And, in the end, this is a very natural hypothesis.

After one gets over the initial shock of its counterintuitiveness, The Central Thesis seems to wrap up the puzzle quite neatly and fruitfully, and that is its best defense. It turns out that the causal nexus has three aspects: the effective, the receptive, and the carriers of its nomic content. The effective and receptive are the two complementary aspects of causation that give natural individuals their nomic content, and these two aspects are carried, ultimately, by the interlocked experiencing of phenomenal individuals within the causal mesh. Physical theory specifies only the effective properties of the basic natural individuals by describing the nomic regularities that hold between their instantiations. It designates phenomenal or proto­phenomenal properties that carry these effective properties. Physicalism is false because the information about receptivity and the carriers is absent from physical explanation: they are further facts about the world over and above those that physical science is committed to. Fortunately, we can draft phenomenal individuals into duty to help on both counts. Since we have independent reason to believe in the existence of phenomenal individuals, the account is as ontologically parsimonious as one could possibly hope for.

The position does have a price of admission, and that price is its implication that some kind of panpsychism is true of our world. Just how widespread experience is remains to be discovered, as the question of which individuals are the natural individuals is a substantial and important empirical question. However, as I argued in chapter seven, we should have expected to arrive at some kind of panpsychism, and it is not quite the evil thing it is commonly thought to be. I think of the panpsychism involved in the view as a benign panpsychism, as experience is likely to be very simple in the vast majority of cases, to be restrained to highly specialized circumstances despite its outrunning cognition, and to be unaccompanied by thought whenever it exists outside of cognitive contexts. Just as I discussed in chapter seven, both the "pan" and the "psychism" are misleading exaggerations. Putting these panpsychic implications aside, the final matter before us is just this. How does all this help with understanding human consciousness?

1. When I speak of "categorical natures", I mean the kind of thing conveyed by an appropriate answer to the question, "What is it to be X?" for the property of being X, or "What is it to be an X?" for the property of being an X.

2. This constitutes yet a third argument against physicalism, distinct from the failures of consciousness and causation to logically supervene. The facts about the natures of the carriers do not logically supervene either.

3. Related arguments from the circularity/schematic nature of the physical to this conclusion are in Fales (1990), pp. 219­220 and Chalmers (1996), pp. 303­304.

4. Sometimes the illusion that colors (for instance) may be "defined" in terms of the relations between them is achieved by importing a color or hue into the system as a primitive, and using it as an anchor for the relational system. This sort of attempt is like importing zero into an axiom scheme as a primitive object, and then attempting to define the natural numbers through a system of relations to it and each other. Zero anchors the structural relations in number space just as a color may anchor structural relations in color space. The importation of an anchor in this way precludes reduction. For instance, when we consider applying the axiom system to sets like the supernatural numbers, the irreducible role that zero is playing in supplementing the structural story becomes apparent. The very need for such anchors shows that we are not reducing the elements in the space to their structural relations alone.

5. Kneale (1949) also uses red/green incompatility as an example of a de re incompatibility between properties, proposing that it might serve as an analogical model for relations between physical properties. My suggestion in the text goes beyond Kneale in several ways, chiefly in taking the panpsychist step of suggesting that they may serve as more than merely a model, and also by proposing the more subtle variety of relations I discuss in the text.

6. This discussion is making partial payment on the promissory note at the end of chapter seven.


Chapter 15

The Consciousness Hypothesis

15.1 Consciousness and high­level individuals

On the sea of individuals, human beings are tidal waves. At least that is the most natural conclusion if one accepts The Central Thesis as the best solution to the carrier problem. The Central Thesis naturally leads to The Consciousness Hypothesis, which is,

The Consciousness Hypothesis­ Each individual consciousness carries the nomic content of a cognitively structured, high­level natural individual.

Arriving full circle ­ The Consciousness Hypothesis brings us full circle. When this book began we were concerned with consciousness, and the problems it presented for our physical understanding of nature. Prolonged reflection on those problems motivated our taking a closer look at causation. Using the apparent problem of epiphenomenalism as our point of departure, we treated causation rather independently of the problem of consciousness, first rejecting the Humean view and then exploring a substantive alternative. Our exploration of non­Humean causation led to the carrier problem, whose solution seemed neatly captured by the postulate that phenomenal individuals carried the nomic content of natural individuals. The analytical flow of our discussion has taken us from consciousness to causation, and then, independently, back to consciousness again. The purpose of this chapter is to tighten the circle a bit by discussing directly how The Consciousness Hypothesis, as a byproduct of Liberal Naturalism, avoids the explanatory failings of pure physicalism.

The chief problem consciousness presented physicalism was its failure to logically supervene upon the physical facts. This failure corresponds to an explanatory gap that has to be interpreted as an ontological gap, since the ontological posits of science are embedded within its explanatory frameworks. Within this version of Liberal Naturalism, it is instructive to see how the more detailed analysis of causation into effective properties, receptive properties, and carriers does and does not evade the kinds of arguments used to undermine physicalism.

Chalmers has nicely summarized the five extant arguments against the logical supervenience of consciousness on the physical. They are (i) The possibility of inverted spectra; (ii) The logical possibility of zombies; (iii) The epistemic asymmetry between facts about consciousness and other facts; (iv) The knowledge argument; and (v) The absence of analysis. I will consider these arguments, as they impact Liberal Naturalism, in reverse order.

The absence of analysis ­ The argument from the absence of analysis is this. For a property to logically supervene on the physical, it must be at least roughly analyzable into categories that will allow the physical facts to present logically sufficient conditions for application of the category. But the anti­physicalist argues that we can give no adequate analysis of phenomenal properties in purely functional and structural terms. Since these are the most directly relevant categories available within the physicalist framework, it seems that consciousness must not logically supervene on the physical.

Within Liberal Naturalism, the primitive carriers constitute a kind of property that is not just structurally or functionally specifiable. Therefore, we can see directly that The Consciousness Hypothesis is not proposing an analysis of consciousness entirely in terms of structure or function. The analysis is that consciousness is a type of primitive carrier of nomic content, where primitive carriers must contain properties that are intrinsic tout court (i.e., not intrinsic to any system, in the technical sense discussed last chapter). These intrinsic properties must have a kind of existence characterized by a circularly structured interdependence on a distinct kind of entity. This other entity acts as a nexus binding them to other carriers. Finally, The Consciousness Hypothesis does not claim that this is all there is to know about the carriers, as the unanalyzibility of a specific carrier's intrinsic character implies that full knowledge of its categorical being, if accessible at all, could come only through acquaintance.

Where does the argument from analysis stand, given the analysis above? Modulo whatever information is available only by acquaintance, the analysis seems to capture everything important about consciousness, including the existence of information available only through acquaintance. The nonreductive character of the theory is represented by the axiomatic status of The Central Thesis, a thesis whose acceptance is justified partly on the grounds that it uses the missing information that must be acquired by acquaintance.

The knowledge argument ­ The knowledge argument was that a person with perfect physical knowledge could not, even in principle, use that basis of physical information to derive the phenomenal information available in experience. It follows that the phenomenal information is extra information, over and above the physical information.

From the conclusion that full knowledge of a carrier requires acquaintance, we can see that the knowledge argument fails when applied to The Consciousness Hypothesis. Imagine Mary is a brilliant neuroscientist locked in a black and white room, and that she wishes to know everything there is to know about causation in our world. Her knowledge would have to include knowledge of the receptive structure of our world, and of the carriers. She could conceivably theorize about the receptive structure given the effective facts of natural science, great talent with inference to the best explanation, and a little luck. From this, she could infer the structure of internal contrasts that hold between the carriers. However, she would still be missing some facts about the carriers, the facts corresponding to their intrinsic character. Hence, she would not have all the facts. The only way to get these facts is acquaintance with the relevant intrinsic natures that fill the carrier role in her world. Hence, The Consciousness Hypothesis predicts that for Mary to have all the facts she would have to have some further experiential facts about intrinsic natures.

Epistemic asymmetry ­ The argument from epistemic asymmetry is this. Our reasons for believing in the physical facts, or other facts logically supervenient upon them, are straightforwardly based on external evidence. If consciousness logically supervened on the physical, external evidence would give us reason to believe in it. However, the external evidence does not give us adequate reason to believe in something like consciousness. Our only reason for believing in consciousness is our direct acquaintance with it.

This argument does go through against The Consciousness Hypothesis, as we obviously arrived at the need for carriers based on speculation about "external" facts concerning causation. However, on its way through it gets defanged, as its strength depends on what is meant by the external facts giving us no reason to postulate "something like" consciousness. This failure, while real for Liberal Naturalism, is attenuated in a way that saves it from being especially damaging.

The key problem the argument presents for physicalism arises because physicalism claims to be a completely reductive account of consciousness. The Consciousness Hypothesis does not occupy a reductive role in Liberal Naturalism. The Consciousness Hypothesis is an instance of The Central Thesis, which has the place of an axiom in the theory. The analysis that justifies introducing The Central Thesis claims to show only that the ultimate carriers must be intrinsic properties tout court, and that they have a categorical structure that logically mirrors that of phenomenal experience (we will say more about the consequences of this in section two below). The analysis in no way suggests that the carriers are reducible to something else. In this sense, the external facts postulated by our analysis do give us reason to believe in the existence of "something like" consciousness in important respects.

In the next section of this chapter I will make this point even more strongly, arguing that the similarities extend much further. The similarities between phenomenal individuals and the carriers of nomic content are so deep that they should potentially allow us to find satisfying solutions to the puzzles and paradoxes concerning the unity of consciousness, the subjective instant, our knowledge of consciousness, its superfluity, the grain problem, and the boundary problem. Ultimately, the only facts that should escape the Liberal Naturalist explanation are those about the specific characters of the qualities, and an explanation of why phenomenal individuals, rather than something else, should carry nomic content (if there really is an alternative). These open questions are adequately respected by the non­reductive component of the theory. So, to the extent that conscious experience escapes the explanation, it is not claimed to be included in it. Since the force of the argument from asymmetry is just that any facts failing to evade it will have to be included non­reductively, The Consciousness Hypothesis meets its demands.

The logical possibility of zombies ­ A zombie is a being which is physically identical to you or I, but that has no conscious experience. The argument for the logical possibility of zombies is this. The fact that it seems consistent that nonstandard realizations of our functional structures might not be conscious seems to imply the same thing for even our own, organic physical structures. There is, after all, no more of a conceptual connection from organic chemistry to consciousness than there is from other sorts of physical structures to consciousness. This strongly suggests that a zombie world is consistently conceivable, and, therefore, logically possible.

Notice that the argument for the possibility of zombies requires an inference from the conceivability of a zombie world to its logical possibility. One strategy for resisting a zombie argument is to argue against the conceivability of the world. That strategy is available to the Liberal Naturalist who endorses The Consciousness Hypothesis. Within the current framework, we may argue away the apparent conceivability of the zombie world as follows: it is either a world with a different receptive structure than ours, but an indiscernible physics (e.g., like the two Life worlds), or a conception that negligently overlooks the carriers.

To see this, consider what an attempted Zombie argument against the carrier theory must accomplish. The argument must establish the conceivability of a world in which all the facts about the effective and receptive structure of causation are the same, but in which there is no consciousness. This is equivalent to the claim that one can conceive of carriers that are not phenomenal individuals. However, the above discussion of the knowledge argument makes it plain that no one can really conceive of alternative carriers in a sense definite enough to make the case that such a world is a conceivable world. In particular, a conceivable world is logically possible just in case its conception is consistent when the intensions on the concepts have been made suitably definite.

I believe that one must have acquaintance with the appropriate intrinsic nature before one may form a definite conception of a carrier. Consider: did Mary have a conception of what red is, before experiencing it, that was definite enough to conceive of its presence in various possible worlds? It seems doubtful to me, as her concept was indirect, and had content only deferentially. Such a concept certainly couldn't be applied to the case of alternative carriers, as no one exists to defer to. Certainly, given the strict requirements on what carriers must be like, the burden is on someone who claims to be able to conceive of carriers that are not phenomenal properties participating in the experience of phenomenal individuals. They at least must be able to convey to other people what they have in mind. In making my case for The Central Thesis, I was able to appeal to something ­ phenomenal individuals and phenomenal properties ­ whose existence we are acquainted with. In the Zombie case at hand, even such first­person ostension is unavailable as a basis from which to bootstrap the conception, and it is very unclear how else one might get the requisite concepts. One might say that, in a sense, carriers get inside the concepts that we may have of them. As a rule, no definite conception of a particular kind of carrier can exist without acquaintance with the carrier itself. Therefore, it seems that we cannot successfully conceive of worlds with alternative kinds of carriers, and so cannot conceive of a Liberal Naturalist zombie world.

These considerations about what it means to conceive of a carrier are enough, I believe, to refute arguments that purport to establish the possibility of a Liberal Naturalist zombie world. Even so, arguing that we must fail in trying to conceive of a zombie world is not the same as arguing successfully against the possibility of such a world. The question remains: could there have been carriers that are not phenomenal individuals? Conceivability cannnot deliver an answer one way or another, and it seems to be an open question.

In our conservative moments, we may lean towards the negative. Invoking Occam's razor, we may note that the only things we know of that answer to the description of ultimate carriers are phenomenal individuals, and we should therefore conclude that these are the only things that could answer to it. Perhaps this is the simplest answer, as it avoids opening up any further explanatory questions.

In our humble moments, we may admit that we cannot rule out the possibility of alternative carriers, and should, charitably, allow them. In doing so, we treat the phenomenal facts about our world as contingent and "implementational" in character. If that is the case, they turn out to be inessential to causation, in a certain sense, as they could be substituted for without changing either the effective or receptive structure of the world.

In our sober moments, I think we should just be agnostic. We should acknowledge that this is a question about the world­making ingredients God might have had available in the jars of his kitchen cabinet. In the face of such awesome questions, we should simply turn away.

The inverted spectrum ­ The inverted spectrum arguments are analogous to the zombie arguments. The inverted spectrum argument is that we can consistently conceive of a world where the physical facts are the same, but in which color perception is systematically inverted. For instance, if a creature sees in black and white, we can consistently conceive the blacks being switched with the whites.

Inverted spectrum cases, in their original context, are meant to work against reductive accounts of consciousness. The Consciousness Hypothesis represents a non­reductive hypothesis, as both the effective and receptive facts logically supervene on the facts about the carriers. The target facts are the facts about effective and receptive causation, and the reduction is to facts about experiential and proto­experiential subjects. Standard inverted spectra cases, if they go through, just show that states of effective and receptive causation may be multiply realized. That conclusion is metaphysically harmless to The Consciousness Hypothesis and Liberal Naturalism, and we can avoid the epistemological danger it presents by appealing to simplicity constraints.

An argument harmful to The Consciousness Hypothesis would need to show that the phenomenal facts cannot play the reductive role they need to in the carrier theory. Notice that, by hypothesis, phenomenal individuals carry nomic content via internal contrasts within and between elements of phenomenal spaces. A truly harmful argument needs to show that phenomenal spaces could not instantiate the requisite necessary relations, and so could not act as carriers in the first place.

That claim is every bit as substantial as The Consciousness Hypothesis itself. It is the very strong claim that the phenomenal facts about the world could be just the same, while the (supposedly) supervening causal facts were different. No argument to this effect has appeared in the literature, and it is not clear how one could be made. It needs to invoke a conception of the totality of phenomenal facts that is more complete than any we seem to have available.

