Natural fabrications: science, emergence and consciousness 9783642295980, 9783642295997, 3642295983

Table of contents :
NATURALFABRICATIONS......Page 4
Acknowledgments......Page 6
Contents......Page 7
1 Overview......Page 9
Part IThe Scientific Picture of the World......Page 14
Part IIEmergence and Consciousness......Page 68
Notes......Page 213

Citation preview

THE FRONTIERS COLLECTION

Series Editors Avshalom C. Elitzur Unit of Interdisciplinary Studies, Bar-Ilan University, 52900, Ramat-Gan, Israel e-mail: [email protected] Laura Mersini-Houghton Department of Physics, University of North Carolina, Chapel Hill, NC 27599-3255 USA e-mail: [email protected] Maximilian Schlosshauer Department of Physics, University of Portland, 5000 North Willamette Boulevard Portland, OR 97203, USA e-mail: [email protected] Mark P. Silverman Department of Physics, Trinity College, Hartford, CT 06106, USA e-mail: [email protected] Jack A. Tuszynski Department of Physics, University of Alberta, Edmonton, AB T6G 1Z2, Canada e-mail: [email protected] Rudy Vaas Center for Philosophy and Foundations of Science, University of Giessen, 35394, Giessen, Germany e-mail: [email protected] H. Dieter Zeh Gaiberger Straße 38, 69151, Waldhilsbach, Germany e-mail: [email protected]

For further volumes: http://www.springer.com/series/5342

THE FRONTIERS COLLECTION Series Editors A. C. Elitzur L. Mersini-Houghton M. Schlosshauer M. P. Silverman J. A. Tuszynski R. Vaas H. D. Zeh

The books in this collection are devoted to challenging and open problems at the forefront of modern science, including related philosophical debates. In contrast to typical research monographs, however, they strive to present their topics in a manner accessible also to scientifically literate non-specialists wishing to gain insight into the deeper implications and fascinating questions involved. Taken as a whole, the series reflects the need for a fundamental and interdisciplinary approach to modern science. Furthermore, it is intended to encourage active scientists in all areas to ponder over important and perhaps controversial issues beyond their own speciality. Extending from quantum physics and relativity to entropy, consciousness and complex systems—the Frontiers Collection will inspire readers to push back the frontiers of their own knowledge.

For a full list of published titles, please see back of book or springer.com/series/5342

William Seager

NATURAL FABRICATIONS Science, Emergence and Consciousness

123

William Seager University of Toronto Scarborough Scarborough, ON Canada

ISSN 1612-3018 ISBN 978-3-642-29598-0 DOI 10.1007/978-3-642-29599-7

ISBN 978-3-642-29599-7

(eBook)

Springer Heidelberg New York Dordrecht London Library of Congress Control Number: 2012939636 Ó Springer-Verlag Berlin Heidelberg 2012 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. Exempted from this legal reservation are brief excerpts in connection with reviews or scholarly analysis or material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work. Duplication of this publication or parts thereof is permitted only under the provisions of the Copyright Law of the Publisher’s location, in its current version, and permission for use must always be obtained from Springer. Permissions for use may be obtained through RightsLink at the Copyright Clearance Center. Violations are liable to prosecution under the respective Copyright Law. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. While the advice and information in this book are believed to be true and accurate at the date of publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be made. The publisher makes no warranty, express or implied, with respect to the material contained herein. Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)

Acknowledgments

It is impossible to thank by name all the people who have helped me on this project in one way or another, but I am very grateful for all the comments and criticism. I do want to acknowledge the extraordinary efforts of my two very able research assistants, Matt Habermehl and Adrienne Prettyman, whose help was invaluable, as well as the aid and encouragement of Angela Lahee of Springer-Verlag. Much of this work has been presented at conferences devoted to consciousness studies, most especially the famous biennially alternating conferences which under the optimistic banner of ‘Towards a Science of Consciousness’, are held respectively in Tucson, Arizona, and various interesting places around the world. All of the organizers, commentators, discussants, and attendees of the TSC and other conferences—where philosophy and science come alive—get my thanks. This project was materially aided by generous grants from the Centre for Consciousness Studies at the University of Arizona and the Social Sciences and Humanities Research Council of Canada as well as travel grants from my own institution, the University of Toronto Scarborough. Finally, I would like to thank my wife, Christine McKinnon, for her unflagging intellectual and emotional support, not to mention supreme patience. Toronto, March 2012

v

Contents

1

Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Part I

1

The Scientific Picture of the World

2

Looking Out . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

9

3

Looking In. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

25

4

Consciousness in the Brain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

37

Part II

Emergence and Consciousness

5

Emergence and Cellular Automata . . . . . . . . . . . . . . . . . . . . . . . .

65

6

Against Radical Emergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

85

7

Emergence and Supervenience . . . . . . . . . . . . . . . . . . . . . . . . . . .

121

8

Generalized Epiphenomenalism . . . . . . . . . . . . . . . . . . . . . . . . . .

155

9

The Paradox of Consciousness . . . . . . . . . . . . . . . . . . . . . . . . . . .

175

10 Embracing the Mystery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

189

vii

viii

Contents

Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

211

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

243

Titles in this Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

259

Chapter 1

Overview

On the face of it, emergence is everywhere. At least, if by ‘emergence’ we mean the ubiquitous fact that most things possess properties which their constituents lack, then emergence is everywhere. The simplicity of this conception masks a great deal of complexity however. There are many forms of emergence, some of which are most certainly present in our world while others are doubtful and clash with our best developed picture of the world. This book is about the nature of emergence, the scientific picture of the world and the ultimate question of whether and how consciousness fits into that picture. The overall arc of argumentation of the book is quite straightforward. Part I aims to sketch the scientific picture of the world against which the entire argument is constructed. To some readers, the material in Part I will not be new. Readers familiar with this material or willing to grant the range, power and comprehensiveness of the scientific picture of the world should thus feel free to skim the first three chapters. But the point of Part I is not mainly to impart scientific information. It is rather to emphasize the fact that science has been astonishingly successful in constructing an account of the world which is comprehensive, unified and yet hierarchical. It is strange but little remarked that science has been so very successful in building deeply connected explanatory theories that appear to reveal a breathtaking unity in nature. These connections reach so deep and far that they are today beginning to encompass the most mysterious feature of the universe: consciousness. In every area of study we find anchor points linking emergent features to the fundamental constituents of our world. Throughout science there are rich interconnections which underpin ever more comprehensive explanatory projects. No permanent dead end has yet blocked our progress and everything seems to hang together remarkably well. All this suggests a view of the world—one I hope readers find to be virtually commonsense—which I call the scientific picture. According to this picture there are fundamental features of the world, which it is the province of physics to reveal, and then a vast and convoluted hierarchy of ‘higher level’ entities, properties, processes and laws of nature. The scientific picture thus endorses emergence but of a particular kind, which I label ‘conservative emergence’. This kind of emergence has gone by

W. Seager, Natural Fabrications, The Frontiers Collection, DOI: 10.1007/978-3-642-29599-7_1, © Springer-Verlag Berlin Heidelberg 2012

1

2

1 Overview

many names: benign, weak, epistemological. Its core claim is that all emergent features are strictly and entirely determined by the nature and state of the fundamental physical constituents of the world. That is, any two worlds that were indistinguishable with respect to all their fundamental physical features, would necessarily be indistinguishable with respect to all their emergent features as well. It is not obvious that the world’s emergence is all and only conservative. Other forms of emergence are perfectly conceivable. Many philosophers and scientists have held that conservative emergence is inadequate to accommodate the diversity, dynamics and organization of the world. I will call the main competitor to conservative emergence ‘radical emergence’ (also known as strong or ontological emergence). Radical emergence denies that the state and nature of the physical fundamentals wholly and strictly determine, by their own nature, all the emergent features of the world. According to radical emergence, the world has some leeway as to its total structure in despite of its basic physical state. That is, any two worlds that were indistinguishable with respect to all their fundamental physical features, could nonetheless be distinguishable with respect to all their radically emergent features. The simplest way to conceive how radical emergence would work is to countenance the presence of ‘laws of emergence’ which are not included in the correct theory of the physically fundamental. Such ‘extra’ laws could then vary from world to world, leading to different sorts of emergent features. Despite the clear distinction between conservative and radical emergence it is surprisingly easy to confuse them. The reason is that the conservative-radical distinction interacts with a second distinction which is of much greater practical importance to working scientists. This is the distinction between accessible, or intelligible, versus inaccessible explanatory structures. Roughly speaking, conservative emergence coupled with accessible explanations of the emergents is reduction. Since reduction is composed of two distinct components it is possible for reductionism to fail even if conservative emergence exhausts all the emergence there is in the world. Many examples of complex systems can be found where there is no doubt that the emergent level is conservative, but there is no hope of explaining or predicting what is happening at that level in terms of lower levels. The simplest example is that of cellular automata, which will be discussed in Chap. 5. Of course, the big question is: is there any radical emergence in the world? Part I of this book assembles a tiny fragment of the vast amount of indirect empirical evidence we have that suggests not. If there is radical emergence, it seems to be very reluctant to reveal itself. Part II begins with more theoretical arguments for the conclusion that conservative emergence is all we need within the confines of the scientific picture of the world. The general concept of emergence is sharpened by an examination of the perfect imaginary playground: John Conway’s game of life. Radical and conservative emergence can be precisely defined in this domain. The general application of cellular automata to the problem of emergence opens up further vistas, all the way to the speculative possibility of digital physics. If the world is digital we can derive something close to a definitive proof that conservative emergence is the only kind of emergence at work in the world.

1 Overview

3

Chapter 6 attempts to undercut arguments in favour of radical emergence more directly, and via a more conventional route. Although some have argued that chaos in dynamical systems supports the existence of radical emergence I find insufficient support for this claim. Without doubt, classical dynamical systems exhibit emergence but conservative emergence seems sufficient to account for it. An important point, which recurs throughout the book, is that the claim that conservative emergence exhausts emergence does not imply any strong reductionist conclusions. Strong reductionism entails that the reduced theory can by jettisoned in favour of the reducing theory in explanatory contexts. In fact, this is seldom if ever true and the discussion of emergence and chaos shows why. Issues of complexity, limits of measurement accuracy and questions of system scale intelligibility all imply that reference to emergent features is indispensable for successful explanation, prediction and understanding. But this vitally important aspect of emergence is entirely compatible with conservative emergence. Chapter 6 ends with a discussion of the issue of whether quantum mechanics introduces such a deep change in our understanding of nature that it mandates acceptance of some form of radical emergence. Again, I think not. QM undeniably introduces new features, most especially the superposition principle and entanglement, which force modifications to our conception of emergence. But I argue that quantum emergence remains within the domain of conservative emergence. Chapter 7 is the final chapter devoted to understanding emergence. Here, I develop a detailed analysis of emergence based on the concept of supervenience. A powerful philosophical tool, the basic idea of supervenience is that of a principled dependence of one domain upon another (so obviously already intimately related to the topic of emergence). The mantra of supervenience is this: domain A supervenes upon domain B when there can be no change in A without a change in B. This simple formulation disguises a complex notion which this chapter explores. Chapter 7 is the most technical part of the book, being somewhat, in the words of Thomas Hobbes, ‘covered over with the scab of symbols’. I think the precision and clarity of logical symbolism is critical, and as deployed here requires only a basic understanding of first order syntax. I have endeavoured in any case to provide a clear textual explication of every formula and its role in the argumentation. This chapter develops a precise general definition of emergence, the radical-conservative distinction and forges an interesting link between the temporal dynamics of system evolution and the supervenience based account of emergence. The book then turns from articulating our understanding of emergence to exploring the consequences of emergence, particularly the consequences of the world being restricted to conservative emergence in accord with the scientific picture of the world. To this end Chap. 8 argues that all emergent features must be, in a sense, epiphenomenal. That is, such features cannot possess distinctive causal efficacy of their own. Perhaps that is not so surprising given that conservative emergence implies that everything about emergent features is derivative from the fundamental physical structure of the world. It also may not seem very threatening given that we can discern a kind of conservatively emergent causation, which derives from the fundamental level of causation in roughly the same way that conservatively emergent laws of nature appear. Once again, it is important to stress that the epiphenomenal character of conservative

4

1 Overview

emergents does not mean they should be ignored or demoted from pride of explanatory place. But the metaphysically real features which drive the world from state to state or which, in the word of C. Lloyd Morgan, provide the ‘go’ of the world all reside at the fundamental level alone. What is the upshot of this generalized epiphenomenalism? I argue in Chap. 9 that it leads to a severely worrying consequence: the paradox of consciousness. In a nutshell, the problem is this. By their nature, conservative emergents have no actual role in the world, apart from being appreciated and deployed by those conscious entities that find them, or concepts of them, helpful in categorizing and organizing their experience. These emergents stand as mere epistemic potentials, awaiting take-up by some appreciative mind. They are the ‘natural fabrications’ of my title. The paradox looms when we realize that consciousness itself, according to the scientific picture of the world, must be merely a conservatively emergent feature. The scientific picture strongly suggests that consciousness is the product of vastly complicated neural systems which appeared in the universe long after its creation. But if consciousness itself is a conservatively emergent feature then it too stands as epiphenomenal and a mere epistemic resource which appears only to...consciousness. There is a worrying whiff of circularity here which I argue is the fundamental and irresolvable problem with the treatment of consciousness in the scientific picture of the world. The paradox of consciousness leaves us in a quandary. The success of the scientific picture cannot lightly be thrown away, but the paradox undercuts its most basic presuppositions. In the absence of any settled solution, Chap. 10 offers a range of possible escape routes which I will sketch here in anticipation of their full development. I call these options: Watchful Waiting, Embracing Emergence, Favouring Fundamentality and Modifying Metaphysics. The first involves simply waiting and hoping that further advances within science will both continue to support the scientific picture of the world and somehow integrate consciousness into that picture without falling afoul of the paradox. Sure to be a popular option, it nonetheless leaves our problem entirely unresolved. If the paradox of consciousness is an essential consequence of the scientific picture then no amount of waiting and pursuing standard research strategies will ever provide a resolution. The other options are thus of necessity quite radical departures from the scientific picture. As its name suggests, Embracing Emergence requires the introduction of radical emergence into our view of the world. One difficulty with this option is the worrying question that if radical emergence is present in the world then why has it not revealed itself heretofore? Those who embrace emergence will have to explain why the world does such a good job of pretending that conservative emergence suffices. That and other problems are explored in Chap. 10. The traditional alternative to radical emergence in the face of problems of integrating recalcitrant phenomena into some theoretical domain is to expand the range of fundamental features. In this case, the option of Favouring Fundamentality will require that consciousness itself, in some form or other, presumably of an extremely primitive, simple and unstructured kind, be added to the roster of the universe’s truly basic or fundamental entities. One such view is panpsychism, but there are other possibilities. Sheer incredulity is a frequent response to panpsychist views, but an

1 Overview

5

interesting case can be made that panpsychism is actually the minimal alteration of the scientific picture of the world which can resolve the paradox of consciousness. The final option, Modifying Metaphysics, will strike many as the most radical of all. It stems from the observation that there is a tacit but vital premise underlying the scientific picture of the world: scientific realism. This is the position holding that science provides the best, perhaps only, route towards knowledge of the ultimate nature of reality. It is the core belief supporting the idea that the scientific picture is and should be our metaphysics of nature. But alternative views of science and nature have been advanced with considerable philosophical sophistication and from a base of excellent scientific knowledge. The rejection of scientific realism forces a radical alteration in the dialectic which leads to the paradox of consciousness. The result is highly interesting and perhaps encouraging, especially insofar as the denial need not undercut the explanatory power of science within its proper sphere. In basic terms, these options appear to exhaust the alternatives. I wish I could prove that one of them offers the one true path to enlightenment. But I fear that the still immature state of our understanding of science, emergence, consciousness and their relationships compels us to leave the issue unresolved. While the problem of consciousness is endlessly fascinating, it will, I think, become ever more pressing and intellectually visible as our knowledge of the anchor points between brain and consciousness grows over the next century.

Part I

The Scientific Picture of the World

Chapter 2

Looking Out

2.1 Lonely Earth Try to imagine a Lonely Earth in a universe very different from ours, where the night sky is almost empty. To the naked eye, there appear only seven celestial objects which slowly move around the night sky in bizarre and complex ways. Some rare nights would be utterly black, others—equally rare—would reveal all five visible planets and the moon. In such a world, the history and practical use of astronomy would have followed a course very different from ours, even far back into ancient times—just think of the effect on early navigators of there being no orienting stars, such as the conveniently located ‘north star’.1 Without the field of fixed stars to provide a background frame of reference, it would have been very difficult for protoastronomers to wrestle order out of the apparently unruly planetary behaviour. And very likely more than astronomy would have suffered. Conceivably, mathematics itself would never have really gotten off the ground and perhaps Plato’s vision of an eternal, unchanging realm of perfect truth and beauty would have been almost unthinkable. But for the sake of this thought experiment, let us instead suppose that at least science and technology advanced in our imaginary world pretty much as they did in ours, so that Galileo’s twin could, in the realm of observation, eventually turn his crude telescope onto the few available celestial objects and, in the realm of theory, could apply Euclid and read the language of mathematics in the great ‘book’ of the natural world. This imagined world would present no impediment to Galileo’s earth bound experiments on motion and gravitation. And it seems that this observer could more or less duplicate many of our own Galileo’s astronomical discoveries, such as the mountains of the moon and the system of Jovian satellites. It is true that without the fixed backdrop of stars it would be a lot more difficult to measure the position of the planets accurately, but it would not be altogether impossible (we can envisage using the horizon and time of year as the basis of a system of celestial measurement). So we might be permitted to go on to imagine figures akin to Brahe and Kepler, who led the way to the mathematical understanding W. Seager, Natural Fabrications, The Frontiers Collection, DOI: 10.1007/978-3-642-29599-7_2, © Springer-Verlag Berlin Heidelberg 2012

9

10

2 Looking Out

of planetary motion via universal gravitation which was perfected by Newton (it is hard to see, though, how—lacking the supreme regularity of the stellar ‘sphere’ as an anchor—the Ptolemaic system would ever be devised in our imaginary world). Now let the people of our imagined lonely Earth continue their scientific development in parallel to actual progress, up to modern times. Visual access to a starry sky does not seem crucial for this development. But as their science moves towards the equivalent of our twenty-first century physics, severe strains would appear with respect to the science of cosmology. Even in the actual world, this is a strange science, with no prospects for direct experimentation and reliant on disturbingly long and complex chains of argumentation linking its hypotheses with the relatively few observations to which they are relevant. Nonetheless, our cosmologists have managed to produce a grand and beautiful account of the overall structure of the universe, from its inception (minus a few picoseconds) to its possible end states. But how could any kind of a scientific cosmology get off the ground on Lonely Earth? We can assume that as their telescopes increased in quality and power, they would discover Uranus, Neptune and Pluto; in fact, they would have the advantage in spotting new planets and dwarf planets (if they thought to look and had the patience) for there would be no field of stars with which to confuse a planet. But even measuring the size of the solar system would be difficult. We did it by using trigonometry and parallax, the apparent displacement of a nearby object against a distant background caused by the motion of an observer (you can readily observe parallax by holding up a finger at arms length and viewing it through each eye in turn). Without the fixed stars, our imaginary astronomers would have had to devise a notional celestial coordinate system, and instruments of sufficient accuracy that reliable assignments of location could be made without the guide of any visible background (notably, it would require exceptionally accurate clocks, which took a long time to develop in our world). At the very least, this would have made their distance measurements far more uncertain than ours were.2 In any case, once they had their solar system mapped out, no further advance in cosmology would seem to be possible. On the theoretical side, the development of relativity theory and quantum mechanics could proceed (on Earth, astrophysical data had very little to do with the creation of either special or general relativity, and nothing to do with the creation of quantum mechanics). As theoretical curiosities— even in our world they were little more than that for some time—general relativists could create the various cosmological models, such as Einstein’s static universe or the Friedmann/Lemaître expanding universe model. But these theoretical exercises would lack any contact with sources of evidence relevant to cosmology. One of the very few (and highly influential) early tests of general relativity was the 1919 measurement by Eddington of how starlight is bent by the gravitational field of the Sun. In our imagined world, such a test would be impossible (the other early test however, the relativistic explication for the behaviour of Mercury’s orbit, could have been accomplished, at least in principle, given that they could measure the position of Mercury with sufficient accuracy). What could the scientists of that world say about the creation and evolution of the universe? Next to nothing. I would venture to guess the best account they could come

2.1 Lonely Earth

11

up with would barely go beyond Laplace’s ‘nebular hypothesis’.3 In the absence of any of the stellar nebulosities which so piqued the curiosity of our astronomers and set the stage for crucial debates about the extent of the Milky Way galaxy and the overall size of the universe, even that would take real creative genius. The nebular theory of solar system formation states that in the beginning was a slowly revolving (at least in the sense that it had some net angular momentum) amorphous cloud of gas and dust which gradually collapsed due to its own internal gravitational attraction. The increase in spin caused by the collapse would produce a disk of material, out of which would form the planets as the sun formed at the disk’s centre. Powerful basic evidence for some such process lies in the two simple facts that all the planets revolve around the Sun in the same direction and their orbits lie pretty much in the same plane. Being in possession of quantum mechanics, our imagined scientists could discover why the sun shines and they would then, perhaps, have a reasonable account of the formation of their solar system from the assumption of a primordial gas cloud. Their chemistry and biology could be as advanced as ours; they would have as much reason as we do to believe that life is a complex chemical process driven by evolution so they might well conjecture that life emerged from an earlier lifeless state. But the whole system would remain deeply puzzling. The initial cloud of gas and dust from which their solar system formed would have to be simply postulated as, in the first place, existing, and then as containing just the mix of elements which they would observe in the current form of their solar system. But then, assuming our imaginary scientists discover how the sun can forge new elements through a host of possible nucleosynthetic pathways, they would have an elegant solution to a non-existent problem. For there is nothing in their world that could generate these elements. Yet I don’t think our Lonely Earth scientists could reject the nebular hypothesis in favour of the conclusion that their universe was eternal and essentially unchanging. Several lines of thought and evidence would lead them to the opposite conclusion. In the first place, they could calculate how long the sun should be able to generate energy via nuclear fusion, given its composition and mass. This would tell them that sun could only be a few billion years old. Second, from the rate of radioactive decay of various elements present within the bodies of the solar system, they would find that none were more than 4 billion or so years old (this would be in concordance with the age limit of the Sun). They could also note that the face of the moon, for example, reveals a violent history of impacts which no longer occur at high frequency, obviously suggesting that the solar system is not in some kind of equilibrium state as would be expected if it were eternal. Finally, even if they allowed for some unknown physics which permitted the Sun to shine indefinitely and which somehow eliminated the evidently finite age of the material in the solar system, very fundamental dynamical considerations would reveal that the system itself was unstable over extremely long periods of time (see Lecar et al. 2001). Thus our imaginary scientists would have good reasons to believe that their solar system had a finite age of 5–10 billion years. This would put them in a difficult position, for the evidently finite age of their solar system would be coupled with no

12

2 Looking Out

evidence whatsoever of any kind of ‘initial state’ from which the solar system could have evolved. Ultra-speculative theory sketches might be invoked to ease their minds. Theorists might argue that the universe is really a system of causally disjoint realms in which every possible configuration exists somewhere (there are stronger and weaker versions of ‘every possible configuration’; for examples of the ultra extreme end of the scale see Tegmark 2007b; Lewis 1986). Such a view is a radical extension of theories that many of our cosmologists hold, but it is quite a stretch to believe in arbitrary creation events which generate isolated nebula, already properly stocked with the right mixture of light and heavy elements so as to generate a life creating solar system. Nor would the so-called anthropic principle be of much comfort, for our imaginary thinkers find themselves not just in a universe suitable for life, they find themselves in an exceptionally bizarre world that, among its many other oddities, also happens to support life. It is not a world that observers would expect to find themselves in, if one takes into account only the necessity for that world to support intelligent observers, which is the most the anthropic principle could license. It might well be that the most reasonable response would be to accept that the particular world configuration in which they discover themselves was somehow an explicit target of a mysterious creation event. Thus, while at first it does not seem difficult to pretend that the universe might consist of nothing but our solar system and indeed such a world does not in itself seem to violate physical law, its existence would be exceedingly puzzling to its inhabitants.

2.2 Galactic Cosmology What a contrast to the situation that we find ourselves in! As we’ll see, in our world, at every turn Nature has provided more clues which lead to a grand and unified account of the structure, and possibly, creation of our universe. So far, we have worked out a still incomplete but nonetheless remarkably detailed and coherent picture of the cosmos. In complete contrast to our imaginary scientists of Lonely Earth, no sudden complete explanatory roadblocks are ever thrown up against our theorists. How curious. No one knows when our ancestors started to think seriously and reflectively about the nature and structure of the world, but it must have been a very long time ago. Evidence of sophisticated astronomical knowledge goes back thousands of years—for example, the Egyptian pyramids are exquisitely constructed to line up with compass directions, with interior shafts that point to particular stars (as they would have appeared at the time of construction). Such knowledge must itself have descended from a long history of observation and thought. Possibly some (literally) hard evidence of this early time comes from various peculiarly carved bones and antlers; for example, the approximately 30,000 year old ‘Blanchard bone’ appears to represent the lunar cycle (see Kelley and Milone 2005, pp. 95–96). These ancient efforts to wring order out of the evident complexity of the night sky were immeasurably magnified through the self-conscious application of mathematical representations of

2.2 Galactic Cosmology

13

planetary motion and the ‘heavenly spheres’. The codification of Ptolemy—the culmination of a long history of Greek geometrical astronomy—managed to encompass the entire visible universe in one rational system, albeit based upon a fundamental heliocentric misconception, and set us on the path to modern scientific cosmology. The result, some 2,000 years later or more, is a spectacularly comprehensive and beautiful account of the universe. In parts it remains tentative and more or less hypothetical, but that befits a cosmology wishing to call itself scientific. Both its grand scope and fully detailed intimate links with all the physical sciences are breathtaking. If I could, I would have every child learn all their science via the study of this cosmology, which could easily be retold with increasing depth at every level of education and would naturally include everything from basic physics to biology. Of course, our cosmology butts up against its limits at various places, and mystery there intrudes. Why does anything exist at all? Why are there laws of nature? Why does mathematics do such an amazing job of describing our world? But, arguably, these aren’t even parts of cosmology as a scientific discipline, although they are of course questions that need consideration. Perhaps they are unanswerable, purely philosophical in the good sense of being ‘ultimate questions’. The status of another question, which is the topic of this book, remains unclear. Why and how does consciousness appear in the universe? Is this question ultimate and philosophical or empirically answerable and destined to be the crowning achievement of our scientific picture of the universe? What seems so striking is how far off mystery can be pushed, especially as compared to the imaginary situation of Lonely Earth discussed above. Let’s work backwards from the situation in which we find ourselves. We have worked out the structure of the Solar System by observation and already quite a bit of impressive direct physical investigation (every planet save Pluto has been visited by at least one robot probe, and the New Horizons spacecraft is scheduled to arrive at Pluto in 20154 ). We find that the Solar System cannot be much more than 5 billion years old. But that’s no great mystery. Our Solar System is embedded in a giant stellar system in which we can now observe the kind of dust and gas clouds surrounding budding proto-stars upon which the nebular hypothesis of solar system formation depends.5 We have discovered a large number6 of extra-solar planets; solar systems are not rare. Ours may be especially well behaved and well endowed to support life, but it is hardly surprising that life will only appear in those solar systems conducive to it, and so, out of the myriad of diverse solar systems in the galaxy, we must obviously expect to find ourselves in one that is suited to our kind of life. Our galaxy—the Milky Way—is also a dynamic and evolving system. While one might have conjectured that it was eternal, it could not be considered as unchanging. If eternal, it must be following some kind of grand cycle of star death and rebirth. For we know that the stars cannot shine forever—they depend on a finite supply of nuclear fuel, mostly hydrogen, which they ‘burn’ through nuclear fusion. Paradoxically, the more fuel a star begins with the shorter its lifetime. The big stars’ own gravitational force compresses the hydrogen fuel, forcing it to burn more vigorously. What happens to a big star that runs out of fuel? Before burnout, it is the pressure from the energy released via atomic fusion that prevents a star from collapsing under its own weight. When its main fuel source is expended, a star shrinks and heats

14

2 Looking Out

by compression, initiating a new wave of atomic fusion, this time not of hydrogen but of many heavier elements, from helium through to iron. Eventually there are no new fusion routes left and the star suffers a catastrophic collapse. The compression and shock wave from the collapse generates a truly immense amount of energy in a very small amount of time. Such supernovas can briefly produce more energy than entire galaxies. But more important from the cosmogenetical point of view is that the exploding star spews forth almost the whole periodic table of elements into the interstellar environment. The twin discoveries of stellar nucleosynthesis and the seeding mechanism of supernova are two of the most significant discoveries of twentieth century astronomy and astrophysics. These discoveries answer one of the questions that would simply stump the astronomers on Lonely Earth. They show how the complex mix of light and heavy elements found in the initial nebula which seeded the Solar System was both formed and put into our galactic interstellar space. Once we have the solar nebula in place and properly stocked, we can look in two directions: forward towards the creation of the planets, or backwards. Looking back, we face the awkward question of the existence of the Milky Way galaxy itself. For the supernova-condensation-supernova recycling of stellar material would eventually lead to most stars being too small to explode. Material would thus be locked away in low mass, cool ancient stars and white and brown dwarfs. A certain percentage of higher mass stars would also lead to such lockup, since they will evolve into neutron stars or black holes. But the galaxy we observe is not like that. It has an interesting and varied mix of stars of all kinds with ages that seemingly do not exceed about 10–15 billion years. A curious feature of the galaxy is that the oldest stars are concentrated in a set of vast spherical collections of stars, called globular clusters, that themselves orbit around the periphery of the galaxy. These stars present two very striking properties: they contain extremely small amounts of heavy elements and appear to be very old.7 The obvious reason for this is that they are largely stars that formed early, out of primitive material that had not yet been recycled through many generations of star formation and explosion. If the galaxy was truly old and somehow could regenerate star formation, we would expect to see the heavy elements spread more or less evenly throughout the stellar population. Since we don’t it is hard to escape the conclusion that the galaxy itself was created somewhat more than 10 billion years ago. After its formation, the evolution of the galaxy proceeded as outlined, with successive waves of star formation (which may be linked in complex ways to the rotation of the galaxy and its internal, spiral arm structure), leading to supernovas which enriched the interstellar environment with heavier elements, which then condensed into later generation stars, one of which is our Sun. But how was the galaxy itself formed? Our best theories are, in essence, upscaled versions of the nebular hypothesis of solar system formation; early stars and other condensations of mass in a universal gas cloud presumably condensed into proto-galactic structures from which the hugely diverse range of galaxies and types of galaxies we now observe were created. There now exist computer simulations of galaxy formation that do a pretty fair job

2.2 Galactic Cosmology

15

of recovering the kind of galactic forms we actually observe from primordial clouds of gas and dust.8

2.3 Universal Cosmology Our knowledge of galaxy formation is far from complete (the nature, role and even the existence of so-called dark matter is just one area of intriguing ignorance), and creation events are hard to observe since they happened so long ago (which is to say, they are very far away and thus faint or obscured). But we can pursue our general question nonetheless. The universe we can observe is a giant system of galaxies, perhaps a quarter trillion or so in total (so roughly speaking there are about as many galaxies in the observable universe as there are stars in our galaxy). Is this system eternal? If so, then there must be some mechanism of galactic ‘recycling’ else all the galaxies should be populated only with very old stars. We observe no such thing. Much more significant than this however, is the fundamental observation that the fainter a galaxy is, that is, the farther away from us it is, the more rapidly it is receding from us. Now, in fact we cannot directly measure the speed at which a galaxy is moving away from us. What is observed is that the light from certain galaxies is reduced in energy or ‘red shifted’, a phenomenon somewhat analogous to the Doppler shifting of the pitch of a sound source that is moving away from us. And there is a systematic correlation between the amount of red shifting and various indications of distance.9 This correlation between distance and recession speed was discovered by Edwin Hubble and Milton Humason back in 1931, though at the time it was viewed with suspicion, not least because Einstein had devised a model in which the universe was static. But other theorists—Friedmann and Lemaître—had already developed models of an expanding universe which could fit the increasingly convincing and accurate determinations of the expansion rate (codified in a number somewhat peculiarly named the ‘Hubble constant’, though it is neither fundamental nor constant). Since the 1930s evidence has mounted in support of the expanding universe hypothesis and against the idea that the galactic structure of the universe is eternal. For example, if we survey very distant galaxies (that is, galaxies that exhibit a very high red shift) we find that the statistics of galactic morphology are notably different from those of closer, older galaxies. In particular, there seem to be very few spiral galaxies in the survey which itself suggests that perhaps the typical spiral arm structure evolves from the evidently less structured elliptical forms. If we take the Hubble constant seriously it is natural to think that all the galaxies were closer together in the past and that if we trace back far enough, all the matter in the visible universe must have been confined in a very small volume. It is then not difficult to conceive of the universe as coming into existence from a very small but, to understate it rather severely, highly energetic event. This hypothetical event came to be called the ‘big bang’ (originally a pejorative term coined by Fred Hoyle, who championed a theory in which the universe was infinitely old). In fact, it is a striking fact that the form of the relation between increasing velocity and distance is

16

2 Looking Out

linear, which has two interesting consequences: no matter where you are it will look as if everything is moving away from you as the centre, and there must have been a time in the past when everything was close together (for a nice discussion of this and other aspects of Hubble’s work see Sandage 1989). However, the expansion of the universe as evidenced by galactic recession is far from proof that the big bang occurred or that the universe is not eternal. Hoyle and others accepted the expanding universe but denied the existence of any creation event (for Hoyle at least, this was in part motivated by an anti-religious sentiment which decried the possibility of something like a ‘creation event’). The cost was to posit the continuous low level creation of matter and some sort of cosmic repulsion that drove expansion. The amount of fresh matter needed to maintain the universe in a Steady State (as the theory of Hoyle, Hermann Bondi and Thomas Gold came to be called) is actually very small—about 10−40 g/cm3 /s—and if uniformly spread out would be completely undetectable. You would certainly not expect to see tangible lumps of matter just appearing out of nowhere every so often. From whence derives the energy which through all eternity drives the ‘matter creation field’ remained an open question. Any genuine scientific theory must posit observable consequences, and the steady state theory implies that any uniform sample of galaxies visible to us will have the same mix of ages. It is the Hubble law which allows us to take such a sample. A large collection of galaxies with the same red shift will form such a uniform sample, and thus should exhibit the same mix of characteristics as a large collection of nearby galaxies. But, as noted, they do not; rather they exhibit characteristics which are highly suggestive of a prior state of galactic evolution. There is a fair amount of other evidence leading in the same direction. The most decisive evidence against an eternal universe turned up before the kind of galactic evolutionary observations we have been discussing were very numerous or robust. This is the famous Cosmic Microwave Background Radiation or CMB.10 The CMB was accidentally discovered (although it had been predicted by some astrophysicists) by Arno Penzias and Robert Wilson of Bell Laboratories in 1964 as they worked on a radio astronomy project with an antenna released from corporate needs. The CMB is a bath of electromagnetic radiation which can be observed emanating from everywhere in the universe—no matter where you point your antenna you will observe it (some of the snow on an empty television channel is caused by the CMB). The CMB is a direct prediction of the big bang model; what is more the prediction is that the ‘structure’ of the CMB should be identical to the electromagnetic emissions of an object in thermal equilibrium at a temperature of about three degrees above absolute zero. This very low temperature represents the remnant of the big bang which has been cooling down for the last 10–15 billion years. The cooling is the result of the billions of years of expansion of the universe in a process somewhat analogous to the cooling of an expanding gas. The thermal nature of the CMB is also one of the most spectacularly accurate predictions ever made. In 1989 the Cosmic Background Explorer (COBE) spacecraft was launched with the aim of measuring the CMB much more precisely than ever before. Figure 2.1 shows the spectacular result of one of the COBE instruments which provides the overall spectrum of the CMB.

2.3 Universal Cosmology

17

Fig. 2.1 Cosmic background radiation spectrum

Actually, the curve portrayed in the figure is the theoretical spectrum for a black body (an object that absorbs all radiation perfectly, emitting only thermal radiation). The COBE data consisted of many data points strung out along this curve, as illustrated. The data points themselves should by rights be invisible since the error in the measurements is far less than the width of the curve at this scale!11 Of course, there are no crucial or decisive experiments in science. Although the CMB is exactly what a hot big bang would lead us to expect, versions of the steady state theory can always be maintained in the face of this evidence. Theorists can postulate that not only matter but radiation is continuously being created and that it is, somehow, being ‘thermalized’ so as to take on the characteristics of black body radiation. For example, a universe in which tiny iron ‘whiskers’ are uniformly scattered about would serve to thermalize background radiation (see for example Narlikar et al. 2003). Such a model seems rather ad hoc, but there remains some observational evidence against the standard big bang model, of which the most interesting (both scientifically and from a ‘science studies’ point of view) is the problem of discordant red shifts. Halton Arp is infamous in astronomy for persistently discovering what he claims to be physical associations between objects that the big bang model asserts cannot be associated, typically a low red shift galaxy seemingly coupled to a high red shift quasar.12 Arp has cataloged a host of discordant red shifts (see Arp 2003) of various kinds and personally believes that some kind of ‘new physics’ is necessary to explain them, and that the standard big bang model must be rejected. He—and a few other mavericks—remains adamant, but nowadays receives little attention from mainstream astronomers. For the big bang model has many other fundamental pieces of evidence in its favour. A rather bizarre one which nicely reveals how seemingly disparate aspects of physical theory interact is the supernova decay dilation effect. Supernovae have been studied in sufficient numbers that we have mapped out certain characteristic patterns

18

2 Looking Out

of light emission that the various types tend to produce. There are two main kinds of supernova (types I and II) and they each have their own light curves. In crude terms, a schematic example of the comparison of a ‘nearby’ type I curve with that of two high redshift supernovae would reveal that the latter curve is stretched out.13 Why should this effect occur? Are the laws of physics different a long way away, or a long time ago? That’s always possible, but there is no other evidence for revising physics in a spatial and temporal way. On the other hand, if the universe is expanding, so that distant supernova have a very high relative velocity to us, then we would expect to see the effects of special relativity. Famously, one of these is time dilation: the faster an object moves relative to us, the slower its ‘clocks’ will run relative to ours. By ‘clocks’ here is meant any physical process by which time could be measured (including ordinary clocks, the human body as it ages—hence the ‘twin paradox’, or any other process of change). The expanding universe hypothesis provides a very elegant explanation of the otherwise anomalous light curves of distant supernova (see Leibundgut 1996). Yet another crucial piece of evidence is the big bang model’s prediction of the ratios of the light elements, especially the hydrogen to helium ratio (see Peebles, Chap. 6). Stars can and do produce helium but the same ratio of hydrogen to helium is found both in galaxies with lots of heavy elements (hence galaxies which have a long history of nucleosynthesis) and those with very small amounts of heavy elements. So the amount of helium observed does not seem to be a product of the gradual creation and dissemination of helium from successive stellar generations. However, under the hot big bang model, we can deduce that there was a period (about 1 s after the big bang event) when neutrons and protons were no longer transforming into each other under various high energy processes, but were instead ‘free’ particles, though still highly energetic, mixing in an extremely hot gas. The initially almost exactly equal numbers of neutrons and protons would begin to tip in favor of the proton as neutrons naturally decay (into a troika of a proton, electron and neutrino), but as soon as the temperature dropped enough there would also be rapid neutron–proton fusion into deuterons, which would in turn fuse with protons into helium three and, by gaining another proton, into helium four (the stable form). The prevalent conditions ensured there was little or no fusion of any heavier elements through incremental processes then at play, though traces of lithium and a few other elements were formed. From the rate of neutron decay and the temperature bounds permitting fusion to begin, the ratio of protons to neutrons in place when fusion begins can be calculated. The ratio turns out to be about 13 protons for every 2 neutrons, which means that once all the neutrons are safely ensconced in helium nuclei (where they are safe from decay), for each helium nucleus—which contains two protons—there will be around eleven hydrogen nuclei (each a lone proton). By mass, helium will end up forming just under 25 % of the total, hydrogen around 75 % with a few light elements and deuterium (the isotope of hydrogen which contains one neutron in addition to one proton) filling up the rounding errors. It is a spectacular confirmation of the big bang model that this does indeed seem to be the primordial ratio of hydrogen to helium (the observed ratio is a little higher because of the contribution of extra helium from fusion occurring in stars since the big bang).