15.2 The puzzles for Liberal Naturalism

In part two of this book we explored a variety of puzzles, paradoxes, and tensions that the rejection of physicalism seems to lead to. The first tension that we explored concerned the fact that some sort of panpsychism was the most likely outcome of the Liberal Naturalist turn. We then explored six different puzzles and paradoxes: a puzzle about the unity of consciousness, a paradox about the simultaneity of the subjective instant, a puzzle about the epistemology of consciousness, a puzzle about its seeming superfluity, the paradox of the grain problem, and the boundary problem for phenomenal individuals. In this section I return to these problems, suggesting ways that The Consciousness Hypothesis may help to illuminate or resolve each of them. The end result is an explicit articulation of a host of striking parallels between consciousness and whatever it is that must carry receptivity and the receptive field.

This is more than just an exercise in solving the mystery of consciousness. It should also help to cement the case for the existence of receptivity, and the truth of The Consciousness Hypothesis. After all, what have we learned about receptivity? The receptive connection is a non­physical aspect of causation, an aspect which is metaphysically crucial to the dynamical structure of the world without being an explicit element of physics. To some extent, the previous chapters have merely gestured toward the idea by giving the logic behind the role of receptivity. This logic itself is abstract, and unusual. Some readers may have a gnawing sense that they still do not have a good idea of what a receptive field really is. The basic discomfort may be that, despite the formal explanation, receptivity and receptive fields still seem somehow alien to our experience of the world.

Given all the information about the physical make up of the world, we could still understand how all kinds of high­level structures would arise, and we could describe the "dynamics" of the "processes." In fact, one could throw in the laws of nature as bare counterfactual constructs, constructs with a logical but not natural basis, and reproduce almost the entire everyday catalog of facts about the world. In this sense, the receptive field is explanatorily irrelevant relative to the physical facts. It becomes relevant only when one is concerned to trace out a naturalistic understanding of the causation that produces the cohesion and identity of the physical processes, creating a more intricate internal structure within nature.

Naturally, one wonders a bit just how to concretely conceive of the kind of entity that I have proposed we add to our view of nature. At the point that understanding by analysis gives out, one would like some reassurance that understanding by either acquaintance or analogy may kick in. Such understanding may lessen the sense of alienness. For instance, when discussing the possibility of panpsychism, I introduced something I called 'the qualitative field'. Two chapters later I refered to the field of experience for phenomenal individuals. Two chapters after that, I introduced the idea of the receptive field of natural individuals. If The Central Thesis is correct, then these are three names for the same thing. I am now going to argue that this consolidation enables us to shed light on the puzzles articulated earlier. Afterwards I will trace the parallels between consciousness and the receptive field in explicit overview.

Ubiquity and fundamentalness ­ Our best reflection led us to believe that the qualitative field, whatever it turned out to be, would be something that (i) would help us to "get under" physics by finding a way to see the physical and experiential as co­equal aspects of a deeper kind; (ii) would be ubiquitous; and (iii) would be fundamental. The construal of the causal nexus formulated in The Central Thesis meets all three expectations: it "gets under" physics by postulating a deeper kind, it shows experience or proto­experience to be ubiquitous, and it explains why experience should be fundamental.

First, the idea of carriers is a way to get under physics because it yields a concrete causal structure and intrinsic character to a world that is incompletely and schematically described by physics. The intrinsic natures in the field are designated by physical predicates, as they carry the effective properties. Their existence presupposes an equally fundamental receptive connection that binds them into an experiential nexus, bringing into play a variety of internal relations between them. Neither "arises" from the other. Instead, they mutually suppose and support one another, cooperatively implementing causation. Second, natural individuals will not correspond only to cognitive systems, although just which kinds of things are natural individuals is still an open question. Whatever the outcome of those investigations, experience will end up being surprisingly widespread. Thirdly, and finally, specific carriers are introduced as instances of The Central Thesis, an axiom of the theory, and are therefore fundamental.

The unity of consciousness ­ I found the intuition behind the unity of consciousness to be difficult to pin down. My best attempt to reconstruct the problem was this. It did not seem as if the experiential elements of consciousness were proper components of consciousness, as their existence seems to presuppose the existence of the experiential manifold they are elements within. The unity of consciousness, I suggested, could be explicated as the claim that the existence of each element was holistically dependent in this way on the existence of a common whole. I ended that discussion by saying that the problem of unity presented us with both a challenge and a puzzle. The challenge was to articulate what the unity of consciousness is in a clearer manner. The puzzle was to explain how a single system may appear to both have and not have this kind of unity.

By answering the puzzle first, The Consciousness Hypothesis may also address the challenge. The most natural way to approach the puzzle is by appealing to the way that the instantiation of an effective property depends upon its receptive context. Recall that an actually existing effective property is identified with the constraint that it imposes within causal nexi. This identity condition is a result of the compositional circularity in the categorical being of effective and receptive properties. Outside of the receptive context, therefore, no effective property exists (although a potential to be effective may exist). In short, the receptive context brings the effective property into existence, and so an effective property may exist only within its receptive context. Upon reflection, this kind of compositional circularity seems to mirror the kind of unity we are after.

Since effective properties do not exist outside of a receptive context, no carriers of effective constraints would exist outside of a receptive context. If The Consciousness Hypothesis is correct, then phenomenal characters carry the effective properties in our world, and the phenomenal field of experience is carrying the receptive field of a high­level individual. It follows that no phenomenal character exists, or could exist, independently of the experiencing of a whole individual. They are brought whole into existence, and leave existence, with the individual manifolds they help to constitute.

The observation above helps to explain the first part of the puzzle, namely, why a system may have the kind of unity that consciousness seems to have. The second part of the puzzle involves explaining why it would seem to also not have that kind of unity. This second part of the puzzle may be solved by noting the partially reductive character of the hierarchy of individuals in the causal mesh. Although no individual is wholly reducible to the lower level individuals bound within it, as each receptivity is unique to it, each higher level individual is also constituted, as it may bind a complex multitude of lower­level individuals that do not depend on it for existence. From the perspective of another system, the high­level receptivity of the individual would be far less striking than the hierarchy of effective individuals it directly and indirectly binds. Particularly, the members of this hierarchy will seem like components.

Together, these observations promise to solve the puzzle. The non­reductive element of the individual, its receptivity, is responsible for the system possessing the appropriate unity. A receptive connection brings effective constraints into existence by providing an appropriate context for them to operate within. The reductive element of the system, its hierarchy of lower­level individuals, make it appear to be a component system. The challenge has been met also, as the kind of unity at issue may be more clearly understood on the model of the compositional circularity holding between the receptive and effective aspects of an individual.

This observation about unity connects directly to the previous speculation about a possible connection between the unity of consciousness, and functional elements within a functional being. The connecting point is that the totality of constraints an individual may present within a nexus corresponds directly to the notion of a functional role within a canonical context. Relative to their natural context ­ the higher­level individual ­ bound elements instantiate functional being. Their effective states, being identical to the effective constraints they present within the nexus, are instantiations of functional states. As such, they only exist relative to their canonical context.

The subjective instant(1) ­ The problem of the subjective instant arose because it seems that the conscious subject occupies a kind of privileged reference frame in which conscious events are all occurring simultaneously. Yet we know that conscious events correspond to asynchronously occurring brain events, events for which there is no privileged reference frame. To sketch a resolution to this puzzle, I need to use some claims that I will support in the next chapter's discussion of space and time. At this point, I will have to ask the reader to take them at face value, at least as provocative claims with an interesting application here. First, each element of a nexus is copresent within it, where this copresence is a kind of immediacy of presence. Second, spacetime is likely not primitive, and we should expect it to be reducible to more fundamental facts about the world's receptive structure. Finally, the facts about space and time are relative to a given frame of reference, and the nature of the causal mesh means that the frame structure is layered. It is spread both vertically and horizontally.

When one elaborates on the structural observations, it turns out that choosing a frame of reference to measure time is equivalent to choosing an individual in the mesh. The vertical structure of the mesh arises from the fact that it is layered: individuals at lower levels become bound by receptive connections into individuals at higher levels. Within this vertical scheme, the state of a higher­level individual has a dual character. It is, at once, an immediate single state determination for the higher­level individual, and a multiplicity of state determinations for the lower level individuals bound within it. Projecting this situation into a coherent spacetime scheme may very well require that an instantaneous state for the higher­ level will correspond to a duration in the existence of the lower level individuals.

To justify this last assertion, I need to ask the reader to consider a causal process like that depicted in figure 15.1. The process connects a series of higher­level individuals i1, i2, and i3, and possesses the following structure. It consists of two unidirectional connections (constituting individuals i4 and i5 at a higher level). One unidirectional connection connects i1 to i2, and the other connects i2 to i3. In notation we would represent them as i1[i2] i4, and i2[i3]i5. Additionally, assume that each individual i1, i2, and i3 is a polydirectional nexus of the form [...ij...]ik , and remember that the overlaps between them ensure that i1 shares some of its members with i2, which shares some its members with i3.

Each individual's state is an instantaneous state for it. It seems, then, that an individual's own state determination will be instantaneous from the reference frame of that individual. This is as true for the states of the individuals i1, i2, and i3 as any other individuals. From the reference frame provided by each, the state determinations of its bound ij elements would have to occur simultaneously, as no internal duration exists to spread them across (durations require unidirectional connections). Moreover, i1, i2, and i3 would form a stream embedded within i4 and i5. From their perspective, a continuous transition would exist, with the state of i1 overlapping with and flowing into i2, and i2 overlapping with and flowing into i3.

Finally, a full articulation of the internal causal structures involved here would also include the structures of the bound ij's. When fully articulated, this could be very complex, involving many, many layers, with corresponding relations of orderliness and fixity. For example, the ij's themselves could be unidirectional nexi (i.e., of the form ih[ig]ij), and possess yet further internal structure. How would such structure be projected into spacetime from the perspective of another nexus? Such complicated, layered individuals could only be represented by projecting temporal and spatial separation between their component individuals. In this way, an internally instantaneous state of a high­level individual may correspond to a duration when this state is projected into spacetime from another frame of reference.

This fact, although counterintuitive, is a consequence of the non­reducibility of higher­level individuals to their components, along with the relativization of spacetime. In short, the solution I am suggesting is that time may be relativized to this extra, vertical dimension of reality. The new relativization is a relativizing of instants, and of what counts as instantaneous. As we move upwards in the hierarchy, states of the individuals become more and more internally complex, and therefore more difficult to compress into instants from other frames of reference. Even so, the receptive nexus of the higher­level individual does provide a privileged frame of reference for the experiencing subject. Its state determinations, from its own point of view, are instantaneous, and require simultaneous state determinations of at least its immediate components.

Such internally complex, high­level individuals may appear differently when one attempts to construct time from other frames of reference. To that individual, were it cognitive, a paradox would result if it tried to internalize an "objective" construction of time, viewing itself "from the outside" without realizing the relative character of instaneous states. In essense, as it became aware of its many layers and tried to find temporal room for those layers, it would find itself losing the instantaneous character of its own state, internally observed. One might think of it this way: as one tries to assimilate the evolution of all layers into a single temporal framework, the internal complexity of the higher­levels forces one to project their instantaneous states onto durations at the lower levels. As one artificially flattens the mesh, the individuals involved in higher­level processes become squeezed, and take on temporal width.

The knowledge paradox ­ The paradox involving knowledge rested on an argument that consciousness, if it fails to logically supervene upon the physical, must be epiphenomenal. Epiphenomenalism was interpreted as causal irrelevance, and the work in the last few chapters has overthrown the intuitions leading to the paradox. Consciousness is no longer, in any sense, irrelevant to causation. We have landed on the other end of the spectrum, now seeing consciousness and experience as the essential elements that carry causation in our world. Some might claim an a posteriori identity between states of consciousness and brain states, just so long as we are careful to remember that the nature of brain states is richer than the physical facts alone convey.

Despite this achievement, a residue of skepticism may remain based on an intuition that crucial aspects of experience, involving receptivitity as it does, are not causally responsible for our brain states. One might have a nagging feeling that causal responsibility hangs only off the effective side of things. In a certain sense, even the qualities that carry the effective dispositions may be viewed as extraneous in that they can be omitted from adequate explanations of behavior. This residue of skepticism can be best diminished, I think, by taking a closer look at the notion of causal responsibility it uses. The question of what kinds of things may claim causal responsibility in our world is subtle.

Imagine that Trey has a date with Carol, and that she is supposed to meet him for dinner at Everybody's Pizza on Wednesday. When the time on Wednesday rolls around, Trey is there waiting for Carol, but Carol fails to show. Trey sits at his table, waiting, and as time goes by he realizes what has happened. At first he feels disappointed, and as he reflects on it he gets angry. Then he leaves.

The question I want to consider is what, in this situation, caused Trey's feelings of disappointment and anger? I suggest that, at least at first blush, the cause of his feelings was the fact that Carol did not show up. Notice what a strange sort of thing that is. It is a negative fact. It is nothing that Trey's brain states interacted with. It was not even an event, unless there are such implausible things as negative events that make negative facts true. The thing causally responsible for those emotions, it seems, was an absence. This at least seems to suggest that causal responsibility may accrue to things other than what most philosophers of mind commonly assume. I am raising this as a puzzle case, and will not argue for a final conclusion on the question of whether negative facts can be causes. Instead, I am going to consider each answer in turn. I will argue that either position one takes on this case undermines the contention that knowledge of consciousness is specially problematic.

Case one: We allow the fact that Carol did not show up to be the cause of Trey's disappointment and anger. If this is true, it shows that causes need not be events, and need not be physical. At least in some cases a cause of an event may be a structural fact about the world involving abstract objects such as the closure conditions necessary to draw negative conclusions. The world's receptive structure, I submit, is just a structural fact of a different sort. The fact that the brain states have the receptive structure and carriers that they do, and no other, is a structuring cause of our behavior. This relationship, I submit, should be intimate enough to be justificatory. At the very least, we need a solid argument that the receptive facts, as causes, are more problematic than negative facts are. Only then do we have reason to worry that facts about consciousness cannot be justificatory.

Case two: We deny causal responsibility to the fact that Carol did not show up. In its place, we must substitute something like the presence of the unfulfilled expectation. However, the negative fact still shows up in our story as the reason that the expectation went unfulfilled. This answer commits us to the position that reasons do not need to be causes.

Notice that the negative fact, as a reason for Trey's feelings, can still play a justificatory role in establishing them as proper in his situation. At the first step they still help to explain his feelings, and, at the second step, they are crucial to establishing those feelings as proper in the overall situation. It follows immediately that the possesion of certain mental states may be justified by facts that need not be causes of those mental states, either directly or remotely.

When we turn to consciousness, the facts about receptivity and carriers will certainly count as reasons for our mental states. After all, even if they do not accrue causal responsibility, they are crucial components supporting our total causal situation (e.g., had these carriers and receptive facts not existed, these mental states would not have existed, either in their physical or phenomenal aspects). As such, they form part of the full explanation of the existence of our beliefs, occurring as reasons for our having the brain states that we do. As reasons, the presence of that structure and those carriers should be able to play a justificatory role in the full story of why we have the mental states that we do.

So it seems that each case promises to leave some wiggle room for us to have knowledge of consciousness. My purpose here is just to deflect the objection by answering the challenge raised by it. I think the problem of giving a positive account of the epistemology of consciousness, like epistemology generally, fast moves into murky area. The discussion above should be sufficient to show that consciousness presents no special problems due to causal irrelevance, but one may still wonder just how the positive story might go.

I do not know exactly how the discussion should play out, but I will try to trace out the structure of the problem in a way that might be helpful. The first step in a proper solution requires recognizing that we have a third type of knowledge, over and above the propositional knowledge expressed by "knowing that..." clauses, and the skillful knowledge expressed by "knowing how..." clauses. This third kind of knowledge is a kind of empathic knowledge, expressed by "knowing what..." clauses. Examples of this kind of knowledge are: knowing what it is like to hear a scream, knowing what it is like to feel an itch, knowing what it is like to have something on "the tip of your tongue", knowing what it is like to be in love, and so forth.

This "knowing what..." is a kind of knowledge by acquaintance. As such, it is not truth evaluable, just as skills are not truth evaluable. It is a basic way of being for conscious subjects, likely presupposed by the other kinds of knowledge. The epistemic puzzle for consciousness does not concern how we may have knowledge in the "knowing what..." sense. This knowledge is knowledge of the basic causal nature of the particular that we are. It is acquaintance with our own nomic content, and is available to us because of the immediate nature of the shared receptive connection that brings consciousness into existence.

The puzzle concerns how this "knowing what..." justifies instances of propositional knowledge. How does the intimacy of acquaintance license the uttering of sentences? This is a deep puzzle for epistemology generally, I believe. I would guess that its solution lies in giving some epistemic value to the particularity of a creature's circumstances, and trying to understand the variety of ways that particularity is responsible for that creature being what it is, in the states that it is in. Beyond this, I am not sure what to say, and am willing to leave the task of completing an adequate positive theory to the epistemologists.

The superfluity of consciousness ­ The problem concerning the superfluity of consciousness concerned the challenge that an epiphenomenal consciousness would pose to scientific realism. If we know that one causally irrelevant entity exists, why should we have faith in Occam's razor? Since carriers are required for causation, the problem of superfluity is solved directly.

The grain problem ­ The grain problem contrasted the homogeneity of experiential qualities with the structural complexity of their supposed physical basis. In the earlier discussion, I suggested that the problem might be soluble by carving off the functional being of the events from their physical basis, focusing only on their level encapsulated causal role within a canonical context.

The view of causation proposed here accomplishes this feat. The hierarchy of natural individuals define the levels of nature, and each natural individual provides a canonical context to refer to in characterizing functional being. The functional being of an entity, within a natural individual, corresponds to the set of effective constraints it may instantiate within its context. Remember that effective properties are non­reductively defined, and, in particular, should be identified with the constraints they present within the nexus.

The formal idea was that the receptive field instantiated a function that mapped prior effective states onto constraints, which were called the posterior effective states. The effective property the individual had within the nexus was then identified with the posterior effective state, so that two structurally different states that presented the same constraint mapped onto the same effective property within the nexus. The differences in internal structure were lost, and, ultimately irrelevant.

The Consciousness Hypothesis maintains that phenomenal individuals carry the nomic content of natural individuals, and that phenomenal qualities (or their proto­phenomenal analysans) carry the effective properties within a nexus. These effective properties are the instantiated constraints. Given this identity, what carrier facts should we suppose correspond to this fact about prior differences mapping onto identical constraints within a nexus? It seems natural to suppose that, for each distinct effective property, there is a single carrier, but no more than a single carrier. If this is so, each component state within a phenomenal nexus must map to some specific carrier according to the constraint it presents. The carrier, then, could not be structured to match the internal structure of the component's state since it must stand ready to carry the constraint presented by a variety of different component states, each of which may be structurally different. It follows that carriers would reflect the functional being, and only the functional being, of component states. The finer grained structure should be filtered out.

Here is another way to see one's way to this same conclusion. Higher­level individuals account for immediate interactions, and they have internal structure. The internal stucture consists of the lower­level individuals bound within it, and the effective differences they make to one another. This structure is intuitively an information structure: the effective state of each individual is identified with the difference it makes to the possible states of the higher­level individual. As the lower­level individuals dynamically evolve the shape and character of this information structure will evolve. Since the receptive field will continue to bind the lower­level individuals so long as their interactions continue, one should expect the topology and character of the receptive field to be responsive to this information structure. Given this view, the receptive field is something whose intrinsic character one would expect to reflect the information structures it instantiates, and just that structure. The carriers, then, instantiate just this information structure. Any further internal structure at the lower­levels is irrelevant to their job.

The boundary problem for natural individuals ­ According to The Consciousness Hypothesis, the natural individuals are phenomenal individuals , and the phenomenal field carries the receptive field. It follows immediately that the boundaries of phenomenal individuals are determined by the facts about the world's receptive structure, i.e., the facts about what is bound to what. Unfortunately, I have only offered an idealized "in principle" solution to the problem, rather than a theory of what kinds of systems count as natural individuals. Still, I believe I have met the challenge laid down earlier, which was just this: identify the kind of fact about the world that grounds the possibility that there could be mid­level phenomenal individuals bounded in just the way human consciousness is. That fact is the presence of receptive boundaries at many different levels of nature.

The remainder of the problem is by far the most difficult part. It consists in adapting the theory of natural individuals to real science to determine the actual world conditions under which shared receptive connections exist. This is clearly a highly non­trivial task. Nevertheless, binding within a receptive field makes for a very plausible condition for natural individuation. Thus, the existence of a receptive field would be the hallmark of the existence of an inherent individual. Such individuals could be contrasted with societies of individuals, and with the conceptual construction of individuals due to an act of abstraction from the organization of a piece of spacetime.

15.3 Complementary strengths and weaknesses

Most people, when first exposed to the idea of receptivity, express intellectual interest along with strong misgivings. What is the root basis of such misgivings? After all, there does not seem to be anything logically or metaphysically out of order with the idea, and the weight of argument seems to require the world to have such a thing, or something very similar. Ultimately, the problems are conceptual and epistemic.

On the conceptual side, what sort of conception may we have of something like receptivity? I believe people who wonder this are using "conceive of" in an unusually strong sense. They mean something like "image it." I do not think this makes the puzzlement unreasonable. Given a failure to conceive it, even in this strong sense, we may rightly question whether we really know what we are talking about.

On the epistemic side, one might wonder how we could ever get access to the supposed facts about receptivity. Since the physical facts underdetermine them, the story about receptivity may seem like rationalist speculation, unfettered. To avoid epistemic obscurity, some will want to go back and question our reasons for thinking the receptive field must exist. From this vantage point, people sharply feel the temptation to explain it away in some fashion.

We can sum up the puzzle in the following way: receptive fields are metaphysically required to account for the causal nature of the world, but they are conceptually obscure and epistemically opaque. Given their epistemic opacity, we vividly feel an urge to explain them away. Why shouldn't we follow this urge?

Before giving in to this urge, we should take one more step backwards to get a perspective on what these last few chapters have revealed about receptivity and the character of the receptive field. Putting it all together, I have argued that receptivity is explanatorily irrelevant from the point of view of physics; it is nevertheless necessary for a full explanation of the causal structure of the world; its exact relation to the effective (i.e., physical) properties is sui generis due to a kind of compositional circularity; the receptive field must have some kind of partless unity; it has a sui generis relation to physical spacetime; shared receptivity defines inherent individuality in nature; the carrier structure of the receptive field will reflect an information structure in the natural world; receptive connections can provide important help in constructing a concrete and detailed account of what information is, and the channels through which it flows. One could also make arguments from the privacy of an individual's receptivity to the privacy of that individual's experience.

The similarity to traditionally problematic claims about consciousness blares forth like the sirens on trucks speeding to a four alarm blazer! All these attributes are precisely reminiscent of the kinds of features consciousness is typically thought to have, and the kinds of explanatory problems it presents. I want to stress, and this is important, that the entire story about the receptive field and its queer nature was motivated and developed independently of the problem of consciousness.

I have summed up the intellectual problems presented by the receptive field in this way: it seems to be metaphysically required as part of the causal structure of our world, but seems conceptually obscure and epistemically opaque. I now want to point out that consciousness presents itself as having many of the same troubling features, but we are tempted to explain it away for very different reasons. It is epistemically transparent, and conceptually immediate, but it seems metaphysically baroque. Not only are the phenomenal qualities that exist within consciousness brute features of nature, but many of the apparent features of that experiential context are also brute and awkward. The entire package is unmotivated by any deeper naturalistic considerations. Why should the world contain such a thing?

With consciousness, the problem is not finding out that it exists, or that it has many of those features. Epistemically, they are presented to us. But it is very difficult to believe what is presented to us because, metaphysically, it seems too queer and unmotivated a kind of thing. It is just a strange "nomological dangler" on an otherwise internally complete and self­consistent machine, a physical machine belonging wholly to the physical world.

Once one steps back and takes the overview, it is apparent that the strengths and weaknesses of the two mysteries complement one another. The existence of the receptive field can be well­motivated metaphysically, but it seems epistemically inaccessible. The existence of consciousness is self­evident epistemically, but completely unmotivated metaphysically. Additionally, the ontological characteristics we are inclined to attribute to each are precisely the same. We could just fight the battle on two fronts, denying both the existence of consciousness and of causation.

I hope that, by now, the alternative is obvious. I would like to propose that it would be much more interesting to accept the existence of both consciousness and receptivity, and follow out the project of developing the resulting view of nature. We can become Liberal Naturalists. After all, we cannot help noticing that each strange phenomenon, consciousness and receptivity, seems to undercut the motivation for eliminating the other. The receptive field is epistemically opaque, yes, but perhaps consciousness can be a model for a real, live caught­in­the­trap receptive field. The qualitative field of consciousness seems too strange to be real, too arbitrary and brute, yes, but perhaps the qualitative field of consciousness is the way it is because that is what the receptive field has to be like. It is not stranger than causation itself, and causation has to be strange.

The parallels are too striking to ignore, and The Consciousness Hypothesis makes perfect sense of them. Perhaps our phenomenology carries the nomic content of a whole individual, both receptive and effective aspects? In any case, we must make a theory of the causal structure of the world. Such a theory will likely involve the receptive field, or something quite like it. We should be very open to the possibility that it will shed light on why our existence has the wonderful experiential and qualitative nature that it does.

At the very least, we have now exposed the false dilemma typically presented to anti­physicalists. When asked if consciousness is a ghostly Cartesian entity mysteriously interacting with the physical world, or an ugly "nomological dangler" irrelevant to it, the Liberal Naturalist may answer: neither. If physicalists want to press the issue, they now must address the hidden premise in their argument. They have a great burden of proof, as we now know that their argument is unsound under at least one substantive view of causation. What is the physicalist's theory of causation, and what are their carriers?

1. The answer to this puzzle relies on an analysis of time that is detailed in chapter sixteen. Readers should profit from it more after reading that chapter, although I hope the gist of the idea should still come through prior to reading that discussion.


Chapter 16

Space, Time, and the Unity of the World

16.1 The direction of time and the unity of the world

Chapter ten criticized the Humean view of causation for the awkward way it handles the metaphysics of space, time, and the unity of the world. In this chapter I return to those themes, and I will use the account of causation from chapter twelve to sketch an explanation of the unity of the world. We will also explore a possible way of reducing the facts about space and time to more fundamental facts about the causal structure of the world. These topics deserve a full, rich treatment of their own, so the purpose of this discussion is just to point at a way that an explanation might go.

This chapter introduces three new concepts. It introduces a vector called an ingression that charts a trajectory from the space of possibility to actuality; the concept of a hit that names the state of an individual after it has become fixed and definite; and an account of cascades that may flow from hits when certain kinds of orderliness are present.

This book is painting a picture of the world as a mesh. Its natural ontological structure is one of interlocking, overlayed causal nexi, each a natural individual, ordered hierarchically into levels, and ordered horizontally across levels. An individual's nomic content contains its causal significance within the mesh. We can analyze nomic content into two components: the individual's place in a structure of shared receptive connections, and the possible values of its effective states. Through the actualization of its nomic content, an individual's presence exhibits direct and indirect constraints on the possible states of other individuals in the mesh. Just as in a mesh, no link can be isolated from neighboring links, with effects of any manipulation possibly propagating in all directions. Causation is therefore conceived of as an operation on a space of possibilities.

Conceiving of causation in this manner raises a contentious issue, because the account is not explanatory unless a possibility space exists for causation to constrain. Some form of robust realism about possibility appears to be an ontological commitment of this view of causation, and, I believe, of any adequate view of causation. One can directly argue for this stronger conclusion by pointing out that causation has no work to do unless there exist real alternatives to actuality. Could the world be so metaphysically thin that only the actual existed? If so, the existence of causal production and constraint would be superfluous. Since the work causal powers perform is plausibly essential to them, then, plausibly, they could not really exist within an actualist metaphysics.

The argument for this conclusion will show that the following two positions,

(i) an actualism in which possibilities are fictional constructs, and

(ii) the position that some real power of causal production exists.

are inconsistent with one another under the usual realist accounts of what it means to be a causal power. Let C be any productive cause. To a first approximation, this means that C raises the probability that some event e will occur, relative to some background conditions. In the limit case, C by itself is sufficient to make e actual, and the background conditions do not matter.

Assume that (i) and (ii) are both true. The concept of C's being a productive cause makes two appeals to possibility. The first appeal is to the probables that do not become actual, but might have. The second appeal is in the implication that the probabilities associated with these probables would have been different had C not existed. If we are realists about these possibilities, we contradict assumption (i). Therefore, the possibilities appealed to must be treated as fictional objects. Since both the probables and the alternative probability distributions are mere fictions, the causal "power" is idle. This contradicts assumption (ii). Both cases lead to a contradiction. Either the power of "production" is idle, contradicting realism about causation, or we live in a world with real, unactualized possibilities, contradicting actualism.

The idea of causal constraint, not causal production, is at the center of the theory we are exploring. This difference in emphasis makes mutual causal significance, rather than asymmetric causal responsibility, the fundamental notion. We could easily construct an argument parallel to the one above showing that the present scheme is equally committed to realism about the space of possibilities. Since any realist theory of causation will use one of these two notions, it seems that any realist theory of causation will be committed to realism about possibility.

By combining the above argument for realism about possibility with the arguments for realism about causation given in chapter ten, one may produce a strong composite argument against actualism. The argument against actualism goes like this. If possibilities were only fictions, then no causal facts could exist. If there were no causal facts, then, as chapter ten argued, there likely could be no world, and we certainly could have no knowledge of the world. Therefore, possibilities are not mere fictions, and actualism is false.

We now have a picture of the natural world as a mesh of causal nexi, each carrying causal significance for the rest. The final metaphysical picture is one of the actual world raised up in relief against a metaphysical background of possibility.The attempt to more deeply understand nature's connection to its metaphysical background presents a rich vein of puzzles. In the end, one cannot pretend to understand the natural world fully without understanding its relations to this seemingly non­natural background of possibility.

Reflections ­ I think this is an interesting picture worth investigating further. I will investigate it below, engaging in some very speculative metaphysics. I do it in the spirit of story telling, spinning a yarn that I hope may be woven into the fabric of things that are. My main intents are to provoke thinking and speculation about these very deep matters, and to do so in a way that is ontologically bolder, and I think more substantial, than the deflationary ideas one often finds in contemporary analytic philosophy.

The speculation I am about to embark on, although fun for some, might be too "metaphysical" for some others. It is likely to leave everyone feeling a little dizzy, as if we left the ground too long ago, and too far behind. While reading what follows, I believe one may ameliorate some of the pie­in­the­sky feeling it may produce by remembering that the story is mostly systematizing a variety of elements that we already have independent, rational grounds for introducing.