2.3 Universal Cosmology

19

The present amount of deuterium in the universe is also significant for the big bang model. It is a highly curious fact that deuterium exists in the universe at all. Deuterium is easily ‘ignited’, which is why it is a favoured fuel for our perennially prospective fusion reactor programs (as the old saw has it, ‘fusion is the energy source of the future…and always will be’), and hence is rapidly consumed by the fusion processes within the stars. If the universe was eternal, there would have to be some process of deuterium creation (the random creation of neutrons and protons would not suffice, since they would not likely be able to combine and free neutrons decay in about 15 min). But if the conditions of the big bang are appropriate, a small amount of deuterium can remain after the helium has ‘frozen out’. There doesn’t seem to be any other recognized process by which deuterium could be created. Observations of inter-galactic dust clouds, which hopefully represent the primordial distribution of matter, suggest that the early universe ratio of deuterium to hydrogen was about 1 to 10,000.14 That deuterium had to come from somewhere, and it is hard to see where except for a period of nucleosynthesis of exactly the sort predicted by the big bang hypothesis. Compare once again the situation of our astronomers and astrophysicists with their counterparts on Lonely Earth. They were driven by the data to embrace a picture of the universe in which the conditions for the development of the solar system simply appeared out of nowhere within a universe in which that appearance made no sense at all. The discussion above highlights how the imaginary scientists would find a kind of ultimate mystery almost everywhere they looked, once they looked deeply enough at any rate. They would find, since we supposed that their Earth and ours were qualitatively identical, that the water in their oceans contains not insignificant amounts of deuterium (about one 1 water molecule per 5,000 would be ‘heavy water’). They would have no explanation for where this deuterium came from—it was just salted in the initial solar nebula. Perhaps they would infer that a beneficent god must have stocked the oceans in preparation for the shift towards fusion power. In spectacular contrast to this, our scientists have found a rather grand, deeply complex and inter-related account of the creation of the solar system as part of an entire cosmos, in which a great many lines of evidence all point in the direction of the hot big bang hypothesis. Mystery remains, of course, as well as any number of scientific puzzles. But it appears that the mystery is being driven into that small corner of the room where rather abstract and philosophical issues get discussed. Why is there something rather than nothing? Why do we have laws of nature, and why did we get just these laws of nature? These mysteries face the Lonely Earth scientists too; it is just that they also have an elephant in the room. As our scientists close in on the big bang via current remnants of the early stages of the universe another feature becomes apparent. I will call this the simplicity of the early universe. All science except for physics ceases to apply to the universe as we move back towards the big bang event. Thus between the creation of the universe and the present time the objects, processes and, perhaps, the laws of all the sciences beyond physics must have emerged out of conditions to which those sciences do not apply. This is a significant fact to which all the lines of evidence we have been considering point.15

20

2 Looking Out

The extremely early stages of cosmic evolution are not well understood because they require physics not yet in our possession. Thus there is one cosmological ‘epoch’ which remains completely mysterious; it takes up the time from 0 to 10−43 s, a duration called the Planck time, and an associated size of about 10−35 m called the Planck length. The density of the universe during this time is so high that gravity would be expected to have significant effects even at the ultra microscopic level of particle physics. But we have no theory which integrates gravitation with quantum mechanics, though there are several candidate approaches (see Smolin 2001; Greene 1999 for popular expositions). However quantum gravity works, it takes almost no time at all for the universe to expand (and thus for its density to fall) sufficiently that gravity no longer has any microscopic effects, as its strength is swamped by the strength of other forces (of which, after a peculiar cascade of emergence we will consider, there are eventually three: the strong nuclear, the weak nuclear and the electromagnetic). We then enter what is sometimes called the GUT epoch. GUT stands for Grand Unified Theory, which sounds impressive. Unfortunately, there is no such theory which has been empirically verified. But it is thought that we know quite a bit about how such a theory should look. Supposedly, this is the period when there are but two forces, gravitation and a single GUT force which incorporates the nuclear strong force, the nuclear weak force and the electromagnetic force. This period lasts for only a very short time, until about 10−35 s after time 0. The GUT ought to be some kind of extension of a theory we do have in place, the electroweak unified theory independently developed by Sheldon Glashow, Abdus Salam and Steven Weinberg (for which they won the 1979 Nobel prize). The most disquieting prediction of GU-theories is that the proton is not stable, though possessed of a half-life trillions of times longer than the age of the universe. Current observations show no sign of proton decay however,16 and the traditional GUT construction is no longer seen as very promising, but there are other candidate theoretical approaches that promise the same kind of unification. As the GUT epoch began another rather bizarre and mysterious process seems to have taken place: cosmic inflation (see Guth 1997 for a first hand account of the development of this idea). This was an almost instantaneous increase in the scale of the universe by many orders of magnitude. The main evidence for inflation stems from both the overall and the detailed structure of the CMB. Overall, the CMB is almost perfectly isotropic, as if the system from which it emerged was pretty much in thermal equilibrium. Without inflation this is hard to fathom. We sit at the notional center of a universe which has been expanding for, say, 12 billion years. So, ignoring some relativistic niceties, photons from the CMB from regions on opposite sides of the universe which we capture today are separated by 24 billion light years. Since no causal influence propagates faster than the speed of light it is impossible for these photons to have been in causal contact since their creation. No matter how far back in time you go, if the universe has expanded uniformly—as the Hubble relation suggests—there could never have been a time when these regions were causally connected. Yet they have almost identical characteristics (to about one part in 100,000). This is called the horizon problem. Inflation solves it on the assumption that a huge expansion occurred after an initial thermal mixing. But this

2.3 Universal Cosmology

21

early state did not, and could not, reach perfect equilibrium and thus whatever residual inhomogeneities remained would themselves have been stretched out during cosmic inflation. As Andrew Liddle rather quaintly puts it: ‘inflation is trying as hard as it can to make the Universe perfectly homogeneous, but it cannot defeat the Uncertainty Principle which ensures that there are always some irregularities left over’ (Liddle 1999). Inflation predicts that, roughly speaking, the fluctuations in the CMB should be very nearly the same at all frequencies. So far, our observing instruments have verified the isotropy of the CMB to a high degree, and have more recently turned to examining the minute fluctuations (notably via the densely acronymic COBE and WMAP spacecraft and DASI instruments as well as the BOOMERANG balloon laden instruments in Antarctica). What we see thus far is quite in line with the predictions of cosmic inflation. Inflation neatly answers another question: why is the universe ‘flat’ (that is, why is the current geometry of the universe very near, if not exactly, Euclidean)? The anthropic principle gives one answer: a ‘typical’ non-flat universe would either collapse too quickly for life to form, or have such a sparse matter density that the preconditions for life (such as stars and solar systems) would not be able to arise. Therefore, it is to be expected that we shall observe ourselves to be in a flat cosmos. Like all anthropic ‘explanation’, this does not in fact explain why the universe is flat. Cosmic inflation does. Just as blowing up a balloon makes regions on its surface more nearly flat, so too inflation’s blowing up of the cosmos make its regions more nearly Euclidean. Without inflation the density of the universe must be postulated (it is not fixed by the big bang itself so far as anyone knows) as being between 0.999999999999999 and 1.000000000000001 times the critical density at the age of 1 s (see Guth 2002). Inflation blew up the scale of the universe by a huge amount, perhaps 70–100 orders of magnitude. One might wonder how come the universe was not left a virtually empty and cold vacuum after blowing up some 10100 times in size. Inflation itself deposited huge amounts of energy into the universe,17 leaving it a fairly dense, roaring hot stew of elementary particles, just as in the basic big bang model (it has thus been said that inflation is thus a ‘bolt on accessory’ (Liddle 1999) to the hot big bang theory). The inflation happened so fast that it did not extend across any significant physical changes. Inflation began in the GUT epoch and ended as the quark epoch began. This marks the time when the temperature fell low enough to permit the appearance of the most elementary of particles, the quarks and their gluons which mediate the strong nuclear force. The quark-gluon plasma characteristic of this epoch has been recreated on Earth (in ultra small measure for a fleeting instant), initially at Brookhaven Laboratories in New York state18 and now at the Large Hadron Collider in Europe. Further cooling reduced the energy of particle interaction and the universe passed through various epochs (cosmologists differ somewhat on the number of these worth distinguishing). First, at about a millionth of a second after the big bang, comes the hadron epoch, in which quarks bound into the class of composite particles called baryons, such as the familiar protons, neutrons as well as a host of mesons. While the laws of nature appear to be almost perfectly symmetrical with respect to the production of

22

2 Looking Out

matter and anti-matter, there is a very slight, and not well understood, asymmetry which led to the overproduction of matter baryons (by convention, what we are made of is called matter rather than anti-matter). Next—approximately 1 s after the big bang—comes the lepton epoch, in which electrons, neutrinos and muons take on an independent rather than the merely ephemeral existence they enjoyed before (during which they were being continuously created and destroyed). At this point neutrinos more or less stop interacting with other forms of matter. The electrons are now free to combine with protons to begin the formation of atoms. This leads to the epoch of nucleosynthesis, as discussed above.

2.4 Cosmology and Emergence Of special interest is a sequence of ‘phase transitions’ that occur through these epochs, each one marking a loss of symmetry and the separation of forces that were heretofore unified. During the initial, almost entirely mysterious Planck epoch at the very beginning of things, all the forces were unified as described by the elusive ‘theory of everything’. The decoupling of the forces occurred via a complex process of spontaneous symmetry breaking and mass creation. This is best understood in the case of the decoupling of the electromagnetic and weak forces. When unified, according to electroweak theory of Glashow, Salam and Weinberg, the weak nuclear and electromagnetic forces have the same strength, and form a single electroweak force. The phase transition occurs via a transformation of a field, called the Higgs field after its inventor,19 which imbues the particles (three kinds of bosons) mediating the weak nuclear force with mass. The range of a force is inversely proportional to the mass of its exchange bosons, and one distinguishing characteristic of the weak nuclear versus electromagnetic force is that the latter has infinite range whereas the former is effective only over very short ranges. The exact mass acquired via the Higgs mechanism is not directly fixed by the theory but depends upon a free parameter (the Weinberg angle) whose value can be measured empirically. But the electroweak theory does tell us how to relate this parameter to the ratio of the expected masses. It was a great triumph of the electroweak theory when the predicted bosons were found—with the proper masses—in 1983/1984 (a first hand account of the complexity and difficulties of this experimental episode can be found in Watkins 1986). It is conjectured that via similar processes all the forces decoupled sequentially as the universe cooled from an initial unified state. Two features of this cascade are particularly important to us. The first is that it seems there will be no way to predict from knowledge of the unified state the exact values of the relevant parameters of the post-unified state. The process of ‘spontaneous symmetry breaking’ involves a random element. By way of analogy, consider the properties of a bar magnet. An iron bar magnet will lose its magnetism if heated above 770 °C (what is called iron’s Curie temperature). The atomic/electronic spins are disordered by thermal agitation above this temperature. As the bar cools magnetism returns and the magnetic

2.4 Cosmology and Emergence

23

field of the iron will be oriented in a definite direction, though there is nothing which forces any particular magnetic orientation. Certain patches of order appear and begin to crystallize, forcing order upon nearby regions, until eventually a single magnet orientation emerges. That is, there is spatial symmetry which is broken by the imposition of a definite magnetic orientation. The laws of nature do not distinguish any spatial direction as ‘special’ even after the magnetic orientation is fixed by accident (or design). The creation of the magnetic field of a bar of iron by spontaneous symmetry breaking is thus a case of emergence which is unpredictable in principle. Similarly, the ratio of the relative strength of the four forces and the masses of the force carrying particles may not be set by nature but emerge through a random process. No matter how much one knew about the symmetric phase, it would be impossible to predict these values. Of course, we do not really know that these values are randomly fixed by the free evolution of the early universe. It is possible that deeper theory will show how they are dictated by physical law (for example, typical GUTs fix the free parameter of the electroweak theory mentioned above, while leaving other parameters unfixed). It is far too early to have any confidence about what the final theory of everything will look like or even whether we shall be able to construct one. The mechanism of spontaneous symmetry breaking discussed here serves to illustrate one possible kind of emergence. More important, the discussion illustrates the status of the postlepton epoch physics. No matter what form the theory of everything takes, a strong constraint on it will be to generate the state of the universe a few seconds after the big bang. Short of a truly revolutionary shift in astronomy and astrophysics, the big bang model of the ‘later’ universe (after 1 s that is) is now a part of standard knowledge. And we have seen how our astronomers, in such stark contrast to the imaginary scientists of the lonely Earth thought experiment, have at every stage found both satisfying answers to current questions about the nature and structure of the universe and Earth’s place within it as well as answers that fit together into an elegant total picture. Picture then the state of the universe when approximately 1 s old. Is there anything missing from our cosmological picture? Considered from the standpoint of the recognized scientific disciplines, almost everything. There is no chemistry (organic or even inorganic), no biology, no psychology and no sociology. Is this surprising? From a purely physical point of view, not at all. Conditions of the universe at this time are, by Earthly standards, very extreme. The temperature was something like one billion degrees and it was quite impossible for atoms to form. The formation of stable atomic nuclei was just becoming possible as the temperature dropped below 1 billion. There was then a small window of opportunity where density and temperature permitted the synthesis of hydrogen and helium nuclei (as described above). Thus it is a natural fact that the only science which applies to the universe at the age of 1 s is physics. As we move back in time closer to the big bang itself, we encounter more exotic, eventually frankly speculative, physics, but it’s physics all the way back. However, as we pursue time away from the big bang towards the present, we obviously enter the domain of application of all the other sciences. Thus, for example, chemical and biological entities and processes (such as molecules, bonding, living

24

2 Looking Out

organisms and natural selection) must be in some sense emergent phenomena. A little more precisely, any phenomenon which appears in a system which heretofore did not exhibit it can be labeled diachronically emergent. The cosmological tale as we now conceive it must be a tale of diachronic emergence. My labeling naturally suggests a second form: synchronic emergence, which can be roughly defined in terms of properties possessed by a system which are exemplified by none of its components. The cosmological story told here supports synchronic emergence no less than the diachronic form. It seems quite evident that the physical entities which exist at 1 s after the big bang have no biological properties. Yet some 12 billion years later we know that at least a few systems possess such properties. The extent of biology in the universe is one of the most interesting unanswered questions facing science today. People have wondered about exobiology for a long time, but before long we will be mounting serious efforts to grapple with this issue. It is on the edge of technical feasibility for us to launch space based observatories that will be able to analyze the atmosphere of the planets of nearby stars. As James Lovelock pointed out some time ago (Lovelock 1965), the presence of life, at least life as we know it, on a planet has profound effects on its atmosphere, maintaining it in a non-equilibrium state that is a signpost of biological activity. As I write this, the presence of methane in the atmosphere of Mars has been detected. Because of inevitable chemical reactions, methane cannot remain for long in a planet’s atmosphere unless it is being replenished. Since Mars shows no signs of active volcanism (one of the few possible alternative sources of methane), it is possible that we have already indirectly detected life on Mars. In any case, the obvious diachronic emergence of biology is coupled with its synchronic emergence: we do not think that the physical components of living things are themselves carriers of biological properties. This assumes that the basic structures of the world have not themselves acquired biological properties at some time after the big bang. Does it seem plausible to suppose that although the protons that existed 1 s after the big bang did not exemplify any biological properties, these very same protons now do carry biological properties? On the contrary, there is no reason at all to think that the properties of protons have changed in the slightest across the billions of years since the big bang.20 There is no need to ascribe the properties of any science save physics to the early state of the universe. It follows that all scientific properties save those of physics are at least diachronically emergent and in all probability synchronically emergent as well. Emergence will be extensively discussed below. The lesson of this chapter is that the overall scientific picture of the universe to which we have been led by abundant evidence strongly supports emergence—in some sense yet to be fully clarified. But before turning to the nature of emergence itself, I want to examine another line of evidence that also supports emergence, but this time in the narrower domain of life and mind rather than the whole universe.

Chapter 3

Looking In

3.1 Jelly Beings Everyone can remember how the Scarecrow would time and again save Dorothy, Toto, the Tin Man and the Lion with his quick wits and resourcefulness. And he did it all despite the apparently serious impediment of entirely lacking a brain. Although his head was stuffed with straw, he gave convincing evidence of a host of mental attributes. Perhaps, as Wittgenstein said, it is enough simply to behave like a living human being to merit the ascription of consciousness (see Wittgenstein 1953, § 281). Such a remark invites a variety of possible thought experiments analogous to that of the lonely Earth in the previous chapter, albeit ones that are still more fanciful. It does indeed seem in some sense possible that we could have discovered that human heads were stuffed with straw, or, a little more plausibly, a completely undifferentiated jelly. Or we could go further. Let’s try to imagine a world where science progresses much as it did on Earth, except for the bizarre changes necessary to support this extension of the Wittgensteinian thought experiment. So far as anyone can tell, the insides of all animals is just a kind of undifferentiated jelly (except for the bones, say). Call these creatures ‘jelly beings’. So, try to imagine the slow advance of science proceeding pretty much as it did here on Earth, with the bizarre additional fact that all animal life forms are mere jelly beings. To be sure, anatomy will never amount to much in our imaginary world, but the physical sciences can progress normally. But as scientists tried to integrate their knowledge into as complete and coherent a picture as possible they would face an anomaly much more severe than the one which faced our imaginary astronomers of Lonely Earth. Though apparently, in some attenuated sense, possible, jelly beings make no sense at all. Very abstract theories, such as thermodynamics and evolution, might be applied to them, but with the devastating addendum that there would be no explanation at all of how these abstract conceptions were implemented within the jelly beings. Perhaps a counterpart of Mendel could discover the statistics of inheritance in this strange world, but how could there be any reasonable account of

W. Seager, Natural Fabrications, The Frontiers Collection, DOI: 10.1007/978-3-642-29599-7_3, © Springer-Verlag Berlin Heidelberg 2012

25

26

3 Looking In

the mechanism of heredity? It is all just jelly inside, amorphous and, by hypothesis, not analyzable in the chemical terms suitable for non-animal nature. Biological explanation would thus run up against a solid brick wall. The cognitive and neurosciences would suffer the same fate. We have stipulated that, within the thought experimental situation, animals, including humans, behave just as they do in the actual world, but unfortunately there would be no detectable causal structures mediating perception or action. I won’t dispute Wittgenstein’s claim that such a discovery would not affect our attributions of mentality to others or ourselves (how could it affect self-attribution in the face of continued consciousness after all). But there would be no theoretical purchase on how consciousness is implemented in the head. Obviously this would give aid and comfort to dualists and other nonnaturalists about the mind. In this fantasy, humans and animals would really possess the anatomical uniqueness so often sought for and claimed, but the price would have been the loss of all hope for any scientific biology, psychology or cognitive science. In the real world, things went very differently. From very ancient times, we have known something about the complex and articulated internal structure of animals, and doubtless there were any number of conjectured links between the internal organs and the mental or spiritual characteristics of things. The role of the internal organs was given at least a proto-scientific veneer with the humoural theory of the ancient Greeks. This account saw the four humours (the bodily fluids of blood, phlegm, black and yellow bile) as working together within a kind of homeostatic system, and so, to a limited degree, caught an important truth. Of course, gross misconceptions abounded. Aristotle famously held that the brain was a peripheral organ whose lowly function was the cooling of the blood and heart, and while such a view would have meshed with the jelly beings scenario, his outlook did not survive for long in the real world. The essential role of the brain in mental functioning has been long recognized, although—understandably—incompletely and inaccurately. Perhaps the Presocratic philosopher Alcmeon was, around 450 BCE, the first to state in at least a quasiscientific way the link between mental functioning and the brain. He is reputed to have performed dissections of the eye and optic nerve which led him to posit the general principle that ‘all the senses are connected to the brain’ (see Barnes 1983, pp. 478 ff.). Somewhat later, around 400 BCE, Hippocrates made the stronger claim that the brain is ‘the organ which enables us to think, see, and hear, and to distinguish the the ugly and the beautiful, the bad and the good, pleasant and unpleasant’ (Gross 1999, p. 13). The importance of the brain was codified by Galen (circa 180 CE) whose medical insights presided over the middle ages (see Gross 1999). Theologically minded medieval thinkers (as they all were) largely agreed that the seat of the soul was in the brain, and indeed they frequently localized it more specifically, favoring the ventricles as likely places where mind (soul) and body communed together (following on this tradition, it is perhaps little wonder that Descartes took as one of his tasks the locating of the exact place in the brain where the immaterial mind interacts with matter; see Zimmer 2004 for an extensive account of the slow recognition of the importance of the brain for mental function). Our jelly beings thought experiment is too crude and extreme to be worth pursuing. But no less than in the astronomical thought experiment, the real world contrasts

3.1 Jelly Beings

27

starkly with the imaginary in the way multiple lines of evidence arise and come together in aid of the construction of a unified account of the phenomena at issue. From an abstract vantage point, the big bang hypothesis describes the way in which physical law was implemented in one particular universe (or sub-universe: the region we call our universe or the observable universe). On the side of biology, the overarching theory is not physics, but evolution, and biology on Earth is the tale of how evolution has been implemented here.

3.2 Cell Theory and Organic Molecules From this standpoint, the nearest thing to a big bang hypothesis in biology is the discovery of the genetic mechanisms of cells. In turn, we can break this down into the cell theory of living organisms and the much later idea of an internal cellular mechanism of heredity (still later found out to be DNA based). Of course, the cell theory is directly at odds with the jelly beings thought experiment. Instead of a solid brick wall through which explanation cannot pass, the cellular structure of living things provides a highway into both the current functioning and evolutionary history of life on Earth, and then joins an expressway of explanation which runs all the way back to the big bang itself. Just as the stellar nature of the Milky Way, the mountains of the moon and the Jovian satellites were discovered as soon as Galileo turned a telescope upon them, so cells were discovered as soon as the microscope was invented and turned upon living things. Robert Hooke’s amazingly detailed illustrations in Micrographia (published in 16651 ) revealed, among many other seemingly greater wonders, the structure of a piece of cork, which Hooke likened to the honeycomb of a bee hive and borrowed the term ‘cell’ to describe it. It was not until very much later that the idea that the cells within organisms were themselves biological units which collectively composed more complex entities. In 1839 Theodor Schwann and Matthias Schleiden advanced the cell theory (under that name), although the idea that all living things were composed of cells, whose individual capacities and organization gave rise to the functions of the organs or properties of the organisms which they constituted, was already quite widely held. The cell theory proposed that most organisms were, in essence, vast colonies of more or less specialized microscopic ‘animacules’. The codification of the cell theory followed hard upon the heels of another decisive event in biochemistry: the synthesis of organic compounds from purely physical ingredients. In 1828 Frederich Wöhler had produced urea in his laboratory from inorganic materials. At the time it was widely held that organic compounds were not reducible to the inorganic because living matter was supposed to contain or exemplify a distinctive ‘vital principle’. Vitalists of the day responded, not very convincingly, that one of the essential ingredients in Wöhler’s process—cyanate— had been derived from blood products and thus he had merely transferred the vital spark from one compound to another. Other vitalists explored the remaining logical option of denying that urea was really an organic compound, being just a breakdown

28

3 Looking In

product of biological metabolism. But of course Wöhler was simply the vanguard of a host of synthesizers. One of his own students, Adolph Kolbe managed to produce acetic acid from indisputably non-organic precursors by 1845 and during the 1850s Pierre Bertholet similarly synthesized a raft of other compounds (including methane and alcohol). Of course, it is logically possible for things to have gone differently. Along the lines of our thought experiments we can imagine a world in which it turned out to be impossible ever to synthesize organic chemicals from purely inorganic precursors. Scientists in such a world could invoke the word ‘vitalism’ to label their problem, but their explanatory project would have ground to a halt at the organic/inorganic divide. A curious footnote to the history of emergence illustrates this fantasy; for a short while it looked like it was coming true. In 1824 Wöhler, working in the famous laboratory of J. J. Berzelius, had made the astonishing discovery that cyanic acid and fulminic acid had quite distinct properties. What was astonishing was that, as Wöhler meticulously worked out, the two acids have identical chemical constituents. Chemical orthodoxy at the time demanded that distinct properties of complex substances required a difference in their elemental composition. Berzelius himself introduced the term ‘isomer’ for compounds with identical chemical constitutions but distinct properties. It was then hypothesized that it was the ‘structure’ of the compounds that differed, but the whole concept of chemical structure was at the time extremely dubious since very few chemists could bring themselves to believe in the reality of atoms. Often atomism meant little more than a belief in uncompounded chemical elements, with no commitment to the idea that such were grounded in discrete, quasi point-like units of matter that went together like Tinkertoys and could exemplify internal geometric relationships (see Brock 1993, Pullman 1998). Various structure theories were developed that could handle isomerism (e.g. Berzelius’s copulae theory or Benjamin Brodie’s ‘operational calculus’). But then came the discovery of the photo-isomers, whose crystals were optically active. That is, polarized light shone through them has its plane of polarization rotated either right or left. Otherwise, such crystals did not appear to differ physically or chemically. This presented chemists of the day with the rather fundamental problem that ‘the number of isomers exceeds the number of possible structures’ (J. Wislicenus, as quoted in Brock 1993, p. 260). It is, once again, possible to imagine that no physical explanation could have been found for this; that these organic compounds simply, and as a matter of brute fact, differed in a way unrecognizable to physical science. Perhaps it would have been said that they differed at the level of their ‘vital principle’ rather than in mundane physical or chemical properties. Instead, in the real world, we have a explanatory triumph that deeply supported the link between the organic and inorganic realms and the emergence of the former from the latter. The first stereoisomer (as they would come to be called) to be noticed was tartaric acid which in 1847 Louis Pasteur tackled as a doctoral student. He grew large crystals of the acid and noticed, by naked eye no less, that there were two asymmetric forms, each the mirror image of the other.

3.2 Cell Theory and Organic Molecules

29

Pasteur’s discovery led Jacobus van’t Hoff to propose that chemical stereoisomers could be explained by assuming that each isomer possessed a distinct spatial orientation of its constituents. This required that one adopt a quite realist attitude to atoms—or at least the chemical constituents of these compounds—and especially the bonds between them which had now to take up ‘real’ angular positions relative to each other, thus permitting the mirror image asymmetry which provided the extra degree of freedom to explain optical activity. Although brilliantly successful and fruitful van’t Hoff’s proposal was not greeted with universal enthusiasm. Kolbe himself launched a vituperative attack against van’t Hoff, for which Kolbe has unfortunately been branded as a stodgy and narrow minded reactionary ever since, impugning van’t Hoff’s credentials as a scientist (for van’t Hoff worked at the time in a veterinary college) and lampooning the metaphysical credulity of those chemists who sought refuge in ‘supernatural explanations’ (see Brock 1993, pp. 262–263). In fact, the discovery of chemical chirality was crucial in the advance of chemistry (and atomism). Chirality has chemical consequences far beyond photoisomerism, as shown by the examples of thalidomide (only one of the stereoisomers of the drug has the infamously disastrous side effects) and penicillin (bacteria are susceptible to its attack through their use of one of the isomers of alanine which our own cells lack). Something of our fantasy world of blocked chemical explanation persists in current creationist writings. It is a curious fact that virtually all life on Earth depends upon amino acids with ‘left handed’ chirality (the use of right handed alanine in bacteria a lucky exception). The mechanism of selection is unknown and mysterious, since production of amino acids in the laboratory generally yields so-called racemic mixtures in which left and right handed forms are equally abundant. Even extraterrestrial amino acids found in meteorites exhibit a preference for the left handed form (see Cronin and Pizzarello 1997).2 But if all physical processes of amino acid production yield racemic mixtures how can we explain the overwhelmingly left handed amino acids within organisms? Creationists can, and do, pounce on this, declaring that only a supernatural process could select out one isomeric form for use in biology. Given the history of explanatory success in grappling with problems like this, this declaration seems decidedly premature. The problem of homochirality, along with a host of others, remains to be solved as we seek a scientific account of the origin of life. I would be disinclined to bet that this is the place, finally, where scientific explanation will fail. Still, at a deeper level, the origin of chemical chirality remains something of a mystery. The quantum physics that ought to lie behind chemistry, at first blush, appears to obey the ‘law’ of conservation of parity. So it seems reasonable to infer that there should be no difference between the properties of the enantiomers (i.e. the mirror image forms) of chiral molecules and no reason for nature to prefer one over the other (once an environment with pronounced chirality forms, there is room, as we have seen above, for this to have very powerful chemical effects). However, as was discovered in 1956, parity is not conserved at the fundamental level in interactions involving the weak force (see Pais 1986, pp. 530 ff.) and the electroweak theory (mentioned in Chap. 2 above) codifies, if not yet explains, this breakdown. Small but measurable differences in the energy levels of chiral molecules stem from weak interaction

30

3 Looking In

effects. Some scientists (e.g. Abdus Salam, see Salam 1991) have proposed that this fundamental asymmetry accounts for why just one form of organic molecule is favored throughout biology. There is little evidence for this hypothesis at present but it would help to explain the apparent universal preference nature has for certain forms of chiral molecules and would, of course, reduce the number of seemingly brute facts in the world while increasing the scope of emergence. Whatever its ultimate origins, chirality presents an excellent example of emergence, since obviously the atoms which ultimately make up, say, amino acids possess no spatial asymmetry unlike the chemical compounds they constitute. The purely geometric differences in stereoisomerism thus nicely illustrate some of the more abstract sources of emergence. Now, to return to the main story. The chemists’ newly forged link between the organic and inorganic aspects of nature provided some highly suggestive evidence that cells themselves were made up of entirely normal matter fully obeying the laws of inorganic chemistry. The major substances found in cells (lipids, polysaccherides, proteins, and nucleic acids) were discovered by 1869 (nucleic acids last) and were found to be ‘normal’ chemical compounds, although most details of their structures and functions remained for twentieth century investigators to elucidate (most famously of course the helical arrangement of nucleotides in DNA and its role in inheritance, another interesting example of emergence via spatial orientation). At roughly the same time, Darwin’s theory of evolution appeared (Darwin 1859), immediately reforming much of biology, and providing some inkling of how life might have progressed from simple unicellular forms to multicellular plants and animals.

3.3 The Neuron Doctrine We begin to approach the domain of cognition and consciousness when we apply the cell theory to the brain. This specific application is called the ‘neuron doctrine’ and it is one of the linchpins of modern neuroscience (see Shepherd 1991 for a detailed history). The first discovery of nerve cells, the large Purkinje cells of the cerebellum, was made in 1837 by the eponymous investigator, actually prior to Schwann’s enunciation of the cell theory. Thus, although it was clear that the brain contained cell like structure, it was not at all clear that the cell theory really, or completely, applied. The multiple tendrils of each nerve cell appeared to connect directly with each other within the nervous system. Instead of a system of intercommunicating but individual cells it was thought that the ‘whole nervous system represents a protoplasmic continuum—a veritable rete mirabile’ (Lewellys Barker, as quoted in Shepherd 1991, p. 65). This unitary network theory stood in opposition to the cell theory. It was the long labor of many through the nineteenth century that established that nerve cells were indeed individual cells and that these cells were the core functional unit of the brain. The two points were quite distinct. The famous Norwegian arctic explorer Fridtjof Nansen was a brain scientist before a Greenland trekker; in his doctoral thesis of 1887

3.3 The Neuron Doctrine

31

he brilliantly deployed the novel method of staining cells invented by Camillo Golgi to defend the view that nerve cells were individual entities rather than ‘concentrations’ in a ‘protoplasmic continuum’. But having established that, he then went off the rails with the additional claim that the nerve cells served a merely nutritive function in support of the activity in the nerve fibers where the real action was. Nansen was not a central figure in nineteenth century neuroscience, but in his demotion of the nerve cells he was expressing a common sentiment; indeed the original network theory of the nervous system remained dominant for some time. The neuron doctrine was finalized and essentially demonstrated to be true by one of the founders of modern neuroscience, Ramón y Cajal, who perfected the Golgi staining method and devoted tremendous energy to the microscopic examination of brain cells. Golgi had discovered in 1873 that nervous tissue first hardened with potassium dichromate then soaked in silver nitrate left individual nerve cells darkly stained (Golgi called this the reazione nera or black reaction). For reasons no one understands, this process selectively stains only a few nerve cells, making them stand out starkly from the background tangle of cells and their processes. At his own expense, Cajal set up an academic journal in 1888 to disseminate his results. He wrote all the articles and sent copies, needless to say unsolicited, to eminent scientists across Europe. To his disappointment—how naive is that—Cajal’s work was not immediately recognized. It was his pilgrimage to a Berlin conference in 1889 where he made his breakthrough. The sheer quality of his slides, which delegates could view through the several microscopes he set up at the conference, made it clear that Cajal had taken a great leap forward in neuroscience. With surprising rapidity, a number of eminences, such as Albrecht Kölliker, came around to the idea that nerve cells were separate entities and that they were the core functional units in the nervous system. How neurons functioned remained controversial for some time, although it had been known since Luigi Galvani’s work at the end of the eighteenth century that the long tendrils that sprang from the nerve cells—eventually categorized as axons and dendrites, for output and input signals respectively—could be stimulated by electricity. Galvani took the natural step of identifying the nerve impulse with an electrical signal, but it was not at all clear that this was correct. One of the foremost physiological researchers of the time, Johannes Müller (he was in fact the first professor of physiology as an independent field), remained highly skeptical of the electrical hypothesis (see Müller 2003; Müller’s attitude towards Galvani’s views is discussed in Nicholas Wade’s introduction). Some followers of naturphilosophie, an influential philosophical movement which maintained a number of mystical and vitalist tenets about living things, regarded the nerve impulse as radically non-physical and possessed of an infinite velocity. Quite dubious theoretical calculations of the nerve impulse yielded wildly varying estimates, some fixing it at more than ten million miles per second, far beyond the speed of light. Müller himself made various attempts to measure the effects of a neural electrical current, but without success (his instruments were simply not sensitive enough). It took a very long time to work out the processes by which electrical signals propagate along axon or dendrite, which do not work in anything like that

32

3 Looking In

of simple conducting cables (for the details of this exceedingly complex but by now quite well understood process, see Nicholls et al. 2001). Nonetheless, the basic fact of the electrical nature of neural signaling was soon established. One of the famous experiments in this field was Hermann Helmholtz’s demonstration in 1850 of the finite, and indeed very modest, speed of nerve impulses. By means of a conceptually simple procedure which was essentially no more than tickling a frog’s leg and waiting for the associated movement, Helmholtz measured the velocity of the nerve impulse to be a mere thirty meters per second, hardly a supernatural accomplishment (Helmholtz had dissected the frog to isolate a ‘nerve-muscle preparation’ and deployed precision instrumentation to mark fine temporal differences). In the spirit of our thought experiments, here is yet another place where nature could have thrown up a roadblock and placed us in a mysterious world unamenable to scientific explanation. But didn’t. It is conceivable that the nerve impulse could have turned out as unmeasurable or possessed of a seemingly infinite velocity and that it never revealed itself as a kind of electrical signal at all. That would have been powerful evidence for the mystical views of vitalists but would have left an unbridgeable gap between brain processes and the rest of nature. Instead, at every step nature and ingenious experimenters came into harmony through nothing more than hard thinking and hard work. From our discussion, we can isolate three crucial points where the nervous system, and especially the brain, is anchored to the rest of the natural world. First, organic material, the makeup of cells throughout living nature, is composed of exactly the same chemical and atomic constituents as the inorganic world. Second, the nervous system is structured like the rest of organic nature, composed of individual cells which are the core functional units. Third, the nerve cells form a network in which communication is effected by electrical signals. This last point has been elaborated in stunning detail over the last 150 years (see Nicholls et al. 2001), culminating in the detailed knowledge we now possess about synaptic transmission and the role of the myriad of different neurotransmitters therein, including more than a glimmering of their psychological import. Though the pioneers of the neuron doctrine could not have guessed at the complex processes that underlie the nerve impulse, their basic idea has been entirely vindicated. It would take us far afield to survey the elaboration of these anchor points over the last century and a half. Suffice it to say here that the initial general hypothesis that physiology would prove explicable in standard scientific terms, and the more specific neuron doctrine have been and continue to be immensely fruitful. The discovery that the neural network is constituted out of ordinary physical material leads to another sort of advance critical to neuroscience’s development, that of effectively monitoring and measuring the brain’s activity. If the neurons are the core functional units and they communicate via electrical signals then measurement of external electrical and magnetic properties will reveal neural activity (albeit of a collective nature). This is the basis of the electroencephalogram, or EEG, and the magnetoencephalogram (MEG). It is also possible to measure the electrical properties of individual neurons directly via the insertion of a probe electrode into or very near to the neuron. Measurement and manipulation are often two sides of a single coin. In this case, it is also possible to influence the neuron by reversing the process

3.3 The Neuron Doctrine

33

with an electrically charged electrode. The famous work of Wilder Penfield (see Penfield 1958) deployed this method and showed, serendipitously since he was actually looking for focal points linked to epileptic seizures, how the electrical stimulation of neurons could lead to the production of various sorts of conscious experiences. In 1954 an electrode placed in the septal area of a rat’s brain revealed the now famous ‘pleasure center’ (see Bozarth 1994 for a review). Robert Heath performed similar— ethically unrepeatable—experiments on human beings in the 1960s (see Heath 1963, Heath 1964) with roughly similar results. The idea of a pleasure center is probably overstated, pleasure is better seen as dependent on a complex and quite extensive reward system within the brain which is modulated by a variety of neurotransmitters and hormones.3 The point here is simply that this modulation ultimately works by changing individual neurons’ propensities to produce their characteristic electrical signal.

3.4 fMRI Anchor Points Recently, sophisticated neural measurement technologies have transformed neuroscience, namely positron emission tomography and nuclear magnetic resonance (though politeness nowadays requires that we drop the black magic word ‘nuclear’). These techniques, especially the latter, have opened a window on the working brain and, in many ways, the mind as well. Magnetic resonance imaging (MRI) exploits some intricate details of the first anchor point mentioned above. The basic MRI signal is the radio emission of protons whose spin axis is precessing at a characteristic frequency in a magnetic field (the target protons are usually those in the hydrogen nuclei of water since it is so abundant in living tissue). An MRI machine is, to a first approximation, just a huge magnet designed to supply a gigantic and fixed magnetic field over a large volume of space with a strength of about 2–4 Tesla (which is 20,000–80,000 times stronger than the Earth’s own magnetic field, which sounds impressive but an ordinary fridge magnet might well attain 0.2 Tesla—over a minuscule volume). The main magnet is what accounts for the size and a lot of the cost of an MRI machine.4 Because they possess electric charge and the quantum mechanical property of spin, the protons in the MRI subject act like little magnets themselves and will tend to align with the main magnet’s field. The alignment is, as a quantum mechanical phenomenon, only probabilistic and the spin axes form a kind of ghostly cone around the direction of the main field. In a process somewhat analogous to the way a top will wobble, these spins will tend to precess around the field’s direction at a frequency—called the Larmor frequency— dictated primarily by the strength of the field. These precessing spins generate a constantly changing magnetic field and hence will produce an electromagnetic signal. However, in the initial state, at equilibrium in the main magnet’s field, the precessing spins will be out of phase and effectively cancel each other out. The trick of the MRI scanner is to get the target protons into a coherent state so that a rotating magnetic field will be generated.

34

3 Looking In

Another complication is worth noting. Quantum spins in any direction can take on two values, call them ‘up’ and ‘down’. Both up and down spins can be in alignment with the main magnetic field, either parallel or antiparallel with the field. We might then expect that all effects will be washed out by the parallel and antiparallel spins canceling each other. However, because there is a small energy advantage to the parallel alignment there will be a small asymmetry or excess in the number of protons which align parallel to the field compared to those aligning antiparallel (the asymmetry, whose value depends on the strength of the main magnetic field, is just a few parts per million in standard MRI machines). The MRI signal is dependent on the unmatched protons which in the end do produce a net magnetic field in the subject aligned parallel to the main magnetic field. it is bizarre to think that when you go into the MRI scanner your body, or parts thereof, is literally turned into a magnet. The first step in acquiring an MRI signal is to focus radio energy at the exactly appropriate, or resonant, frequency onto the subject. This energy is absorbed by the target nuclei, causing a good number of them to undergo a ‘spin flip transition’ into the higher energy state and come into phase with each other. This effect, called nuclear magnetic induction, was discovered in 1946 (the original article is Bloch 1946). There are various ways to regard what is happening, but we can conceive of it as the net magnetic field ‘tipping’ out of alignment and precessing around the direction of the main magnetic field. This in turn generates an electromagnetic signal identical to the induction signal. Now, when the induction pulse is turned off there is a short period of time—the relaxation time5 —during which the generated signal continues on its own and can be detected, until the proton spins fall out of phase and the system returns to equilibrium with the net magnetization once again aligned with the main magnetic field. The complexity of the whole system is dizzying (and I am only providing the merest hint of a sketch of how this all works), which matters because it emphasizes what I called anchor points—the way that our fundamental physical understanding meshes with processes which reach all the way up to the psychological. Extracting information from this signal is just as complex as its generation. The trick is to exploit the fact that the induced signal will subtly vary because the magnetic field is not homogeneous and is differentially affected by the nature of the tissue at any particular location. The signal would be undecipherable however unless it was restricted to or regimented within a spatial region. To achieve spatial localization the operators of the machines also carefully modulate the magnetic field in which the subject is immersed. Thus, MRI machines have a set of secondary or gradient electromagnets (of much less intensity than the main magnet) which can be manipulated to alter the overall magnetic field with high spatial and temporal precision. The result is that the received signal is composed of a large number of slightly different components which reflect, in complex ways, the nature of the substance emitting them and their location. These can be teased out by a mathematical analysis (Fourier analysis) at which modern computers are very adept and which can then be used to produce an image.

3.4 fMRI Anchor Points

35

The ‘standard’ MRI we have been discussing is medically most useful for imaging static structures (see for example Chap. 4, Fig. 4.2). MRI machines used in this way are just a kind of jumped up x-ray device. Starting in the early 1990s a new kind of imaging was developed that could track real time metabolic changes6 ; this was called functional MRI (fMRI) and has been intensely applied to brain imaging over the last decade. Metabolic processes within cells are sustained by a host of inputs, of which perhaps the most crucial is oxygen. Neurons are prodigious energy consumers and require a rapid and large supply of oxygenated blood (something approaching one liter of blood is pumped through the brain every minute). The brain is also capable of directing oxygenated blood to those regions which are most active (a nineteenth century finding, see the classic article Roy and Sherrington 1890), and this ‘haemodynamic response’ is what can be measured via fMRI. Just as in ordinary MRI, the extraction of an image depends upon variations in the magnetic field. It turns out that oxygenated blood and deoxygenated blood have quite distinct magnetic properties (oxygenated and deoxygenated haemoglobin are diamagnetic and paramagnetic respectively) which results in differential distortions in the local magnetic field as blood is moved into the brain and its oxygen is taken up in cell metabolism. This in turn causes changes in the relaxation times of brain tissue depending on whether or not it is receiving extra oxygenated blood—that is, depending on how active the neurons within it are.7 While these differences are very small, they are detectable. The primary use of fMRI, as well as all other brain scanning techniques, is of course for practical medical diagnosis and research. That is not what we are interested in here but luckily there are a number of researchers with access to scanners looking at the neural substrates of mental states and processes. One extra difficulty that intrudes here is that, by and large, the brain is a seething hive of activity in which it is very difficult to trace neural activation that is related to specific cognitive states. Put another way, the brain has a great many jobs to do and it is always on the job. Every task involves many common functions, and the distinctive activity associated with a given task is usually very small relative to the overall activity of the brain. Thus most ‘cognitive imaging’ involves taking averages over many trials and many subjects, and also taking such averages during a control or rest state. The difference between the two averages is taken to be perhaps an indication of the brain activity distinctively associated with the cognitive state being investigated. We are a long way away from real-time, single-subject mind-reading machines, although we are already at the stage where a serious neuroscience article can begin with these words: ‘[a] challenging goal in neuroscience is to be able to read out, or decode, mental content from brain activity’ (Kay et al. 2008). MRI machines are constantly improving in sensitivity and it would be very rash to deny the possibility of mind-reading via fMRI and successor technology, or to suggest that it won’t happen in, say, the next fifty years.8 This is another place where nature could have refused to cooperate, but did not. It is possible to imagine that all the experimental design and mathematical manipulations applied to the fMRI signal simply could not generate any coherent data linking haemodynamic activity to neural processes, still less to cognitive states. Such a possibility is far less radical than our other thought experiments and it is thus significant

36

3 Looking In

that to the contrary, nature is, despite the primitive level of current instrumentation, already providing us with robust results from MRI machines. We’ll start to examine some of these in a moment, but first a brief digression on the anchor points. As we have seen, there are multiple places where the activity of the brain is anchored to more basic features of the world. It is natural to ask about the relations between the anchor points. These relations themselves ought to weave a web of explanatory interconnections. There are also more practical scientific issues about calibration of new measurement devices and about what it is exactly that is being measured. One place where the philosophical and scientific considerations come together is the relation between the haemodynamic response measured in fMRI and the activity of neurons. The neuron doctrine assigns to the individual neuron and its output signals the central functional role in the brain. It is assumed that the haemodynamic response matches increased neural activity. This crucial hypothesis is what links fMRI to the basic functional units in the brain. It is only quite recently that this linkage was tested by Logothetis et al. (2001) whose difficult experiment managed to ‘show unequivocally that a spatially localized increase in the BOLD (i.e. ‘Blood Oxygen Level Dependent’) contrast directly and monotonically reflects an increase in neural activity’ (p. 154).9 Of course, the ultimate object of the exercise is to discover—in some detail befitting the ongoing exponential explosion in our knowledge—something about the neural substrate of mental states, especially states of consciousness, and to this we now turn, beginning with a fanciful tale.