In particular, we already have rational grounds for being realists about causation; for being realists about possibility as a consequence of our realism about causation; for believing that facts about causation will allow us to reduce the facts about the direction of time; for thinking that space and time are inseparable and relative; for introducing a story about the receptive and effective aspects of the causal nexus; and for wanting a story that presents a unified treatment of the natural world and its non­natural background of possibility. All those demands are already on the table, each rationally motivated, as I undertake the weaving of the tale below.

Once rational inquiry has taken us this far, it is a little late to balk at metaphysical speculations on the grounds that they are metaphysical. Some story has to be told about the direction of time, and about how the natural world is connected to its non­natural background. It would be strange if such a story did not seem a bit strange when viewed from the ordinary, common sense perspective from which we started our inquiry. I believe things are always this way in science and philosophy: common experience presents various phenomena and a need to explain them in an integrated way, and the end of that inquiry leads us to a terrain that looks very different from the one at which we started. While every new terrain deserves criticism, criticizing any of them simply for being disorienting, while tempting, is unwarranted.

A second way to alleviate those ill­ at­ease feelings is to remember that the story below is tentative and illustrative, intended merely to point in a direction of possible inquiry. The key question it is trying to resolve concerns how effective facts may become fixed relative to a given individual, and thereby yield a direction to time. Other stories should be possible. One alternative story with elements analogous to those in modern physical theory postulates random fluctuations in possibility space, within the active constraints. Essentially, individuals sometimes become actual as a result of this randomness. If we postulate, as before, that individuals have an orderly effective and receptive structure, then perhaps we can reproduce cascades to give a direction to time. While I believe that this kind of story is less satisfying in many ways than the one I give below, it has the advantage of incorporating ideas that we have precedent for. Which kind of story one likes better matters much less than the idea that a story space exists to explore.

A third way to soothe concern is to note some surprising, but potentially fruitful, consequences. One very interesting consequence that emerges from the discussion so far is that nothing in the theory requires that every individual at every level end up with a definite effective state. If the mesh indeterministically constrains an individual, then that individual, consistently with everything that we have said, may hold onto all of its remaining potentialities. This has a surprising implication: worlds may exist where higher­level individuals may achieve a level of definiteness that differs from the definiteness achieved at lower­levels.

How can this be? Recall from the discussion of high­level individuals that the effective properties of those individuals are not identical with the simple structure produced by the effective states of their constituents. Instead, the effective properties of individuals are identical with the constraint its effective structure presents within higher­level individuals. A high­level receptive field may instantiate a many­one realization function that maps a class of lower­level effective states onto a single higher­level effective property. Perhaps a bit surprisingly, this notion of effective properties allows for the logical possibility that the effective state of a high­level individual may become fixed and definite even though the effective states of its constituents remain indefinite.

Here is how that could happen. Imagine that an effective state S of a high­level individual becomes fixed and definite, and this effective state is realizable by either of two states, s1 or s2, of its constituents. The actuality, the definiteness, of the higher­level state S is compatible with its constituents remaining in the indefinite state: s1 s2. Before explaining why this is, the notion of an indefinite "disjunctive" state requires a special understanding. It becomes easier to grasp disjunctive, indefinite states once one realizes that a disjunctive state such as s1 s2 is logically equivalent to the conjunctive state: potentially s1 potentially s2 not potentially any other state. Since we are forced into being realists about possibility anyway, nothing is inherently objectionable about a conjunction of potentials.

The moral here is deep, and deserves more reflection. The indefinite state of an individual is equivalent to a definite state of that individual understood as a pluralistic selection from its space of possibilities. The example teaches us to refrain from treating individuals as just actualities by reminding us that individuals have deep roots in the metaphysical background. What exists in the actual world is just the tip of an ingression, the hit, but everything in nature is thoroughly modal (the concepts of ingressions and hits will be discussed extensively below). The true nature of an individual stretches along its ingression. While actuality and possibility are distinct, they are not separable. In the extreme case, the entire high­level history of a world could enter a perfectly definite cascade of states even though no lower­level individual ever possesses a definite effective state.

Even in the face of this possible indeterminacy, the states of higher­level individuals continue to logically supervene on the states of lower­level individuals (given the non­supervenient fact that the higher­level individual exists at all). The fact that the lower­level individuals are in the indeterminate state s1 or s2, and that s1 and s2 realize the same high­level effective state S, logically determines that the high­level individual is in state S. Ask yourself, "What other state could the higher­level individual be in?" The phenomenon does not undermine the sufficiency of low­level states, but the necessity of their definiteness. This result is extraordinarily counterintuitive, and deserves further investigation. For one thing, it violates certain naive assumptions about "constitution", and the relations between constituting entities and constituted entities.

Despite this usurpation of common sense, it is a straightforward logical consequence of the theory. Every way one might remove it seems ad hoc. If we had not already discovered quantum mechanics, perhaps the result would be completely unacceptable. As it is, it is very counterintuitive, but no more counterintuitive than one already well­tested theory about the effective nature of our world. It is even strangely resonant with some aspects of that theory.

16.2 Ingressions, hits, and cascades

The unity of the world ­ These tight relations between possibility and actuality, along with the need to be realist about possibility, raise problems for the task this chapter is taking up. To give an account of the unity of the world or the direction of time, no matter how sketchy that account is, we must at least begin down an avenue of speculation about these relations between the world and its metaphysical background, no matter how few steps we take. To be honest, I have to admit up front to being unsure how to proceed. I am inclined, at least initially and tentatively, to follow Whitehead here, and so that is what I will do.

Following Whitehead, the suggestion I will begin to develop is that the actual world is connected to its metaphysical background by a process of becoming. One can represent this process of becoming as a kind of vector called an ingression from possibility to actuality. Of course, this raises the very difficult and important question of what the metric is on the space of possibilities. I will not be able to discuss the issue in depth here. With Whitehead, I am inclined to think it is something like a degree of definiteness, or determinateness. Still, it is not an easy task to articulate what that is. These issues really require extensive independent treatment.

For the schematic purposes of this chapter, I suggest thinking of definiteness as corresponding to the degree that an individual's nomic content has been made complete. That is, to become fully definite an individual must have all of the slots in its receptive field saturated by other individiduals, the saturation must make its effective state definite, and it must take a fully specific place in the receptive mesh by saturating the slots of other individuals, if such slots are available to it. In short, all of its relations to other individuals must be fully specified. Definiteness is full immersion in a context. In contrast with this, indefiniteness is a kind of context independence. An indefinite individual is one which may exist in multiple, more definite contexts. An individual with unsaturated slots, or an unsettled spot in the mesh, is less definite than it can be. Definiteness and indefiniteness admit of degree. An actual entity is an entity that is fully definite. A possibilia, then, is an indefinite individual or circumstance, and a possibility is a way of making some possibilia fully definite. A possible world is a way of making a fully definite world. I do not think of "ways" as entities. They are more like intentional projections regarding possibilia, a projection that goes at least partway into a fictional ingression for that possibilia.

What role do ingressions play in this picture? We may think of ingressions in either of two ways. On the one hand, an ingression charts a trajectory through the space of possibility for that individual, into the actual world. On the other hand, we should not separate possible from actual natures, for the actual entities are the possible entities made fully definite. They are not set apart, as a thing does not cease to be itself when it becomes actual. Given this, an ingression also represents the complete structure of the individual's nature along a line of indefiniteness to definiteness.

Figure 16.1 is a schematic illustration of an ingression. I call the point where an ingression intersects with actuality a hit. I suggest thinking of a hit as the point of maximal definiteness for that ingression. If we do this, we can define the actual world as the fully interrelated collection of all hits within the larger space of possibilities. For what follows it will be important to avoid identifying an individual's nature with the hit, which is simply the tip of its ingression. An individual's nature is not simply its actuality. The nature of an individual should be thought of as spread or stretched along the length of its ingression. Equivalently, ingressions should be thought of as schematic ways of explicitly drawing out components in the nature of an individual, where these components have various degrees of context independence. In this way, natures are taken to be complex entities containing indefinite, context independent components as well as definite, actual states with a complete context. The indefinite elements contain the individual's potentials. As we move up an ingression, away from the hit, we traverse an expanding well of potential.

Once we have this picture, how the world gains its unity becomes obvious. Within a level, the overlaps between the slots in distinct receptive connections produces the unity of that level. The binding relation that holds between higher­level individuals and the individuals of the unified lower levels produces the unity between levels. For an ingression to hit the world, its unsaturated slots must fill with individuals from that world, and it must find a place at its own level. The world is simply the closure of the mesh across and within levels.

16.3 Space and time

I believe the explanation in section 16.2 provides an adequate view of the unity of the world. How can we understand the direction of time? Unfortunately, accounting for the direction of time is more difficult than accounting for the unity of the world because we must treat space and time together. To approach this problem, I suggest that we do away with space and time as basic entities. The strategy I will pursue involves reversing the common way of viewing these things. I will not assume that the causal mesh exists in space and time. Instead, I suggest that space and time may be a construction out of the topology of the causal mesh. In trying to illuminate the construction of space and time, I will focus on the concepts of fixed properties and unidirectional connections, introducing a more concrete account of an ingression to help flesh out the explanation. I will explain the story's basic elements first, and then illustrate it using the Life world. As a causal structure for the Life world, I will use the causal schema from chapter twelve's suite three, diagram (a).

Remember from the Life example that we should expect the individuals of an orderly world to possess an equally orderly pattern of receptivity. The task before us is this: connect the regular pattern of receptivity to the ingressions of the instantiated individuals, then reconstruct the facts about space and time from this structure. To begin, observe that an ingression has the following character: it possesses two poles. One pole corresponds to the hit, which is the point of maximal definiteness, and one pole corresponds to its origin in the space of possibility, which is the point of maximal indefiniteness. For an individual with a receptive field, in between these poles are various degrees of incomplete saturation in the receptive field: some of the slots in its receptive connection do not contain individuals at all. This gradation of indefiniteness exists along the length of the ingression prior to the hit. At these points of incomplete saturation, we can think of higher­level individuals as riding piggyback atop the lower­level individuals that partially saturate them. As mentioned earlier, to become hits these individuals must become fully saturated, take on a definite effective state, and take a definite place within the mesh.

The entity's ingression maps the process of becoming, where the entities may have a variety of possible structures. To understand time, imagine that some of these ingressing individuals contain unidirectional connections. Unidirectional connections are asymmetric, and it is a small step to speculate that the direction of time might supervene upon a series of overlapping unidirectional connections. Let us call a series of individual ingressions with overlapping unidirectional connections a cascade. The hypothesis here is that the direction of time follows the direction of asymmetric constraint within cascades.

We can understand this suggestion in the following way. Imagine a world with cascades, and an individual i1 that is an element bound within a cascade. As a consequence, it is both asymmetrically constrained and asymmetrically constraining. Its immediate future may be organized relative to the element it is asymmetrically constraining, and its immediate past may be organized relative to the element asymmetrically constraining by it. The individuals in i1's cascade form a set of pivot points for organizing the spatiotemporal facts relative to it; an individual's past, for instance, should be constructed as a function of the facts that are fixed relative to it, both directly and transitively.

Given a regular enough pattern of instantations for receptive fields, the facts about the cascade should naturally partition themselves into an ordered set: each element in the cascade would correspond to a different order in a temporal history, and would belong to a different time slice. The difficult challenge is to obtain the facts about spatial organization by placing the individuals in these partitions of temporal facts in relation to other individuals. To accomplish this goal, we must somehow use the causal relations between members of the cascade and other members of the total causal mesh.

A natural idea is that the facts about causal relevance, centered around i1 and the cascade that it belongs to, may yield a projection of the mesh's causal topology into a spatiotemporal geometry. The intuitive idea is that neighbor relations in space should correspond as closely as possible to the facts about the degree of causal relevance, likely corresponding to facts about direct and indirect connection. In other words, there is a causality condition on locality, not a locality condition on causality.(1)

How might the construction go? The membership criteria for other individuals that belong to i1 's time slice may be given recursively using its causal relations to other individuals as follows,

An individual i2 belongs to the same time slice as i1 just in case i2 is immediately bound to i1 within a polydirectional connection, or i2 is polydirectionally bound to some other individual that is in i1's time slice.

This rule yields a set containing all the individuals that i1 is polydirectionally connected to, and all the individuals those individuals are polydirectionally connected to, and so on. The same rule, when applied to i1's ancestors and descendents in the cascade, yields the individuals that are members of their time slices.

The recursive rule yields a set of individuals that occupy the same time slice as i1. The task is then to reconstruct the spatial neighbor relations between i1 and the other individuals in its time slice using just the facts about the structure of their causal relations. To simplify the model, I will assume that space is a discrete grid (in a model with continuous space, the concept of "neighbor" would need to be more technically specified). To construct spatial relations, we will try to enforce locality by projecting individuals that share a time slice with i1 (and are thus directly causally relevant to one another) into space using the following three rules. Then,

(1) If individuals i1 and i2 share a receptive field with one another, and there is no individual i3 that shares a receptive field with one but not the other, then i1 and i2 occupy the same point in space.

(2) Any individuals in i1's time slice that are, like i1, directly constraining i1's immediate descendent are neighbors for i1 if they do not meet condition (1).

(3) Otherwise, the individual is not a neighbor of i1.

Rule one grounds the possibility of pointlike entities with many properties, such as electrons whose properties of mass, charge, and spin are all present at a point. Rule two allows extension through space by establishing neighbor relations. Rule three prohibits assigning i1 neighbors that are not immediately causally relevant to its immediate future, which provides the reference frame. Starting with a given individual, one may find its neighbors by application of rules (1)­(3). By reiterating that basic procedure on the neighbors themselves, one may reconstruct a spatial geometry for the time slice. Furthermore, as one moves backwards in time one must include more and more individuals whose effective states could make a counterfactual difference to the frame of reference, and the expanding structure looks much like a light cone.

In practice, we should treat these rules as "soft" constraints. The "correct" projection is the one that best balances the need to preserve locality (treating direct causal significance as either a spatial sameness or neighbor relation), and the desire to obtain a simple and orderly geometry. One might give up some locality to make a gain of simplicity in the geometry, but ideally only a little would be acceptable.

16.4 Spacetime in the Life world(2)

The suggestion above is admittedly speculative. To get the clean, simple spacetime geometry of our world, the receptive structure of the world would have to be very special and orderly. Maybe our world is like that, and maybe not. At this early stage, all we are after is a possibility proof for the basic idea. Hopefully, the proposal at least promises to show how spatial and temporal facts may be a projection from more fundamental facts about the mesh of causal nexi. One reason for optimism is the observation that the proposal implies that, to reconstruct spacetime, one must first position oneself in the mesh by specifying a given individual as a frame of reference. Thus, if the suggestion is correct, it explains in nonepistemic terms why the spatio­temporal facts would not be absolute.

To illustrate the concepts just introduced, as well as to win some assurance that they can do the job, I will derive the spacetime structure of Life from one of the causal structures given in suite 11.3 of the tutorial. The one I will use is from diagram (a), and figure 16.3 re­presents that diagram to help the reader. Figure 16.4 separates and schematizes the types of higher­level individuals that exist in the structure represented by the diagram. Intuitively, since we have backwards engineered that scheme from given facts about the geometry of spacetime in Life, we should be also be able to forward engineer the geometry of spacetime from those facts. A quick glance shows that we can. Each cell is a member of a cascade that orders time, and is connected to an eight cell receptive field that constrains it and can provide neighbors for its ancestors and descendants.

The engineering problem is trivial. The challenge lying ahead is to tell a metaphysical story about the nature of ingressing individuals, and laws governing their

ingressions. The ingressing individuals must produce this causal mesh from which we can derive an appropriate spacetime structure. The individuals underpinning an appropriately structured mesh will be complex, spanning many hierarchical layers. At their indefinite poles, each of these individuals will contain unsaturated connections whose saturation grounds its existence in the Life world. We will deduce the structure of these complex individuals from the "ground up", specifying component individuals at layers zero and one first. We will then fit them together into the top­level individual. Before embarking on this task, I want to emphasize that this explanatory priority is pragmatically motivated, and I am not suggesting that it reflects a logical or metaphysical priority.

To begin, consider an individual "on" or "off" property that ingresses into the world. I will treat this as the ingression of a basic Life cell. Let's call this cell i1. The first hypothesis is that the ingression of each cell like i1 is accompanied by a background of higher­level individuals piggybacking into the world on its ingression. At level one, this background consists of eight type (a) individuals, and the structure of this background is shown in figure 16.5. Each of these type (a) individuals binds i1 within the specific slot of an eight­place receptive connection. Figure 16.5 also depicts these individuals, and the structure of relations between them. At the far pole of the ingression, the pole furthest from the hit, this background has three distinctive features,

(i) Each strip contains seven unsaturated slots. The saturated slot in each binds the ingressing cell, i1.

(ii) Although each level­one individual binds i1, i1 occupies a different slot in each. If we look at figure 16.5 and read it from H's neighborhood to E's, we see the cell i1 occupying the diagonal of the matrix.

(iii) These cells overlap with one another at the places specified by the letters A­H. For instance, the letter E's occurrence at slot two of the first strip, slot three of the second strip, slot five of the sixth strip, and slot seven of the eighth strip represents that these receptive connections overlap with one another at those slots. Slot two of the first strip overlaps with slot three of the second, slot five of the sixth, and slot seven of the eighth, and so forth. The overlaps here mean that the corresponding fields must end up sharing the same individual in those slots when the ingression hits the world. One way to envision the overlapping structure here is to picture it like a molecule with a complicated, folded topology. Think of the receptive connections as "folded" at the overlaps in their slots like a complex molecule with atoms sharing electrons on its looping branches and folds.

Notice that each strip reflects the neighboring relations of one of i1's neighbors. The neighboring relations are encoded in the left­to­right ordering of receptive slots. The reader should be able to recover them by projecting this ordering onto a clockwise ordering of the appropriate cell's (i.e., H, G, F, etc.) neighbors. The letters in the fields reflect places for i1's potential neighbors, as depicted in the matrix in the upper left corner. When saturated, the individuals in the slots labeled A­H will become i1's neighbors. To make the neighbor relations to i1 easier to see I have suppressed the outer shell (some of the neighbors of i1's neighbors) in figure 16.5, highlighting just those that share i1 as a neighbor. Figure 16.6 depicts the full set of neighboring relations. The topology of the "molecule" is clearly complicated, as there are many mutual overlaps.

One might think that this scheme is too complicated and indirect. Instead of having this large background "molecule" piggyback into the world with i1, none of whose cells represent i1's neighborhood, why couldn't i1 directly bring its own neighborhood into the world? Couldn't just a single eight­slot individual do the job? Two problems confront that approach. First, i1 is not part of the eight­slot individual containing its neighborhood. Not being bound within that individual, it would share no other natural relation to it. This leaves us without any obvious mechanism to allow the higher­level individual to piggyback onto i1's ingression. Second, nothing about the simple eight­slot structure guarantees that the individuals who fill those eight slots will instantiate with 'neighbors' that have i1 in their 'neighborhoods'. To achieve the symmety in the neighboring relation the structure of the overlaps in the proposed background will be important. It helps produce a cohesive spatio­temporal interpretation of the causal mesh.

The alternative we are pursuing is that i1 brings each of its potential neighbor's neighborhoods with it. In exchange for this service, i1 has sneakily nestled itself comfortably into the appropriate place relative to each potential neighbor. Combined, this background represents a kind of mirror that reflects i1 from all directions. Having this mirror will help us place i1 at the center of these neighbors when the slots become saturated. To construct the appropriate neighborhoods, the individuals that end up bound to the unsaturated slots in i1's background must also have backgrounds that mirror i1, and one another, in the appropriate way.

To achieve the required coordination between cell backgrounds, we are going to allow the backgrounds of different cells to merge fields as they ingress. A merger of a field can be represented by having two lines of ingression intersect, and then follow the same path into the world. This is depicted in figure 16.7.

By allowing elements of an ingressing cell's background to merge with elements from the backgrounds of other cells, we reduce the total background that the world must support. In a merger, two or more zero level individuals come to share the same eight­slot individual as an element in their backgrounds, so their backgrounds come to overlap. In the Life case, we can think of it as a unification of two less definite type (a) individuals into a single more definite individual.

As a parsimony condition, we will stipulate that whenever a merger can occur, it does. Furthermore, we will govern when a merger can occur using just an integrity condition. The integrity condition requires that the merger preserve the molecular topology of both cell's backgrounds. If this condition is met, we will say the merger is non­creative. To preserve the topology of the background, no single slot in the background that either field belongs to prior to the merging will merge with two or more slots from the background that the other field belongs to.

What is the substance of this condition? Because each slot represents a place for an individual, two slots from one field unifying with one slot from another field could only occur if it created a new overlap (or destroyed one). That kind of merger would obviously be creative. How can someone picture this condition in terms of the actual difference it makes to i1's ingression? As we move down along the ingression, towards the tip which is its hit, the unsaturated slots in i1's background become saturated by other cells. The merger condition dictates that when an ingressing individual saturates a slot in i1's background, every field in i1's background that overlaps at that slot will merge with some field from the new cell's background. The backgrounds for i1 and the new cell become unified at just those fields. If all ingressions meet this condition, the totality of ingressions will create a set of neighborhoods like those depicted in figure 16.6. Running a computer simulation of the process will verify this. I provide Prolog code for such a simulation in appendix B.

We now have a scheme that provides all the right neighborhoods, but no centers for them to be neighborhoods of. To finish the job, we must introduce the other higher­level individuals into the ingression. Figure 16.8 depicts these individuals. Part (a) of the figure depicts the unidirectional connection from the ingressing cell, i1, to its ancestor, slot K. I1 and the slot for its ancestor constitute a level one individual, i2, that is unidirectionally bound to one of the neighborhoods, call it i3. Together, i2 and i3 form a level two individual, i4. I4 is the master individual whose ingression we have been building towards.

To ensure that i3 is the correct neighborhood we have to postulate a law governing the choice of binding partner for i2 within i4. The law must be keyed to the ancestor cell that ends up occupying the unsaturated slot of i2 because we want i3 to contain that ancestor's neighbors. Let us call this ancestor K. The law requires that i2 must bind to a neighborhood field that (i) binds only individuals from K's background, but (ii) does not bind K itself. Only one unique neighborhood satisfying these conditions will exist, and it will be K's appropriate neighborhood. Figure 16.9 depicts the structure of the total ingressing individual.

Once these temporal facts are set, the two dimensional spatial facts fall out of the causal relations. The single constraining ancestor cell, and a single neighborhood field for that ancestor, will jointly determine the state of each individual cell (perhaps modulo an initial random ingression of neighborhoods). These neighborhoods will mirror one another and that single cell in a special way. In fact, reconstructing the causal relations within a two­dimensional spatial structure organizes them most simply. Relative to a given receiving cell, this structure will have the single ancestor cell at its center, with the cells in the neighborhood field arrayed clockwise in a square around it. We may construct arbitrary elements of the spatial layout in a time slice by considering each of the other cells in that slice, and consistently interpreting all of its neighbor relations in this way. As an end result, we can recover the two­dimensional spatial structure of the entire time slice. From the fact that they are all part of the same partition as the ancestor cell, we can recover the fact that it is an immediately previous time slice.

Our final task is to regulate motion in the Life world. We want the basic effective properties to be able to "move" from slot to slot, but movement of the slot structure itself should not occur. For instance, if a cell slot A at time t2 holds the constrained element in a unidirectional link, and the other end of the link becomes saturated by A* in the immediate earlier time t1, then the identity of the enduring individual will follow the link from A* to A. Now, imagine that a cell slot H* is part of A*'s neighborhood at t1. We want the analagous cell slot H in A's background at t2 to have H* as its immediate ancestor. Otherwise, A will have a different neighbor than A at that place. In principle, neighbor movement like this is a perfectly acceptable thing. However, the causal structure we have provided for the Life world is not compatible with it.

We could govern movement in several ways. Most directly, we can simply postulate a law that forbids it. A little more subtly, we could add a parameter to the model such as an orientation on the ingressing individuals of type (b). These orientiations "point" the unsaturated slots in definite directions within the developing mesh. We would have to give a sense to "orientation" that is not spatial, but that should be achievable. The larger moral is that in an orderly world ingressions themselves will be highly ordered. The order will at least be a lawlike consequence of some deeper structural or metaphysical facts. From these kinds of facts, facts about causal relations alone, it seems that one can reconstruct the structure of a world's spacetime. If this is correct, a world's spatio­temporal structure is projectible from a set of more basic facts.

This model implies a few other interesting things about spacetime. First, as mentioned earlier, spacetime must be relative. One can easily see that the organization of facts in spacetime will be relative to the place that one occupies in the causal mesh. To know where something is, one must first anchor oneself within the world and ask, "Where is it from here?" and then negotiate the structure of causal links relative to that place.

Second, space and time are inseparable because one cannot fix the spatial facts simply from the information contained in the eight­cell neighborhood fields. Neighborhoods must have centers, and the pivotal center is the ancestor of the frame of reference itself, bringing time into the picture in the determination of space.

Third, it seems plausible that the only worlds that are possible are ones in which spacetime is either infinite, or finite and unbounded. The reason is that in a finite, bounded world the "corner" individuals could not complete their ingressions, since their backgrounds could find no individuals to fill their unsaturated slots.

16.5 "Fixed" facts and the puzzle of unidirectional connections

An important conceptual problem that remains is the challenge of understanding how an effective fact may be "fixed" in the way the story requires. This is a problem for cascades, but no analagous problem exists for ingressions. Ingressions are vectors. Various points on the vector may correspond to various metaphorical "events", like mergings, that occur before an ingression hits actuality, but this appeal to "events" is not vicious because it is metaphorical. The ordering in an ingression is not a temporal ordering, as it exists orthogonally to actual time. Indeed, a worry like this seems to stem from taking the representation of an ingression too literally. An ingression is just a representation of the structure of an individual's nature. It does not truly represent a dimension in which things "occur."

Cascades, alternatively, do occur in the actual world, subserving time. The real problem is accounting for what it means for some facts to be fixed relative to others without assuming time. Taking the difference between polydirectional and unidirectional connections as primitive is inelegant. The metaphysics would be cleaner if we could analyze unidirectional connections.

Perhaps we can answer this challenge by introducing a kind of priority parameter within a nexus. Since we are conceiving of causation as a process of multiple constraint satisfaction, it is not too much of a leap to introduce priorities among the constraints. We can postulate that receptive connections contain priorities associated with their slots. The set of priorities are parameters of the field. A unidirectional connection is one where the slot occupied by the receptive individual has lower priority than the other slot in the connection. In general, within a nexus an individual A is 'fixed' relative to B just in case A occupies a slot of higher priority than B.

While this complicates our picture in one way, it also simplifies it in another way, and perhaps a more important one. We can now assimilate unidirectional and polydirectional connections to a single ontological category: receptive connections with priority parameters. A polydirectional connection is a receptive connection in which all priorities are equal. A unidirectional connection is a connection with favored slots. The values of the priority parameters are different in the two different connections, but that is the only difference. It also opens the door to other kinds of nexi, providing the possibility of an even richer world.

Armed with a coherent idea of why some individuals would be fixed relative to others, one can tell the story about how cascades may occur in a way that avoids the concerns above. Just as the length of an ingression is not temporal, these priority relations are not temporal. Still, their asymmetry promises to yield the temporal ordering relation we were looking for.

1. Brian Cantwell Smith (1996) has expressed this idea well. He says, "Distance is what there is no action at." (p. 190)

2. This section is very difficult, and may be skipped. By appealing to our model world of Life, it illustrates in technical detail the suggestions in the previous section. It also contains some wilder ideas about ingressions, and provides some fun food for thought for those who take on the challenge, and willingly engage the material.


Chapter 17

Conclusion

17.1 The sliding tile puzzle revisited

I began this book with a promise that I would treat the problems of understanding consciousness and causation like they were the last two pieces in a sliding tile puzzle. Part of that promise was that I was going to challenge the established order in our picture of nature, and undo that understanding in a series of regressive moves. The second part of that promise was that the old order would eventually reappear, but this time in a context that allowed a more complete ordering of the entire puzzle, with consciousness and causation at least well on their way to becoming nestled into their proper places.

In the intervening chapters I have tried to make good on that metaphor. I argued at length in support of the anti­physicalist arguments. By doing so, I challenged and rejected perhaps the chief scientific and philosophical hypothesis of the twentieth century, the hypothesis that the basic physical facts are all the basic facts there are. After kicking away this metaphysical support, I lingered over a series of immediate problems that seem to arise, and that seem to violently threaten our view of ourselves as natural objects. I argued at length that some sort of panpsychism could be true, and likely was. I wondered how to make sense of the idea that a single system could be both unified in the way consciousness seems to be, and also fail to have that unity in the way that the brain seems to fail to have it. I worried about how to make sense of the subjective instant, and how to avoid the dualist dilemma of having to choose between epiphenomenalism and interactionism. I paused to note the structural problems created for our image of nature by the grain problem and the boundary problem for phenomenal individuals. Finally, I challenged the almost universal assumption that a predictively adequate physical theory was also causally complete.

By the time I had reached that point, the regressive part of my project was done. The task of the last third of this book has been to look over the motley pieces in the now jumbled sliding tile puzzle, disordered and ugly, and figure out a strategy for returning them to their old places of comfortable honor. The strategy I chose was to rethink causation itself, focusing on what a causal connection might be. Turning to an ancient conception, I thought of the world as a mesh of interconnected individuals, each with effective and receptive aspects. Taking a modernist view on these old notions, I attempted to model their logical relations in as much detail as possible, using a toy physics. These efforts led to the introduction of yet a third aspect of causation that I called carriers, and I named the resulting picture of our world's causal character "The Carrier Theory of Causation." If the models introduced are on the right track, the idea of understanding consciousness as a special kind of carrier, one involved in cognitive contexts, seems very promising.

The novelty of the proposal involved two elements. The first novel element concerned focusing on, and carefully modeling, the receptive aspect of individuals. I argued that we can best understand physical causation as describing the nomic mosaic, which specifies the regular and lawlike ways that the world's basic effective properties instantiate. The existence of effective properties seems to logically imply that individuals have receptive aspects or properties also. To remedy an apparent neglect of the receptive connections between individuals, I proposed models for understanding the effective and receptive sides of individuals, and the relations between them.

The second novel element of the account was an argument that these effective and receptive sides of individuals cannot exist as pure dispositions, but must be carried. I gave arguments that carriers must have structural and natural features that precisely mirror those consciousness seems to have. The duality of causation is that it divides into effective constraints, and the reception of these constraints. The duality of consciousness is that it divides into phenomenal quality, and the experiencing of that quality. I argued that we may find a place for consciousness by postulating that it exists as a kind of carrier for the effective and receptive sides of natural individuals.