Chapter 4

Consciousness in the Brain

4.1 From Phrenology to Brain Imaging Jill is worried about the sincerity of Jack’s affection. Young, beautiful and the founder of a NASDAQ leading, multibillion dollar biotechnology firm she has developed a natural fear of predatory suitors. Too many times the semblance of true love has proven temporary sweet illusion. Enough is enough. But how can Jack’s tempting and oft professed love be tested? In the old days, a persistent knight could be set on an impossible quest that only a true hearted suitor, guided by the indisputable power of love plus a considerable amount of luck, would have the ghost of a chance of completing. But this sits ill with the contemporary temper (and in any case dragons have become distressingly rare). Jill’s thoroughly modern instinct is to seek a technological quick fix: a mechanical gauge of Jack’s affection. How about functional magnetic resonance imaging? Thus Jill requires that Jack’s quest shall be to lie very quietly within an fMRI chamber, patiently examining images of an assortment of lovely women, among which are occasionally seeded those of Jill herself as well as some of Jack’s previous loves. Short of acute claustrophobia, this is not a quest to shrivel the heart of a good knight, loyal and true. But dissemblers should fear it, for the fMRI machine cannot be fooled by sweet words and heavy lidded gazes. And so Jack is slid into the long dark tube of the MRI machine, like a submarine’s torpedo readied for firing. Operators, and Jill, anxiously watch a bank of monitors as a garishly coloured image of Jack’s brain appears before them. Sure enough, whenever a picture of Jill is presented to Jack there is a strong and distinctive response from deep within his brain. A certain pattern of activation spreads over particular small regions; a marked deactivation occurs in another region. After enough runs to guarantee statistical significance, the sages gravely confer and agree: this is a sure sign of true love. Jill is ecstatic. Jack may be beginning to have some doubts, despite his now publicly certified feelings for Jill. Let’s look more closely at this, for it contains or leads to almost everything we need to think about when we ponder the problem of linking the fundamental functional W. Seager, Natural Fabrications, The Frontiers Collection, DOI: 10.1007/978-3-642-29599-7_4, © Springer-Verlag Berlin Heidelberg 2012

37

38

4 Consciousness in the Brain

Fig. 4.1 Basic brain anatomy

units of the brain—the neurons—with mental states and, ultimately, consciousness. First, I’m not (quite) making this up. fMRI is a real step towards mind reading, which at least in a crude and, so to speak, collective form may turn out to be easier than anyone would have imagined. Recent work (Bartels and Zeki 2000; Bartels and Zeki 2004) foreshadows a scientific version of my modern fairy tale. Indeed it is the case that people in love show an easily identifiable pattern of brain activity in certain regions of the brain. The medial insula, parts of the anterior cingulate (more on these brain regions below) and zones of the striatum (a part of the brain in which resides the so-called ‘pleasure centre’) are all characteristically activated by the sight of the loved one, while a part of the prefrontal cortex tends to deactivate. This pattern of activation and deactivation is not idiosyncratic or confined to any one individual but was observed over a number of subjects very consistently. It is no accident that these are the parts of the brain implicated in our love-test. To set these particular regions into focus, we should begin with an overview of the brain. We can divide the brain into three basic components: the cerebrum, cerebellum and brainstem. The cerebral cortices make up the cerebrum, the largest and evolutionarily newest section of the brain which wraps fully around the brainstem and dwarfs the cerebellum, which hangs below the cerebrum. The external appearance of the brain reveals four of the cerebral lobes (named after the bones of the skull under which they lie) and the cerebellum, along with a portion of the brainstem (see Fig. 4.11 ). Though mightily enhanced by modern imaging evidence, the idea that mental functions might be closely associated with delimited or localized brain regions is of course not new. Its first at least quasi scientific appearance is with the phrenology of Franz Gall and Johann Spurzheim in the early nineteenth century. Although phrenology is nowadays a kind of paradigm case of pseudo scientific quakery, Gall himself was no quack and was perhaps only guilty of an overly optimistic hope that the physiological correlates of mental function would be easy to find. He is hardly the only one in science guilty of such optimism, which may be partly excused as simply

4.1 From Phrenology to Brain Imaging

39

a methodologically virtuous love of simplicity in one’s hypotheses. Gall, however, may have violated Einstein’s dictum: theories should be as simple as possible, but no simpler. But, after all, it is not absolutely inconceivable that variations of the skull could correlate with mental functioning, and what else did Gall have to go on? Scientists, if not philosophers, must match their aspirations to available technology. And Gall wasn’t completely off track. Researchers nowadays use cranial casts of our hominid ancestors to look for the traces of brain structure left in fossilized skulls, and have discovered, for example, some evidence that Homo erectus may have had language, or proto-linguistic, abilities of some kind by finding a marked asymmetry between left and right brain in the region of Broca’s area (of which more below) in an ancient, though not precisely dated, skull found in Indonesia (see Broadfield et al. 2001). Phrenologists also deployed what seems to modern ears a curiously quaint scheme of mental categories, including amativeness, conscientiousness and mirthfulness. But then, the current idea that there is a part of the brain devoted or specialized for recognizing faces may come to seem no less preposterous, whether or not some kind of modularity of cognitive function is maintained (maybe this is already happening; see Gauthier et al. 2000). The concept of functional localization admits of a more abstract definition than the obvious one of correlating a brain function with activity in a particular part of the brain. It seems reasonable to allow that an anatomically scattered set of neural systems could form a ‘functional unity’ devoted to a particular cognitive task. It might be better to think in terms of a multi-dimensional categorization involving more or less delimited regions of the brain and the degree to which these regions are specifically devoted to a particular task. Maximum functional localization would then be a matter of maximal specificity conjoined with minimal dispersion in the brain. It is far too early to say how cognitive functions will map onto such a categorization. And with respect to knowledge of the basic neuro-cognitive functions which underlie the mind, we really aren’t all that much better off than the phrenologists. Another way to put this is to distinguish between the relatively low-level, brain structural hypothesis of localization and the relatively high-level, cognitive functional hypothesis of modularity (see the classic Fodor 1983 for the core ideas of modularity theory2 ). Modules are functionally discrete but they do not have to be localized. Furthermore, our mental functions might be the product of the interaction of many modules devoted to sub-mental activities—perhaps that is the most likely scenario. A healthy skepticism about localization of mental function is thus warranted. But of course that should not and will not stop efforts to correlate mental states or processes with neural activity, though it favours caution in the interpretation of such correlations. In general, there are no specific fine-grained functional correlates to the gross anatomical divisions illustrated in Fig. 4.1. For example, the occipital lobe is devoted to vision, but if one regards the recognition of objects by sight as part of our visual abilities then vision extends far beyond the occipital region. The brain is extremely densely cross connected throughout, and signals from any part of the brain can be sent to any other part. Many of the bizarre neurological syndromes, which traditionally offered the only window into brain-mind linkages, are as much

40

4 Consciousness in the Brain

Fig. 4.2 Brain bisection (MRI scan)

the result of loss of coordination between distinct parts of the brain as of damage to particular functional regions.

4.2 Communication Breakdowns The most spectacular example of these communication breakdowns is the result of severing the massive, high bandwidth, connection between the two cerebral hemispheres called the corpus callosum, shown in cross section in Fig. 4.2. There is an extremely rare congenital syndrome, agenesis of the corpus callosum (ACC) in which the organ never develops at all. Although usually involving severe behavioural, cognitive and developmental difficulties, ACC is occasionally almost entirely asymptomatic. It seems likely that in the absence of the corpus callosum pre-existing subcortical pathways, whose functions are replaced with callosal connections as the brain matures, remain functional in victims of ACC (see Lassonde et al. 1986). The separation of the hemispheres is, however, more often the result of deliberate surgical intervention. Brain bisection or commissurotomy is done in the last resort to relieve dire and otherwise untreatable forms of epilepsy. A successful procedure prevents a seizure from spreading from one hemisphere to the other. But such operations produce some incidental symptoms that forcefully suggest the creation, or conceivably even the pre-existence, of two quite distinct consciousnesses residing in one brain (for an account of the early work on this see Gazzaniga 1970). A basic consequence of brain bisection, plus the curious fact that sensory and motor connections are crosswired,3 is that speech mechanisms, which are usually located on the left side of the brain are isolated from any visual information that happens to be restricted to the brain’s right side. For example, if bisection patients

4.2 Communication Breakdowns

41

are briefly presented with a short and meaningful phrase like ‘key ring’, where the little space between the ‘y’ and the ‘r’ is set as the fixation point at the centre of their visual field, they will be prepared to say that they saw only one word: ‘ring’. Despite the direct, and fully conscious, report that they saw ‘ring’, if asked to select an object corresponding to the presented word, their left hand (under the control of the right side of the brain) will select a key rather than either a ring or a key ring. Nor is this behavioural divisiveness restricted to and only revealed in special laboratory conditions. There are many reports of commissurotomy patients suffering from frequent and distressing ‘intermanual conflict’ in ordinary life, sometimes for an extended period of time after the operation (Joseph Bogen 1998 writes that ‘[a]lmost all complete commissurotomy patients manifest some degree of intermanual conflict during the early postoperative period’) occasionally of a quite violent nature. There are certainly two centres of ‘behaviour control’ here; are there similarly two centres of consciousness? I don’t think anybody knows. Philosophers have of course weighed in on the issue. Thomas Nagel provided an early, fascinating and inconclusive discussion (see Nagel 1971). Charles Marks and Michael Tye both defend the single consciousness view of bisection (see Marks 1980; Tye 2003). In contrast, Roland Puccetti argued for the dual consciousness view, along with the yet more audacious claim that even normal brains have two centres of consciousness within them, in Puccetti (1973). A somewhat elusive and obscure alternative in which split brain phenomena illustrate ‘partial unity’ of consciousness (roughly defined in terms of states A, B and C being such that A and B are co-conscious with C but not with each other) has been explored by Susan Hurley (see Hurley 2003). A good overview of the vast general topic of the unity of consciousness can be found in Brook and Raymont (2010). Be that as it may, the disruption in behavior which frequently at least appears to involve dual consciousnesses is occasioned by a purely physical intervention which can be straightforwardly investigated. Although nature has not been kind to those who have been bisected, she has been once again surprisingly friendly to those who seek to explain what is going on in the brain.

4.3 Blindsight Another interesting oddity of consciousness that stems from the intricacies of the brain’s internal communication system is blindsight (for an authoritative overview, as well as integration with other deficits, see Weiskrantz 1997), an affliction in which people who sincerely declare themselves to be utterly blind in part of their visual field (called the ‘blind field’) and who generally act fully in accordance with these assertions nonetheless are able to obtain and use visual information presented in the blind field. Blindsight is the result of damage to a specific part of the occipital (or visual) cortex, a part of the brain devoted to the processing of neural signals from the eyes, in a region called V1 (see Fig. 4.3) which is the first area of the visual cortex to receive input from the retina (the resulting blindness is called cortical blindness

42

4 Consciousness in the Brain

Fig. 4.3 More brain regions

to distinguish it from much more common peripherally caused forms of impaired vision). Thus, more or less damage in V1 will cause a greater or lesser zone of blindness, or scotoma, in the patient’s visual field (on the side opposite to the damage in accord with the twisted wiring of the brain). Seemingly, it is ‘obviously’ true that if the first link in the chain of neural processing necessary for seeing is broken no vision can result. Nonetheless, people with blindsight can respond to visual stimuli. For example, if a bright light is flashed in the blind field, a blindsighted subject will deny seeing anything but can nonetheless guess with extremely high accuracy the location of the flash. Much of what we know about the neural causes of blindsight comes from studies on monkeys, who possess a visual system very similar to our own and can have blindsight intentionally imposed upon them (see Stoerig and Cowey 1997). Monkeys with one or both of their visual cortices excised give clear indications of blindsight. It does seem however that monkeys do much better than humans at adapting to the condition so as to use their blindsight in everyday activity. It is sometimes hard to tell the behavior of a monkey fully blindsighted by removal of the visual cortices of both hemispheres from the behavior of a normally sighted monkey (see Humphrey 1984, Chapter 3; Stoerig and Cowey 1997, pp. 549 ff.). Perhaps this is only because of the extremely extensive and rigorous training to which monkeys can be subjected (for example, Nicholas Humphrey worked at getting a blindsighted monkey named Helen to see again for 7 years), but I wonder whether human introspective consciousness in a way interferes with such training in humans since it is so ‘evidently’ pointless to even try to see when one ‘knows’ that one is blind. What is more, it is possible for information presented to the blind field to influence conscious thought. In an experiment somewhat reminiscent of the split brain research noted above, a blindsighted subject hears (and typically also sees in their non-blind

4.3 Blindsight

43

visual field) an ambiguous word, say ‘bank’, immediately following solely visual presentation of a disambiguating clue, say ‘fish’, to their blind field. Such cueing seems to affect the interpretation of the word that is consciously heard despite the apparent—to the subject—complete unconsciousness of the cue (see Marcel 1998).4 Subjects are as perplexed about their own abilities as anyone else and, very curiously, never begin to use their blindsight intentionally even though this would seem to involve only forcing themselves to guess about what is around them. They strongly report that they see nothing in their blind field; one subject, when asked how he could explain his ability to guess correctly the existence or non-existence of a target replied that perhaps he saw ‘a different kind of nothing’ (Stoerig and Cowey 1997, p. 551).5 The ability to access, in any way whatsoever, visual information without consciousness seems very strange, probably because our richest and perhaps most vividly conscious sensory experiences are those of colour, form and observed motion. But, on the other hand, we know that many things influence our behaviour without our being conscious of them and, in light of the many other connections in the brain that deliver retinal information perhaps we should not be so surprised at blindsight. One hypothesis about how this works is the dual path theory of David Milner and Melvyn Goodale (2006) which asserts that visual consciousness involves a set of dorsal neural pathways but that another, ventral, set of pathways provide additional, or at least other, visual information processing as well. Oversimplifying, blindsight is the result of degradation in the dorsal path with preservation of the ventral path. It is hardly surprising that the working brain would have a host of interactive systems which could be selectively lost or damaged without a total loss of function. Milner and Goodale provide a fascinating account of the blindsight abilities of a human victim of a tragic ‘natural experiment’ in Goodale and Milner (2004).

4.4 Neuroeconomics Despite the features of global communication characteristic of the brain, it is thus evident that we can, albeit crudely, mark out some gross functions associated with the main anatomical regions. The frontal lobe (or lobes, since they come in pairs, one for each hemisphere) subserves various ‘high level’ functions: thinking, deciding, integration of memory, thought, emotion and value. Recently, much attention has been devoted to this last topic because of its corrective perspective on the role of emotions in rational thought. Far from being unsettling distractions interfering with the dispassionate assessment of our plans and goals, emotional responses are crucial elements of a coherent, and indeed rational, life. The famous, and terrible, story of Phineas Gage well illustrates the absolute need for emotional engagement with thought and planning. In 1848, a railway construction accident severely damaged a large section of Gage’s left frontal cortex (prefrontal in fact, just in front of the temporal/frontal junction)—a tamping rod, almost four feet long and more than an inch in diameter at the wide end, was driven right through his brain by a prematurely exploding charge of dynamite! Gage did not lose consciousness at the scene of the

44

4 Consciousness in the Brain

accident and somehow survived the inevitable massive infection. But, although he was left in apparently full possession of his cognitive faculties, his personality was utterly changed, and most definitely for the worse. A previously steadfast, dependable, staid and reserved man, Gage seemed to lose control of himself and could never again hold down a steady job or live a normal life. In Antonio Damasio’s (Damasio 1994) examination of this incident, from a stomach churning account of the accident itself to its pathetic and unhappy outcome, a strong case is made that Gage’s fundamental problem was the destruction—via physical damage to a specifiable region of the brain—of appropriate emotional response to and/or evaluation of ordinary life situations.6 There is no claim here that the frontal cortex is the ‘seat’ of the emotions. A great number of brain systems, most of them evolutionarily older and more ‘primitive’ than the cortex, are involved, and involved in different ways, in many distinct kinds of emotions and in our overall emotional consciousness. Gage’s problem is more a problem of integration of emotional response with what might be called the value of the event triggering that response. It is common for people with damage similar to Gage’s to be unable to make the most trivial decisions; they will dither and endlessly ponder the pros and cons of, say, whether to leave by the front or back door. Of course, such value is relative to the situation in which one finds oneself. The man in ‘The Lady or the Tiger’ could not really be faulted if he gave in to hesitant deliberation over which door to choose (especially if he had thought a little deeper about the situation and its emotional as well as motivational freight), and while exasperated parents may not be able to understand how their teenage daughter can spend an hour deciding which dress to wear to a non-event, this is evidently no insignificant matter to her. The significance of an event is gauged by our emotional response to it. Such responses frequently occur entirely within the realm of contemplation as evidenced by our bouts of fretting, wondering or worrying. The mythical purely rational being, untroubled by any emotions whatsoever, would end up doing absolutely nothing since nothing would matter to it. Thus, the original Star Trek’s Mr. Spock is revealed to be full of emotions, though generally of a rather high-minded sort: loyalty, love of knowledge and justice, and the like. Without these, motivation would be utterly lacking and he would never even bother to show up for duty. Nonetheless, our culture’s deep sense of the conflict between reason and emotion is not motiveless. One area of life in which rationality and emotion are very closely intertwined is finance and here no less than anywhere else the constitutive role of the neural substrate of cognition is currently generating lots of data and even some public notice. The nascent field of neuroeconomics is now popular front page news as the July 5 2004 issue of Newsweek’s article entitled simply ‘Mind Reading’ (via fMRI of course) attests (see Glimcher 2004 for more background). It is now quite possible to witness the brain basis of the perennial conflict between cold rationality and hot emotion in the arena of monetary exchange. In the so-called ultimatum game, two participants, A and B say, play under these rules: A will be offered a sum of money and A must decide how much of this initial sum to offer to B. B can take it or leave it, but if B refuses A’s offer then neither

4.4 Neuroeconomics

45

player receives anything. So, the obviously rational thing for A to do is to offer B some minimal amount (perhaps some amount that can be regarded as effectively non-zero if we suppose that some amounts are so small they aren’t even worth thinking about). And it is equally obvious that the rational thing for B to do is to accept whatever A offers, since—surely—something is better then nothing. What is frequently found however is that unless A makes a pretty substantial offer, B will ‘cut off his nose to spite his face’. B will walk away with nothing but the satisfaction of knowing that A is being punished for making a lowball offer. This is true even in one-shot encounters where there is no prospect of ‘training’ one’s opponent to share fairly. Sanfey et al. (2003) have used fMRI to reveal the neural dynamics underlying this theoretically odd, but deeply human, behavior. Remarkably consilient with Damasio’s account, two brain areas are especially active during the ultimatum game, the prefrontal cortex and another area known to be associated with emotional response, particularly negative emotions such as disgust—the insular cortex (to be discussed in more detail below). The more active the insula, the more likely player B is to turn down A’s lowball offer. Those of a phrenological bent might exclaim: here is the organ for ‘righteous indignation’7 ! It is tempting to speculate about the origin of this seemingly innate irrationality and there is no shortage of Darwinian tale spinning from evolutionary psychologists (see Pinker 1999 for a compendious and entertaining introduction to this field; for a collection of important scholarly articles see Buss 2005). It is an easy game to play. Back in the EEA (that is, the ‘environment of evolutionary adaptedness’, presumably the African grassland of half a million years ago or so) we all lived in small and highly socialized tribes in which everybody knew everybody and must have faced many daily encounters of at least a quasi economic nature. A nicely tuned sense of disgust towards unfair exchange would serve us well so long as sharing would lead to greater reproductive success overall, which surely it would under conditions of rapidly shifting resource ownership in hunting and gathering societies. Of course, judging fairness requires no less rational acumen than sharp practice so the need for rationality is not at all lessened in a tribe of fair sharers. And, of course, under conditions of repeated play, it is rational for us to punish non-sharers if that increases the chances they will share the next time we play. Players have to be smart enough to figure all this out however. We might also expect that a certain amount of ‘random’ cheating could be, more or less infrequently, advantageous as well, thus encouraging a certain sly deviousness that we can still recognize in modern humans. We could even invoke the mighty powers of sexual selection with the hypothesis that women— being more socially astute than men—would favor those men given to fair dealing and hence help fix the genetic basis for the brain dynamics Sanfey et al. (2003) observed in action. (Of course, I am just making this all up as I go; for a powerful criticism of the whole project of evolutionary psychology see Buller 2006.) In any event, it is interesting to observe that chimpanzees are, in this regard, rather more rational—in the blinkered purely economic sense—than human beings. It is possible to train chimps to play an analogue of the ultimatum game. Amongst chimpanzees, the ‘dispenser’ in the game tends to fork out a very small amount, and the ‘receiver’ strongly tends to take whatever is on offer (see Jensen et al. 2007).

46

4 Consciousness in the Brain

This may well reflect the cognitive sophistication, possessed by humans but lacking in chimpanzees, necessary for the complex interactions characteristic of social exchange. Either that, or chimps just see things more clearly than we do. It would be very interesting to compare the neural processes occurring in the apes as compared to those reported by Sanfey et al. (2003).

4.5 Language and Recognition Of course, the human trait which most strongly contributes to our sociability is language and the links between brain and language have been long known. It is highly suggestive that this newest addition to the capabilities of animals on Earth and the most distinctively human trait has a neurological foundation which seems to be unusually localized in important respects. Since the nineteenth century, two areas have been identified as crucial for linguistic cognition and behavior. One, Broca’s area, resides (almost always) at the rear of the left frontal lobe (see Fig. 4.3). It seems to specialize in the sequencing of speech into grammatical and fluid forms.8 The characteristic syndrome caused by damage to this area—Broca’s aphasia—is difficult, halting ungrammatical speech with much circumlocution, but with considerable retention of basic comprehension. Another sort of aphasia, and its associated region of the brain, is also named after a pioneer of neuroscience, Carl Wernicke. A kind of inverse of Broca’s aphasia, victims of Wernicke’s aphasia (or ‘fluent aphasia’) retain easy and relatively grammatical speech, but their talk is utterly senseless. Consider the ‘Cookie Theft Picture’ used in the Boston Diagnostic Aphasia Examination. This picture, readily viewable on the internet, presents a domestic scene with some emotionally intense content. The picture shows a mother calmly washing the dishes while water pours out of the overflowing sink in front of her and her children steal some cookies from the cupboard, with one child clearly about to fall off the stool he stood on to get access to the cookie jar. Here are two commentaries on the picture (from Avrutin 2001): [1] B.L.: Wife is dry dishes. Water down! Oh boy! Okay Awright. Okay … Cookie is down … fall, and girl, okay, girl … boy … um … Examiner: What is the boy doing? B.L.: Cookie is … um … catch. Examiner: Who is getting the cookies? B.L.: Girl, girl. Examiner: Who is about to fall down? B.L.: Boy … fall down! [2] H.W.: First of all this is falling down, just about, and is gonna fall down and they’re both getting something to eat … but the trouble is this is gonna let go and they’re both gonna fall down … but already then … I can’t see well enough but I believe that either she or will have some food that’s not good for you and she’s to get some for her too … and that you get it and you shouldn’t get it there because they shouldn’t go up there and get it unless you tell them that they could have it. And so this is falling down and for sure there’s one they’re going to have for food and, and didn’t come out right, the uh, the stuff that’s uh, good for, it’s not good for you but it, but you love it, um mum mum (smacks lips) … and that so they’ve … see that, I can’t see whether it’s in there or not. Examiner: Yes, that’s not real clear. What do you think she’s doing? H.W.: But, oh, I know. She’s waiting for this! Examiner: No, I meant right here with her hand, right where you can’t figure out what she’s doing with that hand. H.W.: Oh, I think she’s saying I want two or three, I want one, I think, I think so, and

4.5 Language and Recognition

47

so, so she’s gonna get this one for sure it’s gonna fall down there or whatever, she’s gonna get that one and, and there, he’s gonna get one himself or more, it all depends with this when they fall down … and when it falls down there’s no problem, all they got to do is fix it and go right back up and get some more.

I think the reader will have no difficulty in making the diagnosis. The familiarity of language localization makes it easy for to us forget the remarkable fact that one could use such commentaries to predict with high accuracy the general location of physical damage in the brain. As we extend our ability to monitor the cognitively active brain our knowledge of such localization is growing rapidly. The temporal lobe of the brain, in which Wernicke’s area resides, is also the residence of the so-called auditory cortex, devoted to various relatively ‘high-level’ aspects of audition. It is an interesting question what this ‘devotion’ consists in however, which is relevant to our project here. It seems very unlikely that brute location has anything special to do with audition and altogether more likely that some features of cortical structure or organization are what is crucial. Some remarkable work carried out in the laboratory of the neuroscientist Mriganka Sur (see Sur et al. 2000) reveals both the amazing plasticity of the brain and reinforces the ideas both of localization of function and some kind of intimate connection between neural organization and mentality.9 Exploiting the fact that in ferrets the neural pathways from retina to brain develop largely after birth, Sur and his colleagues managed to induce new born ferrets’ optic nerves to project to the auditory cortex. Astonishingly, in the face of visual stimulation and typical environmental interaction, this part of the ferret’s brain took on an organizational structure very similar to that of the normal visual cortex. Sur et al. also showed that mature rewired ferrets would respond to visual stimuli received in the revamped auditory cortex just as standardly wired ferrets had been trained to respond to visual stimuli processed by the normal visual cortex. Despite the very unusual, and probably unique in mammalian natural history, neural substrate involved it is very tempting to endorse the idea that these ferrets enjoy visual experience with genuine visual phenomenology (at least insofar as ferrets possess any conscious phenomenology at all, a thesis I am prepared to accept).10 In humans, parts of the temporal lobes are specialized for tightly constrained recognition tasks. For example, an area at the junction of the right temporal and occipital lobes seems to be necessary for the identification of faces and damage to it leaves people utterly unable to recognize anyone by face alone.11 They retain the ability to see faces, of course, but there is no longer anything distinctive about particular faces. The condition is known as prosopagnosia and forces sufferers to devise intricate alternative strategies of recognition, which might depend upon the way someone walks, the shape of their legs in a particular sort of clothing (such as jeans) or some other distinctive feature (for a fascinating first person account of prosopagnosia see Bill Choisser’s web page (Choisser 2007); for a striking display of the specialization of face recognition see Goodale and Milner (2004), plate 5, discussed on pp. 57ff.).

48

4 Consciousness in the Brain

Another area deep within the temporal lobe (in both hemispheres this time) is dedicated to places and, among other things, houses. As with the story of Jack and Jill, it is possible to observe these areas in action with fMRI scans. A new twist to this now familiar story is that it is possible to monitor brain activity in a way that discriminates what is conscious from what is not. This works by exploiting binocular rivalry, which is the rather uncommon experience of consciousness-switching that occurs when each eye is presented with distinct inputs (if you have a telescope, or some binoculars, at hand, you can experience rivalry by keeping one eye open while the other looks through the scope (or through just one eyepiece of the binoculars)). In a very elegant experiment, Tong et al. (1998) (see also Kanwisher 2001) presented to subjects in an fMRI chamber dual visual stimuli: houses to one eye, faces to the other. Binocular rivalry results in the subject’s consciousness flipping fairly rapidly, every ten seconds or so, between the two presentations. The fMRI scans—after the usual collection and statistical reduction of the data—clearly reveal the alteration in consciousness. Every few seconds the face area lights up, then fades down as activation in the place/house area takes its turn. While current imaging techniques are generally far too crude to allow real time observations of this effect12 —it takes intensive after the fact analysis of the fMRI signals to reliably spot the alternating activation—there can be little doubt that with time it will be possible to perform the basic mind-reading involved in knowing whether someone is conscious of a face or not. Perhaps someday employees will be monitored to verify that they are not spending too much time daydreaming about their lovers. Or, more optimistically if less plausibly, perhaps the skull will be sanctified as a strict boundary of privacy through which no agency can legally intrude. An important qualification needs to be emphasized once again. Although I think it is true that we could detect what a person is conscious of via possible extensions of the brain monitoring techniques discussed here, there is no implication that the ‘seat’ of consciousness of faces, for example, is the face area of the brain. As I have already tried to stress, consciousness is more likely a phenomenon requiring integration of activity across a great many brain systems.13

4.6 Temporal Lobe Religiosity The temporal lobes also seem to be implicated in a much more bizarre phenomenon for which there is no common label. We might very loosely, and unfortunately rather phrenologically, call it ‘religiosity’. There is evidence that unusual stimulation of the temporal lobes results in feelings of ‘supernatural presence’, mystical communion and the like. Since this is a highly distinctive and significant state of consciousness, it is intensely interesting to find even the first hints of a neurological foundation for it. Victims of temporal lobe epilepsy have sometimes noted that immediately prior to a seizure, and a sure sign of one impending, is a strange, powerful and wonderful ‘feeling’. The most famous and articulate sufferer was Fyodor Dostoevsky, who

4.6 Temporal Lobe Religiosity

49

presented the condition occasionally, but significantly, in his fiction. In The Idiot Dostoevsky has Prince Myshkin meditate upon his own affliction of epilepsy: …there was a moment or two in his epileptic condition almost before the fit itself…when suddenly amid the sadness, spiritual darkness and depression, his brain seemed to catch fire…his sensation of being alive and his awareness increased tenfold at those moments which flashed by like lightning.…all his agitation, all his doubts and worries, seemed composed in a twinkling, culminating in a great calm, full of serene and harmonious joy and hope, full of understanding…He often said to himself that all those gleams and flashes of the highest awareness…were nothing but disease, a departure from the normal condition. And yet he arrived at last at the paradoxical conclusion: ‘what if it is a disease…what does it matter that it is an abnormal tension’…Those moments were…an intense heightening of awareness…and at the same time…the most direct sensation of one’s own existence…If in that second—that is to say the last conscious moment before the fit—he had time to say to himself, consciously and clearly, ‘Yes, I could give my whole life for this moment’, then this moment by itself was, of course, worth the whole of life… (Dostoevsky 1869/1955, pp. 243–244)

The link to religious feeling and mysticism is clear, the disease playing a curiously redemptive role. It is perhaps possible to stimulate something like this mystical sense artificially. As explored in a lengthy series of articles, Dr. Michael Persinger and a number of co-researchers (see Persinger 1983 for an early example) has devised a machine that, he claims, can induce mystical or religious experience. The machine operates by immersing the brain in a feeble but complex dynamic magnetic field which mimics and interacts with the natural electromagnetic activity of the temporal lobes, a technique generally referred to a transcranial magnetic stimulation (TMS). TMS is another beautiful anchor point which links basic physical properties and processes to the high level phenomena of mental functioning and consciousness. TMS operates by electromagnetic induction: a rapidly changing magnetic field will induce an electric field which affects the way neurons generate their output signals thus affecting neural function (generally by promoting neural activity) and, via a cascade of emergent processes, eventually altering the subject’s mental state. Examples of the effect of TMS include interference with judgments of moral responsibility (Young et al. 201014 ) and the temporary creation of a kind of Broca’s aphasia (Stewart et al. 200115 ). Persinger’s work purports to involve the more outrè mental state of ‘mystical communion’. Wearing a rather bizarre looking modified motorcycle helmet outfitted with a set of solenoids, a subject is settled comfortably in a darkened room while Persinger’s device plays a subtle symphony of magnetic fields designed to resonate with and hopefully amplify certain patterns of activity within the temporal lobe. The machine is reputed to have weird effects. After fifteen minutes of exposure or more, subjects often report a very strong sense of presence, as if an unseen ‘entity’ was with them in the experimental chamber (such a feeling of presence is highly typical of mystical experience). The well known consciousness researcher Susan Blackmore, a psychologist noted for her debunking of claims of the paranormal, was terrified by Persinger’s machine and reported feeling something ‘get hold of my leg and pull it, distort it, and drag it up the wall…Totally out of the blue, but intensely and vividly, I suddenly felt anger…Later, it was replaced by an equally sudden attack

50

4 Consciousness in the Brain

of fear’ (Blackmore 1994). Despite her experience, I have the impression that those of most subjects is not unpleasant, even if disturbing, but rather somehow especially ‘significant’. The effectiveness of this ‘experience machine’ is also highly dependent upon the subject. A Canadian journalist did not have such ‘good’ luck as Blackmore, apparently because his temporal lobes are insufficiently sensitive or labile, but even he managed a minor experience (see Hercz 2002). Some researchers, including Persinger, would like to explain our susceptibility to mystical and supernatural experiences, and hence the associated beliefs, to ‘inappropriate’ activation of the temporal lobe (perhaps brought about by local anomalies in the Earth’s magnetic field). Famous haunted houses might become scientifically explicable as zones of anomalous geomagnetic fields. In a more sinister and, if this is possible, in a still more speculative vein, Persinger has written a warning that the power levels required for inducing the requisite brain activity are not particularly high, leading to the possibility of large scale remote manipulation of our brains (see Persinger 1995, with the Strangelovian title ‘Electromagnetic Remote Control of Every Human Brain’). The paranoid sci-fi fantasy of satellite mind control creeps into sight. After all this, it is only fair to state that the sole attempt at a fully independent replication of Persinger’s results failed miserably, with the authors opining that the ‘god helmet’ works by the good old psychological mechanism of suggestibility rather than direct transcranial magnetic manipulation of the temporal lobe (see Granqvist et al. 2005). An obvious and intriguing—if somewhat irrelevant to the problem of consciousness—philosophical question here is whether such research ought to undermine religious belief.16 A surprising number of people who have discussed work like that of Persinger’s unite in denying any such implication, preferring to believe that the physical causes of these experiences are irrelevant to their status as evidence. After all, the possession of a detailed neurophysiological account of visual perception which allowed for the technical ability to generate visual hallucinations would hardly suggest that we are all deluded in trusting our eyes. In fact, perhaps one could turn discoveries such as Persinger’s around, to make them stand as positive evidence for religious belief. The argument would go like this. There is no Darwinian explanation of why we should be able to sense divine presence or have mystical experiences, since these have little positive bearing upon survival and reproduction (in fact, rather the opposite, if the legendary solitary and anti-social life of the mystic is anything like a typical result of the divine vision) and thus the fact that we possess such a ‘sense’ at all is like a maker’s mark, a sign that this faculty is telling us something important about the world. On the other hand, it seems to be a feature of mystical experience that it is best obtained by creating decidedly abnormal states, e.g. by lengthy fasting, long periods of isolation, etc. which might of course lead one to suspect that mystical revelation is more akin to hallucination than perception. It could be replied that exactly because of its potential for anti-Darwinian results the divine sense is and should be difficult to engage. But there is a philosophically more fundamental problem here. In order for this divine sense to operate there must be stimulation of certain regions of the temporal

4.6 Temporal Lobe Religiosity

51

lobe. This is, so far, analogous to visual perception: in order for visual perception to occur there must be stimulation of regions of the occipital lobe. The difference between veridical visual perception and mere illusion or hallucination, is the correlated source of these stimulations. When I see a pink elephant, the ultimate test of whether I am hallucinating or not is whether a pink elephant is in fact causing the activity in my visual cortex. Similarly, to tell whether the ‘God-sense’ is veridical or not we have to look at the cause of the sense of divine presence or mystical experience. It is abundantly clear that these causes are going to be perfectly natural: perhaps unusual geomagnetic activity, perhaps Persinger’s machine, perhaps, most often, internally generated stimulation of the temporal lobe due to some form of physical or psychological stress. If Susan Blackmore feels a strange invisible presence tugging at her leg while wearing Persinger’s helmet, we are not much inclined to count this as evidence of invisible leg-pullers, and one reason is that we have an obvious alternative cause of this bizarre experience ready to hand. In every case of mystical experience mediated by temporal lobe function, we will have such an alternative explanation because there has to be something in the natural world which sparks the anomalous activity of the temporal lobe. As in our earlier discussion, this is not an a priori knowable fact about the world. It is always conceivable that some cases of temporal lobe excitation will have no link to the rest of the natural world, that no explanation in terms of local conditions (either within the brain or in the state of enveloping electromagnetic fields) will be possible. But the very possibility of Persinger’s machine suggests otherwise. It suggests that here we have another anchor point connecting quite basic physical features of the world, in this case electromagnetic fields and the electro-physiology of neurons, to the very high level feature of the state of consciousness which constitutes the mystical experience.

4.7 Limbic Tone And, what is more, we know that temporal lobe disturbances can cause peculiar disorders of consciousness which no one could attribute to the creator. One of the most interesting of these also reemphasizes the importance of the interaction of various regions of the brain, and moves us into a new region of the brain. Deep within and surrounded by the cerebral cortex lies a set of structures sometimes called, since it seems to have appeared only with the rise of the mammals, the mammalian brain, or the (core part of the) limbic system (a rather vaguely defined functional unit also encompassing certain parts of the cerebral cortex, occasionally also called the ‘limbic lobe’). It includes the hippocampus, thalamus, hypothalamus and amygdala, all perched on top of and somewhat enveloping the brain stem while being in turn completely surrounded by the cerebrum (see Fig. 4.4). The limbic system serves a variety of fundamental functions, particularly in the assignment of emotional ‘tone’ or ‘value’ to experiences (as well as the control of glandular response through hormone release) and is crucial for the formation of long

52

4 Consciousness in the Brain

Fig. 4.4 Limbic system (image courtesy NIH)

term memory. Intuitively, of course, there is a very close connection between these aspects of cognition: emotionally intense experiences are hard to forget17 and have unmistakable strong physical effects on the rest of the body. It is part of the evolutionary function of emotion to tell us what is important and hence what to learn. As with any aspect of the brain’s functioning it is safe to say that the interconnected operations of the components of the limbic system are at best imperfectly understood. It seems that the thalamus serves as a kind of gatekeeper and clearing house, shuttling important signals to other parts of the brain for analysis and assessment (all sensory signals, save those of smell, pass through the thalamus). The hippocampus (despite its Latin name only very slightly like a seahorse in appearance) is crucial for the formation of memories, though not for memory storage. People unfortunate enough to have lost or damaged their hippocampi (as usual, there are mirror image versions of the organ on each side of the brain) simply cannot lay down any new explicit memories.18 They do retain the ability to learn new motor skills and there are curious indications that experience is stored in some non-explicit way, especially if the experience is emotionally charged. Such unfortunate subjects will form aversions to people with whom they have regularly unpleasant interactions, although they have no idea why they have such feelings (see Damasio 1999, Chapter 2). The almond shaped amygdala, which resides just in front of the hippocampus, is notably focused on the assignment of negative value to experiences and serves an especially distinctive function in assessing stimuli as fearful. Subjects with damaged amygdala show such symptoms as a lack of appropriate response to fearful stimuli, a marked lack of the proper reserve in dealings with strangers and even the inability to recognize expressions of fear in others (something at which we are normally extremely swift and good at; in fact, fearful faces appear to be among the most salient of emotional displays, see Yang et al. 2007).

4.7 Limbic Tone

53

The interaction of the temporal lobe’s recognitional abilities with the limbic system’s function of emotional assessment leads to a very bizarre disorder of consciousness, known as Capgras syndrome (first reported as a distinctive condition in Capgras and Reboul-Lachaux 1923). Sufferers fall under the delusion that those closest to them—spouses, parents, sometimes even pets—have been replaced by impostors (as it may be evil twins, clever actors, disguised aliens, robots). A plausible account of Capgras syndrome has been offered by H. Ellis and A. Young (1990); see also Stone and Young (1997) which exploits the diversity of function within the brain and the attendant possibility that ‘complex’ states of consciousness require these functions to interact in particular ways. In this case, Ellis and Young endorse an idea in Bauer (1984) which posits dual information pathways in the brain, a cortical path underpinning conscious recognition ‘by similarity’ and a second path involving limbic structures which adds ‘emotional recognition’ (note the conceptual similarity of this suggestion to the explication of blindsight of Milner and Goodale discussed above). Thus we have the functions of recognition (especially, as we shall see, facial recognition), emotional response as a component of recognition and the integration and assessment of that emotional response into appropriate feelings, thoughts and ultimately beliefs. It seems that someone suffering from Capgras syndrome retains the ability to recognize people by the standard physical clues of appearance, voice, etc. In particular, the specialized face recognition circuitry, as discussed above, remains pretty much intact (frequently, and probably significantly, there is usually some impairment in the ability to recognize faces in Capgras patients). But the problem lies in the distinct neural pathway involving the limbic system which fails to assign appropriate emotional tone to incoming stimuli, and hence to the presence of loved ones. Thus the patient can see that someone who at least appears just like a loved one is present, but the patient has an anomalous lack of appropriate conscious emotional response. The brain’s attempt to integrate perceptual recognition with lack of what might be called ‘emotional recognition’ leads to the belief that the person is not really the loved one at all, but rather some kind of impostor. Sufferers are not, of course, inclined to say that they lack an emotional response to familiar people or animals, rather they assert that their friends, pets and lovers ‘look funny’ or ‘not right’. It seems the tendency to place fault anywhere but in ourselves is not just a feature of normal human psychology. Such a theory of Capgras syndrome raises some intriguing points. About consciousness itself, it suggests that we experience the emotional tone of the world around us as a ‘part of’ that world, not a post-facto and relatively independent assessment of an initial, and neutral, perceptual experience. That is, it seems that we represent things around us as possessing emotional, or in general, value attributes in addition to the more commonplace perceptual attributes (see Seager 2000a, 2002). At the level of general cognition, the syndrome raises a disturbing question about rationality, one which could be asked about many of the bizarre deficits that result from localized brain damage. The question is why are the victims of this sort of damage prone to formulating (or as it is frequently labeled, confabulating19 ) such wild hypotheses to account for their strange feelings, and why do they end up actually

54

4 Consciousness in the Brain

accepting these hypotheses? I think one might naively be inclined to suppose that if, for example, a patient was told that they had Capgras syndrome (and, let us say, they were also provided with good case studies of other instances of the problem, and as comprehensive a neurological account of it as we now possess) then the patient ought to accept that diagnosis and thence give up the wild belief that a set of impostors has been implanted in their home. While this purely intellectual acceptance likely would not restore the emotional response, it would, one might think, vastly improve the lives of Capgras victims. Now, there may be some people who sustain the kind of brain damage that results in delusional syndromes such as Capgras who do come up with more reasonable hypotheses. We do not hear about such cases in the literature. Of the cases we do read about, there is little indication that the delusion can be cured by ‘cognitive therapy’. Why is that? There is some evidence that the processes that underlie belief fixation themselves are disturbed in victims of ailments such as Capgras syndrome (that is, delusional states in general). Stone and Young (1997) report that delusional people tend to exhibit several biases in their reasoning, such as jumping to conclusions, failing to take adequate account of possible alternative possible explanations, failing to give sufficient weight to background beliefs, an abnormal willingness to embrace probabilistic judgments upon unusually little evidence (they will, for example, form an opinion as to how many of the balls in an urn of red and white balls are, say, white more quickly than average20 ). These are all faults of reasoning that might tend to lead delusionals toward more rather than less extravagant explanations of their plight. We must be careful, though, in our use of the loaded words ‘disturbed’, ‘fault’ and ‘bias’. It may be that the reasoning styles of delusional people are well within the normal range. There need be no implication that whatever organic problem caused the delusional state had any sort of direct and detrimental effect upon the subject’s reasoning skills (although of course this may often occur). It may be, for example, that in a case of Capgras syndrome a pre-existing style of reasoning is, as it were, exploited by the breakdown in the emotional recognition system to generate the characteristic bizarre delusions. The proverbial Missourian skeptic who suspiciously demands ‘show me’, if beset by damage to the emotional recognition pathway might have difficulty accepting the true identity of a spouse, despite non-sensory evidence, the pronouncements of authorities and background knowledge, given that loved ones simply do not seem to be who they claim and look to be.21 Still, the rarity of Capgras syndrome along with the scattershot nature of accidental brain damage makes it quite plausible that reasoning itself might be affected, perhaps minimally in amplifying or emphasizing certain preexisting ways of thinking. However, this amplification need not necessarily be the result of physical damage to or malfunction of some core ‘reasoning component’ of the brain. It could, for example, simply be the effect of a sudden, forced greater reliance on reasoning in contexts where reason was previously entirely unnecessary and quite out of place. One normally does not have to come to a reasoned rejection of the hypothesis that one’s spouse has been replaced by an almost perfect replica—such hypotheses just do not arise except in exceptionally bizarre circumstances, and/or the philosophy seminar room. The Capgras victim is

4.7 Limbic Tone

55

in the extreme position of persistently getting evidence from a sensory system that remains generally reliable that the bizarre hypothesis is a live option. Thus it might be expected that distinct forms of functionally localized brain damage could interact with idiosyncratically varying modes of reasoning already in place to yield an affinity for particular syndromes. Perhaps someone whose general willingness to believe a novel proposition tends to depend more upon intuition, that is, the ‘feel’ of a situation and its associated emotional response, and less upon abstract logical thinking and acceptance of external authority would be more susceptible to Capgras syndrome. Worse, there is every reason to think that beliefs so founded could be, paradoxically, confirmed by ‘testing’. Once the basic delusional hypothesis of Capgras syndrome has suggested itself, as Stone and Young put it (rather comically despite the tragedy of the situation), ‘apparently confirmatory evidence is rapidly forthcoming, because the more carefully relatives are observed, the stranger they seem, and they will also start acting unusually in response to the increasingly bizarre behavior of the person now experiencing the full-blown Capgras delusion’ (Stone and Young 1997 p. 344). It is also interesting to speculate that if and insofar as reasoning styles are in part a matter of cultural milieu (see Nisbett et al. 2001) we might also find the prevalence of the various delusional disorders to correlate with culture in some way. Now, there is of course a trivial dependence of delusional states upon culture insofar as forms of knowledge, theory and indeed concepts themselves are cultural constructions. Thus it is impossible for someone unacquainted with modern western culture to be under the misapprehension that they are the victims of alien abduction, or that Lady Gaga is secretly in love with them. However, it is easy to see that analogues of such delusions could be present in any culture (delusions of demonic possession, voodoo witchcraft and the like can readily stand in for alien kidnapping and CIA mind control). Some delusions which it seems could or should be equally prevalent across cultures are not, and these are more interesting cases. In south China, Singapore and Malaysia is found the delusion known as Koro—the fear that one’s genitals are shrinking, which while far from common there is extremely rare elsewhere. It is far from clear that there is any kind of culturally mediated cognitive explanation for the distribution of Koro. The idea is not unthinkable however.22 Nisbett’s research suggests that western modes of thought tend to focus upon selecting highly salient particular features or individuals out of the environment, relatively ignoring the context in which they occur. This might seem tailor made for the Capgras delusion, in which, it appears, one super-salient fact about a highly significant individual seems to drive the delusion. In contrast, those from Asian cultures (and it is the culture that matters here, not the genes, Japanese Americans reason as westerners according to Nisbett) reportedly pay a lot more attention to the context of an event and the relations between the elements of the environment. Perhaps this would make Capgras less likely in Asian cultures. I have no idea, and since according to Nisbett part of the Asian ‘way of thought’ is a greater ‘tolerance of contradiction’, Capgras could sneak back in. It would be interesting to see if Capgras syndrome does correlate with culture in any way.

56

4 Consciousness in the Brain

4.8 Cingulate Consciousness It is past time to return to our tour of the brain. Roughly speaking, the inner regions of the cerebral cortices, surrounding what I have called the limbic system (although the nearby cortical regions are often included within the limbic system) are involved with the integration of sensory information, emotional response and thought. One broad area, the cingulate cortex (which lies within the fold between the two cerebral hemispheres just above the corpus callosum), seems to bear a special relation to consciousness. For example, without a proper functioning cingulate, pain loses its affect, though it can still be felt. Such a loss of affect, the felt significance of thoughts, perceptions and sensations, resulting from damage to parts of the frontal lobe has been noted for some time. Many of the unfortunate victims of mid-twentieth century attempts at psychosurgery, those insulted by the removal, destruction or disconnection of the prefrontal lobes (for which the inventor—Antonio Moniz—received a Nobel prize in 194923 ) suffered a general loss of affect and would sometimes report that pain no longer bothered them (by and large, very little either bothered or pleased them). Cingulectomy, a closely focused and highly constrained descendant of the lobotomy aimed at the cingulate cortices (see Fig. 4.2), remains an option of last resort for severe intractable pain but is now a very precise and carefully circumscribed operation. When this operation succeeds patients can still feel their pain, but it no longer hurts. The awfulness of pain can even be brought under a kind of mental control. A remarkable study (Rainville et al. 1997) employed hypnotic suggestion to reduce (and also to increase) the ‘unpleasantness’ of the pain of putting one’s hand in very hot water. But that is not the remarkable thing about this experiment. Subjects were asked to put a hand in water of about 47◦ C (117◦ F) and rate both the intensity and the unpleasantness of the pain on a scale from 0 to 100 (for the sake of comparison, the average ratings for unhypnotized control subjects were around 78 for intensity and 61 for unpleasantness). Subjects were then hypnotically suggested to either increase or decrease the unpleasantness of the pain. The unpleasantness ratings of the pain changed significantly (ratings were now an average of 81 and 45 respectively). What is remarkable is the brain confirmation of the changed ratings and the precise localization of the change in brain activation corresponding to increased or decreased unpleasantness. Using PET scans, the study clearly shows greatly increased activity in the anterior cingulate cortex upon hypnotic suggestion to increase unpleasantness, and markedly decreased activity in the very same region given the suggestion to decrease unpleasantness. The cingulate cortex is by no means merely or solely devoted to registering the affect of pain. It is involved in a number of tasks related, roughly speaking, to the monitoring and assessment of ‘internal’ states: visceral sensations, feelings (of which pain is of course an especially important representative) and emotions. Nor is its functioning invariably associated with consciousness. Sleepwalkers show unusual activity in the cingulate cortex during their somnambulism (in which consciousness—even the kind associated with dreaming—is usually absent) but it is interesting that episodes of

4.8 Cingulate Consciousness

57

sleepwalking go with activation of the posterior cingulate, whereas conscious states and processes are associated with activation of the anterior cingulate (for example, the kind of sleep that is correlated with dream states, the so-called rapid eye movement (REM) sleep, coincides with a marked increase in anterior cingulate activity (see Braun et al. 1997). It has been postulated that the anterior and posterior portions of the cingulate cortex are functionally differentiated, with the posterior part playing a more passive, monitoring role while the anterior takes on a more active ‘executive’ function (perhaps we can see here a connection between activation of the anterior cingulate and the imperative motivation of pain).24 Let me remind the reader yet again that this is not to be taken as a claim that the ‘seat’ of consciousness (or even that of conscious painfulness or other ‘intrinsically attention demanding’ experiences) has been found in the anterior cingulate cortex. My view is that trying to localize consciousness is as fundamentally mistaken as trying to localize a hurricane’s destructiveness at some definite zone within its swirling winds, or to nail down the beauty of a painting in some particular region of it. But this does not mean that we cannot discover that certain brain regions figure in the generation of consciousness, or kinds of consciousness, while others do not (and, of course, we should expect there to be gradations on a scale of significance for consciousness generation across various brain regions or systems). The anterior cingulate is at the top of the brain, at the base of the cleft between the two hemispheres (the longitudinal cerebral fissure) just above the corpus callosum (again, see Fig. 4.2). If we followed the ‘gray matter’ (composed of the neurons themselves that make up the processing units of the brain according to the neuron doctrine) along the convoluted surface of the cortex from the cingulate upwards to the surface of the frontal lobe and then down towards the temporal lobe we would find a kind of fjord leading into the brain. This is the lateral sulcus or Sylvian Fissure (marking a boundary between frontal and temporal lobes). Buried within this cleft is the insular cortex or insula, a curious zone of tangled gyri and sulci (see Fig. 4.5). The insula is another cortical area primarily involved in emotional response and the generation of feelings. It has been specifically linked to recognition of facial expressions of disgust (see Phillips et al. 1997)25 and is part of the general limbic complex which deals with ‘hot’ and disagreeable emotions, of which fear is the prime example. But the insula also seems to play a role not frequently associated with the cerebral cortex, namely in the monitoring and control of fundamental physiological functions. The insula responds differentially and uniquely to changes in cardiac rhythm and blood pressure.26

4.9 Self Representation In the development of his general theory of personal consciousness, Damasio postulates that the insula is part of the cortical structure which encodes the ‘most integrated representation of the current internal state of the organism’ (Damasio 1999, p. 156), p. 156) thus forming a crucial component of what Damasio calls the proto-self. It is

58

4 Consciousness in the Brain

Fig. 4.5 The insular cortex (image courtesy John Beal)

extremely important to stress how our value assessment of the world (including our internal milieu)—which is the basic function of the emotions—is an essential part of our representation of the world. Moving upwards across the brain we come to the parietal lobe and the somatosensory representation that is laid out across it from top to bottom (see Fig. 4.3 above). Readers have doubtless seen a diagram of the grotesque ‘somatosensory homunculus’ who represents the amount of brain power devoted to all your body’s internal sensory channels (there are in fact a huge number of such ‘maps’ in a number of places in the brain, especially in the visual cortex representing various types and patterns of stimulation on the retina). The distortion in the homunculus is due to the fact that certain parts of your body are much more densely covered with sensory receptors than others. For example, and not surprisingly, our hands are highly over represented compared, say, to our arms; the tongue is huge compared to the ear. The somatosensory homunculus has its motor twin in a neural strip just in front of the somatosensory homunculus, across the divide between the frontal and parietal lobes (the central sulcus). The differences between the two homunculi are interesting and intuitively reasonable, diverging exactly in those places that differ in sensory receptiveness versus motor control (no doubt readers can readily think of some examples). The somatosensory cortex is intimately related to consciousness, providing as it does a core part of our representation of the current state of our body and its immediate relation to the environment. There are some spectacular and instructive breakdowns in the integration of one’s awareness of one’s own body caused by damage in either the somatosensory or motor cortices. The well known, but nonetheless bizarre, phenomenon of the phantom limb is, in part, a result of a disconnection between the somatosensory cortex and the rest of the body. When a limb is amputated it leaves behind, at the very least, a location in the brain where sensation from it was registered and represented. If that part of the brain is stimulated (perhaps because of stimulation of nearby areas of the somatosensory representation, which need not be a nearby area of the body since the physical layout of the somatosensory representation does not

4.9 Self Representation

59

fully preserve anatomical relations) it is probable that there will occur a sensation as if in the missing limb. The phantom limb is a beautiful example of what philosophers call the intentionality of sensation. Intentionality, in general, is the feature of mental states (or any other representational state) of ‘being about’ something. Pictures, words, diagrams, computer programs are all examples of intentionality. It is a nice question what the difference is between intentionality and merely carrying information, but there must be such a difference because everything carries information about other things, but very few things in the world are representations (and representations need not represent everything they carry information about). For example, the moon carries a lot of information about the history and early conditions of the solar system but does not represent that information. Textbooks about the moon and the solar system actually represent those things as well as—we hope—carrying information about them, as well as carrying information about many other things not represented at all. Thus it seems reasonable to suppose that intentionality is to be explained as carried information plus something else. What a representation represents is a subset of what it carries information about, and somehow what determines something to be a representation is what determines which subset of information is represented (or is supposed to be represented). Presumably it is correct to say that the somatosensory cortex really does represent the state of our body (at least the component having to do with tactile sensation) whereas it does not represent various facts about our species’ evolutionary development (even though, I suppose, one can find out a lot about the latter from the structure of the former). It is then very tempting to say something along the lines of: it is the biological function of the somatosensory cortex to provide information about the tactile state of the body, and it is having this function that makes the states of the somatosensory cortex representational states (and, of course, it then follows that what it is these states have the function of providing information about naturally constrains the field of information to what is supposed to be represented). It may be alright to give into this temptation, but such weakness immediately leads to the demand for an account of functions. And while there are lots of these on offer, no one knows which, if any, is correct (see Ariew et al. 2002). In any case, whatever account of representation (especially mental representation) one favours, a vitally important feature of intentionality is that it is possible to represent things that do not exist. Thus the word ‘unicorn’ represents a creature which does not, never did and in all probability could not exist (recall that only virgins can see unicorns, and how that is supposed to work I don’t have the faintest idea). Similarly, the phantom limb that gives pain after amputation simply does not exist (if one could pick something up with one’s phantom limb that would be different). Nonetheless, the limb is represented as existing by the mind. This is important for understanding consciousness. The representation of the tactile body is available to consciousness. But it is not always, nor all of it, in consciousness. Activation of the somatosensory cortex is not by itself sufficient for consciousness. However, that part of the somatosensory representation which is presented in consciousness is a representation of the body, which is represented in all the distinctive ways we have of feeling: emotional as well as sensory. In general, I tend to endorse the idea that

60

4 Consciousness in the Brain

consciousness just is a form of representation, of which the consciousness of the body is but one component.27

4.10 Intentions Another fascinating aspect of the parietal lobe is worth mentioning: a particular area of the posterior parietal called the parietal reach region (PRR). As usual, most of what we know about this brain region comes from invasive studies on monkeys, but there is some fMRI work which supports the unsurprising conclusion that we share the PRR with our simian cousins (see Connolly et al. 2003). What is especially intriguing about the PRR is that neural activity within it seems to correspond to high-level, occurrent (hence conscious) intentions to reach out for or point at things. The PRR is active when one (or at least a monkey) thinks about reaching for something even if one is in the dark and whether or not the movement is actually executed. Some recent work involving individual neuron monitoring in monkeys (Musallam et al. 2004) reveals our growing ability to decode the mental state of a subject based upon analysis of the neural activity of the subject’s brain. It is an interesting question how to set up an experiment to verify that you have correctly decoded an intention that is not acted upon. In this case, the problem is ingeniously solved by first teaching the monkeys to reach for where a light appeared on a display by memory, a couple of seconds after the light has gone off. Several neurons in the PRR are monitored while the monkeys work on this task until the neural code is cracked. Then the experimental paradigm is altered: reward the monkey for reaching for where the computer predicts it will reach on the basis of neural state. Over time quite high accuracy of prediction is attained. Such results are still quite crude, and may suggest to the suspicious reader that the experiment more likely indicates that monkeys are smart enough to learn how to manipulate their own brain’s activity in order to receive a reward rather than a straightforward case of mindreading from a naturally occurring mind-brain correlation, but there is no doubt that linkages are being forged from neural activity to mental states. The ultimate goal is the development of ‘neural prosthetics’ by which the disabled may be able to control computers, machines or more standard prosthetic devices via a real time decoding of their intentions’ reflections in the brain. It is striking, and more than a little sobering, to find the authors bold enough to state: ‘recording thoughts from speech areas could alleviate the use of more cumbersome letter boards and time-consuming spelling programs, or recordings from emotion centers could provide an online indication of a patient’s emotional state’ (Musallam et al. 2004, p. 258). It will not be very long before prosthetic limbs will be directly activated by decoded neural signals (in the more distant future such linkages will extend beyond the body to institute the long held dream of independence from physical instrumentality). Be that as it may, and seeing how we have circled back in the direction of our original lovelorn thought experiment, it is time to draw our cursory summary of brain and consciousness to a close. We could go on in much greater depth about

4.10 Intentions

61

the exponentially expanding knowledge of relationships between brain systems and consciousness (for overviews see Koch 2004 or Zeman 2002) but the point should be clear enough. In a way strikingly parallel to the modern vision of cosmology, there is abundant evidence of the dependence of high level phenomena upon the neural activity of the brain. Even as galaxies resolve themselves into stellar interactions and stars resolve themselves into atomic processes so the brain, and seemingly also the mind, resolve themselves into the ceaseless biochemical activity of the neurons. The linkage from the sub-atomic realm to the foundations of consciousness remains murky and tenuous, but at the same time already richly interconnected and exceedingly fruitful. As in the first chapter, at every step of the way we can easily imagine an explanatory roadblock being thrown in our path. Instead, nature has been generous to prepared minds, steadily revealing more and more of the underpinnings of consciousness. There is no doubt that consciousness is now within the purview of neuroscience. When Galileo first turned his home made telescope on the heavens he had an ontological epiphany: What was observed by us…is the nature or matter of the Milky Way itself which, with the aid of the spyglass, may be observed so well that all disputes that for so many generations have vexed philosophers are destroyed by visible certainty, and we are liberated from wordy arguments. For the Galaxy is nothing else than a congeries of innumerable stars distributed in clusters. (Galilei 1610/1989, p. 62)

Will some future neuroscientist, with some new kind of instrument of which we now have but the most meager idea, have a similar experience which itself reveals that consciousness is ‘nothing else than’…what? The story I’ve told so far may make that in some sense possible or even to seem a juggernaut of likelihood, but the story thus far begins and ends with apparently non-conscious components. This means that consciousness must be an emergent feature of the world. So, what is emergence?

Part II

Emergence and Consciousness

Chapter 5

Emergence and Cellular Automata

5.1 The Game of Life Despite its portentous name, emergence seems to be the most natural thing in the world. Everywhere we look we find things which possess properties which their components or constituents lack. Trees are made of leaves, trunk and root, but none of these things are trees in their own right. If anything about the structure of nature seems obvious and irrefutable it is that nature possesses a hierarchical structure within which ‘higher level’ entities have properties lacked by ‘lower level’ entities. This is what I labeled synchronic emergence in Chap. 1. Nature no less obviously appears to evolve or develop in time, so that new entities and new features appear over time. This was labeled diachronic emergence. In the last three chapters we have been exploring a small part of the coastline of the natural hierarchy, from the ground floor level of the sub-atomic realm to the dizzying heights of consciousness itself. This is a hierarchy of both synchronic and diachronic emergence and thus we have been exploring and implicitly defending a grand system of emergence. Although familiar and ubiquitous, there are some philosophical problems that arise from emergence. As often happens in philosophy, what seems obvious and unproblematic in normal use reveals some deeper mysteries when looked at more closely. The best way to approach the puzzles of emergence is with a basic example. My example is admittedly rather hackneyed, but that is only because it is the perfect beginning point in the examination of emergence. The example is John Conway’s ‘Game of Life’ (henceforth simply ‘Life’) or, more generally, cellular automata.1 Life was invented by Conway in 1970 and was widely popularized by Martin Gardner in some of his famous ‘Mathematical Games’ columns in Scientific American (see Gardner 1970). The concept of a cellular automaton (CA) goes back somewhat further, to work of John von Neumann, Stanislaw Ulam and Arthur Burks. The statistical mechanics model of Ernst Ising in 1925 (for an account of the model and its history see Brush 1967) is an even earlier version of the essential ideas behind the cellular automaton, though one that seems to have had little influence on the subsequent development of the mathematical concept (see Hughes 2000 for an interesting W. Seager, Natural Fabrications, The Frontiers Collection, DOI: 10.1007/978-3-642-29599-7_5, © Springer-Verlag Berlin Heidelberg 2012

65

66

5 Emergence and Cellular Automata

philosophical investigation of the Ising model). The most compendious and ambitious exploration of the cas must surely be found in Wolfram (2002), wherein the reader will discover that—apparently—cas provide the answer to life, the universe and everything. It is hard to know what to make of Wolfram’s book. Some reviewers seem to have had something like the experience once had by Dorothy Parker: ‘this is not a [book] to be tossed lightly aside; it should be thrown with great force’. My own opinion is simply that it is indeed possible, although apparently unlikely, that there could be something like a cellular automaton model of the fundamental physics of our world. This possibility is of great significance for the topic of emergence. The game of Life is ‘played’ in an imaginary world in which space is two dimensional and quantized: the world is a grid of positions, and these positions can have but one binary property: that of being ‘occupied’ or ‘unoccupied’ (on or off, alive or dead). An initial position is simply a set of occupied positions, or cells. The dynamics of Life are also very simple, consisting of but three ‘laws of nature’. Here’s how Martin Gardner explained them back in 1970: Conway’s genetic laws are delightfully simple. First note that each cell of the checkerboard (assumed to be an infinite plane) has eight neighboring cells, four adjacent orthogonally, four adjacent diagonally. The rules are: (1) Survivals. Every counter with two or three neighboring counters survives for the next generation. (2) Deaths. Each counter with four or more neighbors dies (is removed) from overpopulation. Every counter with one neighbor or none dies from isolation. (3) Births. Each empty cell adjacent to exactly three neighbors—no more, no fewer—is a birth cell. A counter is placed on it at the next move (Gardner 1970, p. 120).

The ‘components’ or denizens of a Life world are simply the cells and they have but two possible states—alive or dead (conventionally represented by a cell being colored or left blank respectively). But Life worlds evolve in complex and interesting ways, exhibiting features that go far beyond the on/off nature of the individual cells.2 This is certainly a kind of diachronic and synchronic emergence. Figure 5.1 presents an example of a possible initial state. Each ‘configuration’ here will, so to speak, persist and form into an identical copy of itself after four iterations of the rules; but the copy will be moved two squares horizontally. The left configuration moves towards the right, the right configuration then necessarily—by symmetry—moving to the left. Therefore a collision is inevitable, the outcome of which is the birth of two instances of one of the more famous denizens of the Life world, called gliders. After 19 iterations we are left with the pattern shown in Fig. 5.2. The two gliders are also 4-iteration persisting patterns and will move off diagonally to the northwest and northeast. Figure 5.3 presents a much more complex initial state. It would be a tedious business to calculate by hand how this will evolve, but, as mentioned in note 1 above, there are many computer implementations of Life. Using one would reveal that this pattern is intrinsically stable with a period of 246 time ticks but in addition steadily gives birth (once every 246-ticks) to gliders which depart to the northwest, as illustrated in Fig. 5.4. The circled glider is proceeding off the grid to the upper left and another one will eventually be spawned in the central ‘interaction zone’ only to follow the first to infinity. Other initial patterns will ‘absorb’ gliders,

5.1 The Game of Life

67

Fig. 5.1 Possible Life configuration

Fig. 5.2 Two gliders

Fig. 5.3 More complex pattern

or ‘reflect’ them. And there are many other well known Life creatures (e.g. ‘glider guns’ such as we have seen that generate gliders, ‘puffer trains’ and a whole host of different kinds of guns generating all sorts of glider like configurations). All such features meet the simple definition of emergence we are currently working with. They

68

5 Emergence and Cellular Automata

Fig. 5.4 Future of Fig. 5.3

have properties which their components lack. Gliders, for example, reconstitute their original form after four iterations; individual cells lack that feature. This is a case of synchronic emergence. Other initial patterns of occupied and unoccupied cells— glider guns—will produce gliders, a feature which did not exist in the original pattern. This is diachronic emergence.

5.2 Digital Physics The ontology and laws of the Life world are very simple. Does this preclude the game of Life from providing any insight into the kind of emergence we seem to find in the real world? Far from it. It has been conjectured that the ultimate foundation of the actual world is a cellular automaton (or cellular automaton-like) system (see Wolfram 2002 for some recent speculation). The first person to investigate this idea seems to have been Zuse (1969).3 The most actual work on the idea appears to be that of Fredkin (1990, 2003, 20044 ). The general program of modeling the world at its most fundamental level as a cellular automaton is called ‘digital physics’ or, more provocatively, ‘finite nature’ by Fredkin.5 The scale at which this hypothetical cellular automaton which ‘computes’ the universe (let’s call it the ultimate cellular automaton or UCA) would be very much smaller than even subatomic dimensions, probably approaching the Planck length. In fact, there is no particular reason to suppose that the UCA works in what we call space. The neighborhoods around the cells of a CA are purely abstract. We might thus hope that space, and time as well, are no less emergent features of the world

5.2 Digital Physics

69

than its more ponderable denizens.6 This would be an amazing vindication of a part of Leibniz’s old idea that the radically non-spatial monads’ systems of perception generate the relational structure we call space. Time is perhaps a more problematic emergent for Leibniz since the monads themselves seem to evolve against a temporal background, although Leibniz might be able to construct time from the internal causal relations of the monads (see Cover 1997). We could perhaps go so far as to consider the ‘instantaneous’ (or infinitesimally temporally extended) states of all possible monads as the fundamental entities of the universe. Then time itself will appear as a special set of relations amongst these states (such an idea has been explored in the context of physics in Barbour 2000 with the possible states of the universe replacing those of the monads). Then instead of saying that the UCA operates at a length scale of about the Planck length, we should say that it is at that length that spatial properties and relations would break down, and the universe would give evidence that it is the product of a CA. Unfortunately, experiments that could probe such a length scale are hard to imagine, but hopefully as digital physics is developed some experimentally accessible implications will emerge. Otherwise the doctrine can only remain pure metaphysics, albeit mathematical metaphysics. No one knows how or if digital physics will pan out. The core problem is to link some CA architecture to the physics we already know. As Fredkin puts it: ‘the RUCA [reversible universal cellular automaton] runs a computation. As a consequence of that process and of appropriate initial conditions, various stable structures will exist in the lattice. For each such stable structure, we expect that its behavior will mimic the behavior of some particle such as a muon or a photon. What we demand of a correct model is that the behavior of those particles obeys the laws of physics and that we can identify the particles of the RUCA with the particles of physics’ (Fredkin 2003, p. 192). The particles (and fields and everything else for that matter) of standard physics would thus all be emergent entities. The task of establishing how standard physics can be linked to some RUCA is immensely difficult however and not yet very far advanced. It might be thought that the continuous mathematics of the calculus which serves as the basic mode of description for all our basic theories, and which has been applied with huge success throughout all domains of science, precludes any serious interpretation of the world as based upon the digital and discrete workings of a cellular automaton. After all, according to standard formulations of physics, many systems are capable of instantiating an infinite and continuous range of states (e.g. the position of a particle can take on any real number value). The situation is not so clear cut however. It is possible to demonstrate rigorously that some CA systems generate behavior asymptotically describable in continuous terms. For example, the Navier-Stokes equations governing hydrodynamical flow can be retrieved from so-called lattice gas models which are types of CA (see Frisch et al. 1986). Of course, I emphasize that no one knows whether all of continuous physics can thus be retrieved or approximated from a CA model, but no one knows otherwise either (see Fredkin 2003, 2004). It is important to bear in mind that the ‘space’ of the CA lattice is not the space of our physical universe (as deployed and described in scientific theory) and the time-tick of the CA is not the time we measure or experience. So the mere

70

5 Emergence and Cellular Automata

fact that the space and time of the CA are discrete does not rule them out even if we grant that the best scientific description of our experienced (and measured) space and time makes them continuous quantities.7 The weird behavior of some quantum systems called entanglement,8 in which two systems that have interacted maintain a mysterious kind of connection across any distance so that interaction with one will instantaneously affect the state of the other, might appear to contradict the adjacency requirement of cellular automata. But, again, it is far from clear that this is a real problem for digital physics and for the same reason. The spatial distance separating the ‘parts’ of an entangled system which makes entanglement seem ‘weird’ need not be reflected in the underlying workings of our hypothetical universal CA. In fact, could it be that the phenomenon of quantum entanglement is trying to tell us that what we call spatial separation is not necessarily a ‘real’ separation at the fundamental level of interaction? CA systems extend far beyond the game of Life. They can be defined for any number of dimensions or number of neighbors. They all share a fundamental finiteness and discreteness. Traditional Life has some deficiencies as a serious model of the foundation of the world. For example, the Life rules are not reversible, which means that information is lost as a Life position evolves. Any number of initial positions will lead to there being no living cell left after a certain period of evolution and it is rather difficult to extrapolate backwards from an empty plane to the particular initial position which led to it! It appears however that the laws of nature we have discovered so far guarantee that information is conserved. This follows from the determinism of our fundamental theories. Nature also seems to abide by certain fundamental symmetries or, equivalently, conservation laws. Some of these emerge naturally from reversible cellular automata (see Fredkin 2004). It is worth stressing that although quantum mechanics is said to have introduced irreducible indeterminism into our most basic picture of the world, the evolution of the wave function of any quantum system is strictly deterministic. Only the measurement process introduces any possible indeterminacy. Under the standard view, measurement introduces a random collapse of the wave function into one of the eigenfunctions of the observable property being measured (that is, if we are, for example, measuring position our measurement will yield a particle with a definite position). Since the wave function usually encodes more than one eigenfunction we have only a probability of getting any particular measurement result. However, a full quantum mechanical description of the world ought to include the measurement apparatus and will deterministically move the wave function of the system under observation plus the apparatus to a new quantum mechanical state. Presumably, a quantum cosmology will incorporate a quantum mechanical description of the whole world and thus will have no need to ever invoke the collapse of the wave function during measurement. All sorts of fascinating questions intrude here, but it seems quite viable to maintain that the quantum mechanics of the whole universe will maintain determinism and thus the preservation of information. Readers may recall the flurry of journalistic interest in 2004 when Stephen Hawking conceded that black holes do not purge information from the world (as he formerly believed) but rather retain it in correlations between what falls into the black hole and

5.2 Digital Physics

71

the Hawking radiation which it emits. The point is expressed very nicely by Lubos Motl who wrote in his blog: When we burn books, it looks as though we are destroying information, but of course the information about the letters remains encoded in the correlations between the particles of smoke that remains; it’s just hard to read a book from its smoke. The smoke otherwise looks universal much like the thermal radiation of a black hole. But we know that if we look at the situation in detail, using the full many-body Schrödinger equation, the state of the electrons evolves unitarily (Motl 2005).

Notwithstanding the foregoing, the game of Life has one extremely interesting positive feature: it is computationally, or Turing, universal (see Berlekamp et al. 1982; Gardner 1983).9 What this means is that a Life position can be set up which can be interpreted as carrying out the computations of any Turing machine. This means as well that the Life rules can simulate any other cellular automaton. Thus the deficiencies noted above are rather notional, although it would be a perverse metaphysics (or physics) that asserted that the basis of the universe was Conway’s game of Life, simulating a reversible cellular automaton (there are any number of CAs which are universal in the sense that they can emulate any Turing machine so the claim that the universe is such-and-such a reversible CA being simulated on an X-type CA would seem to be a paradigm case of underdetermination of theory by data; nor could we rule out a, perhaps infinite, chain of CAs, each simulating the one ‘above’). Theoretically speaking however, this fact shows that any kind of emergence which is embodied in some CA architecture will also occur in some Life world.

5.3 Hypercomputation It is also true that any finite CA (or finite ‘section’ of a CA) can be simulated by some Turing machine (or our own personal computers, which are Turing universal modulo memory capacity). This means that there is a potentially straightforward test of whether the universal CA hypothesis is correct. If some aspect of nature is correctly and essentially described in terms of uncomputable functions, then the universe cannot be running on an underlying CA.10 A large number of proposals for at least the possibility of such hypercomputational processes (to somewhat extend a term of Jack Copeland’s) have been advanced (see Copeland 2002; Stannett 2006; Ord 2006 for surveys). Roughly speaking, these proposals involve coming up with a way in which an infinite amount of information can be processed in a finite time. Two important variants of these schemes appeal respectively to, first, the effect of parts of nature embodying non-computable real valued parameters (such parameters exist with infinite precision) and, second, to possible natural processes that allow for infinite computations to occur in a finite time. A simple example of the second kind is the ‘accelerating Turing machine’ of which many variants have been proposed. The core idea is simply that of a Turing machine (or any other standard, Turing universal, computer) whose execution speed doubles for each operation. If an accelerating Turing machine takes 0.5 seconds

72

5 Emergence and Cellular Automata

to complete its first operation, 0.25 seconds for the next operation and so on, then evidently it can complete an infinite set of operations in one second. Such a machine could compute the standardly uncomputable halting function11 in the following way. The machine begins by simulating the action of some Turing machine (given some particular input). If that machine halts then the simulation will discover this at some stage of processing and then it will write a ‘1’ in its output and halt. If the simulated machine never halts then that ‘1’ will never be written but, after 1 s, we will know that the accelerating machine has simulated an infinite number of operations. Hence if there is no ‘1’ in the output then we know that the simulated machine does not halt with the given input.12 An example of the first scheme can be found in Siegelmann and Sontag (1994) (see Copeland 2002 for a brief description). A specially constructed neural network in which some of the connection weights are uncomputable real numbers can itself compute Turing uncomputable functions. Perhaps that’s no surprise. In an analogue of the famous GIGO law of computer programming, we might not be surprised to find ‘noncomputable numbers in, noncomputable numbers out’ (see Davis 2006 for this point and a blistering attack on the whole idea of hypercomputation). Are any physical implementations of such hypercomputational systems really possible? A fundamental difficulty intrudes. The schemas of hypercomputational machines rely upon continuous physics being the correct description of the world; they are developed within that framework (there is no reason to expect that uncomputable numbers will play a role in a nature which is not continuous). There is of course nothing wrong with that, but it does mean that they do not tell us much about the real possibility of such devices, unless and until we know that nature really is correctly characterized by continuous physics. Since the UCA hypothesis explicitly denies this, the use of continuous physics to generate the plans for hypercomputational devices comes perilously close to begging the question against the UCA hypothesis. And perhaps it is not much of a surprise that the mathematics of the continuous physics we have developed permits uncomputability to appear when sufficiently prodded by human ingenuity. Pour-El and Richards (1981) have shown that certain equations familiar in physics can have uncomputable solutions even if the initial condition of a system obeying the equation is Turing computable. But as Penrose has pointed out (Penrose 1989, p. 188), such results do not seem to describe any real physical system; this ‘pathological’ mathematics does not seem applicable to the real world. Is that an illusion? Is it the result of our own bias in favor of the computable? Warren Smith (2006) provides another interesting example of hypercomputation. Smith shows that the trajectory of a set of point massive particles acting in a Newtonian world is not always computable which fact could, in principle, be deployed to solve the halting problem. Of course, we know that our world is not Newtonian and there are many bizarre features that arise in Newtonian models of physical systems, frequently dependent on the presumed continuous nature of space and time, the assumption of complete determinacy of object properties and the allowance of point masses within the models (for some examples see Earman 1986 and Earman and Norton 1993, §3).

5.3 Hypercomputation

73

In Smith’s model, the possibility of point mass particles approaching each other infinitely closely (or exploiting infinite energy) generates the uncomputability via a ‘singularity’ in which particles attain infinite velocity (and thereby travel an infinite distance) in a finite time.13 Smith makes the interesting general point that these Newtonian models necessarily will outrun the Turing limit. If positions can take on real number values with infinite precision (as certainly befits a point particle in a genuinely continuous space) then the number of trajectories particles can track is uncountably infinite whereas only a countable number of trajectories are computable by Turing machines (this is, I suppose, another case of our ‘law’: uncomputable in, uncomputable out). Smith’s results go further since the singularities can arise for Turing computable initial data. In purely abstract terms, we can think of all possible physical systems and the mathematical functions which describe them and their evolution. The set of computable functions forms a vanishingly small subset of the set of all possible functions, ‘almost all’ of which are uncomputable. We might then expect that almost any system we picked would be correctly describable by an uncomputable function rather than a computable one (it might well be approximately describable or describable within some midrange of values by some computable function of course). Is that why physics is never completed, but must renew itself periodically as the deficiencies in our computable ‘approximation’ to ultimately uncomputable nature becomes apparent? The appearance of the widespread computability of nature is then no more than a reflection of our mathematical predilections and limitations. Such reflections may increase our doubts about the conception of the universe as ultimately some kind of cellular automaton. But this is not perfectly clear. If we allow nature to embody uncomputability into its very fabric in some thought experimentally manipulable way then it is easy to make cellular automata that cannot be simulated by any Turing machine.14 Imagine a cellular automaton whose update rule is this: update cell (n, m) with the standard Liferules if the halting function H (n, m) = 1, otherwise update cell (n, m) with X-rules (where these are just some other well defined set of updating rules). The evolution of this CA is as well defined as the halting function, that is to say, perfectly well defined. But it cannot be simulated by any Turing machine. One can also imagine a kind of cellular automaton in which the update rate varies from place to place in the grid, so that, for example, one could have arbitrarily large regions that acted essentially like an accelerated Turing machine. This region could ‘pass on’ information it has computed to the rest of the CA. We could call such cellular automata hyper-CAs. I cannot define very precisely this concept in the absence of complete catalog of all possible extensions of the basic CA idea. The crucial feature though is the maintenance of the core CA concept. It is an interesting question whether for each proposed hypercomputational device there is a computationally equivalent hyper-CA. For some of the proposals this is obviously the case. The way the accelerated Turing machine could be used to compute the halting function could evidently be emulated by the kind of regionally divided, variable clock rate CA just discussed.

74

5 Emergence and Cellular Automata

The issue is interesting because hyper-CAs share their conceptual simplicity with standard CAs. If we knew that hyper-CAs could incorporate the kinds of hypercomputation we have been discussing as a potential real world phenomenon, and if we also accept that the retrieval of standard physical theory from the theory of cellular automata is a viable project, then we would know that any issue of emergence that might arise in the world of continuous physics could be discussed in the conceptually simpler world of the CA. It is also possible that hypercomputation is an illusion, fostered by the kinds of continuous mathematics within which we have developed our fundamental theories. Certainly, no one has provided any convincing evidence (or so much as a hint actually) of real world hypercomputational processes at work. Furthermore, to the extent that the appeal to hypercomputationalism traces Turing machine transcendent abilities to properties described by fundamental physics, the lessons we learn about emergence from cellular automata will apply to a hypercomputational world as well.

5.4 Conservative and Radical Emergence So let’s look at the topic of emergence from the vantage point provided by cellular automata. Clearly we have versions of both synchronic and diachronic emergence, as discussed above for the particular case of the Life CA. It is equally obvious that the emergent feature of any CA are completely determined, both with regard to their existence (or coming into existence) and their properties by the bottom level structure of the CA’s grid (its pattern of ‘live’ and ‘dead’ cells plus the updating rules). The general concept of determination at issue here is called by philosophers ‘supervenience’. In Chap. 7 we will look at this much more closely (for a recent compendious guide to supervenience see McLaughlin and Bennett 2008), but a simple preliminary characterization of supervenience is simply that domain A supervenes upon domain B just in case there can be no change in A without a change in B. A quick test of your intuitions about any particular case uses what I call the Dali-test. Salvador Dali is reputed to have once said ‘The only difference between me and a madman is that I’m not mad’. If this strikes you as impossible, albeit somewhat pregnant with meaning, then you believe that the property of being mad supervenes on some other features. Contrast with this case: the only difference between an electron and a positron is that the positron has positive charge. No problem there; charge does not supervene on the other properties of the electron. A trivial example is the supervenience of chess positions upon the arrangement of the pieces. A position cannot be changed from a checkmate to a non-checkmate without altering the position of at least one piece on the board. Rather more interesting and indeed a somewhat philosophically controversial example is that of an object’s aesthetic qualities, which seem to supervene upon the physical structure of whatever work of art is at issue. Could it make sense to say that the only difference between painting A and B is that B is ugly? Arguably, you could not turn the Mona Lisa into an ugly painting without making some physical change to it. Or could you? If you hold that beauty is, as such, essentially some kind of social or cultural phenomenon

5.4 Conservative and Radical Emergence

75

then you might well believe that the Mona Lisa could become (or could have been) an ugly painting via changes in the surrounding society. Very well then—you do not believe in the supervenience of beauty upon the physical form of the object in question. This sort of example is not really very strange. After all, one can become an uncle or a widow without there being any change in oneself. A currently much more contentious example in which this debate rages would be the supervenience of the mental upon the physical, which would assert that no change could be made in the mental states of some subject without a concomitant physical change in that subject. It is important to bear in mind the larger metaphysical picture here. A supervenience thesis always comes with a specified foundational or ‘subvenient’ domain. The ultimate subvenient domain is the microphysical state of the universe and the core issue of emergence is how to explicate the relation between the microphysical details and the macrophysical emergent features. The cellular automata model is very helpful in getting clear about what is at stake here. It is evident that within the system of cellular automata there is supervenience of the emergent features upon the basic structure or properties of the CA. For any emergent feature you care to name, say a glider gun in Conway’s Life, it is impossible to alter its properties without making some change in either the configuration of cells which make up the gun, or—rather more drastically—altering the rules of the CA itself. There is no case of any emergent features in a CA whose behavior can be changed, or whose interactions with other emergent features can be changed, except by alteration of the cell configurations which underlie these emergent features and their interactions (or by alteration of the updating rules of the CA). But it is easy to imagine CA-like systems that do not obey the principle of supervenience of emergent features upon the underlying structure and process of the automaton. Consider the game of Life with one extra rule: whenever a glider forms and survives for, let us say, twenty iterations in order to ‘mature’, it attains invulnerability and cannot be destroyed. Rather, any living cell it touches dies with no effect on the glider. So some gliders are, once formed, partially immune to the usual rules of the Life world and have a new power, that of destroying any living cell which they touch. Gliders continue to obey the Life rules internally, so to speak, so that they propagate across the grid normally. This new Life-like cellular automaton is perfectly well defined, and could easily be simulated in a computer.15 It is important to bear in mind that the simulation would be only that, a simulation of this new and bizarre kind of quasi-Life world. The computer code of the simulation or its operation would not exhibit the kind of temporal emergence I’m trying to illustrate here, for the simulation would have to work via some kind of time-counter that kept information in a local record of how long a glider had survived. Inside the world being simulated, rather than the simulation, it is the irreducible temporal fact about how long a glider has survived that governs how it will behave. Note that this shows we can simulate this sort of radical emergence in a system that does not possess any such emergence itself. Of course the proper understanding of this new quasi-Life world would require modification of the rules of Life. The modification would make reference to a certain

76

5 Emergence and Cellular Automata

Fig. 5.5 How will this evolve?

configuration and the time the configuration persists, and such reference would be ineliminable. This extra rule fails to obey two important features of a ‘good’ cellular automaton, which we may call locality and ‘instantaneousness’. The first demands that a cell evolves only according to the state of its neighbor cells (though remember that neighborhood is abstractly defined and is not necessarily a literally spatial notion). The second demands that the history of the automaton is irrelevant to its further evolution, or, to put it another way, the state at time t+1 supervenes upon the state at t.16 My new rules make the state of a cell depend upon non-neighbors if these are part of a glider shaped configuration, but only if that configuration has already persisted through sufficient iterations. So I could present you with a configuration and it would be impossible for you to tell how it would evolve. For example, the one illustrated in Fig. 5.5. It crucially matters how long the glider approaching the block at the southeast has persisted. As a matter of stipulated fact, it has not yet reached the critical lifetime of twenty iterations and will, alas, dissolve into nothingness, taking the block with it. Figure 5.6, a very simply modified glider gun, represents the initial condition from which this glider emerged. If you observed the evolution of this pattern, you would find nothing out of the way; it would quickly (after 230 time steps) lapse into a stable configuration, almost unchanging save for three ‘bar oscillators’ (a column or row of three cells will endlessly oscillate between its vertical and horizontal form). However, if we simply moved the southeast-most block of four live cells a little further to the southeast, then the glider produced by the evolution of this initial configuration will have persisted for more than 20 iterations by the time it encounters the newly placed block. We would then observe—to our astonishment—that the glider does not suicidally destroy itself and the block but rather plows right through it (leaving nothing behind). If we suppose ourselves to be scientists trying to infer the underlying laws of the Life world based upon the observed events, we would face a difficult problem here. We might be forced

5.4 Conservative and Radical Emergence

77

Fig. 5.6 Glider gun

to give up the intuitive and beautiful twin ideas that everything that happens at t+1 is the result of the configuration at t plus purely local rules of interaction. We might be forced to say that there is something special about gliders as such; they are not just another configuration of cells held together, as it were, only by the underlying laws of Life. One way to highlight the extent of the change envisioned here is to note that the standard Life world beautifully instantiates a very strong form of mechanism, a useful definition of which was provided by Charlie Dunbar Broad: Thus the essence of Pure Mechanism is: (1) a single kind of stuff, all of whose parts are exactly alike except for differences of position and motion; (2) a single fundamental kind of change, viz, change of position. Imposed on this there may of course be changes of a higher order, e.g. changes of velocity, of acceleration, and so on; (3) a single elementary causal law, according to which particles influence each other by pairs; and (4) a single and simple principle of composition, according to which the behaviour of any aggregate of particles, or the influence of any one aggregate on any other, follows in a uniform way from the mutual influences of the constituent particles taken by pairs (Broad 1925, pp. 44–45).