Within the framework I developed, I gave reasons for believing that we may best understand conscious qualities as carriers of effective constraints, and that we may understand the experiencing of those qualities as resulting from their shared receptivity within the higher­level individual. If this is true, consciousness cannot be understood on purely physical terms, and yet it avoids both causal irrelevance and interactionism. The resulting view is a panpyschist view very reminiscent of the sort proposed by Whitehead, in which the fundamental entities are processes composed of internally linked, experiential events. Some of these events would have to be irreducible state determinations of high­level individuals.

By introducing this analysis of causation, I may have found a place for consciousness. Have I restored the old order, though? Can science, and the scientific understanding of the world, survive within a Liberal Naturalism that endorses The Carrier Theory?

17.2 Science and The Carrier Theory

To wonder if science can survive the carrier theory is to wonder in the wrong direction. A more legitimate question is whether the carrier theory may survive an encounter with real science. The Carrier Theory of Causation is put forward as an analysis of what causation is, and any analysis of causation that makes the science of our world seem impossible or false must be rejected. In the last third of this book, The Carrier Theory was developed always as an articulation of some very general and intuitive principles, and these articulations occurred within the framework of one toy physics or another. Its points of contact with real physics were tenuous at best, mostly suggestive, and always implicit. The real challenge for The Carrier Theory is this: can The Carrier Theory grow up? Can we identify real physical conditions that betray the receptive structure of the world, if not logically then at least through a process of inference to the best explanation?

In chapter thirteen I produced an argument that a theory including the facts about the world's receptive structure would contain physics as a proper subset, and that alone seems to guarantee that a finished carrier theory would preserve physical science in something essentially like its traditional place. The problem that really confronts us is to use our best science in an attempt to bootstrap our way into the total theory of the world's causal structure. Should we be able to do this, we should be able to make predictions concerning what kinds of physical circumstances signify the existence of natural individuals. That, in turn, should yield testable predictions about the physical character of the neural correlates of consciousness, finally yielding criteria for validating or falsifying The Consciousness Hypothesis introduced in chapter fourteen.

Is the task just outlined hopelessly difficult? I do not believe so. Even though the world's receptive structure is strictly irrelevant to physical explanation, that does not mean we cannot make reasonable guesses about how to interpret physical theory in a way that involves receptivity. For instance, if the arguments in chapter sixteen were correct, we have good reason to believe that the direction of time supervenes on asymmetric receptive connections. Although physical theory struggles to explain the direction of time, we nonetheless know which direction it flows in. We can reasonably postulate, given this knowledge, a receptive structure that is heavily asymmetric in that direction.

Similarly, certain puzzles in quantum mechanics, puzzles like the EPR paradox, seem to betray the existence of receptive connections belonging to higher­level individuals. We can examine the mathematical characteristics of such systems and try to extract the formal characteristics that betray such global properties (e.g., the non­additivity of vectors). Quantum coherent systems also are good candidates for higher­level individuals. Finally, the way that individuals decohere from one another will likely yield the strongest clues regarding the world's receptive structure. In examining this decoherence, it should prove fruitful to concentrate on multi­level interactions so as to get a picture of the stratified layers of our world's ontology, and also to concentrate on systems where we see causal amplification of effects from one layer to another.

All these suggestions are speculative, and the technical problems they pose are large. However, they constitute a clear empirical research program, which is something more than Liberal Naturalists have had to this point. One can expect that, should the program prove fruitful, the payoff would be enormously satisfying. It would be nothing less than a simultaneous insight into the deepest truths about both ourselves and the world we are enmeshed within. Now that this day's work is done, and we may start looking forward to the next day's work, the proper attitude we should take towards the difficulty of the challenges that lie ahead should be one of excitement, not of being put off. After all, no one ever really thought that solving the hard problem of consciousness would be easy.


Appendix A

David Lewis' Similarity Metric

To see how far the problem of accounting for the unity of the world has taken the Humeans from their original empiricist motivations, we can briefly consider one quasi­Humean view that, if it worked, might avoid the problem. David Lewis (1973) suggests that we can reduce causal facts of the form "A caused B" to counterfactual facts of the form, "If A had not occurred, then B would not have occurred." The counterfactuals are supposed to refer to sets of possible worlds ordered by their similarity to the actual world. Counterfactuals like the one just stated are true just in case some world where neither A nor B occurs is closer to the actual world then any world where A did not occur but B did.

Lewis' account faces the problem that it seems often to misidentify effects as causes. Imagine an event C that causes E, and that must cause E given the laws of nature and the circumstances. Lewis considers the objection that the counterfactual "Had E not occurred, then C would not have occurred." would seem to be true in such a situation, thus making E a cause of C.

Lewis denies that the problematic counterfactuals are true in the imagined situations. He suggests that such counterfactuals have an inherent direction. Lewis claims it is less of a departure from the actual world if we get rid of the effect by changing something after the cause occurs rather than changing the occurrence of the cause. That maneuver, claims Lewis, leaves the counterfactual world similar to ours for a longer period, and therefore more similar tout court.

Lewis' suggestion has many problems. At a specific level, it is open to many obvious counterexamples. Consider the situation where Trey shuts his eyes, causing his visual field to go dark. By Lewis' criteria, what is the status of the counterfactual, "Had Trey's visual field not gone dark, his eyes would not have shut?" It seems very clear that the closest world where Trey's visual field does not go dark is the one where we imagine that Trey never shut his eyes. It is certainly not easier to change some event that occurs after Trey shuts his eyes. For instance, a world where Trey suffers from some neurological disorder that causes his visual field to persist after his eyes are closed is not closer to ours than one where he simply never shuts his eyes. Thus, Lewis' account entails that Trey's visual field going dark was, after all, a cause of his eye's shutting. That should be false, yet on Lewis' view it is true.

On a more general level, Lewis' assertion that these counterfactuals have an inherent direction can fail on a large scale. As Tooley (1990) points out, many logically possible worlds exist where the similarity relations Lewis uses would not hold. The reason is that Lewis' account implicitly relies on the fact that causes typically have more effects than effects have causes. Therefore, changing an earlier event typically has more "ripples" in the orderliness of the universe than changing a later event does. Only under these circumstances does it pay to wait if one wants to keep the world as similar as possible.

This is a contingent feature of our world, though. Consider a Newtonian world with an initial condition of very high entropy, and with velocities and positions of its basic particles set so that entropy decreased through time. This is a kind of 'inverse' world to ours. To visualize such a world, imagine a film being played backwards. The increases in orderliness that pick out the interesting regularities in this world would routinely have unique future events with many antecedent "ripples" leading to it. Future events in the "inverse" world have the structural properties of past causes in our world.

In this "inverse" world, maintaining the requisite similarity relations by keeping the future constant as long as possible would be easier. This is a world where, according to Lewis' view, causation would be occurring in the opposite direction of time. Generally, worlds where the direction of causation is mixed to whatever degree desired are also logically possible according to Lewis' account. Thus it does not seem to be an adequate analysis of the concept of causation. The account has many further problems. For instance, Bennett (1987) shows how Lewis' account of counterfactuals can lead to contradictions that can only be avoided in implausible ways.

The problems with Lewis' view also reveal problems with the view that the direction of time can be reduced to the direction of entropy. The neurological activity involved in Trey's vision, for instance, corresponds to a reliable, repeated, local decrease in entropy. Yet such decreases do not correspond to a local reverse in the direction of time. Furthermore, Tooley's inverse Newtonian worlds are provably coherent, meaning that the "reductions" of the direction of time to the direction of entropy seem to be relying either on metaphysical neccessities or primitive identities, and inherit all the problems my earlier discussions revealed with those views.

Appendix B

Prolog program for the Life world's ingressions

/* main predicate that ingresses nine individuals, using individual 1

as a pivot. */

saturate :-

individuals([1, 2, 3, 4, 5, 6, 7, 8, 9], [[Pivot, Vars]|Rest]),

!,

bind_templates(Pivot, Rest, Result),

name_slots(Vars),

matrix_print(Vars), keybreak,

nl,

pretty_print(Pivot).

/* unifies remaining vars with the names of the unsaturated slots

they represent */

name_slots([]).

name_slots([[A, A]|Rest]) :-

!,

name_slots(Rest).

name_slots([_|Rest]) :-

name_slots(Rest).

/* Using one element as a pivot for the ingression of the others,

bind each one into the background provided by the pivot. */

bind_templates([_|PivotTemplate], Ingression, FinalBackground) :-

neighbors(PivotTemplate, PivotBackground),

bind_templates_aux(PivotBackground, Ingression, FinalBackground).

/* Cycling through each ingressing individual in turn, extract its

background and find a slot for it in the existing background,

merging its background with the existing background where possible. */

bind_templates_aux(P, [], P).

bind_templates_aux(PriorBackground, [[[Ingressor|Ingression], IVars]|Others], FinalBackground) :-

neighbors(Ingression, IngBackground),

ingress(Ingressor, PriorBackground, IngBackground, NewBackground),

bind_templates_aux(NewBackground, Others, FinalBackground).



/* Find a way to ingress an individual's background into the

prior background. */

/* succeed if the ingressor's background is empty */

ingress(_, F, [], F) :- !.

/* succeed if the ingressor's background can be ingressed */

ingress(Ingressor, PriorBackground, IngBckgrnd, NewBackground) :-

var_frequency(PriorBackground, VarsByRank),

member([_, Slot], VarsByRank),

/* tentatively slip the ingressor into the slot; causes the ingressor

to appear in the prior background. */

Ingressor = Slot,

/* extract all the fields the ingressor now exists in */

extract_ingressor_presence(Ingressor, PriorBackground, RowsToMerge),

/* attempt to find fields from the ingressors background to

merge with these fields from the priorbackground. Return any fields

from the ingressor's background that were not merged */

ingress_aux(RowsToMerge, PriorBackground, IngBckgrnd,

IngBckgrnd, FailedMerges),

/* create the new prior background by appending the failed merges to

the old background */

append(PriorBackground, FailedMerges, NewBackground).





/* if no more fields from the prior background need to be merged with,

succeed. */

ingress_aux([], _, _, FM, FM) :- !.

/* if each field from the prior background that the ingressor

now appears in can non-creatively merge with some candidate

from the ingressor's background, succeed. */

ingress_aux([PriorField|PBRest], PriorBackground, IngBckgrnd,

CandidatesToMerge, FailedMerges) :-

ing_aux2(CandidatesToMerge, PriorField, PriorBackground,

IngBckgrnd, RemainingCandidates),

!,

ingress_aux(PBRest, PriorBackground, IngBckgrnd,

RemainingCandidates, FailedMerges).



/* if this field from the prior background can merge with some

field from the ingressor's background, succeed. */

ing_aux2([IngField|Tail], PriorField, PriorBackground, IngBckgrnd,

Tail) :-

/* They must at least be able to unify */

not not IngField = PriorField,

/* the unification cannot be creative relative to the background

that either field belongs to. */

not creative(IngField, PriorField, PriorBackground),

not creative(PriorField, IngField, IngBckgrnd),

/* the unification would not be creative, so do it. */

PriorField = IngField,

!.



/* if some other field from the ingressor's background can

non-creatively merge, succeed. */

ing_aux2([IngField|Tail], PriorField, PriorBackground, IngBckgrnd,

[IngField|Remainder]) :-

ing_aux2(Tail, PriorField, PriorBackground, IngBckgrnd, Remainder).



/* PB is a list of lists. [[....], [....], [....]]. This procedure

counts the number of occurrences of a variable in the body

of the lists. vf_aux cycles through the list of lists. vf_aux2

cycles through the elements of the embedded list, counting

the variable occurrences. */

var_frequency(PB, RankedByOccurrence) :-

vf_aux(PB, [], [], FinalCount),

sort_on_keys([-1], FinalCount, RankedByOccurrence).

vf_aux([], _, FC, FC) :- !.

vf_aux([H|Bck], CountedSoFar, BufferCount, FinalCount) :-

vf_aux2(H, Bck, CountedSoFar, NewCSF, BufferCount, Buffer2Count),

vf_aux(Bck, NewCSF, Buffer2Count, FinalCount).

vf_aux2([], _, CSF, CSF, Count, Count) :- !.

vf_aux2([H|Body], Bck, CountedSoFar, NewCSF, BufferCount, FinalCount) :-

var(H),

not already_bound(H, CountedSoFar),

occur_in_bck(H, Bck, 1, Count),

!,

vf_aux2(Body, Bck, [H|CountedSoFar], NewCSF, [[Count,H]|BufferCount], FinalCount).

vf_aux2([_|Body], Bck, CountedSoFar, NewCSF, BufferCount, FinalCount) :-

!,

vf_aux2(Body, Bck, CountedSoFar, NewCSF, BufferCount, FinalCount).

/* returns the number of V's occurences in the background */

occur_in_bck(_, [], C, C) :- !.

occur_in_bck(V, [H|T], Count, FinalC) :-

already_bound(V, H),

!,

NewCount is Count + 1,

occur_in_bck(V, T, NewCount, FinalC).

occur_in_bck(V, [_|T], Count, FinalC) :-

occur_in_bck(V, T, Count, FinalC).



/* return all the rows in the prior background that the ingressor

has been tentatively bound to */

extract_ingressor_presence(I, [], []) :- !.

extract_ingressor_presence(I, [PF|PB], [PF|RTM]) :-

already_bound(I, PF),

!,

extract_ingressor_presence(I, PB, RTM).

extract_ingressor_presence(I, [_|PB], RTM) :-

extract_ingressor_presence(I, PB, RTM).



/* removes the first occurence of its second argument from its

first argument, returning the remainder. Upon failure, it returns

successive members of the list. */

remove([H|L], M, L) :- H == M.

remove([H|L], B, [H|L2]) :- remove(L, B, L2).

/* The unification of two fields is creative if it causes any *one* of

the original slots from either field to unify with *two* slots from

the other's background. For example, if a variable A from one field

unifies with a variable B from the other, and then also a variable C.

This would collapse the prior distinction between B and C in the second

field. The ontological side-effect would be to create an 'overlap'

between the background's receptive fields where none existed before.

The routine treats its second argument as an 'aggressor', trying to

unify with the second argument, which it treats as passive. The third

argument is the background of fields that the passive argument belongs

to. It checks this background for side-effects of the unification,

succeeding if the unification would be creative for that background.

This predicate checks to see if any of the individuals (or variables)

that unification might introduce into the background already exists

there. If so, then proposed 'new' element must already be bound to

some distinct slot in the Background, and the proposed unification

creates a new overlap between slots. It is therefore creative. */



double_binding(BindingPairs) :-

member([I1, V1], BindingPairs),

member([I2, V2], BindingPairs),

I1 == I2,

not (V1 == V2),

!.

/* The unification is creative if any element of the aggressor would

be bound to two distinct elements within the passive field. */

creative(Aggressor, Passive, Background) :-

proposed_bindings(Aggressor, Passive, BindingPairs),

double_binding(BindingPairs),

!.

/* The unification is creative if the aggressor would bind onto

two elements of the passive field. The clause above has

failed, so it is not double binding within the local field. It is

still creative if it is double binding relative to the

background that the field belongs to. */

creative(Aggressor, Passive, Background) :-

proposed_bindings(Aggressor, Passive, BindingPairs),

agg_intrusions(BindingPairs, ProposedIntroductions),

member(A, ProposedIntroductions),

element_of(A, Background),

!.

/* Returns the aggressor's half from a set of binding pairs.

These are the elements that would be introduced into the

passive background. */

agg_intrusions([], []) :- !.

agg_intrusions([[H, _]|Tail], [H|Rest]) :-

agg_intrusions(Tail, Rest).