It is remarkable that digital physics embodies an even more stringent form of mechanism than Broad proposed. There is no change of position or velocity in the Life world—these are emergent features that arise from the ‘blinking’ on and off of the automaton’s cells. There are only two states of the fundamental constituents (alive or dead as we call them). There is no pairwise ‘influence’ over distance; the state of any cell is a trivial function of the state of its neighbors. And yet there is an obvious kinship between Broad’s stringent vision of mechanism and digital physics which is most evident in the discussion of emergence. Now, ordinary gliders count as emergent phenomena by our lights but doesn’t the newly defined kind of ‘super-glider’ represent a sort of emergence which deserves separate recognition? Traditionally, emergentists would not have been impressed with the weak notion of emergence we have thus far developed. Emergentists such as Conwy Lloyd Morgan, Samuel Alexander, John Stuart Mill, George Lewes and C. D. Broad wanted emergentism to be a more radical doctrine.17 Roughly speaking, these emergentists would have regarded our super-gliders as genuinely emergent

78

5 Emergence and Cellular Automata

phenomena but standard gliders would not have been worthy of the term. The reason is twofold. First, an ordinary glider’s behavior (and conditions of creation) is entirely dependent on the laws governing the underlying features of the Life world (i.e. individual cell life and death). Second, the powers of a glider, its ability to interact with other features of the Life world (emergent or otherwise) is similarly dependent on the basic laws of Life. It is clear that any system that met the conditions of Broad’s pure mechanism would not allow for radical emergence (of course, it was in part to make this point that Broad chose his definition of mechanism). It seems equally clear that the digital physics of the Life world is similarly bereft of radical emergence. Let us then officially baptize this evidently more serious and exciting, but possibly non-existent, form of emergence ‘radical emergence’ as opposed to the ‘conservative’ emergence that seems commonplace and uncontroversial. The core feature of radical emergence is that the emergents should make a difference to the way things happen, over and above, or in addition to and possibly (or occasionally) in a kind of interference with, the way things would happen without them. The appearance of radically emergent properties or entities is supposed to stem from laws of nature but these laws are laws of emergence which are not listed among nor consequences of the fundamental laws of the non-emergent base features of the world. While it is undeniable that everywhere we look we find emergence, the real question is whether there is any radical emergence in the world. Often, the classical emergentists defined radical emergence in terms of explicability. Only if it was impossible to explain the behavior and powers of an entity purely in terms of the basic laws of nature governing the underlying stuff of the world was that entity truly (or radically) emergent. But the emergentists had a very pure, ethereal and, as it were, non-epistemological sense of explanation in mind here. Normally, explanations are like stories about or accounts of phenomena that make them intelligible to human investigators. In that sense, explanation cannot outrun human intelligibility which, in all honesty, doesn’t really run that far or fast. It would be commonplace to say that it is impossible to explain, say, differential geometry to a four year old child to whom one could explain why it’s a bad idea to pull a cat’s tail. In this sense, explanation is relative to cognitive capacity—quantum mechanics is absolutely inexplicable to a goat. It is an interesting empirical question how far human cognitive abilities extend into the explanatory domain, and what particular features or capacities of our minds set our ultimate intellectual limits (at least one philosopher claims that the explanation of consciousness transcends our intellectual powers—see McGinn 1989). I tend to believe that the capacity-relative sense of explanation is its core meaning, but the classical emergentists seemed to think of explanation as something perfectly objective or as relative to any conceivable mind, of any finite level of intellectual ability. They were seeking to free the idea of explanation from its inherent relation to inquirers. That is really an incoherent goal. What they were really driving at was not an issue of explicability at all, but rather the idea of supervenience or the dependence (or lack thereof) of all emergents—their existence, properties and powers—upon the fundamental features of the world and the laws governing those features. In any event, as I intend it, radical emergence is

5.4 Conservative and Radical Emergence

79

an ontological doctrine about the ultimate nature of the world, not about the powers of minds to understand that world. At other times, emergentists stressed predictability or the lack of it. If it was possible to predict the behavior and powers of something entirely in terms of fundamental laws (and initial conditions) then that thing would not be radically emergent. As with the concept of explanation, predictability is relative to the capacities of the predictor. Modern astronomers can predict the positions of the planets to much higher precision than Ptolemaic astronomers, who in turn did much better than earlier stargazers. Our science still finds earthquakes quite unpredictable, but that may well change in a hundred years. But of course the classical emergentists were not concerned with the ephemeral details of current scientific predictive power. As with their appeal to explanation, they meant predictability in principle, given the correct account of fundamental science and unlimited intellectual resources. Once again, I think they were aiming beyond any cognitive activity; they were aiming at those features of the world that ‘drive’ the world from state to state or fix its dynamics. Talk of prediction is a way to vivify this rather abstract idea.

5.5 Simulation However, the issue of predictability raises an interesting and novel point in the context of cellular automata. In a certain, quite limited, sense the Life CA meets the test of intrinsic unpredictability (see Bedau 1997 for an interesting philosophical discussion of this feature). We have already noted that Life is Turing universal. Any Turing machine can be simulated by some Life configuration. Suppose there was an efficient way to predict what would happen in any Life world, where by ‘efficient’ is meant some method which shortcuts the direct simulation of the Life configuration. Then we could use Life to solve the Halting problem. For any Turing machine (with some definite input), simply emulate it as a Life configuration and use our presumed efficient prediction method to check whether or not it will halt. So, in general, there can be no way to predict what a Life configuration will do.18 Of course, in a great many particular cases, it will be easy to see how a configuration will evolve, but there can’t be a universal method for doing this. On the other hand, it is not difficult to simulate, at least in principle, the evolution of any finite Life configuration. Thus we can conclude that, in general, the only way to predict what a Life configuration will do is to simulate it. Now we come to the rather curious question: does simulation count as a kind of prediction? The obvious answer is: ‘yes, if the simulation runs faster than the system being simulated’. The somewhat less obvious answer is simply ‘yes’. In light of the attempt to spell out radical emergence as an ontological issue, I think the latter answer is the correct one. C.D. Broad liked to talk about a ‘mathematical archangel’ who could deduce how a system would behave (knowing the initial condition of the system) even if the calculations involved far transcended any human abilities (Broad 1925, p. 70). The modern study of complexity and computation has revealed just how powerful such an angel would have to be to get anywhere significant. We

80

5 Emergence and Cellular Automata

know, for example, that a great many conceptually simple and entirely finite problems are probably computationally intractable.19 Examples include such favorites as the ‘traveling salesman’ problem (i.e. find the shortest route that visits all cities in a region) and Boolean satisfiability (i.e. given a compound statement formed from ‘and’, ‘or’ and ‘not’ is there a way to assign true and false to the variables so as to make the whole statement come out true). Computational intractability arises when the number of operations required to solve a problem increases exponentially with the size of the problem. Lloyd (2002) has calculated that if the whole visible universe is regarded as a computer then it could have performed ‘only’ some 10120 elementary operations since the big bang. If we assume (implausibly of course) that we would need to perform at least one operation for each route in the traveling salesman problem, then 10120 operations would solve the problem for no more than about 80 cities.20 From the point of view of a universe presumed to have continuous quantities, both spatial and temporal, this is a woefully inadequate computational base; perhaps this is some sort of evidence in favour of digital physics (see Landauer 1991 and note 5 above). But I don’t think that the time needed for the simulation matters to the kind of abstract points about the nature of emergence that we are interested in. It just is the case that cellular automata are entirely predictable (simulatable) from an understanding of the rules governing individual cell life and death plus the initial configuration. Broad’s mathematical archangel, calculating at supernaturally high but finite speed would have no trouble figuring out how a Life configuration would evolve (which is not to say that the archangel could solve the halting problem). But notice that even the archangel, given only the basic Life rules, would necessarily fail to simulate accurately the evolution of one of our super-gliders. In order to succeed, the archangel would need to know the ‘laws of emergence’ which I arbitrarily imposed upon the Life world. Thus it is fair to say that our modified Life exhibits or at least is a model of a kind of radical emergence. Notice that the example reveals that, in a certain sense, radical emergence is a relative notion. Relative to the local rules of Life our modified game possesses radical emergence. If we add to the archangel’s stock of knowledge an awareness of the special historical and configurational law we have imposed then there is no radical emergence. The general question at issue is whether the world exemplifies radical emergence relative to the correct laws of fundamental physics. If digital physicists such as Fredkin or Wolfram are right about the ultimate basis of the world, then there is exactly as much radical emergence in our world as there is in the Life world, which is to say none at all. But of course, there is still emergence. And it is an extremely interesting fact about our world—even if the advocates of digital physics are right—that it is possible, profitable and even mandatory in any practical science (not to mention day to day living) to deploy theories and laws that operate at the level of the emergents. This fact is often marked by another distinction within the category of emergence: that between a metaphysical or ontological form and an explanatory or epistemological form of emergence. We can terminologically identify metaphysical or ontological emergence with what we are calling here radical emergence.

5.5 Simulation

81

The second form of emergence deserves some further remarks. Conservative or explanatory emergence is not the claim that emergents completely fail to be explicable in terms of the fundamental features of the world. Presumably, in the absence of radical emergence, there is an acceptable sense of explanation (explanation ‘in principle’) in which all emergents are thus explicable (and if they are not then we are just back to radical emergence).21 Rather, conservative emergence embodies the claim that there is an unavoidable need to use the non-radically, but nevertheless emergent features of the world in our science or, speaking even more broadly, simply in our understanding of the world. If there is conservative emergence then it is impossible for us to grapple with the complexity of the world without invoking a host of emergent entities and processes. I think something like this is the viewpoint forcefully advocated by Robert Batterman (Batterman 2002, see especially Chap. 822 ; see also McGivern and Rueger 2010). This can be and is cheerfully conceded by those who deny there is any radical or ontological emergence. They are perfectly happy to endorse the idea that reality is composed of ‘levels’ or a hierarchy of more or less independent systems of emergent entities. But this notion of independence is simply that within each level there is a set of laws over the emergents at that level which do a pretty fair job of describing, predicting and explaining how things happen at that level. The fact that we can successfully grapple with the complexity of the world only by appealing to these emergent levels does not falsify the claim that the entities and laws of each level remain completely supervenient upon the ultimate basis of the world, whatever it might be. It seems also to be true that conservative emergence can unify phenomena that do not possess any common microphysical structure or constitution. For example, the ubiquity of phase transitions across extremely diverse domains of phenomena can be explicated in terms of emergent features, as stressed by Batterman. Yet across these cases we have every reason to believe that criticality arises from the underlying entities and their properties, as evidenced by our success at simulation which invoke nothing but these underlying features (for one of innumerable examples, see Matsumoto et al. 2002).

5.6 Reductionism More than three decades ago, in a now classic paper, Philip Anderson provided a very interesting characterization of the difference between radical and conservative emergence (though not in those words) in ‘More is Different’ (Anderson 1972). It is worth remembering that this supposedly anti-reductionist article begins with this remark: The reductionist hypothesis may still be a topic for controversy among philosophers, but among the great majority of active scientists I think it is accepted without question. The workings of our minds and bodies, and of all the animate or inanimate matter of which we have any detailed knowledge, are assumed to be controlled by the same set of fundamental laws … (Anderson 1972, p. 393).

82

5 Emergence and Cellular Automata

This is what Anderson means by reductionism. We can gloss it in terms of the supervenience of all higher level features of the world upon the fundamental physical structures and processes. For Anderson, reductionism is an ontological doctrine which does not entail anything in particular about the explanatory relations amongst phenomena. In fact, the conflation of ontology and epistemology, which is all too common in philosophical discussions of reductionism, is the core problem examined in Anderson’s article.23 Anderson deploys two very old terms to explicate his view: analysis and synthesis. Ontological reduction he describes in terms of the universal analysis of the high level in terms of the physically fundamental. The first three chapters of this book were devoted to the case for analytic reduction as Anderson understands it. But, says Anderson, analysis does not imply synthesis. The fact that we have good evidence for analytic reduction does not entail the possibility of generating explanations of the high level which restrict themselves solely to the physically fundamental. Of course, in purely ontological terms, analytic reductionism does imply that the world itself is generating all phenomena entirely from physically fundamental phenomena. Anderson’s point is that this does not license belief that there are synthetic explanations of all high level phenomena. Synthesis can fail for a host of reasons. The most obvious (and ubiquitous) is the complexity of large systems, but we have also noted that the analytical reductionist viewpoint will miss unifying and universal features of the world which operate in abstraction, so to speak, from their fundamental constitution. Anderson goes so far as to say that ‘the more elementary particle physicists tell us about the nature of the fundamental laws, the less relevance they seem to have to the very real problems of the rest of science, much less to those of society’ (Anderson 1972, p. 393). He also makes some further remarks which, in light of the foregoing, seem quite obscure: the constructionist [i.e. synthetic] hypothesis breaks down when confronted with the twin difficulties of scale and complexity. The behavior of large and complex aggregates of elementary particles, it turns out, is not to be understood in terms of a simple extrapolation of the properties of a few particles. Instead, at each level of complexity entirely new properties appear, and the understanding of the new behaviors requires research which I think is as fundamental in its nature as any other (Anderson 1972, p. 393).

While the first half of this quotation is admirably clear, the idea that ‘new properties’ appear sounds more like radical emergence and thus can seem to be in some tension with his endorsement of ontological reductionism. I do not think Anderson is being inconsistent here. Rather, I think he means by talk of new properties, new models of complex behavior which deploy concepts new and distinct from those used in fundamental physics. Then it makes perfect sense to predict that the implications of these models, both in terms of their mathematical consequences and physical interpretation, are indeed the product of truly fundamental, creative scientific work. Anderson’s picture, at least as I see it, seems very attractive, but his remark about ‘new properties’ can stand proxy for a deep question. For the real issue is whether there is any radical emergence in our world. We have just seen that if the digital physics hypothesis is correct the answer is ‘no’. It would however be an understatement to say that this hypothesis is highly controversial. The thesis is very

5.6 Reductionism

83

speculative and really on the borderline between science and metaphysics. What I mean by the latter is that it is very hard to see how the claim that the world is fundamentally a kind of cellular automaton could be tested. Well actually, as things stand now, the testing is very easy since the digital physics thus far developed generates ‘grossly incorrect physics’ (Fredkin 2003, p. 209). In defense of the digital physicists, it is fair to say that the theory is simply too immature yet to face the tribunal of experience. Given the staggering success and accumulated amount of continuous physics over the last 400 years, digital physics is a spectacularly audacious hypothesis. Considerable time will be needed for it to be sufficiently developed, at first just to correspond to known physics and then to generate possible experimental tests. Fredkin notes that the digital physics hypothesis which he favors entails, in contradiction with the theory of relativity, that the events occurring within the universe do so against an absolute fixed reference frame. This is a consequence of the demand that properties such as energy and momentum be locally encoded: ‘information that represents the energy and the momentum of a particle is represented by digital information that is associated with the particle’ (Fredkin 2003, p. 209). Thus we might someday expect an experiment which could test for the existence of such a fixed frame of reference at a level of precision concordant with the fundamental length and duration stemming from the emergence of space and time. Of course, as things stand there is no hint of such evidence, but it is hard to prove a negative; the famous Michelson–Morley experiment was, despite its fame, hardly a decisive refutation of the ether. In general, we might look for a variety of what have been called lattice effects, which are divergences from the predictions about a system made on the basis of continuous mathematics caused by the underlying discrete nature of the system (for an interesting example in a biological context see Henson et al. 2001; for a speculative one from physics see note 7 above). But in the end, I suppose that it would not be unreasonable to reject the digital physics model and with that reject the straightforward deduction that the world contains no radical emergence. Perhaps the lessons of the Life world are just irrelevant to any debate about emergence in the real world. Let us then turn our attention to a possible line of argument in favour of radical emergence.

Chapter 6

Against Radical Emergence

6.1 Autonomy and Realization Where should we look for evidence of radical or, as it is also called, ontological emergence? It might be thought that classical physics is not a very promising starting point. Yet there is a natural link between emergence and complexity. There is no doubt whatsoever that complexity engenders conservative, or epistemological, emergence. There is also an argument that the classical but peculiar phenomenon of chaos in dynamical systems implies that radical or ontological emergence may be a genuine feature of our world. I have my doubts, but the subject is intrinsically interesting and cannot fail to shed light on the general topic of emergence. The argument has been advanced in some detail, though also with some caution, by Silberstein and McGeever (1999) which I propose to use as an initial guide. Silberstein and McGeever’s argument that dynamical systems arguably exhibit a kind of ontological emergence, rather than the uncontroversial epistemological emergence which is granted on all sides, and which I take for granted here, depends upon the idea that ontological emergence is the best explanation for the ‘striking multi-realizability (universality) exhibited by chaotic and other non-linear systems’ (Silberstein and McGeever 1999, p. 195). This conclusion is supposed to be reinforced via an appeal to a second feature of these dynamical systems, dynamical autonomy, which they explain as these systems’ ability ‘to maintain their dynamics not only across several physically dissimilar substrates, but also in spite of constant changes in the micro-states of the fundamental constituents that make them up’ (this idea is borrowed, as they note, from Wimsatt 1994). The argument seems to be that some explanation of both multiple realizability and autonomy is necessary, and that ontological emergence provides the best explanation. Discussing the particular case of psychology, they state that ‘if there is nothing physically common to the “realizations” of a given mental state, then there is no possibility of any uniform naturalistic explanation of why the states give rise to the same mental and physical outcome’ (p. 196). In general, the question is ‘why would so many disparate systems yield the same dynamics?’ (p. 196). Ontological emergence promises an answer since it is W. Seager, Natural Fabrications, The Frontiers Collection, DOI: 10.1007/978-3-642-29599-7_6, © Springer-Verlag Berlin Heidelberg 2012

85

86

6 Against Radical Emergence

possible that disparate realizations of some higher level feature will give rise to the same ontological emergents, thus providing the common element which accounts for the basic similarity of the physically diverse set of realizations. While this much ontological emergence might suffice to explain multiple realizability, it would not by itself be a sufficient explanation of dynamical autonomy. For that, as Silberstein and McGeever go on to argue, some ‘downward causation’ from the emergent to the constituents of the realization would have to be added to ‘discipline’ those constituents or constrain them to persist as a realization of the high level feature in question. I will argue to the contrary that, once some unclarity about the notions of realizability and autonomy are cleared up, there is no need to invoke emergence to account for this pair of properties. Indeed, there would be something almost incoherent about the attempt to use emergence for this purpose since multiple realizability and, especially, dynamical autonomy are precisely the best candidates for being the most significant emergent features of the systems at issue rather than features which could be explained by (other?) emergent features. Several rather large issues are raised by the concepts of realizability1 and autonomy. I think realizability should be primarily understood as a relation between a model, usually but by no means always a mathematical model, and some actual device which then realizes that model. The clearest cases, and the one from which the term arises I believe, are computational systems. As alluded to in Chap. 5, Turing (1936) famously devised a purely mathematical model of a generic computing device, which invoked a very small number of very simple operations (reading, writing and erasing a finite set of symbols onto an infinitely long tape), as well as a presumed system-structure of arbitrary complexity in which rules for carrying out sequences of these simple operations were encoded (described by the machines’ arbitrarily large but finite ‘state tables’). These now familiar ‘Turing machines’ are not real devices, but it was soon evident that real, physical machines could be made that almost perfectly mimic the computational capacities of these abstract devices (and in fact Turing was a pioneer in the transformation of his notional machines into real physical devices). Such mimicking is what is meant by the term ‘realization’. Every feature of the model has a corresponding, or ‘targeted’, feature in the realizing physical system— of course the converse does not hold—and for every possible state transition in the model there is a corresponding possible physical transition that preserves the mapping of model feature to corresponding targeted physical feature (thus the idea of realization is closer to the mathematical concept of a homomorphism than to that of isomorphism). To the extent that we can construct a given Turing machine in a number of different ways and using different materials, we have multiple realizability of Turing machines. There is no doubt that if we have realizability of such abstract devices at all, we have multiple realizability. But do we have realizability? Obviously, no real physical device will exactly realize any Turing machine, since the real device is subject to environmental influences and eventually inevitable internal breakdowns. In terms of the characterization of realization just given, this means that the mapping from the transitions in the model to possible physical transitions is imperfect—the physical

6.1 Autonomy and Realization

87

device has possible transitions which violate the mapping. For example, a trivial Turing machine could be devised that simply writes a google (10100 ) 1s on its tape and then stops. No physical device could realize this Turing machine since we can be sure that at some point long before this staggeringly large number of symbols was printed a breakdown would occur and the transition specified by the model would not occur in the realization. Modern day computers are similarly imperfect realizations of what can be described as Turing machines (though, in the interest of speed and efficiency, their architecture is vastly different and much more complex than Turing’s fundamentally simple scheme). So one question that arises is what counts as a realization of a model, given that perfect realization is impossible? One natural answer is that we should use the design of the realizing system to gauge whether it is a realization of a model or not. The design of a computer assumes that the physical devices out of which the realization is constructed will operate as specified, and so we might say that so long as the physical components do so operate, the device will realize the model. We could then define the realization relation in terms of this attempted design or the intentions of the designers. But now we have another level of realization to worry about. The construction of the physical components so that they will perform some abstractly specified function within the overall economy of the device raises exactly the same issue of realization that we face regarding the system as a whole. Nonetheless, I think there is a pretty clear sense in which we could talk of a physical device realizing a model if and so long as the device operates as designed. However, it is important to notice that we have transformed the realization relation from a relation between an abstract model and some physical device into a relation between two models: the original model now is put into correspondence with a model of some ‘properly functioning’ device. But even if this, to borrow the term of Daniel Dennett (1971), ‘design stance’ theory of realization is acceptable for artifacts, natural systems are not designed. For biological systems which have evolved by natural selection we might be tempted to invoke metaphorically a similar notion of design. We might, that is, imagine that Mother Nature had some abstract model in her mind, so to speak, against which we could compare the actual physical devices we find which imperfectly realize that model. But this is really to take the metaphor too far. At best, we can be permitted to regard Mother Nature as a tinkerer, with no clear idea of what she might come up with, although perhaps with some very general and vague goals always at the back of her mind. The problem is that to say that a biological system realizes some model imperfectly presupposes that there is some robust sense to the notion that there is the plan of nature against which we can measure the imperfection of the realizations. This seems rather perverse. For example, it is pointless to wonder whether the human eye is an imperfect realization of some ideal vision system, or a perfect realization of a vision system under a certain set of constraints, or a more or less imperfect realization of something in between these two extremes. Or again, is the heart an imperfect realization of a blood pumper because it is subject to certain diseases or is this weakness a matter of external influence that ought to be considered irrelevant to the quality of the realization? Real answers to such questions would

88

6 Against Radical Emergence

require that there be a specific model against which we could gauge the similarity of the ‘real world performance’ of the mechanisms which enable the biological functions of interest to us with the ‘desired performance’ as given by the model. We can devise many such models but it is absurd to think that nature somehow had one of these and no other ‘in mind’ during the evolution of these biological functions. While some loose characterization of ‘nature’s design’ is no doubt harmless, the specificity of design necessary to ground the notion of realization is simply absent in the case of biological systems. Thus it is incorrect to suppose that whatever has a function must be a realization of some model of a device which serves that function. Although this is of course how we normally produce devices that serve a function, to impose it as a requirement upon natural systems would threaten the already precarious grasp we have of the idea that natural systems genuinely have functions. In any case, Silberstein and McGeever profess to find multiple realizability at the level of dynamical systems in general, independent of whether or not they are literally designed or are the product of biological evolution. It is extremely unclear whether at this level of generality it makes much sense to speak of imperfect realizations in anything but the most tenuous metaphorical sense. Consider a mathematical model of the Earth’s weather system. Every day a wide variety of such models are being run on many computers, some being used for real time weather forecasting, others for long term climate modeling and still others for purely educational purposes. Even the best of these models are not particularly accurate representations of the atmospheric dynamics of the Earth and differ greatly in their internal structure. Is the atmosphere an imperfect realization of some such model? It was certainly not designed to realize any of these models, either intentionally or by evolution. Invoking the notion of realization seems explanatorily backwards in this sort of case. The models here come after the phenomena, and aim to represent certain especially significant features of the phenomena which can be more or less tractably modeled with current technology. To turn things around and claim that the weather realizes imperfectly these models misconstrues the epistemological relations here. It seems more reasonable to say that the weather system acts, in some respects and to a lesser or greater degree, rather like realizations of some atmospheric models. Why is there this match? That’s an intricate question. Part of the answer is that the models were consciously devised with much effort to make just that relation true. A perhaps deeper point is that whatever it is that the universe is doing in the vicinity of the Earth’s surface, certain aspects of it do present phenomena to us that look somewhat like the ‘virtual’ phenomena generated by certain of our models. How can nature do that? In the limit, this is just the question: how is science possible? Probably, its answer has more to do with the ingenuity of modelers than some exciting fact about nature, although of course the fact that nature is such as to permit the existence of modelers is of some local significance to us and, as emphasized in Part 1 we do find the world curiously generous in satisfying our epistemic desires. So the idea of realization in the case of natural systems—whether biological or not—simply comes down to the more or less extensive applicability of some model. This is a very much weaker notion of realization than we find in the field of computation and, in general, intentionally designed systems. If a model can be applied

6.1 Autonomy and Realization

89

to some phenomena with some success then we say that the phenomena realize that model, although always more or less imperfectly. But imperfect realization is now a fact about the applicability of the model, not a fact about the system being modeled (unlike in the case of systems which are really designed to realize some model). Somewhat less grandiosely, an examination of a simple model and some of its realizations might reveal the source of multiple realizability and give us some hints how nature manages to realize humanly devised models (to the extent she does and in the somewhat perverse sense just discussed). Here is a familiar mathematical equation which can be interpreted as a model of a very large number of physical systems: g d 2θ (6.1) = sin(θ ) dt 2 l This simple dynamical system is an oscillator; for example it is (the model of) an ideal, frictionless pendulum, where θ is the angle of the pendulum’s bob from the vertical, g is the acceleration of gravity and l is the length of the bob’s support. From this description of how the acceleration of the bob depends upon its position we can, more or less, mathematically deduce the famous law of the period of a pendulum:  t = 2π l/g

(6.2)

Notice that we already have a primitive but significant form of dynamical autonomy even in such a simple example: the period is independent of the amplitude of the pendulum’s swings (at least for small amplitudes). Thus a pendulum which is gradually losing energy, for whatever reason, will maintain the same period (a very useful property for clocks in a world full of friction), as will a pendulum that is given a little shove. If you forcefully push a pendulum and thus speed it up temporarily it will, of its own accord so to speak, return to its normal period as soon as you stop abusing it. But more important is the lesson for multiple realizability. Obviously, it is possible to make a pendulum-like system out of a virtually unlimited range of materials and in a rather large number of distinct configurations (for example, the bob can be made of any number of things and suspended by a rigid rod or a string, or a vine in a jungle). In fact, from a more abstract point of view, pendulums fall under the class of oscillators which are ubiquitous in modeling and seemingly in nature as well. However, the simplicity of the example shows us the source of multiple realizability in a clear form. The model works by providing a dynamical relation between a set of parameters within an implicitly specified context in which only these parameters affect the behaviour of the system. The model is very pure or ideal in this respect and we do not expect nature to isolate any system in this way, but the applicability of the model requires only that these parameters are so dominant as contributors to the behaviour of the system that the other unmodeled and perhaps unknown influences can be ignored, at least for the level of accuracy at which the modeler aims. In the case of the pendulum, the parameters are the length of the bob’s support and the acceleration of gravity. All that is required for a real physical system to realize a

90

6 Against Radical Emergence

pendulum—which, as we now see, means only that this model is applicable to it—is that it have components which correspond to these parameters which are ‘sufficiently’ isolated from the non-parametrized aspects of the environment. It is trivial to see how a string attached to a weight on Earth (with sufficient space to swing) provides the resources to mimic l and g and hence will act quite a bit like our model ideal pendulum. Now, if we ask why pendulums are multiply realizable, is there any reason to think that ontological emergence has any role to play in the answer? Is the length of the vine an ontological emergent? Is Tarzan’s weight? Hardly. What explains multiple realizability is the way the model sets up its parametrization which imposes no constraints beyond those of the model itself. Dynamical autonomy fares no better in the matter of requiring ontological emergence. This autonomy is a feature of (and at least in our example case deducible from) the model and any system containing features which can be mapped onto its parameters (including the implicit condition of those parameters being the sole—or at least dominant—influence on the behaviour of the system) must of course exhibit the autonomy found in the model. In fact, the multiple realizability and dynamical autonomy of this family of oscillators can be seen to arise from very abstract features of the model. As discussed in Batterman (2002), the model’s structure can be deduced from extremely general considerations which totally abstract from all the details of construction and material constitution. This follows from the highly plausible assumption that the particular units of measurement which we happen to employ are of no physical significance and thus ‘the physics should be invariant across a change of fundamental units of measurement’ (Batterman 2002, p. 15). From this assumption plus the free hypothesis that the relevant physical parameters are length, mass and acceleration a mathematical technique called dimensional analysis can generate, as if by magic, the mathematical model of the pendulum. The basic concepts of dimensional analysis have a long history, dating back at least to work of Jean Fourier, and are mathematically very sophisticated. Modern methods trace back to a theorem proved by Buckingham (1914; see also Strutt 1915). The essential idea is to reformulate a desired but unknown functional relationship of a number of ‘dimensioned’ variables, such as length (with conventionally labeled dimension L), mass (dimension M) or time (dimension T ), in a dimensionless form. The latter can be very roughly explicated by considering the trivial problem of determining how far an object, d, will fall in a certain time, t, subject to a certain gravitational force, g. Let us naively conjecture that what matters is t, g, and the mass, m, of the object (thus implicitly assuming that the object’s colour, or the day of week, as well as innumerable other possible factors are irrelevant). The dimension of d is L, that of g is L/T 2 (distance per second per second, that is, an acceleration) and of course m has dimension M. Can we write an expression that makes the exponents of our dimensions go to zero or ‘nondimensionalizes’ the problem? Well, this will evidently do the trick: U = t 2 gm/dm 2 ; U is called a dimensionless product. The fact that we had to put m in both the numerator and the denominator might suggest to us that it is really irrelevant to the problem so let’s eliminate it, to obtain: U = t 2 g/d which we can rewrite as d = κt 2 g, where κ is an unknown constant. We can now

6.1 Autonomy and Realization

91

determine the constant, which holds for all falling objects, by measuring any falling body’s progress (we know that it is in fact 1/2). Of course, by no means all interesting physical problems yield easily (or at all) to dimensional analysis. However, it succeeds here and in the slightly less trivial case of the pendulum and in many other problems, though most frequently in practical cases by reducing the number of variables rather than eliminating them. This clearly demonstrates that there is at least no need to postulate any form of ontological emergence to account for the autonomy or multiple realizability of the myriad of systems to which such models are applicable. The argument can be summed up as follows. Unlike in the definitional case of computational systems, realization of a model by a natural system amounts to nothing more than the applicability of such a model to the system. There is no ‘fact of the matter’ about whether a certain natural system is ‘genuinely realizing’ model A rather than model B (at least if both A and B are applicable to the system). Multiple realizability stems from the model rather than the world; it follows from the relatively trivial fact that many otherwise quite distinct natural systems can have their components mapped onto what I called the parameters of various models. If this mapping can be accomplished for any natural system then the system will necessarily behave more or less as the model predicts, and thus there will be multiple realizability. Similarly, dynamical autonomy is a feature of the model and so any natural system to which the model applies will automatically exhibit it. Silberstein and McGeever make much of the non-linearity of the models which are supposed to exhibit ontological emergence. So it is important to emphasize that non-linearity is not a necessary condition of either dynamical autonomy or multiple realizability. To take one interesting example, the general equations of hydrodynamics (e.g. the Navier–Stokes equations) are non-linear and of course can be used to produce many chaotic models. But a consequence of these equations is the particular, ideal case of ‘irrotational flow’. Irrotational flow is, roughly, a flow of fluid in which the ‘particles’ of the flowing liquid are not themselves rotating, or in which ‘small’ regions lack any overall rotation. The vortices formed by water swirling down a drain can be quite accurately modeled as irrotational flows (once they are in existence). The irrotationality of the flow should not be confused with the fact that there is of course rotation around the drain itself—irrotationality is, so to speak, a micro-property or local property of the flow (any closed path that does not include the drain itself will have no overall rotation or, as it is called, circulation). Conversely, there can be rotational flow in which there is no circulation. Viscous (that is, relative to the model under discussion, non-ideal) fluids flowing along a straight pipe will tend to ‘stick’ to the wall of the pipe, setting up a gradient of velocity and hence local rotation, which leads to usually unwelcome turbulence in the flow. A common textbook example may make irrotationality clearer. Imagine putting a very small (relative to the size of the vortex) floating paddlewheel into your bathtub as you drain the water. The paddlewheel as a whole will swirl around the drain but the wheel itself will not turn. If you were caught in Edgar Allen Poe’s maelstrom, and it was irrotational, you would not be spinning around as you slowly were sucked down into that terrible vortex.3 In general and in reality, vortices will not be

92

6 Against Radical Emergence

irrotational but, perhaps surprisingly, the vortices generated by both tornadoes and bathtub drains can be modeled quite accurately as irrotational flows. Such models are only approximately correct; the real world flows are not truly irrotational—the models have this property since viscosity is ignored (in terms of the discussion of models above, there is no parameter for viscosity so to the extent that a natural system is significantly affected by viscosity, to that extent the model under discussion here will fail to apply to it). In fact, if one ignores viscosity, there is no way to explain how vortices could ever form at all (without viscosity it would be impossible to stir your coffee—the spoon would just move through the water without setting up currents). Von Neumann disparagingly dubbed the subject of such ideal fluids, the study of ‘dry water’. But these models are nonetheless good enough to account for many features of commonly occurring vortices. What’s of interest to us here is that the equations governing irrotational flow are linear equations (see Feynman et al. 1963, vol. 2, pp. 40 ff.). The only explicit parameter of a model of irrotational flow is the velocity field, which must meet the ‘no local circulation’ condition but an implicit, and also ideal, assumption is that the flowing liquid is incompressible (pressure can thus be ignored). Multiple realizability follows just as it did in the case of the pendulum: any physical system with features corresponding to these parameters (including implicit ones) must act like the model. Many situations involving the flow of real world fluids come reasonably close to fulfilling this condition. Even flowing air will suffice in situations where external pressure changes are not significant (thus, perhaps surprisingly, irrotational flow models do a pretty good job of explaining why airplanes stay up in the air). The dynamical autonomy of irrotational vortices, which is quite striking inasmuch as such vortices will ‘strongly resist’ destruction and can be manipulated as ‘objects’ (you can, for example, make them wiggle and twist in predictable ways without destroying them), is also a consequence of the model and can, in brief, be traced back to the conservation of angular momentum. Thus irrotational models can model the ‘structure’ of tornadoes and bathtub drain vortices quite well. Such models naturally exhibit both multiple realizability and dynamical autonomy no less than their non-linear brethren. Since it is very doubtful that anyone (and evidently not Silberstein and McGeever) would want to claim ontological emergence for the high level features of models based upon linear dynamical equations, it does not seem possible to use either dynamical autonomy or multiple realizability, which are common emergents of both the linear and non-linear systems, as evidence for ontological emergence in the non-linear case. With the red herring of non-linearity out of the way, and given our initial discussion, what remains the basic line of argument deployed by Silberstein and McGeever? They claim that there must be some general answer to the question ‘how is multiple realizability/universality possible?’ (Silberstein and McGeever 1999, p. 196). Their answer, which favours ontological emergence, is that it is the appearance of high level, emergent, causal forces (or higher level ‘entities’ which generate these forces) that explains both multiple realizability and dynamical autonomy. It seems clear however, that both autonomy and multiple realizability are a simple consequence of the applicability of a model to a natural system. I take it that Silberstein’s and

6.1 Autonomy and Realization

93

McGeever’s argument goes thus: very disparate ‘submergent’ bases will generate the same emergents and then these emergents, being the same across the set of disparate submergent bases, will provide a natural explanation for at least multiple realizability. I don’t quite see how this explains dynamical autonomy unless the claim is that the emergent features are just identical to the dynamically autonomous or stable features. But then this does not seem to explain dynamical autonomy at all. It simply assumes autonomy at the level of the emergents. Why should emergent features have this kind of stability? Many things that are intuitively thought of as emergent are not very stable and quickly break up, and various intuitive emergents have their own characteristic ‘life-times’. Special or characteristic instabilities of emergent features ought to have an explanation no less than the stable features associated with dynamical autonomy. In any case, I agree that we do need an explanation of these features of emergent phenomena, but Silberstein and McGeever seem to be offering a definition of emergence in terms of dynamical autonomy and multiple realizability rather than an explanation. As I said above, multiple realizability and dynamical autonomy are the emergent features of the systems under consideration. Given an explanation of autonomy which only requires conservative emergence it is otiose, an unnecessary multiplication of entities, to posit an extra layer of radical emergence. We saw above that at least for simple examples, multiple realizability and autonomy are in the first instance properties of the models—the restricted and abstract mathematical description. The explanation of these features in real systems stems entirely from the applicability of the model to the system, which depends only upon the possibility of setting up the requisite mapping from the abstract description to aspects of the real systems. These explanations, to the extent that we can find them, depend upon how we ‘derive’ the emergent feature from the properties of its constituents in the model. Because the model will be, in general, much simpler than the real phenomenon at issue it will be relatively easy to understand how emergents emerge. To the extent the model applies to some real system, the explanation of emergence will carry over automatically. To continue with the example of the vortex, consider smoke rings (which are not instances of irrotational flow and in fact are very complex). The smoke ring itself is a ‘high level’ entity that presumably emerges out of the interactions of myriads of interacting molecules in the atmosphere. Such rings are highly multiply realizable and occur in a wide range of circumstances, and can form from a wide range of material. What explains the dynamical autonomy and multiple realizability of these beautiful structures? Not the high level features themselves I would maintain. What does explain it is the fluid mechanics of certain abstract models which are applicable to a wide variety of physical systems.4 Fluid mechanics, basically, is just Newtonian classical mechanics applied to continuous substances. It does include some basic parameters of its own, notably pressure and density, and it is a kind of field theory (with, in fact, deep analogies with classical electrodynamics) which naturally regards the properties of continuous substances in terms of continuous distributions of those substances’ properties. The equations of fluid mechanics can model very complex situations, in which there is non-linearity,

94

6 Against Radical Emergence

chaos etc. But it seems clear that a supervenience thesis appropriate for continuous substances holds in fluid mechanics (recall the discussion of supervenience in Chap. 5). Given the state of the field at each point (these are the analogues of the parts of a classical mechanical system) then the equations of fluid mechanics as expressed for the particular system at issue entirely determine the dynamics of the system. No explicit reference to high level features is necessary in principle. We have exactly the kind of predictability in principle that we have for the other deterministic models we have looked at. The stability of smoke rings is a pretty direct consequence of a theorem in ideal fluid mechanics—Kelvin’s theorem—which states that if angular momentum is conserved then the circulation of a velocity field is constant (as above, the circulation is kind of a measure of how much ‘local rotation’ there is in the velocity field of a flowing fluid). This seems to be a case where it is possible to give an explanation of multiple realizability and dynamical stability in terms of the underlying micro structure (in this case the structure of the velocity field). Of course, the explanation applies in the first instance to the model but, as noted above, it will carry over to those systems which realize the model. Another example worth looking at is Edward Lorenz’s famous non-linear but still conceptually fairly simple model of atmospheric convection (see Lorenz 1963; the canonical popular exposition of dynamical chaos is Gleick 1987 and for an excellent philosophical discussion see Smith 1998). This example raises issues beyond those of multiple realizability and dynamical autonomy even as it illustrates them.

6.2 Numerical Digression Before discussing Lorenz’s model directly, I want to digress briefly to consider predictability and dynamical systems more generally. As noted above, when we use the term ‘dynamical system’ we might be referring to a real world phenomenon such as, for example, a pendulum, the solar system or the weather. But we might instead be referring to the mathematical models (sets of interlocking differential equations for example) which may, to a greater or lesser extent, correspond to real world phenomena or may be purely mathematical constructions with no counterparts in the real world. It cannot be a mystery why those systems which do correspond to certain mathematical models share key features with those models. The mystery resides in the question of why there is a world which allows itself to be so modeled by mathematics simple enough for human beings to actually devise.5 The mathematical models themselves exhibit various interesting but purely mathematical properties, including sensitive dependence on initial conditions. It does not follow that nature exhibits these properties unless we assume that the models are in some sense ‘accurate’. This is a substantial assumption since we know the models are ultimately inaccurate (for example, they break down at certain length scales when applied to the real world). Furthermore, there should not be much trouble defining

6.2 Numerical Digression

95

the notion of determinism for these mathematical models themselves even though the concept of determinism tout court is complicated and difficult (see Earman 1986). First recall the notion of the phase space of a system, which is the space of possible states of the system expressed in terms of the basic parameters of the system. For a system of classical mechanics the parameters are the position and momentum of each particle. A one particle system moving in one dimension has a two dimensional phase space of position and momentum (six dimensions for motion in three spatial dimensions). The mathematical characterization of the modeled system is then essentially a constraint on the possible trajectories through phase space that the system can traverse. For example, the phase space trajectory of an ideal (frictionless) pendulum is a circle, so the allowable trajectories are such that p 2 + x 2 = C, where p is momentum, x position and C is some constant fixed by the initial condition of the system. Determinism can be simply defined for such models: a model is deterministic if and only if any point in the phase space of the system is in at most one trajectory (there might be loops—as in the case of the ideal frictionless pendulum, and many regions of phase space might be inaccessible to the system). In this sense, the Lorenz model is clearly deterministic: there are no ‘intersections’ of phase space trajectories from any set of initial conditions. This is simply a mathematical fact. What is amazing about such systems is that they never settle down into a looping trajectory. Predictability of such a deterministic model requires that it be possible in principle to derive a ‘final state’ (usually specified by fixing a parameter interpreted as time) from the equations of the model plus a given initial state. By ‘in principle here’ I mean predictable under ‘relaxed computational restraints’, by which I mean to allow the use of a (thought experimental) computer with arbitrarily large but finite memory and arbitrarily short, but non-zero, clock cycle time. If the defining equations of a model allow for an exact analytic solution then we ought to have predictability. For example, the simple differential equation: df = t + f (t) dt

(6.3)

happens to have an exact solution, namely: Cet −t −1, where C is a constant. So, once we are given an initial state of the model (which fixes the constant) we can directly calculate the final state for any time. We might call this absolute predictability. Many ideal mathematical models of real world systems have absolute predictability: the two-body problem (which applies very well to binary star systems for example), the frictionless pendulum (see Eq. 6.2 above, which applies pretty accurately to well constructed pendulum clocks), or, in general, a huge set of different kinds of oscillators and, at the other end of the scale of simplicity, models of black holes, of which the cosmologist Subramanyan Chandrasekhar made the stirring (if ultimately somewhat dubious) remark that ‘in my entire scientific life…the most shattering experience has been the realization that an exact solution of Einstein’s equations of general relativity, discovered by the New Zealand mathematician Roy Kerr, provides

96

6 Against Radical Emergence

the absolutely exact representation of untold numbers of massive black holes that populate the universe’ (Chandrasekhar 1987, p. 54). But many models of dynamical systems do not admit of any exact analytic solution and thus resist absolute predictability. Some of these are just hard to solve, but some— almost all complex systems of interest probably—are provably unsolvable (that is, it can be shown that the solution is not finitely expressible in terms of standard mathematical functions). Is there a way to predict the final state of the system, given an initial state, despite this? Just as in the case of the Life world, we can turn to simulation in place of absolute predictability. It is possible to approximate the evolution of the system from its initial state and mathematicians have devoted much ingenuity towards inventing and improving these numerical approximation methods over the last 350 years or so. In essence, these methods work by discretizing time, that is, by replacing the continuous time parameter with a ‘ticking clock’. Given that the rule by which the system evolves is clearly stated (as in a differential equation) it is then possible to calculate, more or less roughly, how the system evolves from discrete time step to discrete time step. Around 1769, Leonard Euler invented a simple method of numerical approximation which is not very efficient but can generate arbitrarily high levels of accuracy for a given system once we relax computational restraints. it is worth working through an example to see how this works. Consider the following ultra-simple ‘model’: df = t − f (t)2 dt

(6.4)

Despite its superficial similarity to Eq. 6.3, this is not easy to solve analytically. However, we have been given the rule by which infinitesimal changes to the state of the system cause it to evolve. Euler’s method just pretends that this rule is valid when applied to the finite jump from one discrete time step to the next. For example, let us suppose we know that the system begins in the initial state f (0) = 1. Let’s suppose we are interested in the state of the system at time 1. If so, the interval from 0 to 1 is divided into n subdivisions. For concreteness and simplicity, although certainly not accuracy, let’s use just three time steps. So we’ll have f (0) (which we know already), f (1/3), f (2/3) and then f (1). f (1/3) is found by applying the rule for changing the state of the system, that is, t − f (t)2 , to f (0), which gives us f

  1 1 = f (0) + (0 − f (0)2 ) 3 3

(6.5)

which—if I’ve done my arithmetic correctly—works out to be 2/3. The extra factor of 1/3 is equal to our step size and serves to scale the intermediate steps in line with our particular discretization of time—we are temporarily pretending that the function f is that of a straight line between f (0) and f (1/3) which introduces errors, but errors that shrink as we take more steps (this idea of treating the function as a straight line between time steps is why Euler’s method is called the ‘polygonal’ method).