/* Returns the all the pairs that a unification of the two

fields would merge. */

/* empty sets do not bind any elements */

proposed_bindings([], _, []) :- !.

/* A binding occurs if the fields possess vars with different

identities */

proposed_bindings([H|T1], [B|T2], [[H, B]|Rest]) :-

var(H),

var(B),

not B == H,

!,

proposed_bindings(T1, T2, Rest).

/* A binding occurs if one of the fields has a variable where

the other has a constant. */

proposed_bindings([H|T1], [B|T2], [[H, B]|Rest]) :-

nonvar(H),

var(B),

!,

proposed_bindings(T1, T2, Rest).

proposed_bindings([H|T1], [B|T2], [[H, B]|Rest]) :-

nonvar(B),

var(H),

!,

proposed_bindings(T1, T2, Rest).

/* Otherwise, a binding does not occur */

proposed_bindings([_|T1], [_|T2], Pairs) :-

proposed_bindings(T1, T2, Pairs).



/* succeeds if the Cell is already Bound within the given

Background */

element_of(Cell, Background) :-

member(A, Background),

already_bound(Cell, A),

!.



/* succeeds if the Cell is a member of the given field */

already_bound(Cell, Field) :-

member(A, Field),

Cell == A,

!.

keybreak :- ifthen(get0_noecho(121),break).

template(Y, [Y, [H, [Y, E, Q, S, X, W, V, G]],

[G, [D, Y, E, H, W, V, U, F]],

[F, [P, D, Y, G, V, U, T, R]],

[D, [N, A, B, Y, G, F, R, P]],

[A, [I, J, K, B, Y, D, P, N]],

[B, [J, K, L, C, E, Y, D, A]],

[C, [K, L, M, O, Q, E, Y, B]],

[E, [B, C, O, Q, S, H, G, Y]]],

[[a, A], [b, B], [c, C], [d, D], [e, E],

[f, F], [g, G], [h, H], [i, I], [j, J],

[k, K], [l, L], [m, M], [n, N], [o, O],

[p, P], [q, Q], [r, R], [s, S], [t, T],

[u, U], [v, V], [w, W], [x, X], [y, Y]]).

/* returns the neighborhoods in a given Template */

neighbors([], []).

neighbors([[_,N]|Tail], [N|Rest]) :-

neighbors(Tail, Rest).



retractall(P) :- retract(P), fail.

retractall(_).

/* instantiates a template for each member of a list of individuals */

individuals([H|T], [[Template, Vars]|Tail]) :-

template(H, Template, Vars),

individuals(T, Tail).

individuals([], []).

/* prints the variable bindings */

matrix_print([]).

matrix_print([A,B,C,D,E|Rest]) :-

write(A), write(' '), write(B), write(' '), write(C), write(' '),

write(D), write(' '), write(E), nl,

matrix_print(Rest).

/* returns the length of a list */

its_length(List, Length) :-

il_aux(List, 0, Length).

il_aux([], C, C) :- !.

il_aux([_|T], Count, L) :-

Newcount is Count + 1,

il_aux(T, Newcount, L).


Bibliography

Akins, Kathleen (1993). "A Bat Without Qualities." in Consciousness, Davies and Humphreys, ed. Blackwell Publishers, Cambridge, MA:258­273.

Alexander, S. (1920). Space, Time, and Deity. Macmillan, London.

Allport, Alan (1988) "What concept of consciousness?" in Consciousness in Contemporary Science, A. Marcel and E. Bisiach, eds. Clarenden Press, Oxford:159­182.

Anscombe, G.E.M. (1971) "Causality and Determination" in Causation, Ernest Sosa and Michael Tooley, eds. Oxford University Press, NY: 88­104.

Armstrong, David (1968). A Materialist Theory of Mind. Routledge and Kegan Paul, NY.

Armstrong, David (1982). "Metaphysics and Supervenience." Critica 42:14:3­17.

Armstrong, David (1983). What Is A Law Of Nature. Cambridge University Press, NY.

Arvidson, P. Sven (1992) "The Field Of Consciousness: James and Gurwitsch" in Transactions Of The C.S. Peirce Society, 28:4: 833­855.

Baier, Annette C. (1994) "How To Get To Know One's Own Mind: Some Simple Ways" in Philosophy in Mind, Michaelis Michael and John O'Leary-Hawthorne, eds. Kluwer, pp. 65­82.

Baruss, Imants (1990) The Personal Nature of Notions of Consciousness. University Press of America, NY.

Bennett, Jonathan (1987) "Event Causation: The Counterfactual Analysis" in Causation, Ernest Sosa and Michael Tooley, eds. Oxford University Press, NY: 217­233.

Block, Ned (1980). "Troubles With Functionalism." in Readings In The Philosophy of Psychology, vol. 1., N. Block, ed. Harvard University Press, Cambridge, MA: 268­306.

Block, Ned (1995). "On a confusion about a function of consciousness." Behavioral And Brain Sciences 18:2:227-47.

Boghossian, Paul (1996) "Analyticity Reconsidered." Nous, 30:3:360­391.

Bohm, David (1980). Wholeness and the Implicate Order. Routledge, New York.

Boyd, Richard (1980) "Materialism Without Reductionism: What Physicalism Does Not Entail" Readings In The Philosophy of Psychology, Vol 1. N. Block, ed., Harvard University Press, Cambridge, MA: 67-107.

Braude, Stephen E. (1991). First Person Plural: Multiple Personality and the Philosophy of Mind. Routledge, New York.

Brink, David (1991) Moral Realism And The Foundation Of Ethics. Cambridge University Press, NY.

Carnap, R. (1928) The Logical Construction of the World, Berlin: Weltkreis; English translation by R.A. George, 1967, Berkeley and Los Angeles, University of California Press.

Carnap, R. (1937) The Logical Syntax Of Language, translated by A. Smeaton, London: K. Paul, Trench, Trubner.

Cartwright, Nancy (1983) How The Laws Of Physics Lie. Clarendon Paperbacks, NY.

Casteneda, Hector-Neri (1984) "Causes, Causity, and Energy" in Midwest Studies in Philosophy IX: Causation and Causal Theories, Peter A. French, et al eds. University of Minnesota Press, Minneapolis, MN: 17-27.

Chalmers, David (1995) "Facing Up to the Problem of Consciousness", Journal Of Consciousness Studies, 2:3:, pp. 200­219.

Chalmers, David (1996) The Conscious Mind. Oxford University Press, NY.

Chisholm, Roderick M. (1955) "Law Statements and Counterfactual Inference.", Analysis, 15:5:97­105.

Chisholm, Roderick M. (1966) "Freedom and Action." in Freedom and Determinism, Keith Lehrer, ed. Random House, NY.

Churchland, Patricia S. (1988) "Reduction and the neurobiological basis of consciousness." in Consciousness in Contemporary Science, A. Marcel and E. Bisiach, eds., Clarendon Press, Oxford:273­304.

Churchland, Patricia S. (1989). Neurophilosophy. The MIT Press, Cambridge, MA.

Churchland, Paul M.(1979). Scientific Realism and the Plasticity of Mind. Cambridge University Press, Cambridge.

Churchland, Paul M. (1989). "Functionalism, Qualia, and Intentionality." A Neurocomputational Perspective. The MIT Press, Cambridge, Massachusetts. pp. 23-46.

Churchland, Paul M. (1989). "Reduction, Qualia, and the Direct Introspection of Brain States" A Neurocomputational Perspective. The MIT Press, Cambridge, Massachusetts. pp. 47-76.

Churchland, Patricia S.; Churchland, Paul M. (1990) "Intertheoretic reduction: a neuroscientist's field guide" Seminars in the neurosciences, Vol. 2. pp. 249-256.

Collingwood, R.G. (1940) An Essay on Metaphysics. The Clarendon Press, Oxford.

Cowey, A.; Stoerig, P. (1992) "Reflections on Blindsight." in The Neuropsychology of Consciousness, A.D. Milner and M.D. Rugg, Eds. Academic Press Limited, San Diego, CA.

Cricke, Francis (1994) The Astonishing Hypothesis. Simon and Schuster, New York.

Das Rajarshi; Mitchell, Melanie; Crutchfield, James (1994) "A genetic algorithm discovers particle based computation in a cellular automata." in Parallel Problem Solving in Nature III, Y. Davidov, H-P Schwefel, and R. Manner, eds. Lecture Notes in Computer Science, Springer­Verlag, Berlin: 344­353.

Davidson, Donald (1967) "Causal Relations". Journal of Philosophy, 64: 426-441.

Dennett, Daniel C. (1981) "True Believers: The Intentional Strategy and Why it Works." in Mind And Cognition: A Reader (1990), William Lycan, ed. Basil Blackwell, Cambridge, MA: 150-166.

Dennett, Daniel C. (1981) "Making Sense of Ourselves" in Mind And Cognition: A Reader (1990), William Lycan, ed. Basil Blackwell, Cambridge, MA: 184-198.

Dennett, Daniel C. (1988) "Quining Qualia." In Consciousness in Contemporary Science, A. Marcel and E. Bisiach, eds., Clarenden Press, Oxford:42­77.

Dennett, Daniel C. (1991). Consciousness Explained. Little, Brown, and Company, Boston, MA.

Dretske, Fred (1986). "Misrepresentation." in Belief: Form, Content, and Function, R. Bogdan, ed., Oxford University Press, NY.

Dretske, Fred (1995) Naturalizing Mind. MIT Press, Cambridge, MA.

Ducasse, C.J. (1926) "On the Nature and Observability Of The Causal Relation." in Causation, Ernest Sosa and Michael Tooley, eds. Oxford University Press, NY: 125­136.

Dunn, Michael J. (1973) "A Truth Value Semantics For Modal Logic." in Truth, Syntax, and Modality, H. Leblanc, ed. North­Holland, Amsterdam.

Dunn, Michael J. (1987) "Relevant Predication 1: The Formal Theory" in Journal of Philosophical Logic 16:4:347-381.

Dunn, Michael J. (1990) "Relevant Predication 2: Intrinsic Properties And Internal Relations" in Philosophical Studies 60:3:177-206.

Dupre, John (1995) The Disorder Of Things. Harvard University Press, Cambridge, MA.

Eccles, J.C. (1986) "Do mental events cause neural events analogously to the probability fields of quantum mechanics?" Proceedings of the Royal Society of London B 227:4111-28.

Edelman, Gerald (1985) "Neural Darwinism: Population Thinking and Higher Brain Function." in How We Know. Michael Shafto, ed., Harper & Row, San Francisco, CA.

Edelman, Gerald (1989) The Remembered Present. Basic Books, Inc. NY.

Estep, Myrna (1996a) "A Critique Of James' Neutral Monism: Consequences For A New Science Of Consciousness" (unpublished manuscript)

Estep, Myrna (1996b) "A Theory Of Immediate Awareness And Knowing How: Self-Organization And Adaptation In Knowing Systems" (unpublished manuscript).

Fales, Evan (1990) Causation and Universals. Routledge, NY.

Fales, Evan (1993) "Are Causal Laws Contingent?" in Ontology, causality, and mind: essays in honour of David M. Armstrong, John Bacon, et al eds. Cambridge University Press, Cambridge.

Feynman, Richard P. (1985) QED. Princeton University Press, Princeton, New Jersey.

Fisk, Milton (1973) Nature and Necessity. Indiana University Press, Bloomington, IN.

Flanagan, Owen (1992) Consciousness Reconsidered. The MIT Press, Cambridge, MA.

Fodor, Jerry A. (1981). Representations:Philosophical Essays on the Foundations of Cognitive Science. The MIT Press, Cambridge, MA.

Foss, Jeffrey, E. (1995) "Materialism, Reduction, Replacement, and the Place of Consciousness in Science." Journal of Philosophy: vol. XCII, No. 8, 401-429.

Friedman, Michael (1987) "Carnap's Aufbau Reconsidered." Nous, 21:4:521-545.

Gazzaniga, Michael S. (1988) "Brain Modularity: towards a philosophy of conscious experience." Consciousness in Contemporary Science, A. Marcel and E. Bisiach, eds., Clarenden Press, Oxford: 218­238.

Goodman, Nelson (1965) Fact, Fiction, and Forecast, The Bobbs­Merill Company, Inc., NY.

Gray, Jeffrey (1996) "The Contents of Consciousness: A Neuropsychological Conjecture." Behavioral and Brain Sciences (preprint)

Gregory, Richard L. (1988) "Consciousness in science and philosophy: conscience and con­science" in Consciousness in Contemporary Science, A. Marcel and E. Bisiach, eds., Clarenden Press, Oxford: 257­272.

Hameroff, Stuart (1996) "Conscious Events as Orchestrated Space­Time Selections", Journal of Consciousness Studies, 3:1:36­53.

Hardin, C.L. (1988) Color for Philosophers: Unweaving the Rainbow. Hackett Publishing Company, Indianapolis, IN.

Harman, Gilbert (1994) "Doubts About Conceptual Analysis" in Philosophy in Mind, Michaelis Michael and John O'Leary-Hawthorne, eds. Kluwer:38­43.

Harre, R.; Madden, E. H. (1975) Causal Powers: A Theory Of Natural Necessity. Basil Blackwell, Oxford.

Hamilton, Edith; Cairns, Huntington, ed.(1961) "The Sophist." in Plato: Collected Dialogues. Princeton University Press.

Haugeland, John (1993) "Pattern and Being" in Dennett and his critics, Bo Dahlbom, ed., Blackwell Publishers, MA: 53­69.

Henderson, David K. (1994) "Accounting for macro-level causation." Synthese, 101:2:129-156.

Hodgson, David (1991). The Mind Matters. Clarendon Press, Oxford.

Horgan, Terence (1982) "Supervenience and microphysics" Pacific

Philosophical Quarterly vol 63, pp. 29-43.

Horgan, Terence (1987) "Supervenient qualia." Philosophical Review 96: 491-520.

Hume, David (1748) Inquiry Concerning Human Nature. Anthony Flew, ed. (1962): 17-164.

Jackendoff, Ray (1987). Consciousness and the Computational Mind. MIT Press, Cambridge, MA.

Jackson, Frank (1982). "Epiphenomenal Qualia." Philosophical Quarterly 32, no. 127:127-136.

Jackson, Frank (1986). "What Mary Didn't Know." Journal of Philosophy 83, no. 5:291-295.

Jackson, Frank (1993) "Block's Challenge" in Ontology, causality, and mind: essays in honour of David M. Armstrong: John Bacon, et al eds., Cambridge University Press, Cambridge.

Jackson, Frank (1994) "Armchair Metaphysics" in Philosophy in Mind, Michaelis Michael and John O'Leary-Hawthorne, eds. Kluwer, pp. 23-42.

Johnson­Laird, Philip (1988) "A computational analysis of consciousness." in Consciousness in Contemporary Science, A. Marcel and E. Bisiach, eds., Clarenden Press, Oxford:357­368.

Johnston, Mark (1992) "Constitution is not identity." Mind, 101.401:89­105.

Johnston, Mark (1996) "It Necessarily Ain't So" unpublished manuscript.

Kafatos, Menas; Nadeau,Robert (1990) The Conscious Universe: Part and Whole in Modern Physical Theory. Springer­Verlag, NY.

Kelso, J.A. Scott (1995) Dynamic Patterns: The Self-Organization of Brain and Behavior. Bradford, Cambridge, MA.

Kim, Jaegwon (1982). "Psychophysical Supervenience" Philosophical Studies, 41:1: 51-70.