6.2 Numerical Digression

97

Anyway, to continue with the example, we now know the value of f (1/3) and we can plug that in to the algorithm to obtain the value of the next time step:     2    1 1 1 1 2 = f + − f f 3 3 3 3 3

(6.6)

This works out to be 17/27. By repeating this process just one more time, we find that f (1) = 1574/2187, a rather ungainly fraction approximately equal to 0.719707. Probably there is no mistake in my arithmetic, but using a computer with a time step of 0.001 yields a value for f (1) of 0.833121. Decreasing the step size to 0.000001 gives f (1) = 0.833383. The original approximation is very poor. On the other hand, the difference in final value between a step size of 0.001 and 0.000001 is relatively modest. If we were to divide the interval between 0 and 1 into 10100 steps we’d end up very close to the true value (and under relaxed computational restraints this would by definition present no difficulty). It is fairly straightforward to see how this method (and the others presented below) can be generalized to apply to systems of interconnected differential equations. Similar methods have also been devised for functions of many variables and systems of partial differential equations (which are the more typical case when trying to model real world phenomena). In the world of real computers and metered electricity (to say nothing of the world of computation prior to even the existence of electronic computers), it is expensive to decrease the step size and this is the primary reason why mathematicians have introduced other methods (a daunting partial list: Taylor, Runge-Kutta, Euler-Cauchy, Heun, Verlet, Kutta-Nystrom, Curtis, Richardson, Merson, Scraton, England, Adams-Bashforth, Milne, Adams-Moulton, Hamming and Numerov). So far as I know these are all, fundamentally, refinements on the Euler method, which derive their mathematical justifications from the Taylor series representations of functions. And at least for the common refinements of Euler’s method it is easy to see that they do not improve the in principle accuracy (the accuracy attainable under relaxed computational constraints) of the resulting approximations. For example, Heun’s method (also known as the mid-point rule) modifies Euler’s method by using a two-pass calculation of the next value of the target function. The first pass, sometimes called the ‘predictor step’, uses Euler’s method to determine an approximate value for the target function, f , just as above. Calling this initial value P for predictor value we have P = f (t) + S f  ( f (t), t)) (here f  is the rule for changing the old value into the new value—the derivative of f which we are given in the definition of our model, and S is the step size). Using Euler’s method we would stop here, accept P as the value of f (t + S) and iterate it back into the procedure. Heun’s method instead uses P to compute the rate of change in f at the end-point of the path from f (t) to f (t + S), compares it to the value of the rate of change in f at the beginning of the interval and averages the two and then uses that average in Euler’s rule. That is, f (t + S) = f (t) + S/2( f  ( f (t), t) + f  (P, t + S)). The two values summed in the latter half of this equation are the putative slope of

98

6 Against Radical Emergence

f at the beginning and end of the step change and it is an evident improvement in the approximation to use the average of these two slopes rather than regarding f as a straight line (if f was a straight line then f  ( f (t), t) and f  (P + t) would of course be equal). It can be shown that Heun’s method is definitely better than Euler’s method, with an error of order S 3 as opposed to the S 2 of Euler’s method (since S is by definition less than one the higher the exponent in the error the less the error). In the real world of unrelaxed computational constraints, such an improvement is well worth the cost of the extra computation at each iteration of the method which Heun’s method entails. If, for example, we use Heun’s method to approximate our example function with a step size of 1/3 as above (actually 0.333333) we achieve f (1) = 0.854803; recall that Euler’s method returns 0.719707 for that step size and our most accurate calculation gave us 0.833121 using a step size more than 300,000 times finer than the original. Heun’s method thus provides a huge improvement. But it is clear from its construction that Heun’s method does not add any new accuracy in principle unavailable to Euler’s method (it is a straight tradeoff of number of computational steps versus step size). The very popular Runge-Kutta method (which dates back to 1901) is a generalization of Heun’s method which deploys more predictor values of the slope of the target function across the step change and adds an adjustable weighting of the contributions of these slope values. The method can also be used ‘interactively’ in which the stepsize is varied during the approximation to minimize error. Jos Thijssen goes so far as to say that ‘the Runge-Kutta method owes its popularity mostly to the possibility of using a variable discretisation step’ (Thijssen 1999, p. 474). He also makes a remark which is interesting with respect to predictability: ‘it is possible to implement the Runge-Kutta method with a prescribed maximum error, δ, to the resulting solution, by adapting the step size automatically’ (Thijssen 1999, p. 474). Heun’s method is a low-level version of the Runge-Kutta method with a particular choice of number of predictors and weightings. The most commonly used version of the method is the Runge-Kutta method of order four (the order number comes from the derivation of the method from the Taylor series expansion of the target function; Heun’s method is of order two) which iteratively calculates four predictor values of the slope of f . As the reader can imagine, this begins to get very complicated, but the payoff is that the error for Runge-Kutta of order four goes as S 5 , a rather large improvement over Heun’s method. To get a sense of how remarkably better the Runge-Kutta method is compared to the Euler method consider that with a step size of 1/3 the Runge-Kutta algorithm gives us f (1) = 0.833543 which differs only in the fourth decimal place from the value obtained using Euler with a step size of 0.000001! Nonetheless, it is once again clear that the intrinsic accuracy or the Euler method under the condition of relaxed computational restraints is exactly the same as that of the Runge-Kutta method. Of course, we have been looking at trivially simple examples. To move slightly in the direction of reality, consider that many physical systems are most directly described in terms of the forces acting on the elements of the system and this naturally gives rise to second-order differential equations and there are methods especially suited to these (though it is mathematically possible to transform such equations

6.2 Numerical Digression

99

into sets of first-order equations), such as Verlet’s or Numerov’s methods. Verlet’s method has a neat ‘leap-frog’ variation in which velocities are calculated at time steps exactly half-way between the time steps at which positions are calculated. Verlet’s method has another nice property, that of symplecticity (see below), which means that the error in the energy of the approximated system remains within certain bounds (that can be adjusted). For all such methods, as a practical matter, we want to use a step size which is compatible with the competing constraints of our computational resources and our desire for an accurate prediction of the future state of a dynamical system as represented by a mathematical model. In general, the bigger the step size the cheaper the computation but the less accurate the result. As we have just seen, clever methods let us get more accuracy out of a given step size as compared to less ingenious methods. It would be nice to know, or at least have some idea, of what an appropriate step size might be for any particular system. Here we begin to make our way back to the topic of chaos, for at this point it looks like a clever mathematical device known as Lyapunov exponents might help. Here’s a rough characterization drawn from Gençay and Dechert (1996),‘Lyapunov exponents measure the rate of divergence or convergence of two nearby initial points of a dynamical system. A positive Lyapunov exponent measures the average exponential divergence of two nearby trajectories, whereas a negative Lyapunov exponent measures exponential convergence of two nearby trajectories’ If we take two points which are possible ‘initial’ states of a dynamical system, and we assume that these points are ‘close to each other’ (a somewhat slippery notion) then the difference in final state of the system is typically much larger than the difference in the initial states. In fact, in a system that exhibits sensitive dependence on initial conditions, the difference grows exponentially as the system evolves from each of the two initial states. (One can see that this concept only makes sense for a deterministic system, as defined above, since a non-deterministic system will not have a well defined trajectory through phase space against which one could measure the accumulating error.) Using a very simple example, one can express how the Lyapunov exponent can help measure the exponential divergence thus: where f t (x) is the state the system is in at time t given initial state f i (x), then | f t (x) − f t (y)| (the absolute value of the difference between the two final states) is approximately equal to | f i (x)− f i (y)|etc . Here c is the Lyapunov exponent for the system in question. Knowing the Lyapunov exponent (or exponents, multidimensional systems have one for each dimension of their phase space) reveals a lot about how ‘chaotic’ the system is, helps us to discover the system’s fixed points or attractors, and tells us something about the qualitative behaviour of the system. Basically, a positive Lyapunov exponent means a system is unpredictable, in the sense that very small initial uncertainty about the state of the system will rapidly balloon into a very great uncertainty. But this, once more, does not suggest in any way that in principle predictability will fail. From the point of view of predictability, a Lyapunov exponent merely gives us some idea of how small a step size might be appropriate in order to achieve a given predictive accuracy for a specified time. Since under relaxed computational restraints we can pick any step

100

6 Against Radical Emergence

size we like we can always defeat the exponentially increasing error for the specified time, to the specified degree of accuracy.

6.3 Hitting the Emergence Wall So it seems reasonably clear that we do have predictability in principle, or predictability under relaxed computational constraints for dynamical systems.6 Let us then, finally, return to our discussion of the famous work of Edward Lorenz. Lorenz’s model can be expressed in just three interlocked differential equations that are worth displaying: d f (t) = 10(g(t) − h(t)) dt

(6.7)

dg(t) = 28 f (t) − g(t) − f (t)h(t) dt

(6.8)

dh(t) = f (t)h(t) −

8h(t) 3

(6.9)

The functions f , g and h represent properties of an ideal ‘convecting atmosphere’: f represents the speed of convection, g the temperature difference between distinct convection currents and h represents a profile of temperature difference measured vertically. Although this model is in some ways very simple, the three functions are interdependent. The system cannot be given an analytic mathematical solution and it gives rise to chaotic behaviour in the usual sense that very small differences in initial input conditions lead to very different final states after quite brief evolution times. It was in fact through the necessity of using numerical approximation techniques that Lorenz discovered that his system was chaotic (the difference in initial conditions engendered by round-off errors in his computer quickly led to entirely dissimilar evolutions). Subsequent investigation led to the discovery that this system is what we might call ‘predictable at the emergent level’. What I mean is that no matter what state the system begins in it quickly settles down to behaviour that is macroscopically similar though microscopically chaotic. The macroscopic structure of the model’s dynamics is immediately apparent from a typical diagram of the system’s evolution, the famous butterfly wings picture of the Lorenz strange attractor (see Fig. 6.1). The diagram is simply a plot of the values of the three functions at different times. No matter the initial state of the system (within reasonable limits) it will trace a path very similar to this diagram, and this gross behaviour is entirely predictable. This is certainly a kind of dynamical autonomy. On the other hand, every trajectory is entirely distinct, with not a single point in common with any other trajectory, and the system will switch seemingly at random from one ‘wing’ of the diagram to the

6.3 Hitting the Emergence Wall

101

Fig. 6.1 The Lorenz attractor

other. The appearance of intersection in the diagram is an artifact of compressing three dimensions into two and the need to draw the trajectory thick enough to be visible; in fact the trajectories of the model are infinitely thin and interwoven with an infinite complexity. This example is entirely in line with the discussion above. If we seek an explanation for the striking dynamical autonomy exhibited here, we need look no further than the model itself. It is a mathematical fact that this model will possess the twin strange attractors around which the system will, in its chaotic way, perpetually orbit. Therefore, any system to which this model is applicable will automatically share the model’s autonomy. And, the model will be applicable to any natural system which permits a correspondence relation between the parameters of the models (the interlocked functions, f, g and h in this case) and some features of the natural system (e.g. temperature, air current velocities, etc.) and which is, as implicitly assumed in the model, sufficiently isolated from other disturbing influences. Multiple realizability will similarly, and equally automatically, hold to the extent that there are many different natural systems for which the appropriate correspondence between system and model can be established. New issues arise when we think about this correspondence relation. Since Lorenz was intentionally attempting to investigate certain aspects of the Earth’s atmospheric dynamics, he devised a model whose parameters correspond somewhat (but only somewhat, for this is a crude model) to natural atmospheric variables, such as temperature and wind velocity. But these are themselves high level features relative to the atmosphere’s micro-constituents. As we have seen, this sort of emergence on emergence means that exactly the same questions about model applicability, multiple realizability and dynamical autonomy arise at the level of the parameters of a model such as Lorenz’s.

102

6 Against Radical Emergence

For example, one of the parameters used by Lorenz was temperature, the emergence of which is quite well understood. Temperature is modeled by the average kinetic energy of the micro-constituents of the atmosphere—that is, the molecules of nitrogen, oxygen, carbon dioxide which (along with a host of lesser constituents) make up the atmosphere of the Earth. These micro-constituents are themselves modeled by statistical mechanics, which treats them as individual particles obeying the mechanistic laws of classical physics. The way that temperature, pressure, density and other thermodynamical features emerge from the properties of the micro-constituents is a beautiful but familiar story which need not be repeated here.7 What I want to discuss is the way that a model such as Lorenz’s will inevitably break down if we take them ‘fully seriously’. What I mean can be best explained from the point of view of the predictability of systems such as Lorenz’s. These are by now classic examples of chaotic systems which exhibit sensitive dependence on initial conditions. Let us suppose that there was some natural system, the weather in some region on Earth as it might be, that we sought to model with a system like Lorenz’s (no doubt it would have to be much more complex than the Lorenz system with many more parameters and equations, but that does not matter here8 ). The point of the exercise is to predict weather with some reasonable degree of accuracy for some reasonable time into the future. We will not be satisfied if our predictions bear no relation to the actual weather, or no more accurate relation than folk weather lore can attain, or if satisfactory accuracy can only be obtained for a few days. The fact that weather models are chaotic does not preclude predictability in principle, at least in the sense of ‘predictability’ which includes simulation that we endorsed above. But in the unavoidable absence of relaxed computational constraints the cost, in terms of time and machine resources, of prediction increases exponentially as we seek more accuracy or want to look further into the future at a given level of accuracy. For the sake of the argument, however, suppose that we have an unlimited ability to gather the extra information about the initial or current state of the system which is needed for better and more accurate prediction and that we are operating under what I called above relaxed computational constraints. From the point of view of the model that is all that is needed—and all that could be needed—for unlimited accuracy of prediction out to a specified time. But from the point of view of the natural system under study, this is not at all true. If we seek ever more accurate predictions we must specify the initial conditions of the system to a finer and finer degree. Suppose, for example, that we need better measurements of the temperature throughout the region of interest (even the whole Earth, if global climate is our concern). It is evident that the conditions for the emergence of the relevant parameters, such as temperature, will ‘give out’ as we seek to measure the temperature in smaller and smaller volumes of the atmosphere. Since temperature emerges from the averaged activity of the molecules that make up the atmosphere, if we seek to measure temperature over volumes in which, on average, no molecules whatsoever are present we cannot expect to get any sensible value for the temperature. And yet there is a definite level of desired accuracy and duration of prediction for which the model would demand temperature data for those

6.3 Hitting the Emergence Wall

103

volumes. Since such values are simply unavailable the model no longer is applicable to its target natural system. In general, there will be many reasons why the conditions for the emergence of a model’s parameters break down. Fineness of spatial volume is a typical example, but there might be as well temporal or energy level breakdowns as well. In our example, the model would tend to break down for reasons such as the intrusion of molecular level forces which are invisible at the level of the model. (I should hasten to mention that this is not a practical worry. Current military global weather models have typical cell sizes on the order of a few kilometres, with variable vertical scales, and even there, of course, the data themselves which are assigned to these ‘cells’ are interpolated. It seems also likely that there is a finite size limit on how small a temperature sensing device could be and this introduces another limitation on the application of the model, which assumes that data are ‘free’ in the sense that data acquisition does not itself interfere with the modeled processes.) Let’s call the level of specificity which undercuts or eliminates the conditions of the emergence of the parameters of our model the ‘emergence wall’. If a problem requires data of greater specificity than the conditions of emergence permit then the model has ‘hit’ the emergence wall. It is an interesting theoretical question, and one for which I have no idea of the answer, what sort of theoretical accuracy in weather prediction we could attain— using current models and assuming that data sensors were as unobtrusive as we liked—before the model hit the emergence wall. Would it be months, years or centuries? This is the question of how quickly error tends to grow in our weather models (recall how this question can be expressed in terms of the value of Lyapunov exponents). There is no doubt that there is some time for which prediction would fail because of this problem. If a model is chaotic then the error in prediction grows exponentially with the uncertainty in the initial conditions and the desired length of prediction. That is, if the initial uncertainty is represented as x then the error in prediction will scale as x · eλt , where λ is the Lyapunov exponent for this model (or, somewhat more accurately, is the maximum Lyapunov exponent from the set of exponents which the model will generate—one for each spatial dimension; there are other ignored complexities here, such as the fact that the Lyapunov exponents will vary across different regions of the system’s phase space). Simply as an exercise, let’s suppose that the critical exponent, λ, is equal to one (which is perhaps not too unreasonable seeing as the corresponding exponent for the Lorenz system is 0.9), that the uncertainty of initial conditions in our current weather model can be expressed in terms of a length scale (the cell size discussed above) which we can rather arbitrarily set to, say, 3,000 m. Finally, suppose that we can already predict the weather with ‘acceptable’ accuracy for 4 days. Given this imaginary data, we can compute the maximum length of acceptable prediction before the model hits the emergence wall, where, again somewhat arbitrarily but conservatively, we set the latter value at roughly the size of a typical molecule, or about 3 × 10−10 metres (a cell size for which it is very obvious that the concepts of temperature, pressure etc. that apply in weather modeling have no meaning whatsoever). I have to admit that I am amazed at the answer. Within the confines

104

6 Against Radical Emergence

of this toy problem, even if we knew the state of the atmosphere at molecular length scales we could make predictions as accurate as our presumed current ones for only thirty extra days beyond the original four days.9 On the other hand, and somewhat more optimistically, this suggests that there is a lot of room for improvement in our standard methods of weather forecasting. Weather prediction is notoriously difficult and nobody expects any accuracy beyond just a few days at most, except of course at those places where the weather tends never to change but that reflects no virtue of the weather models. The domain which represents the opposite end of the accuracy scale is celestial mechanics, which many cultures have cultivated and which for thousands of years has made stunningly accurate predictions of events far into the future.10 It is thus interesting and somewhat disturbing that the planetary orbits are in fact chaotic (see Murray and Holman 2001) and, at least for the outer planets, are basically unpredictable on timescales of a few tens of millions of years. The problem of the stability of the Solar System goes back to Newton, who wondered whether God might have to step in every so often to nudge the planets back into their proper orbits.11 The worry seemed real as in the century after Newton the orbits of Jupiter and Saturn would not come into line with astronomical prediction. But this ‘great inequality’ was solved by Pierre Laplace who took into account the 5/2 orbital resonance between the two giant planets, an achievement that confirmed his determinism and helped lead to his famous anti-Newtonian remark that his system had no need for the hypothesis of God. This particular intellectual pendulum swung back with the work of Henri Poincaré on Solar System dynamics in the late nineteenth century which led to a nice statement of the phenomenon that came to be called chaos: ‘…it may happen that small differences in the initial conditions produce very great ones in the final phenomena. A small error in the former will produce an enormous error in the latter. Prediction becomes impossible…’ (Poincaré 1908/1960, p. 68). This is not to say that our Solar System is going to fall apart any time soon (if you can call a few tens of millions of years ‘soon’). Although the orbital position of the planets is chaotic the basic structure of the system is quite stable. For example, the time before Uranus is likely to be ejected from the Solar System under the influence of its much more massive neighbors, Saturn and Jupiter, is something like 1018 years (Murray and Holman 2001, p. 77812 ). Long, long before that the Sun itself will have burnt out. There is a point of some interest here. Although both the weather and the Solar System are chaotic dynamic systems, the timescale on which chaos reveals itself in the latter case is so long that we can preserve the illusion that the Solar System is easily predictable. The same would be true of the weather if we wished to use our models to make predictions for the next five minutes (though of course it will take longer than five minutes to get any ‘predictions’ out of our weather models). Predictability is a relative notion: relative to the timescales natural to human observers, relative the natural timescales of chaos of the systems in which we are interested and relative to the time it takes for our models to generate their predictions. It is important to emphasize that the location of the emergence wall is thus also relative to the model under consideration. A model of the solar system faces an

6.3 Hitting the Emergence Wall

105

emergence wall no less than a model of the weather, but the relationship between the order of accuracy and length of prediction is of course radically different for solar system modeling than it is for weather modeling. Nonetheless, because the solar system is in fact a chaotic dynamical system, the model of more or less spheroidal (the ‘more or less’ can be mapped out in some detail as well) masses in mutual gravitational interaction will give out if predictive demands are pushed far enough (though perhaps the model’s accuracy with respect to the evolution of the real solar system will never fail internally, as we are discussing, but rather more spectacularly in the destruction of the solar system itself by some unmodeled force or event).13 While my weather model example was entirely made up and has no connection to real weather forecasting models, its upshot is nonetheless relevant and sobering. The emergence wall is real. However, the issue only arises when we have a good deal of knowledge about the conditions under which whatever high level feature we are interested in emerges. In order to calculate, even very roughly and approximately, the domain in which the emergence wall threatens a model, we need to have a good idea of how the emergence of the model’s parameters comes about. Unfortunately, for a great many high-level features of the world we have very little knowledge about the conditions of their emergence. It is also necessary to have a reasonably good theory of the emergent features themselves, at least good enough to define initial conditions in ways that constrain the evolution of the system sufficiently to gauge how changes in the initial conditions affect the system’s development. Consider, for example, the issue which is really driving the emergence debate: the emergence of mental states, especially states of consciousness. Here, we are pretty much completely in the dark on both counts. Notwithstanding all the various kind of evidence we possess which links the mind to activities in the brain, a minute fraction of which was discussed above in Chaps. 3 and 4, we know very little about the conditions for the emergence of mental states. We have no theory of psychology which is powerful enough to allow for even rough specifications of initial conditions with enough ‘bite’ to yield predictions of behaviour which we could contrast with the behaviour of alternative, ‘nearby’ initial conditions. We can not apply an analysis anything like that of the weather models to the mind unless and until we have some theory of the relation of mental states to some underlying features for which we also possess some reasonably well worked out theory. It is easy, tempting and possibly correct to speculate that mental states emerge from the activities of whole neurons which stand in more or less particular, essentially discrete, network structures (as discussed in Chap. 3). If so, we might guess that the emergence wall arises somewhere near the length scales of individual neurons and temporal scales characteristic of these neurons’ network activity (that is, their firing rates or ‘spiking frequencies’). It is quite conceivable that we will someday have a theory which provides a good dynamical understanding of the state evolution of biological neural networks. Even in that neuroscientific utopia, we would further require a much better psychological theory than any we possess now. In general matters of the prediction of human behaviour, the best theory we have got is still our ancient standby, folk psychology, and for all its evident strengths it is very weak on

106

6 Against Radical Emergence

assessing the differences in behaviour that arise from assigning closely similar initial conditions to psychological systems. Say that Jill loves Jim ‘very much’ or just ‘a lot’—what’s the difference in behaviour predicted by these different initial states, even allowing that ‘all else is equal’? Perhaps a theory pitched at the emergent level of psychology but explicitly modeled on dynamical systems theory might offer some hope here. The so-called dynamical systems approach to cognitive science aims precisely for this (for examples of the approach in action see the various articles in Port and van Gelder 199514 ; a more abstract philosophical account of the idea can be found in Horgan and Tienson 1996). In such a theory, psychological states would be represented as, or determined by, interacting ‘cognitive’ forces driving the evolution of the system’s behaviour. If such a theory could be constructed and if it allowed for the more or less accurate measurement of these putative psychological forces, we might have a theory suitable for discussing the emergence of psychology from neural activity, and thus the further possibility of locating the emergence wall of psychology. Be that as it may, we can ask a more general question. What, if anything, does the existence of the emergence wall tell us about emergent features? The first consequence is that whatever emergent features we might be concerned with, the theory which deals with them is of limited applicability. There is, of course, a trivial sense in which this is true: particular theories of emergents only apply to systems which support the associated emergent features, and part of the point of invoking the notion of emergence is that such systems are more or less rare. But the case of dynamical systems such as we have been looking at reveals a more serious problem of limited applicability. At least some theories of emergent features will fail within their domain of applicability, or, it would be better to say, within what ought to be their domain of applicability. Following the toy analysis of the emergence wall for the weather model given above, that model would be unable to give, say, a fifty day prediction because of an intrinsic failure. Any attempt to give such a prediction requires us to leave the domain of emergence altogether leading to the total breakdown of the model’s applicability even though we ultimately seek information that is entirely within the purview of the model. The general condition for such a breakdown is some characteristic ‘scale’ of emergence. Emergent features necessarily emerge from activity that occurs on scales finer (in some not necessarily spatial sense) than the scale of emergence, and the detailed behaviour of the emergents themselves remains dependent upon that activity. If the relevant structure or processes at the fine scale fall apart or cease to occur then the emergent features themselves will also disappear. In the weather model, the scale was simply length but other scales are possible (time and energy suggest themselves immediately; abstract scales of complexity—if we knew how to measure it—provide more esoteric possibilities). Prior to drawing any philosophical conclusion about ontological emergence from the fact of the emergence wall, it is worth briefly examining the generality of the phenomenon. It is often said that Newtonian physics is a ‘low-energy’ approximation of relativistic physics. We can see that this comes very close to being equivalent, if it is not strictly equivalent, to saying that Newtonian phenomena (that is, objects,

6.3 Hitting the Emergence Wall

107

events and processes that to which Newtonian models are applicable) are emergent features, springing forth from some other domain. For example, Newtonian masses are constant; they do not change merely because the massive object happens to be moving. Thus a model of the acceleration of a Newtonian mass is quite simple: a constant force applied to a mass will accelerate it at a rate inversely proportional to its mass (Newton’s second law), and this rate will be constant. It follows that the velocity of such a mass will increase in direct proportion to the time the force is applied to it. Of course, relativity tells us that this picture is far from true when velocities become very large relative to the speed of light. We can locate the Newtonian emergence wall once we fix how accurate a characterization we seek as well as fixing the parameters of a certain problem (as in the weather example above). Let us say that we are modeling the time it takes to accelerate something to 100,000 km/s under an acceleration of 10 m/s2 . This is a trivial problem for the Newtonian model: the time we seek, t, is just v/a; in this case t is about 116 days. Although the calculation is not quite so trivial,15 the correct special relativistic model gives an answer of about 118 days. If it really mattered for some reason exactly when (or even to within a day or two) the object of interest reached the target speed (from the point of view of the observer) then our Newtonian model would have hit the emergence wall. A much less trivial example, and one with great practical import, is the corrections to the Newtonian viewpoint necessary for the gps satellite navigation system to work. A number of relativistic effects, derived from both the special and general theory of relativity, including motion induced time dilation, the effect of the gravitational field on local time and effects generated by the Earth’s rotation, have to be included in the gps models. Without them, gps positions would degrade at the rate of some ten kilometres per day and would thus quickly become useless (see Ashby 2003). The Newtonian mechanics of mass and acceleration which the gps system reveals as inadequate is only a tiny part of a much larger general problem of emergence: the appearance of a ‘classical world’ out of the non-classical physics—quantum mechanics as well as relativity—which underpins it. In general, we can say that the classical world emerges from the underlying quantum and relativistic structure of the universe. This is to say, because of the nature of the quantum world, the relativistic world plus the particular state of the universe in our general vicinity, classical models become applicable to a wide range of phenomena across a wide range of spatial, temporal and energetic scales. This problem connects to many of the core problems in the philosophy of physics. The measurement problem in quantum mechanics stems from the putative mismatch between the observed ‘classical’ world of definite objects and properties as opposed to the ‘smeared out’ world of state superpositions evidently predicted by quantum physics. In general, one of the constraints on any interpretation of quantum mechanics is that it somehow recover the sort of world which we observe ourselves to inhabit. The kind of chaotic dynamics discussed above is a hallmark of classical physics but there is some question of how anything like it can appear in a world that is fundamentally quantum mechanical in nature (see Ford 1989, Belot and Earman 1997).

108

6 Against Radical Emergence

While this is an extremely large, complex and by now highly technical issue, there are some elementary, well known and important links between the ‘correct’ physics and the classical world which emerges from it. The toy example above shows that for speeds much less than the speed of light, classical mechanics will be highly accurate at predicting accelerated motion. Of course, our dealings with the world, even in a scientific context, overwhelmingly involve such velocities. A more interesting link, and now one between quantum and classical physics, is provided by the so-called Ehrenfest equations. These state that the expectation values of the position and momentum of a quantum system will obey laws which are close analogues of the corresponding classical laws. In classical mechanics we know that momentum is just the product of mass and velocity, p = mv, which can be rewritten in the form of a relation between momentum and position as p=m

dx dt

(6.10)

The corresponding Ehrenfest equation is this:  p = m

d x dt

(6.11)

where x and  p are the expectation values of the position and momentum operators. Roughly speaking, these are the mean values you would expect to get if you measured position or momentum for a system in a given state many times (which is not to say that you would ever obtain exactly the value x in any particular measurement). Although it is interesting that quantum mechanics duplicates the classical equation, it means little unless we also consider the uncertainty in position and momentum against which the expectations values are defined, for we would hardly find ourselves in a classical world if objects with quite definite velocities did not seem to have any particular location! Everyone knows that the famous Heisenberg uncertainty principle puts limits on how well we can measure both position and momentum and it is generally agreed that this is a feature of the world rather than a mere epistemic limitation.16 Quantum uncertainty is relative to the mass of the objects involved. The uncertainty relation itself is simply this: x · p ≥ , where  (Planck’s constant, h divided by 2π ) has the incredibly tiny value of 1.054 ×10−34 (in units of energy×time). We could therefore know an object’s position to a precision of more than a million billionth (10−15 ) of a centimetre while retaining an even larger order of precision in the object’s momentum (a billion-billionth (10−18 ) of a centimetre-gram per second). Obviously, for objects in the ‘classical domain’, the relevant masses are so large that there is, for all practical purposes, no difference between the expectation value and the values we get for each individual measurement. The second Ehrenfest equation is the quantum mechanical analogue of Newton’s second law, F = ma, or d2x (6.12) F(x) = m 2 dt

6.3 Hitting the Emergence Wall

109

The quantum version once again simply replaces the definite classical values with expectation values, thus: d 2 x F(x) = m (6.13) dt 2 Once again, this tells us little unless we consider the uncertainties involved. It can be shown that for macroscopic objects the uncertainty is beneath consideration, which means in the first place that the following somewhat subtle identity holds: F(x) = F (x), and this means that the equation above falls exactly into line with the Newtonian version. Then we can deploy the same analysis as for the first Ehrenfest equation to show that for large objects the uncertainty in position will be so small that there will be no practical difference between the expectation value of a set of measurements and the values of the individual measurements. That may seem reassuring inasmuch as such a tiny uncertainty at ‘classical’ levels will presumably not lead to any untoward quantum intrusions into the classical world. In a certain sense, of course, this must be true since we do after all observe a pretty good simulacrum of a classical world, but the issue is in fact very complex (see Belot and Earman 1997, Belot 2000). As we have seen above, we can ask about the emergence wall in terms of possible prediction times. Normally, this involves asking how long the emergent level model will provide accurate (enough) prediction before our quest for accuracy forces us to descend to an explicit account of the (or a) submergent domain. We can also ask a kind of inverse question. Supposing we begin within the submergent domain, how long will our predictions remain accurate? This ought to have a trivial answer: forever. But it might not in particular cases, if there is, for example, a hierarchy of levels of emergence and we thus expect to hit another emergence wall as we attempt increasingly accurate predictions over longer time periods. Another more interesting possibility is that it might be the case that the known stability of the emergent domain will put constraints on the submergent account. How can this be? One radical answer would appeal to some kind of genuine, if rather magical, emergent ‘downward causation’. But one must be careful here to spell out clearly exactly what this view amounts to. Judging by the literature, it is apparently tempting and all too easy easy to slip into the posture of radical or ontological emergentism, and then to retreat under pressure to the idea that emergentism is just an epistemic necessity, and that there is no downward causation in the genuinely metaphysical sense of the term. One instructively unclear writer who at least courted this confusion was the famous neuropsychologist Roger Sperry who made the seemingly radical suggestion that the study of mind and consciousness ‘requires a shift to a new form of causality, a shift specifically from conventional microdeterminism to a new macromental determinism involving “top down’’ emergent control…’ (Sperry 1991, p. 222).17 But when we get to the proposed explanation of emergent control it seems less ontologically exciting than it sounds at first. In truth, it seems no more than the claim that the laws of microphysics have to be supplemented with some ‘initial conditions’ in order for any state sequence to be determined. Sperry’s favorite example is how the molecules

110

6 Against Radical Emergence

of a wheel are ‘controlled’ by its emergent circular shape. But either this just means we have to take into account the initial arrangement of molecules plus those in the environment or else it lapses into highly dubious claims about ‘the emergent physical properties and laws for the wheel as a whole’ (Sperry 1991, p. 225). Of course, we can accept the explanatory importance, perhaps indispensability of a sort deeper than the merely practical, of such macro descriptions, but this does not entail radical emergentism. The deflationary reading is encouraged by Sperry’s acceptance of the core claim of the non-radical emergentist that ‘emergent interactions are accomplished without disrupting the chains of causation among the subentities at their own lower levels, nor in their upward determination of the emergent properties’ (Sperry 1991, p. 225). At this point it is tempting to reduce away top down causation by a simple argument based on the transitivity of determination or causation. It may be worth noting a certain similarity between Sperry’s example and Hilary Putnam’s discussion of the square block and round hole (Putnam 1975). Putnam rightly points out that the intelligible explanation of why a the block won’t pass through the hole is not going to appeal to the fundamental physical state of the block but will operate at a more everyday level. That however does not undercut the claim of full determination of this effect by the micro-physical state. And Putnam is consistently clear that the issue is one of explanation. More prosaically, we have here simply a kind of test of the adequacy of the theory of the submergent domain. In the special case of retrieving classical behavior out of quantum mechanics we might ask the question in this way: how long would we expect a classical (emergent) system to retain its—at least approximate—classicality? We know that the answer is ‘indefinitely’ and if the quantum account cannot achieve this then it is at fault and not the classical world. Such derivations can thus provide a test of the theory of the submergent domain.

6.4 Quantum Emergence An especially interesting example of this sort of thing is the investigation of the ‘quantum stability’ of the Solar System undertaken by Wojciech Zurek (this work stems from a research program of Zurek and Paz developed in a series of papers; see for example Zurek 1991, Zurek 2002, Zurek and Paz 1994 and, for the particular case of the Solar system, Zurek 1998). In yet another instance of an emergence wall, the very small quantum level uncertainties in the position and momentum of macroscopic bodies will interact with the deterministic chaos of certain dynamical systems sufficiently to have effects within a reasonably short time (where ‘short’ is relative to the lifetime of the system at issue).18 While in classical models, the growth of uncertainty is merely epistemological and results only in an inability to know the future state of the system, the quantum model of the evolution of such uncertainties basically requires that the system go into a coherent superposition of all the states compatible with that degree of uncertainty. Chaos tells us that this uncertainty is going to grow exponentially quickly which

6.4 Quantum Emergence

111

ought to result in a kind of ‘smearing out’ of position over time. But, one might be forgiven for hoping, the ‘smearing’ will require quite a long time. Unfortunately, Zurek calculates that for the planetary bodies within our Solar System the time at which Newtonian behavior should break down—with the positions of the planets ‘smeared out’ to the size of the system itself such that it makes no sense to speak of their orbits or positions—is less than one billion years! Since the Solar System is more like four and a half billion years old, and the Earth at least seems to have retained a pretty definite position and orbit, there is evidently something wrong with the Zurek calculation. This is a very beautiful problem in the way it unites the two hallmark features of the classical and quantum worlds: deterministic chaos and nonepistemic uncertainty, in a deep paradox. In general, chaos plus quantum mechanics leads to the conclusion that the classical world is impossible. Zurek’s own solution is that the constant interaction between the Solar System and the galactic environment of dust and gas ‘force’ the system to behave classically. In effect, the environmental links prevent the system from entering or remaining for long in the superpositions of states that would be produced by the normal evolution of a quantum state, a phenomenon known in general as decoherence. This is a ubiquitous feature of quantum systems. They are in constant interaction with a huge and complex environment which acts, so to speak, as a continual observer. This manner of speaking suggests that the quantum wave function describing the system in question is continuously ‘trying’ to go into superposition states but the influence of the environment is perpetually forcing the collapse of the superposition into just one of its terms. There are indeed some theories that posit an actual dynamical process of wave function collapse (for an excellent overview of the Dynamical Reduction Program see Ghirardi 2008; Roger Penrose develops a distinct theory of wave function collapse in Penrose 1989, especially Chap. 8).These theories explicitly introduce new fundamental physics for which there is at the moment not a trace of empirical evidence. Why are they needed? Why not let ordinary processes of decoherence ‘collapse’ quantum wave functions in the standard way? One obvious problem with taking the easy way out is immediately apparent when we step back and regard the entire universe as our system of interest. If our theories are intended to provide the true account of the nature of reality, then there seems to be no objection to the idea that the entire universe is itself an entity that ought to fall within the purview of physical theory. But the entire universe does not interact with any environment and thus what we might call ‘total superpositions’ ought to remain uncollapsed. There is no reason to deny that such total superpositions will involve macroscopic, ostensibly classical objects. But then there is no way to escape going all the way to a view in which the universe is in some incalculably complex superposition of everything that could have happened since the beginning of time. This picture of the universe generally goes by the name of the ‘many worlds’ interpretation of quantum mechanics, first proposed by Hugh Everett (Everett 1957), according to which the universe contains infinitely many alternative histories corresponding to all events—or, better, all possible sequences of events—permitted by the laws of quantum theory and the initial quantum state of the universe. This does not seem to be progress. But there is an ongoing theo-

112

6 Against Radical Emergence

Fig. 6.2 Double slit experiment

retical framework in which it can, pretty much, be shown that the overwhelming majority of the worlds that make up the many worlds universe will appear classical (i.e. things will appear to have definite attributes not smeared out and weird superpositional properties). This approach to the emergence of the classical sometimes goes by the name of ‘decoherent histories’ and there is already a huge literature, both scientific and philosophical, exploring its prospects and consequences (Joos et al. 2003 provides a technical book length treatment; see also Tegmark 2007a and for a good philosophically oriented overview and discussion see Bacciagaluppi 2007). The basic idea is that within the quantum jumble, environmental decoherence will select certain states which will become effectively isolated from each other, in the precise sense that they will not ‘interfere’ with each other in the quantum mechanical meaning of interference. This comes close to the claim that they will act like classical states. Sequences of these states can be re-identified over time and generate the worlds—and the observers within them—which make up the total universe. The superpositions do not disappear. They become so to speak manageable, without overtly observable effects (save in special cases where quantum coherence is preserved, as for example in a physics laboratory for a short time). Let’s put this aside for a moment to lay out some basic features of quantum mechanics that will help clarify the claims made by the decoherent history approach about the emergence of the classical domain. Probably the best way to do this is to consider the famous double-slit experiment. Imagine a beam of particles (in principle it does not matter what kind or size) directed towards a barrier in which only two slits allow passage. The distance between the slits has been very carefully calibrated to the size of the impinging particles. At some distance behind the barrier is a detector plate that can register the particles that make it though the slits. The setup is schematically illustrated in Fig. 6.2. Quantum mechanics tells us that, assuming that no one (or no instrument) is looking at the slits, the particles go into a superposition state which is the sum of the

6.4 Quantum Emergence

113

two possible paths they can take towards the detector screen. We can write this as follows: 1 (6.14) ψ = √ (ψt + ψb ) 2 In this formula, ψt and ψb stand for the state of a particle which traverses the top slit and the bottom slit respectively. The total state is an equal sum of both possibilities (the root sign arises as a technicality—we have to square the state to get the actual probabilities of measurement). This is a state quite distinct from either of its components and most certainly does not merely represent our ignorance of which path any given particle might take. Now, suppose we are interested in knowing where our particle will be detected. The whole point of quantum mechanics is to provide an algorithm which allows us to transform descriptions of states such as provided in Eq. 6.14 into a number which represents the probability of any particular measurement we might wish to make on the particle. Of course, in this case we are interested in the position on the detector screen. Let us say we wish to know the chance the particle will end up in region R on the screen. The formula for this probability is written thus: ψ | PR ψ

(6.15)

If you think of ψ as like a vector, PR ψ is the projection of that vector onto a certain axis (which represents a certain measurement value of some observable property) and the procedure is like measuring the ‘length’ of the projection of ψ onto this axis. The details don’t really matter here; what matters is that this ‘inner product’ will generate the probability value we seek. But consider that we know that ψ is actually a somewhat complex state so Eq. 6.15 should really be written out, rather painfully, as:   1 1 (6.16) √ (ψt + ψb ) | PR √ (ψt + ψb ) 2 2 The√mathematical rules of the inner product allow us to factor out and multiply the 1/ 2 and combine the states rather as if we were just multiplying two sums to obtain: 1 [ψt | PR ψt  + ψb | PR ψb  + ψt | PR ψb  + ψb | PR ψt ] 2

(6.17)

Notice the last two terms in this monstrous formula. These are the interference terms. The first two terms represent the probability of the particle being detected in region R when the particle takes a ‘normal’ path through just one of the slits. If we consider them alone we get a nice distribution on the detector screen with peaks corresponding to the two slits, which is just what we would expect if the particles were indeed little BBs or marbles being shot through two slits. The extra interference terms however modify the probability and reveal the truth: the distribution of hits

114

6 Against Radical Emergence

on the screen is a pattern reminiscent of the way waves flow through and around a barrier, as illustrated at the right of Fig. 6.2. Although it is no great feat to perform the double slit experiment with real waves and we have all observed the interference patterns formed on water in a boat’s wake, it is only in recent times that it has been possible to do the experiment with electrons (very particle like entities but difficult to work with on a one-on-one basis). Nonetheless, a team of researchers at Hitachi managed to perform the electron double slit experiment and have produced a beautiful video of the gradual buildup of the interference pattern produced by a beam of electrons (see Tonomura 1989).19 It is important to note that the electrons were produced one by one and did not pass through the experiment in bunches. This is perpetually mind boggling. If the electrons are going through one by one, how do they know how many slits there are in the vicinity? If there’s only one slit there won’t be any interference at all. It is very weird that particles should behave like that—bbs and marbles certainly don’t. However, if one has a way to tell which slit the particles are traveling through then the interference patterns disappear and the particles behave more the way particles ought to behave. In terms of the quantum mechanical description, this requires that the interference terms somehow be suppressed or eliminated. How would observation do that? It is actually not hard to see how this works (given the occasional mathematical pronouncement from on high). So, suppose we have got a device which tells us which slit a particle passes through on the way to the detector. I am going to assume here that we have a perfect device which does not disturb the particle in any tangible way and with 100 % certainty delivers the truth about the path of the particle. Still, it must be the case that after the particle passes by, the detector changes state and since the particle in in a superposition then the detector will also be in a superposition. How should we represent the act of measurement? The basic idea is that the particle and the detector become correlated but this correlation is quantum mechanically special. Formally, we can represent a post measurement state, say the measurement of the particle as passing through the top slit, as ψt ⊗ Dt where Dt indicates that the detector registers the particle as passing through the top slit and ⊗ is mathematical entity called the ‘tensor product’ which we can simply consider as representing the joint state of the particle and detector. Seeing as the particle can go through either slit and emerges in a superposition of these two possibilities the system consisting of particle plus detector will also go into a superposition, the representation of which is analogous to Eq. (6.14): 1 ψ j = √ [(ψt ⊗ Dt ) + (ψb ⊗ Db )] 2

(6.18)

This total state,ψ j , is one in which the states of the particle and the detector are entangled.20 Entanglement is a fundamental feature of quantum mechanics but one that has some bizarre consequences which we will consider below. In this case, it seems quite reasonable that the superposition should encode joint states of particle and detector. At one level, it simply expresses how an (ideal) detector ought to work.