Kim, Jaegwon (1984) "Epiphenomenal and Supervenient Causation." in Midwest Studies in Philosophy IX: Causation and Causal Theories, Peter A. French, et al eds. University of Minnesota Press, Minneapolis, MN: 257-270.

Kinsbourne, Marcel (1988) "Integrated field theory of consciousness" in Consciousness in Contemporary Science, A. Marcel and E. Bisiach, eds., Clarenden Press, Oxford: 219­238.

Kirk, Robert (1994) Raw Feeling Clarendon Press, Oxford.

Kneale, W.C. (1949) Probability and Induction. Oxford Press, Oxford.

Koch, Christof (1996) "Towards the neuronal substrate of visual consciousness" in Towards a Science of Consciousness: The First Tucson Discussions and Debates, S.R. Hameroff, A.W. Kaszniak, A.C. Scott, eds. MIT Press, Cambridge MA.

Koons, Robert C. (1997) "An Exact Theory of Causation: The Deterministic Case." (unpublished manuscript).

Koons, Robert C. (1997) "An Exact Theory of Causation: The Indeterministic Case." (unpublished manuscript).

Kripke, Saul (1971) "Identity and Necessity." in Meaning and Reference, A.W. Moore, ed., Oxford University Press, NY:162­191.

Kripke, Saul (1972). "Naming and Necessity." in Semantics of Natural Language, D. Davidson and G. Harman, eds., Dordrecht:254­355.

Lahav, Ran; Shanks, Neil (1992) "How to be a Scientifically Respectable 'Property-Dualist'." The Journal of Mind and Behavior, 13:3:211­232.

Langsam, Harold (1994) "Kant, Hume, and Our Ordinary Concept of Causation." Philosophy and Phenomenological Research, 54:3:625­647.

Levine, Janet (1986) "Could Love Be Like A Heatwave?:Physicalism and the Subjective Character Of Experience." Philosophical Studies, vol. 49, No. 2:245-261.

Levine, Joseph (1993) "On Leaving Out What It's Like" in Consciousness, Davies and Humphreys, ed. Blackwell Publishers, Cambridge, MA: 121-136.

Lewis, David (1973). "Causation." The Journal of Philosophy, 70:17:556­571.

Lewis, David (1981). "Are we free to break the laws?" Theoria, 47:3:113­121.

Lewis, David (1983). "Extrinsic Properties," Philosophical Studies, 44:197­200.

Lewis, David (1988). "What Experience Teaches," Proceedings Of The Russellian Society, University of Sidney. Reprinted in Mind And Cognition: A Reader, William Lycan, ed. (1990) Basil Blackwell, Cambridge, MA: 490-498.

Lewis, David (1995). "Should a Materialist Believe in Qualia?" Australasian Journal of Philosophy, 73:1:140­144.

Locke, John (1690) An Essay Concerning Human Understanding,Peter H. Nidditch, ed. (1975), Clarendon Press, Oxford.

Lockwood, Michael (1989) Mind, Brain, and Quanta. Basil Blackwell, Cambridge, MA.

Lycan, William G. (1987). Consciousness. The MIT Press, Cambridge, MA.

Lycan, William (1990) "The Continuity of Levels of Nature" in Mind And Cognition: A Reader, William Lycan, ed. Basil Blackwell, Cambridge, MA: 80-96.

Lycan, William G. (1996). Consciousness and Experience. Cambridge, MA.

Mackie, John L. (1965) "Causes and Conditions" in Causation, Ernest Sosa and Michael Tooley, eds. Oxford University Press, NY: 33-55.

Mackie, John L. (1974) The Cement of the Universe: A Study of Causation. The Clarendon Press, Oxford.

Marcel, Anthony (1988) "Phenomenal Experience and functionalism." Consciousness in Contemporary Science, A. Marcel and E. Bisiach, eds. Clarenden Press, Oxford:121-158.

Marks, Charles E. (1981) "Commissurotomy, Consciousness, and Unity of Mind." MIT Press, Cambridge, MA.

Martin, C.B. (1993) "Power for Realists" in Ontology, causality, and mind: essays in honour of David M. Armstrong: John Bacon, et al eds. Cambridge University Press, Cambridge.

McDowell, John (1994) Mind And World. Harvard University Press, Cambridge, MA.

McGinn, Colin (1989) "Can We Solve The Mind­Body Problem?" Mind 98:349-66.

McGinn, Colin (1993) "Consciousness and Cosmology: Hyperdualism Ventilated" in Consciousness, Davies and Humphreys, ed. Blackwell Publishers, Cambridge, MA:155-177.

Mellor, D.H. (1995) The Facts Of Causation. Routledge, NY.

Menzies, Peter (1993) "Laws of Nature, Modality and Humean Supervenience" in Ontology, causality, and mind: essays in honour of David M. Armstrong: John Bacon, et al Ed. Cambridge University Press.

Metzinger, Thomas (1995) "Faster than thought: Holism, Homogeneity, and Temporal Coding." in Conscious Experience, Thomas Metzinger, ed. MIT Press.

Millikan, Ruth G. (1984) Language, Thought, and Other Biological Categories. MIT Press.

Millikan, Ruth G. (1993) White Queen Psychology and Other Essays for Alice. MIT Press.

Minsky, Marvin (1986). The Society of Mind. Simon & Schuster, NY.

Moss, Edward (1995) "The Grammar of Consciousness" St. Martin's Press, NY.

Nagel, Earnest (1961) The Structure of Science. Routledge & Kegan Paul Ltd., Britain.

Nagel, Thomas (1974) "What Is It Like To Be A Bat?", Philosophical Review,83:4: 435­450.

Nagel, Thomas. (1986) The View From Nowhere. Oxford University Press, NY.

Nelkin, Norton (1993) "The Connection Between Intentionality And Consciousness." in Consciousness, Davies and Humphreys, ed. Blackwell Publishers, Cambridge, MA:197­223.

Nemirow, Laurence (1990) "Physicalism and the Cognitive Role of Acquaintance" in Mind And Cognition: A Reader, William Lycan, ed. Basil Blackwell, Cambridge, MA: 490-498.

Newcombe, Freda (1985) "Neuropsychology of consciousness: a review of human clinical evidence." in Brain and Mind, David A. Oakley, Ed. Methuen, London and New York.

Newman, James (1996) "Reticular­thalamic activation of the cortex." (Unpublished manuscript).

Oakley, David A.; Lesley C. Eames (1985) "The plurality of consciousness" in Brain and Mind, David A. Oakley, Ed. Methuen, London and New York.

O'Connor, Timothy (1997) "Ultimate Explanation and the Necessity of Natures: A Cosmological Argument Revisited." (Unpublished manuscript).

Owens, Virginia Stem (1983). And The Trees Clap Their Hands. Wm. B. Eerdmans Pub., Grand Rapids, MI.

Parfit, Derek (1984) Reasons and Persons. Oxford University Press, Oxford.

Penrose, Roger (1989). The Emperor's New Mind. Oxford University Press, NY.

Plantinga, Alvin (1974) The Nature of Necessity. Oxford University Press, London.

Poland, Jeffrey (1994) Physicalism. Clarendon Press, Oxford.

Posner, M.I.; Rothbart, M.K. (1992) "Attentional Mechanisms and Conscious Experience" in The Neuropsychology of Consciousness, A.D. Milner and M.D. Rugg, eds. Academic Press Limited, San Diego, CA.

Poundstone, William (1985) The Recursive Universe. William Morrow and Company, Inc. NY.

Prigogine, Ilya; Stengers, Isabelle (1984) Order Out of Chaos. Bantam, London.

Putnam, Hilary (1960) "Minds and Machines" in Dimensions of Mind. S. Hook, ed., Springer-Verlag, Berlin.

Putnam, Hilary (1967) "The Nature of Mental States." in Mind And Cognition: A Reader (1990), William Lycan, ed. Basil Blackwell, Cambridge, MA: 47-56.

Putnam, Hilary (1973) "Meaning and Reference." The Journal Of Philosophy, 70, 19: 699­711.

Putnam, Hilary (1987) The Many Faces of Realism. Open Court, Lasalle, Ill.

Putnam, Hillary (1984) "Is the Causal Structure of the Physical Itself Something Physical?" in Midwest Studies in Philosophy IX: Causation and Causal Theories, Peter A. French, et al eds. University of Minnesota Press, Minneapolis, MN: 3-16.

Quine, W.V. (1963) "Two Dogmas Of Empiricism," in From a Logical Point of View, Harper & Row, New York.

Quine, W.V. (1992) The Pursuit of Truth. Harvard University Press, Cambridge, MA.

Robinson, William S. (1982) "Causation, Sensation, and Knowledge." Mind, vol. XCI: 524­540.

Robinson, William S. (1996) "The Hardness Of The Hard Problem." Journal Of Consciousness Studies, 3:1: 14-25.

Rosenberg, Gregg H. (1996) "Rethinking Nature: A Hard Problem Within The Problem." Journal Of Consciousness Studies, 3:1:76­88.

Rosenberg, Gregg H. (1997) "Circularity and the Sorites" (unpublished).

Rosenthal, David M. (1993) "Thinking That One Thinks." in Consciousness, Davies and Humphreys, ed. Blackwell Publishers, Cambridge, MA:197­223.

Russell, Bertrand (1961) "A Free Man's Worship." in Ethical Theory, Louis P. Pojman, ed., (1987), Wadsworth Publishing, Belmont, CA.

Ryle, Gilbert (1949) The Concept of Mind, Hutchinson's University Library, London.

Salmon, Wesley (1980) "Probabilistic Causality", Pacific Philosophical Quarterly, 61:1: 50­74.

Salmon, Wesley (1980) "Causality: Production And Propagation", in Causation, Ernest Sosa and Michael Tooley, eds. Oxford University Press, NY: 154­171.

Salmon, Wesley (1984). Scientific Explanation and the Causal Structure of the World. Princeton University Press, New Jersey.

Sanford, David (1984) "The Direction of Causation and the Direction of Time." in Midwest Studies in Philosophy IX: Causation and Causal Theories, Peter A. French, et al eds. University of Minnesota Press, Minneapolis, MN: 53­75.

Scott, Alwyn (1995). Stairway to the Mind. Springer­Verlag, NY.

Searle, John (1980). "Minds, Brains, and Programs," Behavioral

and Brain Sciences, 3:3:417­458.

Searle, John (1983) Intentionality: An Essay in the Philosophy of Mind. Cambridge University Press, Cambridge.

Searle, John (1992). The Rediscovery of the Mind. MIT Press, Cambridge, MA.

Sellars, Wilfrid (1963) "Philosophy and the Scientific Image of Man." in Science, Perception, and Reality. The Humanities Press, New York.

Sellars, Wilfrid (1963) "Some Reflections on Language Games." in Science, Perception, and Reality. The Humanities Press, New York.

Shoemaker, Sydney (1975). "Functionalism and Qualia." Philosophical Studies, 27:5: 291­315.

Shoemaker, Sydney (1996) "Causal And Metaphysical Necessity" (unpublished manuscript).

Siewart, Charles Peter (1994) "Understanding Consciousness" Philosophy dissertation, University of California, Berkeley.

Skyrms, Brian (1980) Causal Necessity: A Pragmatic Investigation of the Necessity of Laws. Yale University Press, New Haven.

Skyrms, Brian (1984) "EPR: Lessons for Metaphysics." in Midwest Studies in Philosophy IX: Causation and Causal Theories, Peter A. French, et al eds. University of Minnesota Press, Minneapolis, MN:245­255.

Smart, J.J.C. (1959) "Sensations and brain processes." Philosophical Review 68:141­56.

Smart, J.J.C. (1993) "Laws of Nature as Species of Regularity" in Ontology, causality, and mind: essays in honour of David M. Armstrong: John Bacon, et al ed. Cambridge University Press.

Sommerhoff, Gerd (1990) Life, Brain, and Consciousness. North­Holland, New York.

Sosa, Ernest(1980) "Varieties of Causation" in Causation, Ernest Sosa and Michael Tooley, eds. Oxford University Press, NY: 234­242.

Squires, Euan (1990) Conscious Mind in the Physical World. Adam Hilger, NY.

Stalnaker, Robert (1968) "A Theory Of Conditionals." Studies in Logical Theory, American Philosophical Quarterly, Monograph Series, No. 2; Nicholas Rescher, ed., Basil Blackwell, Oxford:98­112.

Stapp, Henry (1996) "The Hard Problem: A Quantum Approach" The Journal Of Consciousness Studies, 3:3: 194­210.

Stapp, Henry (1996) "Chance, Choice, and Consciousness: A Causal Quantum Theory of the Mind/Brain" (unpublished manuscript).

Stich, Stephen (1981). "Dennett On Intentional Systems." in Mind And Cognition: A Reader (1990), William Lycan, ed. Basil Blackwell, Cambridge, MA:167-184.

Strawson, P.F. (1959) Individuals. University Paperbacks, London.

Swoyer, Chris (1984). "Causation and Identity" in Midwest Studies in Philosophy IX: Causation and Causal Theories, Peter A. French, et al eds. University of Minnesota Press, Minneapolis, MN:593­622.

Tooley, Michael (1984) "Laws and Causal Relations." in Midwest Studies in Philosophy IX: Causation and Causal Theories, Peter A. French, et al eds. University of Minnesota Press, Minneapolis, MN: 93-112.

Tooley, Michael (1987) Causation: A Realist Approach. Oxford University Press, Oxford.

Tooley, Michael (1990) "Causation: Reductionism Versus Realism", Philosophy And Phenomenological Research (Supplement), 50: 215­236.

Tye, Michael (1995) Ten Problems of Consciousness: A Representational Theory of the Phenomenal Mind. The MIT Press, Cambridge, MA.

Tye, Michael (1996) "The Function of Consciousness." Nous, 30:3:287­305.

Van Gulick, Robert (1993) "Understanding the Phenomenal Mind: Are We All just Armadillos?" in Consciousness, Davies and Humphreys, ed. Blackwell Publishers, Cambridge, MA:137-154.

Vendler, Zeno (1962) "Effects, Results, and Consequences", Analytical Philosophy, 1st ser., Butler, Ronald J., ed., Oxford University Press:1­15.

Wallace, W. (1974) Causality And Scientific Explanation, Vol I & II. The University of Michigan Press, Ann Arbor.

Weber, Marcel (1996) "Fitness Made Physical: The Supervenience Of Biological Concepts Revisited." Philosophy of Science, 63:411-431.

Weinberg, Steven (1992) Dreams Of A Final Theory. Random House, NY.

Weiskrantz, Lawrence (1988) "Some contributions of neuropsychology of vision and memory to the problem of consciousness." in Consciousness in Contemporary Science, A. Marcel and E. Bisiach, eds., Clarendon Press, Oxford:183­199.

Whitehead, Alfred North (1929) Process and Reality (corrected edition). Griffin, Donald Ray; Sherburne, Donald W. eds. The Free Press, NY 1978.

Wilkes, Kathleen (1988) ", yishi, duh, um, and consciousness" Consciousness in Contemporary Science, A. Marcel and E. Bisiach, eds., Clarenden Press, Oxford: 16­41.

Yablo, Stephen (1984) "Identity, Essence, and Indiscernability" Journal of Philosophy, 84:6: 293­314.

Yablo, Stephen (1993) "Is Conceivability a Guide to Possibility?" Philosophy and Phenomenological Research, 53:1: 1­42.

Yablo, Stephen (1993) "Cause and Essence" Synthese, 93:3: 403­449.