6.4 Quantum Emergence

115

But, the reader must be wondering, how does this bear on the question of why, when the detector is present, the interference pattern disappears? To understand this we need to think about how we can find out the probability of the particle being detected in a certain region of the screen if the initial state is ψ j . There is no way we can disentangle the detector from the particle, and the presence of the detector at least makes a clear mathematical difference. The algorithm for probability determination remains the same. To calculate the probability that the particle will ‘hit’ in region R we apply the same procedure as before. However, this time we need an operator which applies to the joint system of particle plus detector. Luckily, operators can be combined in pretty much the same way states can be. Also, we don’t want to make any measurement on the detector itself, we only want to know about the particle’s position. We can then replace PR with PR ⊗ I where I is the identity operator which is such that for any state I (ψ) = ψ. The probability we seek is then given by

ψ j | (PR ⊗ I )ψ j



(6.19)

Obviously, this is going to get very messy √ when written out in full, but, it will take the form of four distinct terms (the 1/ 2 will move to the outside of the whole expression as above and I’ll ignore it from now on). The first term will look like this (using mathematical properties of the inner product and the tensor product): ψt ⊗ Dt | (PR ⊗ I )(ψt ⊗ Dt ) = ψt ⊗ Dt | PR ψt ⊗ I Dt  = ψt ⊗ Dt | PR ψt ⊗ Dt  = ψt | PR ψt  × Dt | Dt  = ψt | PR ψt 

(6.20) (6.21) (6.22) (6.23)

The last line follows because the inner product of a state with itself is equal to 1 (thinking of the states as vectors the overlap of a vector with itself is perfect). So in this case the probability that the particle will be found in region R is not affected by the detector’s presence. But consider one of the cross terms: ψb ⊗ Db | (PR ⊗ I )(ψt ⊗ Dt ) = ψb ⊗ Db | PR ψt ⊗ I Dt  = ψb ⊗ Db | PR ψt ⊗ Dt  = ψb | PR ψt  × Db | Dt  =0

(6.24) (6.25) (6.26) (6.27)

The contribution of this cross term to the total probability has been eliminated. Why does this follow? Because the two detector states are ‘orthogonal’ which is to say they represent completely distinct outcomes. Using the vector image, they are at right angles to each other and have absolutely zero overlap. Of course, the other cross term would devolve to 0 in exactly the same way. The only terms that would make a non-zero contribution to the probability of the particle being found in region R are the two terms which represent the particle taking the ordinary path through one

116

6 Against Radical Emergence

of the two slits. That is, we get particle like behaviour with no interference pattern simply because of the presence of the detector.21 This is an example of decoherence—the initial (coherent) particle superposition state has been effectively transformed into a state in which the particles are acting ‘classically’. Of course, the more compendious system of particle plus detector remains in a superposition or, as we might say, quantum coherence has spread out a bit into the environment. To make the problem more vivid, imagine that the detector is a conscious observer (somehow able to perceive and track the particles in the experiment, but let that pass). According to this analysis, this observer ought to end up in a superposition of ‘seeing’ the particle pass through both of the slits, whatever that would mean. One can scale up thought experiments like these all the way to the infamous Schrödinger’s cat. However, the trick can be repeated. We can view the external environment of the detector as a kind of additional measurement device which will effectively transform the particle plus detector super-system into one that appears classical in the same way the presence of the detectors makes the particles ‘act classical’ from their— the detectors—point of view (the mathematics would be identical to the above save for even more unwieldy formulas).22 Ultimately, coherence is smeared out into the entire universe and, it is hoped, the tiny subsystems of the universe that constitute conscious observers such as ourselves will have experiences of a mostly classical world. This provides both the explanation of why the worlds in the many-worlds of the Everettian interpretation of quantum mechanics are of the sort we perceive as well as the principle for distinguishing these worlds from each other as separate ‘branches’ in the vast total state of the universe.23 We are now in a better position to investigate the most frequently made claim that quantum mechanics shows that some kind of radical emergence must exist in our world. The bizarre phenomenon of entanglement (and quantum mechanical superpositional states in general) is sometimes taken as conclusive evidence for some kind of ontological emergence (see Silberstein and McGeever 1999; Humphreys 1997b; Teller 198624 ). We have already seen examples of entanglement, in the coupling of measured system and detector. But entanglement is a very general feature of quantum mechanics that goes far beyond the context of measurement. It arises from the linear superposition principle, the fundamental feature of quantum mechanics that any two states of a system can be ‘added’ to form a new state. Given certain other core features of quantum mechanics, this leads to the following possibility. Consider a system which creates particles in pairs with some conserved property, such as spin, which can take either a positive or negative value (we can ignore the magnitude and just label these + and −). Assuming we start with zero total spin then the spins of the individual particles (call them A and B) must sum to zero to preserve the conservation of angular momentum. There are two ways that this can happen, namely, if A has spin + and B has spin −, or the reverse. There is no way to control the polarity of a particular particle’s spin during pair production, so when, for example, A is created it is in a superposition of + and − spin, and similarly for B. But since the total spin of the

6.4 Quantum Emergence

117

system has to be zero, if we measure A and find it has spin + then we immediately know that B has spin − (or at least will, if measured, give a guaranteed result of −). This state, known as the singlet state, can be expressed thus: 1 √ [(A+ ⊗ B − ) − (A− ⊗ B + )] 2

(6.28)

where as above the ⊗ symbol (tensor product) represents the ‘joint state’ of particle A and B.25 There is no way to decompose this complex superposition into a form in which the A states and B states are separated, hence the use of the term ‘entanglement’ to describe such states. Entanglement has many peculiarities. Notoriously, since measurement in effect forces a state into one of its components, it entails that no matter how far apart A and B are, upon measuring A to be + (−), it is instantaneously fixed that a measurement of B will give − (+). Also, it seems clear that quantum mechanics is endorsing, or perhaps revealing, some kind of holism about entangled states; they are not reducible to purely local states of the component particles. Furthermore, there is no way to devise any local properties of A and B which can ‘carry’ the observed correlations between possible spin measurements (though this is possible if the properties can exchange information instantaneously across any distance, a possibility that conflicts with our current understanding of the physical world). So entanglement is very weird. However, our question is about whether this phenomenon gives any support to the claim that modern science provides examples of radical emergence. The answer seems clearly to be no. The situation does not seem to be that different from the case of mass, which if we take specific mass properties, is actually quite interesting. For example, the specific mass property exemplified by a hydrogen atom is 1.007825037 amu. The mass of a proton is 1.00727638 amu and that of the electron is 0.000548579867 amu. The atomic mass is not quite the same as the sum of the masses of the components because the energy which binds the electron to the proton must be taken into account according to the relativistic principle of energy-mass equivalence. This latter is an empirical law which governs the causal interaction of the proton and the electron. The empirical principle involved could have been different. But, crucially, the principle is a physical law which operates over purely physical properties. No one would think that the mass of the water molecule is a radically emergent phenomenon. Just as the complexities of mass composition depend logically upon the laws governing mass and energy (which are basic physical laws), so too the complexities of entanglement are a logical consequence of the laws governing the basic properties and interactions of the quantum world. In some ways, the analogy is quite close. One would be severely misguided to think that one could simply add up the masses of the constituents of an object, considered independently, to compute the total mass of the object. The mass of an object is, in a sense, not reducible to the mass properties of the individual components. Their interaction has to be taken into account. But this does not show that an object’s mass is radically emergent from the mass of its constituents, because the interactions are governed by the fundamental laws at

118

6 Against Radical Emergence

the level of the constituents themselves. So too, the superposition principle and the formation of joint states are fundamental principles of basic quantum physics. The singlet state is a state fully described by the fundamental laws of quantum mechanics. The entangled state is a predictable result of the basic laws of the quantum systems and their interactions. In fact, it was predicted, by Schrödinger, who introduced the term ‘entanglement’, right at the birth of quantum mechanics (Schrödinger 1935). There is no hint of radical or ontological emergence here, although the oddity of entanglement is also emphasized by this analogy—unlike in the case of mass, there is apparently no (current ongoing) interaction between the entangled particles!

6.5 Fusion Emergence One of the most carefully worked out accounts of putative radical emergence, and one which again leverages entanglement, is that of Paul Humphreys (see e.g. Humphreys 1997a, b). In fact, Humphreys regards the formation of entangled states of quantum systems as paradigm examples of what he calls ‘fusion’ (see Humphreys 1997a). According to this theory, fusion is the appearance of new properties or, more accurately, instantiations of properties never before manifested in the world. Such fusions bear the general hallmark of emergence of course, but what suggests that they imply radical emergence is that fusions have new causal powers and that the precursor entities which generate fusions lose their independent identity or even cease to exist (hence it would be wrong to consider such precursors to be constituents of the fusion). Humphreys suggests that the canonical form of fusion creation can be expressed in terms of a temporal operation which transforms precursor property instances (which might also be thought of as events if events are taken to be instantiations of properties at particular times by particular objects) into new forms. More formally, if Pmi (xri )t1 is an instantiation of property Pm by xr at time t1 (and where the superscript letters indicate the ‘level’ in the natural hierarchy of emergence) then the fusion of two such events bring about fusion at time t2 is expressed as [Pmi ∗ Pni ][(xri ) + (xsi )](t2 ). In line with the idea that fusion involves novel instantiations and the loss of identity of the precursors, this can also be written as [Poi+1 ][xti+1 ](t2 ). The second form highlights that both the property and the object are new rather then mere composites. The idea of fusion is intriguing and evidently coherent. Is it real? There have been serious philosophical criticisms of the viability of the fusion approach to emergence, which question both whether it can deliver the advantages Humphreys claims (see Wong 2006) and whether fusion really is the best account of domains where it ought to apply, such as chemistry (see Manafu 2011). But leaving aside particular philosophical objections, it seems unlikely that fusions will embody radical emergence if entanglement is the paradigm case, simply because of the arguments given above. Something like fusion seems to be a real feature of nature, and one which is important to emphasize. However, the failure of part-whole reductionism or the fact that nature strays widely from obeying a purely compositional or additive metaphysics does not mean there is radical emergence.

6.5 Fusion Emergence

119

Instead, we find that the mechanism of fusion (or entanglement) is predicted by fundamental theory, which theory in fact requires the features which engender entanglement to successfully predict the activity of even the most basic entities. We also find that the causal powers of the fusions are all drawn from the same family as those that characterize lower level entities (such powers that derive from mass, energy, motion etc.). No fundamentally new observables are introduced through entanglement. No new laws of nature beyond those of fundamental theory are required to specify the states which result in fusion; these are instead strict consequences of the states of the precursor entities. Although it cannot be over stressed that nature is much stranger than the old mechanists would have or even could have dreamed, entanglement (and presumably fusion in general) is nonetheless an emergent feature of the world which does not rise to the level of radical emergence.

Chapter 7

Emergence and Supervenience

7.1 Completeness, Closure and Resolution The metaphysical relation of supervenience has seen most of its service in the fields of the philosophy of mind and ethics. Although not repaying all of the hopes some initially invested in it—the mind-body problem remains stubbornly unsolved, ethics and aesthetics not satisfactorily naturalized—the use of the notion of supervenience has certainly clarified the nature and the commitments of so-called non-reductive physicalism, especially with regard to the questions of whether explanations of supervenience relations are required and whether such explanations must amount to a kind of reduction (a good discussion of these issues can be found in Kim 2005). I think it is possible to enlist the notion of supervenience for a more purely metaphysical task which extends beyond the boundaries of ethics and philosophy of mind. This task is the clarification of the notions of emergence and emergentism, which latter doctrine is receiving again some close philosophical attention (see for example McLaughlin 1992, Kim 1993, Bedau and Humphreys 2008, O’Connor and Wong 2009, Clayton and Davies 2006, Corradini and O’Connor 2010, Macdonald and Macdonald 2010). I want to try to do this in a ‘semi-formal’ way which makes as clear as possible the relationships amongst various notions of supervenience as well as the relationship between supervenience and emergence. I especially want to investigate the impact on our ideas of supervenience of an explicit consideration of a very familiar but under explored notion which is crucial to our scientific understanding of the world: the temporal evolution of states. It will turn out that the impact is significant and extensive. I do not pretend that what follows is fully rigorous, but I do hope the semiformality makes its commitments and assumptions clear, and highlights the points of interest, some of which I think are quite surprising. For readers uncomfortable or impatient with such formalism, I provide explications of all the formulas which should render the discussion easy to follow. I want to begin with a series of definitions.

W. Seager, Natural Fabrications, The Frontiers Collection, DOI: 10.1007/978-3-642-29599-7_7, © Springer-Verlag Berlin Heidelberg 2012

121

122

7 Emergence and Supervenience

D1. A theory, T, is total if and only if it possesses completeness, closure and resolution. These are jointly defined as follows: Completeness is the doctrine that everything in the world is a T-entity or, in principle, has a non-trivial T-description and as such abides by closure and resolution. Closure entails that there are no ‘outside forces’—everything that happens, happens in accordance with fundamental T-laws so as to comply with resolution. Resolution requires that every process or object be resolvable into elementary constituents which are, by completeness, T-entities and whose abidance with T-laws governing these constituents leads to closure. For the particular example of physics (the only theory that could have any chance of becoming a total theory) these definitions become: Completeness is the doctrine that everything in the world is physical (has a non-trivial physical description1 ) and as such abides by closure and resolution. Closure entails that there are no ‘outside forces’—everything that happens, happens in accordance with fundamental physical laws so as to comply with resolution. Resolution requires that every process or object be resolvable into elementary constituents which are, by completeness, physical and whose abidance with physical laws governing these elementary constituents leads to closure. It may be worth reemphasizing here that this is not an endorsement of ‘part-whole reductionism’, though it is consistent with it. We know from quantum mechanics (see the discussion above in Chap. 6, pp. xx ff.) that the states of ‘wholes’ are not simply functions of the states of their parts considered in isolation but this does not tell against the characterization given in the text. Quantum mechanics is a celebration of how the fundamental interactions of things can be understood—rigorously understood—to yield new features. It is, if you like, a mathematically precise theory of emergence, but one that obeys the strictures of resolution. The kind of emergence endorsed by quantum mechanics is a form of conservative emergence. it is also worth noting that conservative emergence has no essential commitment to micro-fundamentalism: the view that all fundamental features of the world reside at the microscopic level. As will be discussed below, it is also perfectly possible to imagine ‘large’ objects which are fundamental or simple. An example is a black hole (at least as classically described) which is essentially a gigantic elementary particle with only three properties: mass, charge and angular momentum. We could distinguish a purely ‘formal’ notion of totality from that defined above. A ‘formally total’ theory is one that would be total if only it were true. It is arguable that the ‘mechanical world view’2 is, or at least was intended to be, formally total (but had no chance of being true of this world). Perhaps ‘classical physics’ (Newtonian mechanics + electromagnetism) was similarly supposed to be a formally total theory. Since I am going to assume that final-physics—whatever it may turn out to be—is true, the notions of formal totality and totality collapse for it. D2. T-possibility: something is T-possible if and only if it exists in some T-possible world, that is, some world that obeys the fundamental laws of theory T. (Example: physical possibility is existence in some physically possible world, that is, a world that obeys the fundamental laws of physics. To avoid making physical

7.1 Completeness, Closure and Resolution

123

possibility epistemically relative we can regard physics to be the true, final physics— whether or not humans ever manage to discover such a theory. We can call this ‘final-physical-possibility’.) D3. Efficacy: a state, F, of system σ is efficacious in producing a state, G, in system π if and only if had σ not been in state F, π would not have been in state G (it is possible, and usually the case, that σ = π).3 A simple example of state efficacy might be the way the mass of an object is efficacious in the energy produced when the object falls from a certain height; if the mass had been less the energy would have been different. By contrast, the color of the object is, presumably, not efficacious in the energy produced by falling. We can also say that F has efficacy if and only if there is a state for which F is efficacious in its production. It is useful to also have a notion of efficacy explicitly relativized to the states of particular theories, so: D4. T-Efficacy: a state, F, of system σ is T-efficacious in producing a T-state, G, of system π if and only if had σ not been in state F, π would not have been in state G (it is possible, and usually the case, that σ = π; the state F may or may not be a T-state). As in D3, F has T-efficacy if and only if there is a state for which F is T-efficacious in its production. As a serious theory of efficacy this has a number of problems. There is an ongoing and vigorous debate about whether supervening realms can be said to have genuine causal power (see the extensive literature that has sprung from Kim’s first articulation of the exclusion argument in Kim 1992). This will be of importance in later chapters. The current notion is intended to be very weak and undemanding. For example, according to D4 it is quite possible that moral properties are efficacious insofar as it is plausible that they supervene on natural properties in ways that permit cases of counterfactual dependency (but see footnote 9 below). Perhaps it would be better to label this ‘prima facie efficacy’, ‘ostensible efficacy’ or just leave it as simply counterfactual dependency but for the sake of simplicity of expression I will stick with the term ‘efficacy’.4

7.2 Supervenience The concept of supervenience is well known and can be rather swiftly characterized, although it is important to distinguish a number of variants (for an overview see McLaughlin and Bennett 2008; for the philosophical development of the notion see Kim 1993, Chaps. 5–9). Supervenience is a special relation of dependence, or at least correlation, of one domain upon another. It is often taken to be a relation between families of properties or states, where a ‘family’ of properties is a set of properties that define the domains at issue. For example, psychological properties form one family while physical properties form another. There is a relation of supervenience of the psychological properties upon the physical properties if, in accordance with the special relation of dependence defining supervenience, we hold that (all instances of) psychological properties

124

7 Emergence and Supervenience

depend upon (instances of) physical properties. It is natural to extend the notion to allow the supervenience of one theoretical domain upon another, in which case the state or property families are given by the theories at issue (as it might be, psychology versus physics). We could also define a supervenience relation between theories by extension of a supervenience relation between their associated theoretical domains (even where these domains might be hypothetical rather than actual). That is, if the U-domain supervenes on the T-domain then we can say that theory U supervenes upon theory T. The exact nature of the dependency relation which defines supervenience is intentionally left rather vague, but one core idea is that there can be no difference in the supervening domain without a difference in the subvening domain, as in the ‘Dali test’ discussed above in Chap. 5. For example, we might claim that there can be no psychological difference without an underlying physical difference, or that there could be no ethical difference between actions without some physical difference, or that there could be no aesthetic difference between two objects without some physical difference between them and so on. A natural way to express this is in terms of indiscernibility with respect to the subvening domain requiring indiscernibility with respect to the supervening domain. Another way is to define supervenience directly in terms of the determination of properties in the supervening family by the properties of the subvening family. It is interesting that these two approaches quite naturally lead to very distinct forms of supervenience. Let’s begin with a catalogue of some basic forms of supervenience, after which I want to introduce a new form that comes in several variants. The three basic forms of supervenience of interest here are strong, weak and global supervenience. The former two notions of supervenience can be formally expressed in terms of families of properties and a direct relation of determination between them. In what follows I will take the line that property families are identified by the distinctive predicates employed by particular theories. The family of chemical properties is the set of properties distinctively referred to by chemistry, the family of physical properties is that set of properties distinctively referred to by physics, etc. I add the term ‘distinctively’ only to indicate that there must be some selection from all properties referred to by a theory since some are inessential to that theory or are deployed only to link various domains to that of the theory at issue (most obviously in the description of measurement instruments and observations). We can also expect that there will be some overlap between theories, but I think we ought to regard this common occurrence as an intrusion of one theoretical scheme into another. In such cases, we ought to assign the overlapping property to the more basic theory. Given a pair of families of properties, we can define a supervenience relation between them in various ways. Strong supervenience is typically defined as: D5. Strong Supervenience: Property family U strongly supervenes upon family T if and only if (∀σ)(∀F ∈ U )(Fσ → (∃G ∈ T )(Gσ ∧ (∀π)(Gπ → Fπ))). This says that it is necessarily true that for any instance of a property in U there is a property in T such that having that property guarantees having the U property. It does not say, though it permits, that some particular T property underlies every instance

7.2 Supervenience

125

of the target U property. If it is the case that every U property supervenes upon one T property then we can say that U reduces to T. The state G is called a ‘realizer’ of F and we say that G ‘realizes’ F. For many domains, especially that of psychology, it is thought that there can be what is called ‘multiple realization’, in which a variety of different T properties subvene the instances of a single target U property. Notice the second necessity operator, which ensures that G subvenes F in every possible world. (That is, in every possible world, anything that manages to exemplify G will also exemplify F, but not necessarily vice versa.) While some G or other in T is required for F to be exemplified, there may well be many G’s that will do the job. Call this set of properties the realizer-set of F (or the T-realizer-set of F). One should wonder about the nature of the internal necessity operator deployed in this definition, as well as the ones to follow (the external operator serves to indicate the non-contingency of these philosophical claims). As we shall see shortly, if one varies the modal strength of this operator one gets different forms of the supervenience relation. The canonical interpretation of strong supervenience is that the internal operator is full-on absolute necessity: there are no possible worlds where an object has G but lacks F. The second form of supervenience of interest to us exploits the possibility of modifying the internal necessity operator to the maximum degree, by eliminating it. Weak supervenience is thus defined as follows: D6. Weak Supervenience: Property family U weakly supervenes upon family T if and only if (∀σ)(∀F ∈ U )(Fσ → (∃G ∈ T )(Gσ ∧ (∀π)(Gπ → Fπ))). Formally, the only difference between strong and weak supervenience is the absence of the internal necessity operator in the latter. Intuitively speaking, the difference between weak and strong supervenience is that although they agree that the supervening domain is determined by states of the subvening domain, the structure of this determination relation can be different in different worlds. We might sloganize: for strong supervenience, once a realizer always a realizer, but this fails for weak supervenience. Just because some G realizes F in some world does not mean that it will realize it anywhere else in logical space. A simple (if somewhat imperfect) example of weak supervenience, presented by Kim (see Kim 1993, Chap. 4, pp. 62– 63), is the supervenience of the truth of a sentence upon the sentence’s syntax. It must be that any two sentences that are syntactically identical have the same truth value (and of course every sentence has a syntactic structure). But we do not expect the truth value to be the same from world to world, as we vary the facts which make the sentences true. We might thus expect that syntactic structure plus a specification of the facts strongly subvenes the truth of sentences.5 The difference between weak and strong supervenience will turn out to be very important for the clarification of various notions of emergence. In fact, it might be better, albeit unconventional, to put a kind of ‘dummy’ modal operator within the formula so we can generate distinct versions of supervenience by, as it were, gradually reducing the modal strength of the internal necessity operator. Strong supervenience employs full-on logical necessity but we can imagine a form of weak supervenience that still demanded that G determine F but only, for example,

126

7 Emergence and Supervenience

across all physically possible worlds. That is, it might be preferable to write the formula for weak supervenience thus: (∀σ)(∀F ∈ U )(Fσ → (∃G ∈ T )(Gσ ∧ (∀π)? (Gπ → Fπ)))

(7.1)

where ? can be replaced by whatever grade of necessity we desire, with the ‘zero’ grade of weak supervenience just being intra-world dependence. A quite different approach to supervenience is also possible. One can express supervenience in terms of indiscernibility rather than property determination (see Haugeland 1982, Hellman and Thompson 1975, 1977). One method of doing this is directly in terms of possible worlds and thus avoids the explicit appeal to modal operators. Supervenience of U upon T would require that, at least, if there is agreement about the assignment of T-states to systems then there is agreement about the assignment of U-states to systems. We might write this as: D7. Global Supervenience: Property family U globally supervenes upon family T if and only if (∀w)(∀w∗ )((w ∼T w ∗ ) → (w ∼U w ∗ )). Where ∼ X is the relation of indistinguishability with respect to the X family of properties. So D7 says simply that any two possible worlds which are indistinguishable with respect to the subvening property family’s distribution (that is, the T properties) are also indistinguishable with respect to the supervening property family’s distribution (the U properties). Despite the elegance of D7 I want to introduce an alternative definition which is closer in form to those of strong and weak supervenience. Obviously, if two worlds are indistinguishable with respect to F-properties then any system with an F-property in one world will also have it in the other, and vice versa (I am going to take it for granted that if the former system does not exist at all in the latter world then that amounts to a difference in F-property distribution). So we can rewrite D7 as: D8. Global Supervenience: Property Family U globally supervenes upon family T if and only if (∀w)(∀w∗ )(∀σ)(∀F ∈ U )((w ∼T w ∗ ∧ Fσw) → Fσw ∗ ). We can also introduce different forms of global supervenience by tweaking the domain of possible worlds. By default, the domain is just all possible worlds—so this might be appropriately labeled ‘logical’ global supervenience—but different types of supervenience arise by restricting this domain in some way. For example, if we restricted the domain to just physically possible worlds then certain scenarios where the supervenience of the mental upon the physical fails might be ruled out. An extreme but uninteresting case would be two worlds that entirely lack physical entities (they are, say, inhabited only by angels). Such worlds are physically indistinguishable but could be psychologically different so supervenience of the mental on the physical would fail. The most significant case, of course, is worlds which are physically indistinguishable from the actual world.6 I need to introduce yet another complication in order to allow for a non-trivial role for the temporal evolution of states. I am going to modify the standard definition of global supervenience by requiring that the T-indiscernibility of worlds be restricted

7.2 Supervenience

127

to indiscernibility up to the time when Fσw obtains. This opens the possibility that global supervenience can fail for properties that depend for their existence at one time on states which occur at a later time, if the possible worlds at issue lack what I will call below T-temporal supervenience (or temporal determination or, loosely speaking, determinism). There are some properties that do have this kind of dependence upon the future. Whether a prediction—a future tensed sentence—is true or not obviously depends upon the future. Less trivially, whether an action, for example, is good or bad might depend upon its consequences. If two worlds which were T-indiscernible up to the time of the action could diverge (with respect to T) after the action then it could be that the action was good in one world but bad in the other. If we were inclined to hypothesize that moral properties and ‘consequences’ supervene upon the physical state of the world up to the time of the action such divergence would represent the failure of that hypothesis. We could distinguish an ‘absolute’ global from a ‘limited’ global supervenience of U upon T, the former involving absolute world T-indiscernibility across all space and time, the latter only indiscernibility up to the occurrence of a given U-state. Fortunately, such a distinction would be of little assistance in what follows, so I shall resist adding yet another kind of supervenience. In any event, the world based formulation reveals an ambiguity in the notion of supervenience (for discussion and an interesting dispute about their relationship see Petrie 1987; Kim 1993, Chap. 5; Paull and Sider 1992). The formulation of global supervenience in terms of worlds, unlike the definition of strong supervenience given above, does not explicitly require that the T-state that subvenes a U-state be a state of the very same system that exemplifies the U-state. This is thus an extremely weak form of supervenience. For example, it permits two worlds that differ only in the position of a single atom somewhere in, say, the interior of the star Vega to have radically distinct distributions of psychological properties—perhaps one is the actual world but in the other there are no minds whatsoever! Since these worlds are physically different, mere global supervenience does not prevent them from being also psychologically different. In the limit, global supervenience is consistent with our world being the only world that contains consciousness if it should be the case that all possible worlds that are in any way physically distinguishable from the actual world are devoid of consciousness. Paull and Sider (1992) argue that it would be hard to understand how the structure of possible worlds could be such, but the point here is that the concept of global supervenience by itself does not dictate the metaphysical structure of modality. It is not difficult to strengthen global supervenience into a form that makes the indiscernibility of particular systems rather than whole worlds the basis of supervenience, a form we might naturally call local supervenience (with the caveat that the term ‘local’ refers to the unity of systems, not to immediate spatio-temporal neighborhood): D9. Local Supervenience: (∀w)(∀w ∗ )(∀σ)(∀π)(∀F ∈ U )(((∀G ∈ T )(Gσw ↔ Gπw ∗ ) ∧ Fσw)) → Fπw ∗ ).7 This adds the condition that it is the systems, σ and π, that are such that if they are T-indiscernible across possible worlds then they will also be U-indiscernible.

128

7 Emergence and Supervenience

Local supervenience is not quite the same as strong supervenience.8 The latter does not require full local indiscernibility as a condition of supervenience but only the sharing of one critical property from the subvening family. Though less weak than global supervenience, local supervenience is still a very weak notion. It permits, for example, two human beings which differ only in the position of an atom somewhere in the interior of their big toes to differ radically in their psychological properties— perhaps one system is me but the other has no psychological properties at all! Problematic examples such as that of Vega or the big toe reinforce the intuitively plausibility of the ‘super-localization’ of strong supervenience, for it seems reasonable to suppose that some T-properties might be irrelevant to the possession of U-properties. For example, in some possible worlds (maybe even in the actual world) there are creatures physically identical to us except that they are made out of antimatter rather than matter. This would, in all probability, seem to be psychologically irrelevant but they would fail the test of indiscernibility since although systems composed of matter and anti-matter can share almost all their physical properties they are obviously physically discernible (have each of them catch an ordinary baseball, but stand way back). Of course, global and local supervenience do not prevent nonidentical systems from possessing the same supervening properties, but we could not use either global or local supervenience to argue for our anti-matter cousins possession of mind, whereas strong supervenience would probably—depending upon the exact range of physical properties we take to subvene mind—provide such an argument. Evidently, strong supervenience implies local supervenience but not vice versa. If we assume strong supervenience and the antecedent of local supervenience we obtain the local T-indiscernibility of σ and π across w and w∗ . By strong supervenience, there is a T-state, G, that σ has which necessitates F. Since σ and π are T-indiscernible, π must also possess G and therefore we must have Fπw ∗ . The reverse fails for we cannot deduce simply from the fact that σ and π share G across possible worlds that σ and π are fully T-indiscernible across the worlds. This argument could fail if we allow—as I think sound philosophy forbids—some very dubious metaphysical chicanery which encodes every feature of a possible world as a property of an individual in that world. An example would be ‘properties’ like ‘exists in a world where the speed of light is 300,000 km/s’, so that any discernibility between worlds will translate automatically into a discernibility between individuals across those worlds. It might thus be natural to think of local supervenience in terms of indiscernibility with respect to intrinsic properties. It is, furthermore, obvious that local supervenience implies global supervenience but that once again the reverse fails to hold (since the assumption of local T-indiscernibility of two systems will not lead to the T-indiscernibility of their entire possible worlds). However, the definitions can be brought together by fiat, if we restrict attention to domains where ‘reasonable’ and ‘plausible’ supervenience relations are local and particular. This restriction is important since it is arguable that efficacy really is both local and dependent upon particular states, and we have a strong interest in domains that are efficacious. An illustration of a non-efficacious and non-local ‘domain’ is

7.2 Supervenience

129

that of financial instruments. Money does not supervene locally upon the physical. Two locally physically identical scraps of paper could differ in their monetary value depending upon, for example, the intentions and social-status of their creators; our authorities intend that only properly minted notes are real money, anything else is counterfeit. But for that very reason, money can’t cause anything as such, but only via its exemplifying certain physical features that cause certain beliefs (in people) or certain other physical states (for example, in vending machines). Of course, this is not to say that appeal to monetary properties is explanatorily impotent (after all, isn’t the love of money the root of all evil), but only that its efficacy is not rooted in its being money, which is invisible, so to speak, to the causal processes of the world.9

7.3 Temporal Supervenience I want now to introduce a new form of supervenience: temporal supervenience, in which the state of a system at one time is determined by the state of the system at an another time (generally speaking, an earlier time if we think of causal determination). Temporal supervenience, as I call it, is simply a familiar notion with an unfamiliar name. But while it is odd to employ the term thus, I use the name ‘temporal supervenience’ to emphasize the analogies between the evolution of the states of systems through time and the kinds of supervenience relations we have already discussed. As we shall see, the two notions have quite deep and somewhat surprising relationships as well. The formalization of this notion unfortunately—from the point of view of ready intelligibility—requires the addition of a temporal term to our definitions but these are in other respects quite analogous to the previous forms. D10. Temporal Supervenience (ts): The states of system σ temporally supervene upon the states of σ if and only if (∀F)(∀t1 )(Fσt1 → (∃G)(∃t0 )(Gσt0 ∧ (∀π)(∀t2 )(∃t3 )(Gπt2 → Fπt3 ))). Here, and below, F and G are possible states, or properties, of a system σ.10 Call F the ‘successor state’ and G the ‘predecessor state’. To avoid clutter, it is not explicitly stated in the definitions but it is assumed that where contextually appropriate the indices on the temporal variables represent time order, so in D10 t0 is prior to t1 and t2 is before t3 . I make no attempt to specify the amount of time there should be between states (so some indulgence on the part of the reader is requested when contemplating the temporal distance between the specified times) or to address the issue of whether time is continuous or discrete. In essence, D10 says that a system’s states temporally supervene upon earlier states if there is an earlier state which determines the later state to occur. This is simply familiar temporal evolution of a system’s states expressed in the language of the supervenience relation, which applies in a remarkably similar way to that of standard supervenience.

130

7 Emergence and Supervenience

D11. Full Temporal Supervenience (fts): The states of system σ fully temporally supervene upon the states of σ if and only if: (∀F)(∀t1 )(Fσt1 → (∃G)(∃t0 )(Gσt0 ∧ (∀π)(∀t2 )(∃t3 )(Gπt2 ↔ Fπt3 ))). The difference between ts and fts is that in fts there is unique temporal determination both backwards and forwards in time (which is not to say that we have backwards causation). One can, that is, as easily foretell the past as the future of the system from its current state. Though it won’t figure much in the discussion below, full temporal supervenience is nonetheless important since, generally speaking, fundamental theories of physics tend to exemplify it. D12. T/U-temporal supervenience: The T-states of system σ temporally supervene upon the U-states of σ if and only if (∀F ∈ T )(∀t1 )(Fσt1 → (∃G ∈ U )(∃t0 )(Gσt0 ∧ (∀π)(∀t2 )(∃t3 )(Gπt2 → Fπt3 ))). D13. Full T/U-temporal supervenience: The T-states of system σ temporally supervene upon the U-states of σ if and only if: (∀F ∈ T )(∀t1 )(Fσt1 → (∃G ∈ U )(∃t0 )(Gσt0 ∧ (∀π)(∀t2 )(∃t3 )(Gπt2 ↔ Fπt3 ))). Note that T and U can be the same theory (or family of states). In the discussion below, intra rather than inter-domain temporal supervenience will figure most prominently. So instead of writing ‘T/T-temporal supervenience’ I’ll just use ‘T-temporal supervenience’. The notions of T/U-temporal supervenience are more useful than the more basic ts and fts since we normally are concerned with the relations of temporal supervenience either within theories or across theories, rather than from an abstract, non-theoretical standpoint. This is especially so when we consider the proper understanding of emergence, with its assumption that the world can be ordered into ontological ‘levels’ which, more or less, correspond to distinct theories. Unsurprisingly, the kinds of modal differences between strong and weak supervenience can be duplicated within temporal supervenience, simply by inserting a modal necessity operator to indicate the strength of the connection between G and F. The generic form would be this: (∀F ∈ T )(∀t1 )(Fσt1 → (∃G ∈ U )(∃t0 )(Gσt0 ∧ ? (∀π)(∀t2 )(∃t3 )(Gπt2 ↔ Fπt3 ))), where, as above, ? represents a ‘variable’ grade of necessity. The different notions of temporal supervenience generated in this way are exact analogues of the difference between weak and strong supervenience as given above. Intuitively, the difference is that weak temporal supervenience requires that every possible world exhibit unique state determination across time (backwards and forwards for full temporal supervenience) but that the particular state-to-state transitions can differ from world to world whereas in strong temporal supervenience the particular determination relation at issue holds across possible worlds. This difference can matter philosophically, as we will eventually see below. One general condition on the above definitions of temporal supervenience should be noted. It is understood that the systems in question are undisturbed systems, where undisturbed is taken to mean that there are no T-influences which are acting

7.3 Temporal Supervenience

131

on the system which are not part of the system. We can allow for ‘approximately undisturbed’ systems where the unaccounted for T-influences are insufficient to much alter the state transitions referred to by the definitions. Also, for cases of disturbed systems, we can always ‘create’ an undisturbed system by letting the boundaries of the system grow to encompass the T-disturbance.

7.4 Top-Down Discipline D14. Top-Down Discipline: A family of states (or theory), U, has Top-Down Discipline (TDD, or U/T-TDD) relative to a family of states (or theory), T if and only if: (1) U supervenes upon T (2) for every U-state, σ, the set of realizing T-states is such that each element can temporally evolve into a realizer of any permitted U-successor of σ and every permitted U-successor of σ is realized by the temporally evolved state of a member of the T-realizer set of σ. D14 is complex and requires elucidation. So, if there is a U-successor state of σ into which σ’s T-realizer cannot evolve then TDD fails. Another way to define the notion of Top Down Discipline is to say that in addition to the supervenience of U upon T it is the case that every element of the T-realizer set of a given U-state, U1 , can evolve into elements of the realizer set of all the permitted U-successors of U1 . Some discussion of the possibilities the definition allows might make this notion clearer. Assume that U supervenes upon T. TDD fails if there is a U-state, σ1 , which has a set of realizers that evolves into a set of T-states that does not form a realizer-set of a U-state which is a permitted (by U) successor of σ1 . A simple abstract example would be this. Suppose that σ1 is multiply realized by the set of T-states {τ1 , τ2 , τ3 }. Suppose further that the laws of temporal evolution in T are as follows: τ1 ⇒ τ1∗ , supervenience). We have τ2 ⇒ τ2∗ , τ3 ⇒ τ3∗ (note we are thus assuming T-temporal   realizes TDD (so far as this example is concerned) if the set τ1∗ , τ2∗ , τ3∗ multiply  ∗ . If, perchance, τ ∗ , τ ∗ realizes one one U-state, which we might naturally label σ 1 1 2   U-state while τ3∗ realizes another, TDD fails (since, for example, τ1 cannot evolve into a member of the realizer set of this latter U-state). Now, consider a case where T-temporal supervenience  fails. In that case instead of {τ1 , τ2 , τ3 } evolving to the determinate set τ1∗ , τ2∗ , τ3∗ we have an indeterminate evolution. For simplicity, let’s confine the indeterminacy to τ3 which, say, can evolve either into τ3∗ or τ3∗∗ (where these states cannot both obtain, no more than can any  ∗pair∗of possible  realizers). Thus ∗ , τ ∗∗ . TDD still holds } {τ to τ , τ , τ , τ , τ T-temporalevolution will lead from 1 2 3 1 2 3 3  if this set, τ1∗ , τ2∗ , τ3∗ , τ3∗∗ , multiply realizes a single U-state. If this set does not multiply a single U-state but rather underlies, U-states, σ1∗ and σ2∗  realize  ∗∗  say, two ∗ ∗ ∗ ∗ ∗ where τ1 , τ2 , τ3 multiply realizes σ1 and τ3 realizes σ2 then TDD fails even in the ‘loose environment’ where T-temporal supervenience does not hold (since, for example, τ1 cannot evolve into a realization of σ2∗ ). But notice that it is possible to

132

7 Emergence and Supervenience

Fig. 7.1 Indeterminate evolution preserving TDD

Fig. 7.2 U-temp. supervenience without T-temp. supervenience

have top down discipline even if the set of T-realizers of some U-state do not evolve to a set which realizes a single successor U-state. This can occur if the T-realizers can each indeterministically evolve to a realizer of any permitted (by the laws of U) successor of the initial U-state, as in Fig. 7.1. Top-down discipline can exist from a supervening domain, U, to its supervenience base domain, T, even if T lacks T-temporal supervenience and U enjoys U-temporal supervenience. In such a case, we could say that there is ‘de-randomization’ of T (see Fig. 7.2). It is possible that the apparent deterministic character of classical (or macroscopic mechanics) is the result of this sort of de-randomization, as the quantum states that realize the macro-states provide top-down discipline for the macro-domain. That is, while there may be intrinsic randomness at the micro-level, it somehow cancels out in the myriad interactions involved in the realization of the macro-states and their dynamics.11 I think that much of the interest in multiple realizability which has been shown by philosophers lies in the possibility of a failure of top-down discipline rather than

7.4 Top-Down Discipline

133

the mere possibility of multiple realization itself. For suppose that the supervenient domain, call it U, enjoyed top-down discipline with respect to its supervenience base, T. If there is T-temporal supervenience (and in much of physics, even, perhaps— given unitary evolution—in quantum mechanics, we seem to have full temporal supervenience) then top-down discipline implies that there is also U-temporal supervenience (see R4 below). This would strongly suggest that the supervenient domain can be reduced to its supervenience base, since there is a set of subvening states that exactly map on to the theoretical relationships of the supervenient domain. That is, there would be a ‘model’ of U in the subvening T-states. While we might expect that some domains, for example that of chemistry, enjoy both top-down discipline and a reductive relation relative to physics, this is not generally the case. Most domains ‘above’ physics, simply do not have (and do not necessarily even want to have) the resources to fully determine state-to-state transitions. If this lack of temporal supervenience in the supervening domain is coupled with the lack of top-down discipline (which I think is the usual case, since we presumably have physical temporal supervenience for the underlying realizing states of the higher-level states), the case for reduction is very weak even though every supervenient state has, of course, an entirely definite, if very large, set of realizers in the supervenience base. This is because there is no model of the supervenient domain within the base. Consider Fig. 7.3. From the point of view of the T-domain situation, the U-state at t1 has to be thought of as a disjunction of the particular T-states that can each realize that U-state. But that disjunction will not ‘act like’ a U-state since it transforms into a disjunction that cuts across U-classifications. This seems to me a powerful reason for regarding supervenience without top-down discipline as a non-reductive relation between domains or theories.12 However, even if there is no reduction of U to T, the T situation nonetheless fully determines the U situation and perhaps even explains why the U laws hold. An interesting example of the failure of top down discipline stems from the relation between thermodynamics and statistical mechanics. Although the mechanical account of thermodynamical properties perhaps comes as close to reduction as any real inter-theoretic relation could, it fails. The reason is that top-down discipline is missing. It is a law of thermodynamics that the entropy of an isolated system never decreases (an isolated system in equilibrium should be able to maintain its entropy at some maximum). However, the thermodynamical properties of a system are not brute intrinsic properties but rather depend upon lower level features of the system’s constituents. The thermodynamical states of, for example, a gas are realized by the particular states of motion of its constituent molecules. It is provable that the mechanical correlate of entropy (defined in terms of the probability of micro-states occupying certain regions of the system’s phase space; see Sklar 1993 for a philosophically based explication) is exceedingly likely to increase with the natural evolution of the molecular states, but it is not certain. This is because there are some possible molecular states which will evolve into lower entropy configurations. If we take an isolated container of some gas and introduce through a small hole extremely hot gas the system will lose entropy via our interaction but will then evolve so that the gas regains equilibrium. But it must be possible to take

134

7 Emergence and Supervenience

Fig. 7.3 A mode of failure of TDD

the state of the gas that has attained equilibrium and, at least in imagination, reverse all velocities of all gas molecules.13 Since mechanics possesses what I have called full temporal supervenience, the gas would evolve to a state where hot gas would stream out of the input hole; it would thus be a naturally evolving system (which it is fair to regard as essentially isolated) with decreasing entropy, contrary to the laws of thermodynamics. In our terms, there are certain particular realizing states that fail to evolve into permitted thermodynamical successor states, which is incompatible with top-down discipline. Does this show that standard thermodynamics is just plain false, or is a kind of ‘special science’ with only approximately accurate, ceteris paribus laws which occasionally break down because of certain incursions ‘from below’? Notice however that it is always possible, in the face of a recognized failure of top-down discipline, to develop, or at least try to develop, a more discriminating U-theory that differentiates the initial U-state at t1 into two states that differ in some heretofore unrecognized U feature. This would permit a U explanation of why the original (now regarded as under-characterized) state could evolve into distinct U-states. This new found difference within the U domain would reflect a partition of the set of T-realizers into two sets, each realizing a distinct U-state. This would restore (so far as our Fig. 7.3 goes) top-down discipline. But it would also strengthen the case for reduction, as we would have succeeded in finding a set of realizers that ‘act like’ U-states throughout their dynamics. There might be some general theoretical pressure to search for such discriminators, at least in some cases.14 However, my own feeling is that most high-level theories are not in the business of giving such a complete description of their domain as to fully constrain its dynamics. The more fundamental a theory is taken to be the stronger this pressure will be; on the other hand, very high level theories, such as psychology, will hardly feel it at all.15

7.4 Top-Down Discipline

135

Fig. 7.4 A mode of failure of STDD

If we take seriously, as we should, the possibility of indeterministic evolution of states, we ought also to consider that there may be definite probabilities associated with possible state transitions. We can easily adapt our notion of top-down discipline to incorporate this refinement. D16. Statistical Top-Down Discipline: A family of states (or theory), U, has statistical Top-Down Discipline (STDD, or U/T-STDD) relative to a family of states (or theory), T if and only if: (1) U supervenes upon T (2) for every U-state, σ, with permitted successors {σ1 , σ2 , . . . , σn } and transition probabilities { p1 , p2 , . . . , pn }, each T-realizer of σ, τ , has permitted successors τ1 , τ2 , . . . , τ j such that each τk realizes one of {σ1 , σ2 , . . . , σn } and if τk realizes σi then the transition probability from τ to τk equals pi . This is unfortunately complex but an example of a failure of statistical TDD should make it clearer. In Fig. 7.4 we see a U-state that can indeterministically evolve into two U-states. The realizing T-states mirror the indeterminacy (they meet the initial definition of TDD given above) and in fact manage to duplicate the correct Utransition probabilities overall (on the crucial assumption that the U-state is equally likely to be realized by the two possible realizing T-states). But the T-transition probabilities do not mirror the U-transition probabilities at the level of the individual T-realizers. So statistical top-down discipline fails. To my mind, the kind of situation illustrated in Fig. 7.4 is one that also counts against reduction of the U-states to the set of realizing T-states. These states do not behave like the U-states they realize at the level of individual states, though their interactions conspire to make the U-state description accurate at its own level. The particular probabilities in this case may seem somewhat miraculous (obviously

136

7 Emergence and Supervenience

they have been rigged to illustrate the desired failure of STDD), but the miracle is less pronounced in cases where high-level statistics emerge out of myriads of lowlevel processes and the high-level statistics reflect various ‘laws of large numbers’. Furthermore, we expect that normally the development of the subvening theory, or at least the recognition of its relation to the supervening theory, will follow on the development of the supervening theory and of course the follow-up theory must be constrained to produce known statistical relationships.16 Normally, we take the possibility of indeterministic state evolution in the domain of a high-level theory to be the result merely of ignorance of—or sometimes unconcern about—underlying determining factors. Such factors may involve unknown, or imprecisely specified, features of the high-level theory or, more often I think, may involve features of lower level theories that ‘intrude’ upon the dynamics of the highlevel theory to determine certain state transitions. For example, it seems entirely plausible to suppose that in certain cases the total psychological state of someone may not determine which action they will perform. Nonetheless, some action will be performed, and there may be no underlying psychological feature that accounts for this. There will, though, be some sub-psychological features which tip the balance. For example, in choosing between the cheese plate or the chocolate cake for dessert, perhaps the level of some neurotransmitter within some critical network of neurons plays a crucial role even though there is absolutely no conscious—or unconscious in a psychological sense of the term as opposed to a merely non-conscious—element of one’s mental state that reflects that neurotransmitter level precisely enough to account for one’s choice. If the underlying realizers of the high-level states are differentiated by features that do not correspond to differences in the high-level characterization, we would expect the probabilities of the state transitions of the realizers to differ from those of the high-level state transition probabilities. Thus we would expect statistical top-down discipline to fail (in the limit of temporal supervenience the probabilities of transition in the low-level theory go to one or zero). As noted above, this would also put a strain on a reductive account of the relation between the high-level states and the set of low-level realizing states. But when reduction fails, we have at least the two options mentioned above: declare the highlevel theory false, or accept that the high-level theory is not in the business of totally constraining the evolution of the systems it describes but is rather focused on elucidating certain typically occurring patterns. Since the range of incursions from below is indefinitely broad and need have no theoretical unity we should not expect any special or unified account of the failures of higher level theories to fully characterize their domains and dynamics.

7.5 Some Consequences The definitions given above result in a number of ‘theorems’ or at least a number of notable relations between them that, in addition to their own interest, can be used

7.5 Some Consequences

137

to clarify a variety of possible views on the nature of emergence. Some additional assumptions are required for certain of these results, which are discussed as necessary. R1. Final-physical-possibility does not imply physical-totality. What I mean by this claim is just that there is no inbuilt necessity that the laws of final-physics will sustain completeness, closure and resolution. It is, I think, the goal of many modern physicists to produce a theory that has totality. The structure of basic physics now in place would appear to be that of one aiming, so to speak, to be a total theory, but the current theory is manifestly incomplete (or worse, incoherent). We do not have a coherent theoretical description of every physically possible situation, even if we permit such understanding to be general and qualitative. For example, we just do not know the physics of processes that essentially involve both quantum and gravitational processes (and the two fundamental theories involved, general relativity and quantum field theory seem to be fundamentally inconsistent with one another). There are several possible approaches to integrating the quantum world with gravitation (for an intelligible overview of some see Smolin 2001), but none are very far advanced at present and it remains far from clear whether any of them will succeed. In fact, it is a possibility that there is no theory which integrates these disparate domains. Nature herself needs no theory and does not calculate the evolution of the universe. There can be no a priori guarantee that there is a unified mathematical description of this evolution. Even something as wild as Ian Hacking’s ‘Argentine fantasy’ cannot be ruled out. Hacking imagines that God did not write a Book of Nature of the sort that the old Europeans imagined. He wrote a Borgesian library, each book of which is as brief as possible, yet each book is inconsistent with every other…For every book, there is some humanly accessible bit of Nature such that that book, and no other, makes possible the comprehension, prediction and influencing of what is going on. (Hacking 1983, p. 219)

While it does seem clear that the research on quantum gravity is intended to complete physics in a way that provides totality, whether physicists can succeed in developing a final theory that is total depends not only upon their ingenuity but also upon the nature of the world itself. It is impossible to say now whether this research will or even can succeed, for it is not given beforehand that there is a single physical theory that can encompass all elementary physical processes. Nor can we yet rule out as incoherent the sort of radical emergence (to be defined below) that denies closure via resolution. Therefore we cannot say in advance that the final physics is a total theory or that the worlds that are final-physically possible are all such as to observe totality. R2. Strong supervenience of U upon T is compatible with the absence of T-temporal supervenience. Failure of T-temporal supervenience means only that there is a T-state, σ, of a system that does not have a predecessor which leads uniquely to σ. Obviously, this does not

138

7 Emergence and Supervenience

prevent U from strongly supervening upon T unless further substantial conditions are met. R3a. Strong Supervenience of U upon T and T-temporal supervenience does not imply U-temporal supervenience. The reason is that Top-Down Discipline of U relative to T might fail. That is, the set of realizers of some U-state(s) might lead to realizations of different subsequent U-states, even though each such realizer T-state has a unique outcome, as illustrated in Fig. 7.3. R3b. Nor does strong supervenience of U upon T and the absence of T-temporal supervenience imply the absence of U-temporal supervenience. This is possible because there could be top-down discipline of U relative to T despite the failure of T-temporal determination, as illustrated in Fig. 7.2 (see the discussion of de-randomization above). R4. Strong Supervenience of U upon T, Strong T-temporal supervenience and topdown discipline of U relative to T implies Strong U-temporal supervenience.17 If we have T-temporal supervenience then U/T-TDD implies that every T-state which realizes some U-state, σ, must evolve to realize a single successor U-state. By strong supervenience, σ must have a realizing T-state. So σ must have a unique successor, which is to say, we have Strong U-temporal supervenience. R5. Strong supervenience of U upon T implies that U-states (probably) have T-efficacy. Suppose that U strongly supervenes upon T and consider some U-state, σ. The state σ has a set of T-realizers {τ1 , τ2 , . . . , τn }. To test if σ has T-efficacy in the production of some state we consider the counterfactual: for some actual system, S, and some actual outcome T-state, τ of system S* , if S had not been in state σ then S* would not have been in state τ .

In the nearest possible world where S is not in state σ we cannot have any of σ’s realizer states obtaining; that is, none of {τ1 , τ2 , . . . , τn }. We can assume that S being in state σ was the outcome of one of these realizing states obtaining (since we need only find one such to reveal efficacy). Since none of these obtain in the counterfactual situation, it is unlikely that S* ’s being in state τ would come about nonetheless. Actual judgments of efficacy would have to depend upon particular circumstances, but it seems that it is very probable that states of strongly supervening domains have (or typically have) efficacy. To take a definite example, suppose that I want an apple and then reach out and take an apple. Was my desire efficacious in this transition? Well, according to strong supervenience we assume that my wanting an apple was realized by some physical state, P, from the set of possible realizers of apple-wantings. In the counterfactual situation of my not wanting an apple, P—along with all the other possible apple-wanting realizing physical states—would not obtain (since if it did, by supervenience, my desire would exist after all). Would I still reach out for the apple? It is certainly possible to imagine situations where this occurs: suppose that Dr. Horrible has trained his ‘action-gun’ upon me and is ready to force my body

7.5 Some Consequences

139

to reach for the apple at the appropriate time should I lack the desire. However, such situations of ‘counterfactual overdetermination’ are, presumably, very rare, and thus we may conclude that strongly supervening states very probably—typically— have efficacy. (If T has full strong T-temporal supervenience then we can say that U-states definitely have T-efficacy. For then S* being in state τ would have a unique predecessor and if that predecessor did not occur then S* would not be in state τ . But then the counterfactual supposition that S is not in state σ would provide the guarantee that the predecessor realizing T-state did not obtain and so S* would not be in state τ as required. This may be of interest since physics appears to enjoy full strong temporal supervenience.18 ) R6a. Strong T-temporal supervenience implies global supervenience for any domain with T-efficacy. Recall that to claim that U globally supervenes upon T is to say that any two worlds that agree on their assignment of T-states to systems will agree on their assignment of U-states. Symbolically, (∀w)(∀w ∗ )(∀σ)(∀F ∈ U )((w ∼T w ∗ ∧ Fσw) → Fσw ∗ ).

(7.2)

Thus the denial of global supervenience would be expressed , after some manipulation, as (7.3) (∃w) (∃w ∗ )(∃σ)(∃F ∈ U )(w ∼T w ∗ ∧ Fσw ∧ ¬Fσw ∗ ). That is, the denial of strong supervenience entails that there are indiscernible T-worlds that differ with respect to the non-supervening U-state, F. To test whether F has T-efficacy we must evaluate the following counterfactual: If Fσ had not been the case then H σ would not have been the case.

Here, H is some outcome T-property of state σwhich obtains in the ‘source world’ (i.e. the world from which we will evaluate the counterfactual, w in the above) and which is putatively brought about by F. To perform the evaluation we consider the T-possible world most like the source world save for the differences necessitated by assuming ¬Fσ. The T-possible world most like the initial world would be one that was identical with respect to T (up to the time when Fσ obtains), differing only with respect to F (and possibly other U-states). We know there is such a world, by the denial of global supervenience (w ∗ in the above). However, by strong T-temporal supervenience, that world evolves over time in exactly the same way as the source world. Therefore the counterfactual is false and F cannot have T-efficacy, contrary to the assumption of R6a. So global supervenience must hold. R6b. Strong T-temporal supervenience implies strong supervenience for any domain with T-efficacy. This argument is slightly less convincing than that for R6a, because we need an additional assumption. Suppose we have T-temporal supervenience but there is a T-efficacious domain, U, that does not strongly supervene upon T. Then by the definition of strong supervenience, there is a T-possible world where there is a system,

140

7 Emergence and Supervenience

σ, and U-property, F, such that Fσ ∧ ¬(∃G ∈ T )(Gσ ∧ (∀υ)(Gυ → Fυ)).

(7.4)

(That is to say, there is no property, G, which subvenes F in the appropriate way.) So we have Fσ and (∀G)(¬Gσ ∨ ¬(∀υ)(Gυ → Fυ)).

(7.5)

Now, this means either (1) that σ has no T-properties whatsoever or (2) there is such a T-property but it does not necessitate F. If the former, then σ is a radically non-T system. Suppose F has T-efficacy. Then the presence of F makes a difference in a T-property. But since F is a property characterizing utterly non-T entities, the presence or absence of F is not marked by any necessary T difference. For while it is perhaps possible to imagine that there might be some kind of a ‘brute metaphysical’ connection between some T-state and the presence of F, this connection is not a T-law (T-laws do not say anything about radically non-T objects). Violation of this connection is thus not a violation of any T-law, and so the world in which this connection is broken is a T-possible world. So, given T-efficacy, there could be two T-indiscernible situations which differed in their outcome because of the difference in F. But this violates strong T-temporal supervenience. That is, since F is not marked by any T-state we can take the F world and the ¬F world to be T-indiscernible (and worlds can’t get any more similar in T-respects than T-indiscernibility), and then use the argument for R6a to show strong supervenience. Now, for the second case, suppose that strong supervenience fails because of (2). Then there is a T-property, G, that σ has but is such that G does not necessitate F. This entails that there is a world in which some system has G but does not have F. We might then try to argue that in every world, G has the same outcome by strong T-temporal supervenience. Thus in whatever world we choose to evaluate the counterfactual which tests for the T-efficacy of F, there will be no T-difference. Therefore F does not have T-efficacy—it cannot make any difference. But this won’t quite work as it stands since it is open to the following worry. The counterfactual test requires that we go to the world most similar to the source world except that ¬Fσ holds. What if this is a world where ¬Gσ holds? Abstractly speaking, this seems to be possible. However, such a world will be quite unlike the source world, since strong T-temporal supervenience requires that Gσ’s predecessor not appear in the test world (else we would get Gσ after all) or else we have a miracle (which immediately violates T-temporal supervenience). That is, the assumption of ¬Gσ propagates other T-changes throughout that world. Thus it is very plausible that a ¬Gσ world is not the most T-similar to the source world. After all, we know that there is a world in which Gσ and ¬Fσ. If this is correct then the test world contains Gσ and hence must evolve to the same successor state as the source world, thus revealing that F does not possess T-efficacy.19 Since strong supervenience implies weak supervenience it trivially follows that strong T-temporal supervenience implies weak supervenience of T-efficacious

7.5 Some Consequences

141

domains. It is also the case that since strong supervenience implies global supervenience we have R6b implies R6a. Furthermore, since strong supervenience implies what I called local supervenience, we also get that strong T-temporal supervenience implies local supervenience. Note also that we have to assume T-efficacy in the above since nothing can rule out the possibility that there are ‘parallel’ domains that do not supervene upon T but rather exist entirely independent of the T-world yet enjoy rich causal relations amongst themselves, a situation that would be approximated by considering Leibniz’s system of monads but without the pre-established harmony. The assumption of T-efficacy forges an essential link between the U and T domains. Such an assumption is reasonable since we have little interest in hypothetical domains that are entirely isolated from each other. In particular, we are not very interested in an epiphenomenalist view of the mind-body relation, though it is important to see that epiphenomenalism cannot be ruled out by any considerations advanced thus far. It is also interesting to note that, given (R5), we have it that strong T-temporal supervenience implies that U is T-efficacious if and only if U strongly supervenes upon T. This highly interesting and perhaps initially surprising result reveals the significance of temporal evolution of states for the metaphysics of dependence. If we have a domain the states of which evolve through time according to the laws of that domain, then there are tight constraints placed upon the states of any other domain which are to have effects within that initial domain. They must ‘ride upon’ the lawful transitions of the initial domain to both preserve those lawful transitions and have their own efficacy, which is to say, that domain must supervene upon the initial domain. R6c. Weak T-temporal supervenience implies weak supervenience for any domain with T-efficacy. The argument for this claim is still weaker since additional assumptions (or modal intuitions) are needed. The argument proceeds in parallel with that of R6b. But when we consider the first horn of the dilemma, that σ might be a radically non-T system, we must consider the counterfactual, if σ had not been F then things would have been T-different. It seems to me that the closest world in which σ is not F is one in which the T-temporal supervenience relations are not altered (since F has nothing whatsoever to do with T, it is hard to see why the T relations would be different in that world).20 If so, F’s T-efficacy would fail. (The alternative idea, I guess, is that because of some kind of pre-established harmony, in the nearest world where σ is not F, the T-temporal supervenience relations must be altered enough to make the counterfactual come out true. But even in such a case, it seems that it is the alteration in T that accounts for the difference in outcome so that intuitively F has no efficacy in the T domain after all.) The other horn of the dilemma leads to the claim that there is an object, π, in the very same world as that in which σ has F such that π has G but does not have F. Then in that very world we have a test of F’s efficacy and—because of weak T-temporal supervenience—within any world the T-temporal supervenience relations are the same. Thus G will lead to the same outcome for system π as for system σ. So F’s T-efficacy seems to fail. If it is insisted that some

142

7 Emergence and Supervenience

kind of ‘singular causation’ is possible then we must use the counterfactual test, and then we can employ the plausibility argument given immediately above. R7. T-Totality implies strong T-temporal supervenience (up to intrinsic randomness of T). Totality is a very strong condition on the nature of the laws of a theory as well as on the ‘metaphysical structure’ of the world as constrained by that theory’s description (roughly, constituent structure with ‘bottom-up causation’ sufficient to yield all phenomena). But is it enough to guarantee temporal supervenience? Let us see. Assume that T is (supposed to be) a total theory but that T-temporal supervenience fails. Then there is a T-property, G, of system σ that does not have a unique outcome (let’s say that in such a case Gσ diverges). If Gσ is a complex state then by the property of totality I labeled ‘resolution’ we can resolve it into a set of elementary T-constituents that act entirely according to T-laws. Therefore, if Gσ does not have a unique outcome this must be because some elementary state does not have a unique outcome. So we might as well consider Gσ to be such an elementary state. It is then impossible for Gσ to diverge because there is a sub-T theory which realizes the T-states and which accounts for the divergence of Gσ. For then not everything that happens would be the result of the operation of T-laws and T-totality would be violated. The only possibility of divergence is if T has some intrinsically random elements within it. That is, if it is a brute fact that for some T-state two (or more) distinct states can ensue. To take a common example, on certain views of quantum mechanics (e.g. those that espouse the ‘uncontrollable collapse of the wave function’ view of measurement) QM-temporal supervenience fails. A particular uranium atom, in state H , may or may not fission. If it does we get, say, state H1 ; if it does not we get state H2 . There is nothing within quantum mechanics to account for this (and, at least on the view of QM we are considering, no hidden variable lurking beneath quantum mechanics either). The fissioning or lack of fissioning at any particular time is intrinsically random. If there is no intrinsic randomness then it seems that totality implies temporal supervenience. We could leave this result there: if there is no intrinsic randomness in the elementary states of T then totality implies temporal supervenience (this is less trivial than it appears since high-level theories can fail to observe temporal supervenience without possessing intrinsic randomness; totality implies that the lack of temporal supervenience must result from intrinsic randomness, not the sorts of intrusions from below that characterize high-level theories). In fact, it implies strong temporal supervenience since totality is a property of the laws of a theory and so naturally sets the conditions of possibility relative to that theory. However, there is more to say about intrinsic randomness. It is important to see that the possible existence of intrinsic randomness does not fundamentally change our result. To take account of this possibility we would have to complicate our definitions considerably, along the following lines. In place of individual states we would have to take probabilistically weighted sets of states. We could then recast our arguments in these terms. Instead of a unique outcome state as the defining characteristic of temporal supervenience we would have a unique statistically weighted set of states. Although this would get very messy I think in the end we would get completely

7.5 Some Consequences

143

analogous results to those obtained when we do not consider intrinsic randomness. A form of statistical temporal supervenience would be defined in terms of predictably weighted ensembles of states. As an illustration, consider a view once defended by Karl Popper and John Eccles (Popper and Eccles 1977). In support of a form of Cartesian dualism, Popper and Eccles hypothesized that perhaps the mind could surreptitiously act under the cloak of quantum mechanical indeterminacy, subtly skewing the intrinsically random processes occurring at the synapses of the neurons. This is conceivable, but it would be experimentally revealed, in principle, by noting that the distribution of outcome states of synaptic conditions did not strictly match the statistics predicted purely on the basis of quantum mechanics (once enough evidence had been accumulated it would be overwhelmingly likely that orthodox QM was failing to predict the correct statistics). In this way, quantum mechanics would be refuted. If quantum mechanics is true (and total, or part of the total final-physics), then the mind can only act in accordance with the statistics predicted by quantum mechanics. This would bear out the statistical version of totality. This reveals that intrinsic randomness within a theory only complicates temporal supervenience but does not destroy its essence. We could then define the statistical efficacy of a state, F (within a theory that allows for some intrinsic randomness) in terms of the presence of F making a difference to the outcome statistics over repeated ‘counterfactual trials’. For example, adding some weight to one side of a die (pretending for the sake of the argument that the die is an example of an intrinsically random system) is statistically efficacious for while it does not prevent any number from coming up it does change the outcome statistics over many trials (perhaps only very subtly). R8. Strong Supervenience of every T-efficacious domain, U, upon T and strong T-temporal supervenience implies T-Totality. Suppose every T-efficacious domain, U, strongly supervenes on T but that T-totality fails. Then either closure, completeness or resolution fails. If completeness fails then there is an entity which has no (non trivial) T-description, that is, an entity which is a radically non-T object. This entity must be from some domain, U. But then there could be a difference in U with no difference in T, for while it is perhaps possible to imagine that there might be some kind of a ‘brute metaphysical’ connection between T-states and the U-states, this connection is not a T-law if U (or that aspect of it relevant to the nature of the entity in question) is a radically non-T domain. Violation of this metaphysical connection is thus not a violation of any T-law, and the world in which this connection is broken is thus a T-possible world. But this violates strong supervenience. Suppose, then, that closure fails. Then for some domain, U (which, here and below, might be T itself), some U-state, σ,21 occurs in violation of some T-laws (say then that σ is a miraculous state or, for short, a miracle). But—by strong supervenience—σ has a realizing T-state, τ .22 By strong T-temporal supervenience, τ has a predecessor state, τ0 , for which τ is the necessary unique outcome. Could τ occur but occur in violation of T-laws? No, for then it would be T-possible for τ not to occur even though its predecessor state does occur. If it is not a matter of T-law that τ0 led to τ then

144

7 Emergence and Supervenience

there is a T-possible world where we have τ0 and but where τ does not occur. But that violates T-temporal supervenience. Therefore, τ ’s occurrence is not in violation of any T-law. Since τ is the realization of σ, σ’s occurrence does not after all violate any T-law, so closure cannot fail. Finally, suppose that resolution fails. Then there is a domain, U, and a U-state, σ, such that either, (1) there is no constitutive description of σ in T-elementary terms or, (2) there is such a description but the presence of a particular instance of σ leads to system behaviour distinct from the behaviour of σ’s elementary T-constituents as they would act under the T-laws governing the elementary T-constituents. Let’s label this possibility the divergence of σ’s behaviour from that of σ’s elementary realizers—the shadow of a significant form of emergence is obviously looming here. The first disjunct violates completeness.23 On the second disjunct, there must be a T-state that subvenes σ, call it τ which is composed of a set of elementary T-features {τ1 , τ2 , . . . , τn } (we know we have this decomposition by way of the assumption that resolution fails via divergence). T-temporal supervenience means that there is a unique outcome of each τi , so {τ1 , τ2 , . . . , τn } has a unique set of elementary T-features as its outcome. Therefore, divergence of σ’s behaviour from that of σ’s elementary realizers violates T-temporal supervenience.24 Since we have assumed that T-temporal supervenience holds, such a σ cannot exist, and therefore resolution holds. So T-Totality follows. R9. Strong T-temporal supervenience implies T-Totality (across domains with T-efficacy). From above (R6a) or (R6b), strong T-temporal supervenience implies Strong T/U supervenience or global T/U supervenience for any domain with T-efficacy. Therefore, from (R8a) or (R8b) the result follows. R10. Strong T-temporal supervenience if and only if T-Totality (across domains with T-efficacy). Various forms of this follow from (R9) and (R6).

7.6 Varieties of Emergence We are now in a position to characterize emergentism in some detail and discuss distinct forms it might take. Emergentism is the doctrine that certain features of the world—features of the emergent domain—emerge out of other features from another domain, call it the submergent domain. We have already seen enough complexity to know that defining exactly what ‘emergence’ is and how it works is not so easy. The simplest view, and one that dovetails with the approach of this chapter, is to regard emergence as relative to theoretical descriptions of the world. A feature is emergent only if it is part of one theoretical description but not another. For example, the valence of an atom is emergent inasmuch as it forms a part of chemical theory but not a part of physical theory (i.e. physics). Or again, the ‘fitness’ of a genome

7.6 Varieties of Emergence

145

is an emergent feature insofar as it is utilized by evolutionary biology but not, for example, by chemistry. Of course, this preliminary criterion is but a part of what it is for a feature to be an emergent feature. We must add a notion of the ‘direction’ of emergence, for while valence is a good example of an emergent feature we are not inclined to call spin an emergent just because spin is not mentioned in evolutionary biology. The ‘direction’ of emergence brings supervenience into the picture in a natural way. For the additional idea is that of determination of the emergent feature by features of the submergent domain. Thus, we find it appropriate to say that valence is determined by physical features, but have no reason at all to suggest that spin is determined by features peculiar to evolutionary biology. It is the nature of this determination relation that clouds the issue of emergentism, and suggests that work on supervenience may be of assistance in its clarification. For example, if we have strong supervenience of U upon T then we have what are in effect ‘laws of emergence’ that are constant across all T-possible worlds. These laws of emergence are expressed in the latter part of the formula definition of strong supervenience (i.e. the ‘(∀σ)(Gσ → Fσ)’ part (where, recall, G ∈ T and F ∈ U ) part of the definition). Notice that this provides another reason for preferring strong supervenience over global or local supervenience—it locates a definite T-state as the base for the emergent properties and this is in line with most emergentist thought.25 If we consider the difference between strong and weak supervenience in terms of emergence, we see that weak supervenience allows for the laws of emergence to vary across submergently possible worlds, which is an interesting and, as we shall see, actually critical component of any serious form of emergentism. One digression. Certain properties can perhaps be called emergent even though they fail to meet our first criterion. Mass, for example, figures in physics, yet the mass of a physically complex object can be thought of as an emergent. This is a ‘compositional’ sense of emergence, roughly characterized as a feature which an object has but which no proper part of the object possesses, although the parts possess ‘cognate’ properties. Thus, ‘having a mass of 1 amu’ is a property of an (ordinary) hydrogen atom, but none of its proper parts have this property. This seems to me rather a degenerate sort of emergence, for the ‘generic’ property—the determinable if you will, in this case ‘mass’, equally applies to both the whole and its proper parts. It is not surprising that a supervenience relation also holds between the submergent properties and the compositionally emergent properties, and usually one that is pretty straightforward and unlikely to lead to any substantial issues of emergentism. This is not to say that the relation between the mass of a composite and that of its constituents is simply additive. It is not. Because of the mass-energy convertibility, the energy of binding amongst constituents is part of the law of emergence for the mass of the composite system. The mass of the composite is somewhat less than the sum of its constituents, which is to say that energy is released through the formation of the composite. But the ‘law of emergence’ in such case follows from the laws of the submergent domain; the laws that govern how massive entities interact to form more complex structures fully determines the mass of the composite entity.

146

7 Emergence and Supervenience

In marking out the central features of emergentism we must begin by contrasting emergentism with dualism. Emergentism is anti-dualist; emergent features are features of objects which always have descriptions—albeit incomplete insofar as they neglect the emergents—from within the submergent domain. Emergence does not generate a realm separate and apart from the submergent domain. A second crucial feature of emergentism is the denial of epiphenomenalism; emergent properties are supposed to be efficacious, their presence makes a difference to the way the world goes. However, the nature of this efficacy is not always clear and can vary from a weak (and generally plausible) to a very strong (and quite implausible) claim about the role of the emergents in the unfolding of the world. We can use the results of this chapter to define the two fundamental types of emergence (along with a rather peculiar and probably useless additional variant). The weakest form of emergence is one which offers no threat to the operation of the submergent domain from which the emergents spring. To put it another way, the existence of such emergents is explicable (in principle, as discussed below) on the basis of the submergent domain. Examples of such emergence are, presumably, the liquidity of water, the shape of macroscopic objects, the chemical properties of substances, the weather, etc. Such an emergence poses no ontological threat— the emergents are clearly features of systems describable in submergent terms. And emergents of this kind can be said to have a kind of efficacy. The view that meets these conditions is what I have called conservative emergence. D17. U conservatively emerges from T if and only if T is a total theory and U has T-efficacy. Some remarks on this definition. It is admittedly highly abstract and rather remote from any real phenomena that might serve as examples. But we have seen such phenomena above in the discussion of dynamical systems and emergence in Chap. 6 in the features called ‘dynamical autonomy’ and ‘multiple realizability’. A more general positive characterization of the kind of autonomy at issue is given by Margaret Morrison as follows: ‘with emergent phenomena we have generic, stable behavior that…is immune from changes to the equations of motion of the system’ (Morrison 2006, p. 886). Such stability will however always be subject to ‘intrusion from below’ and this is a kind of symptom of low level determination. If T is a total theory and U has T-efficacy, then U strongly supervenes upon T (by R6b and R10), so we know that features at the T-level completely determine those at the U level and do so in terms of the constitutive structures. Because of resolution we can expect there is, at least in principle, an explication of the origin of emergent properties based upon the elementary T-features into which every U feature can be resolved. That is not to say that where there is conservative emergence we should eschew higher level explanations in search of lower level ones. In fact, the higher level features might be precisely those that generate understanding of the systemic properties.26 Nonetheless, the subvening level must be such as to enable generation of the stable higher level structures whenever it attains the appropriate state. Such emergents can have efficacy in the way that complexes of elementary T-features can have efficacy. This, in turn, will allow such emergents to pass the counterfactual test

7.6 Varieties of Emergence

147

of efficacy, and hence they will meet the definition of efficacy given and used above. Nonetheless, everything that happens, including the combinations of T-elementary features that underlie the emergents, happens in accord with the laws of T. Furthermore, it is worth recalling the discussion of Chaps. 5 and 6. When I say that under conservative emergence we would have an in principle explication of emergence in terms of the submergent domain I do not mean that the explication would be simple or in any sense practically attainable. It might be of such complexity that it will remain forever beyond our full comprehension. Generally speaking, these explications will proceed on a case by case basis, by the deduction from T-states and T-laws of all the behavioural capacities of U-states as well as the deduction of U-laws as springing from these behavioural capacities. We already know enough about complex systems to be quite sure that the detailed explanation of many emergents will be beyond our best efforts. However, even in the absence of detailed accounts of conservative emergence we might well have a pretty fair general idea of how the determination relation works in many cases. A good recent example illustrates both the nature of conservative emergence and the need for an ‘in principle’ clause (I draw the example from DiSalvo 1999). We have known for a long time how to perform thermoelectric cooling, in which electric current is directly converted into a temperature gradient. The effect was discovered in 1834 by Jean Peltier (you can now buy specialty picnic coolers and the like that operate thermoelectrically). The advantages of such cooling include compact size, silent operation and no moving parts, but applications have been limited by the low efficiency of current materials. Thermoelectric cooling operates at the junction of two different conductors. Passing a current through the junction causes charge conductors to diffuse away from the junction, taking heat with them. While this is an extremely over simplified and highly schematic explanation, it reveals how thermoelectric cooling is conservatively emergent. The efficiency of the process is critically dependent upon the nature of the conductors forming the junction however, and is expressed in a parameter known as zt. Known materials have a zt of about 1; if materials of zt around or above 4 could be found, thermoelectric cooling would vie with conventional methods of refrigeration for efficiency. Unfortunately, there is no general and practical way to accurately predict the zt of a substance. DiSalvo explains the situation thus: Understanding electrical carriers in crystalline solids is one of the triumphs of modern quantum mechanics, and a theory of te [thermoelectric] semiconductors has been available for about 40 years. This transport theory needs one input: the electronic band structure. More recent advances in determining the band structure, based on density functional theory and modern computers, give acceptable results. The main input to band theory is the crystal structure of the material. Known compounds can be sorted into a much smaller group of crystal structure types. A given structure type may be adopted by many compounds, and by comparison, we can often predict which elemental compositions will have this same structure because of similar atom sizes and average valence, for example. However, many new ternary and quaternary compounds adopt new structure types which cannot be predicted beforehand, and without the crystal structure, electronic band structure cannot be calculated. Not only is the inability to predict crystal structure (and thus composition or properties) the main impediment to predicting which new materials will make better te devices, this

148

7 Emergence and Supervenience

inability is most often the limiting factor in obtaining improvements in most other materials applications. (DiSalvo 1999, p. 704)

This inability to predict conservatively emergent properties stems from a number of problems: the incompleteness of our grasp of theory, our inability to perform extremely complex calculations, lack of knowledge of details of atomic structure, and so on. But there is no real question that a ‘mathematical archangel’—to once again borrow Broad’s evocative term—unfettered by limitations of computational speed or memory capacity could deduce zt from quantum mechanical principles and knowledge of the basic physical structure of the candidate materials. More abstractly, if we have a total T-theory then we can in principle explicate the behaviour of any system of any complexity from a knowledge of its elementary T-structure. We know from totality, that all systems have such a structure and closure guarantees that such predictions are in principle possible (they may, of course, yield only statistical results depending upon the nature of the T-theory).27 So, conservative emergence is the model of emergence one must adopt if one accepts that physics is (or will be) a total theory. It may be the natural view of emergence from within the ‘scientific view of the world’, since that view is taken by very many thinkers to include the claim that physics, which provides the fundamental description of the world, is a total theory. But I would like to remind the reader that, as noted above in the discussion of (R1), no one knows for certain whether or not the final physics will be a total theory, and hence no one knows if the fundamental structure of the world is ‘physically total’ either. Whether or not the world is total is an empirical matter, and cannot be decided by any a priori metaphysical arguments. The original emergentists, which include Mill (see Mill 1843/1963, especially Bk. 3, Chap. 6), Lewes (who introduced the term ‘emergentism’ in Lewes 1875), Morgan (1923), Alexander (1920) and Broad (who, in 1925 articulated and vigorously defended the coherence and general plausibility of emergentism although in the end did not fully endorse it), would not have been satisfied with mere conservative emergence (for an excellent general discussion of their views, see McLaughlin 1992). They wanted more, and in particular they wanted their emergents to possess both a stronger form of efficacy and a more mysterious and portentous relation to the submergent domain than conservative emergence allows. Furthermore, although the move from submergent to emergent was to be mysterious it was to be a part of the natural order, not a mere accident or lucky chance. That is, the presence of an emergent feature was supposed to be in principle unpredictable even given a completely precise specification of the submergent domain and a complete theoretical understanding of that domain. A sign of this kind of emergence is, as Broad put it, ‘…that in no case could the behaviour of a whole composed of certain constituents be predicted merely from a knowledge of the properties of these constituents, taken separately, and of their proportions and arrangements in the particular complex under consideration’ (Broad 1925, p. 63). The point of talking of ‘prediction in principle’ is to provide a natural way to erase the epistemological constraints which can cloud metaphysics. The claim of impossibility of prediction of U-states on the basis of fundamental T-state even in

7.6 Varieties of Emergence

149

principle is the denial of determination or strong supervenience of U upon T. It is conceivable that this venerable way to approach the ever present gap between epistemology and metaphysics which links in principle predictability with strong supervenience masks another distinction, a distinction between predictability (in any sense) and determination. If so, the deployment of the idea of prediction in principle would become (even more of) a metaphor for the determination of all properties but those of the submergent domain. But I take it that Broad and the other emergentists did intend to speak of a lack of determination or supervenience when they talked of a lack of predictability in principle and I will follow them in this. Following Chaps. 5 and 6 above, let us call this hypothetical, new form of emergence radical emergence. It is obvious that radical emergence implies that the submergent domain is not total (or that the theory of the submergent domain is not total). The failure of totality can be further diagnosed as a failure of closure. Completeness can hold, since the emergents are not new substances; and resolution can hold in the sense that complexes that possess emergent properties can be resolved into elementary constituents of the submergent domain. But the behaviour of these complexes is—most emphatically—not given by the concerted behaviour of those elementary constituents as they act, or would act, solely under the laws of the submergent domain. Thus closure must fail. We also know, from R10, that the failure of totality implies that we do not have strong T-temporal supervenience. So if radical emergence is true then physics is not total. This could obtain in two ways. The first is that physics, as a theory, could be merely formally total. That is, physics could have the form of a total theory but be false of the world. Right now, given the pretensions of physics and its structure, this seems to be the only way radical emergence could be true. It is from this viewpoint that a severe tension is generated between radical emergence and physical theory. But the other way totality can fail is, I think, more promising. It is possible to imagine physics just giving up its efforts to be total and resting content with describing the nature of the ‘ultimate constituents’ of the world with no implication that this description will implicitly fully constrain all of the world’s ‘behavioural possibilities’. It will, that is, be possible to resolve every complex physical entity into ultimate physical constituents, but not possible, even ‘in principle’, (and not thought to be possible) to recover the behaviour of every such complex merely from the interactions of the constituents as they act according to the laws of fundamental physics. This would of course be to embrace radical emergentism. This would indeed be a radical departure from our usual understanding of the aim of physical theory, for it requires a physics that is essentially ‘incompletable’, one admitting that the transition from elementary physical activity to the activity of complex physical systems is not entirely governed by fundamental physical law. Thus it feels instantly implausible to modern sensibilities. And this implausibility may be grounded in more than emergentism’s unfashionable opposition to the current physicalist zeitgeist, since emergentism may contradict some of the very general principles upon which our modern physical understanding of the world is based. But it is difficult to decide whether radical emergence actually requires the violation of such principles. For example, does radical emergence entail the violation of the principle

150

7 Emergence and Supervenience

of the conservation of energy? It seems that it might not, and there are at least three ways to arrive at this conclusion. However, one of these ways, due to McLaughlin (1992), reveals the almost irresistible urge back towards the totality of physical theory and the consequent demotion of radical emergence to mere conservative emergence. McLaughlin’s suggestion applies to a system with emergent features acting in a way that appears to diverge from the action we would expect based on the physical understanding of the constituents of the system, thus violating the conservation of energy. Energy conservation can be reclaimed by positing a new sort of potential energy field which the emergent features can, so to speak, tap. The difficulty with this solution is that this potential energy field will naturally be counted as a new and basic physical feature of the world, which restores totality to physics and with it predictability (in principle) of the behaviour of complex systems from a knowledge limited to all the fundamental physical features of the system in question. An example to illustrate this problem is the bizarre Casimir effect, which at first sight may seem to offer an instance of radical emergence. If two flat metal plates are placed very close to each other (but not touching) there will arise a (non-gravitational) force between them, pushing them together ever so slightly. Is this the radical emergence of a new force emerging from certain quite particular macroscopic configurations of matter? No. The standard explanation of the Casimir effect is, roughly, that there is energy in the quantum mechanical vacuum which is, because of the nature of the metal plates and arcane details about the possible distributions of virtual photons and their characteristic wave lengths between and beyond the plates, slightly greater outside the plates than between them. The point here is that the explanation spoils the appearance of radical emergence, for the ‘potential energy’ locked in the vacuum is explicable in elementary terms, and thus the force between the plates is predictable (in principle) just from basic physical features (and of course it was predicted, by Hendrik Casimir, in 1948 and unambiguously experimentally observed in 1997). McLaughlin’s proposal, then, is a general method of transforming radical into conservative emergence, by the postulation of new potential energy fields which can be regarded either as stemming from or as themselves constituting new elementary physical features of the world. That is, these fields might be explicable in more elementary physical terms (rather as in the example of the Casimir effect) or they might be new brute facts about the basic physical structure of the world. The second proposal retains the radical nature of emergence but requires that there be a high-level ‘conspiracy’ to balance the energy books. That is, the defender of radical emergence can believe that energy is created or destroyed, as needed, ‘out of the blue’ when certain complex configurations are realized but that somehow an equal amount of energy disappears, or appears, from the universe ‘somewhere else’ whenever these configurations arise. This is not logically incoherent, but would of course be utterly mysterious from the point of view of basic physics. With respect to this ‘defense’ of energy conservation, it seems the defender of radical emergence might do better to simply allow the conservation of energy to lapse on the grounds that it is better to have one mystery rather than two (the second being the odd and oddly coordinated disappearance/appearance of energy). After all,

7.6 Varieties of Emergence

151

if we are allowing radical emergence there is no reason to deny that energy itself can radically emerge. But a third method of saving energy conservation is perhaps more in line with radical emergentism and its assertion that fundamental physics, conceived of as a total theory, is incompletable. The main idea here is that energy conservation is a system relative property, and those systems exhibiting emergent properties will abide by energy conservation as systems, with no implications about the processes involved in the system’s coming into being. What I mean can best be explained by an example. Physical systems can often be described mathematically in terms of a certain function, called the Hamiltonian, which encodes relevant properties of the system as well as forces acting on the system. The simplest case of use to us is the classical Hamiltonian of a single particle constrained to move in a single dimension subject to a field of force. The mathematical expression is this: H (x, p) =

p2 + V (x) 2m

Here, p represents momentum, m mass and the function V (x) represents a ‘force field’ in which the particle moves. From this equation one can deduce the functions governing the position and momentum of the particle over time. Most significantly, the Hamiltonian function is an expression of the energy of the system, and it can be shown that the time rate of change of H (x, p) is exactly 0, i.e. that the energy of the system cannot change over time. But notice that this description of our system says nothing about the nature of the ‘particle’ involved, and nothing about the nature of the force which governs its motion. So a system with emergent properties could instantiate this description at the level of the emergent features. The radical emergentist regards as another matter altogether the issue of whether, or how, the constituents (entities or processes) of this system unite or combine to create the whole system. Thus we are free to regard energy conservation as a constraint only upon systems as such. For if radical emergentism is true, there is no way to understand the creation of complex systems entirely in fundamental physical terms. Simple, non-emergent systems will obey the principle of the conservation of energy and so too will complex systems with emergent properties. The transition from simple, non-emergent to complex, emergent systems is not explicable by basic physics and is thus not bound by principles restricted to fundamental physics. Although radical emergence denies the totality of the submergent domain, it is an open question whether we could allow strong supervenience within our radical emergence, since while non-T-totality implies non-T-temporal supervenience, it does not imply that strong supervenience fails. However, it is easy to argue that strong supervenience is at odds with radical emergence, for it would make the emergent features objectionably unexplanatory or, in a way, epiphenomenal. Consider that the lack of T-temporal supervenience entails, at least, that it is possible for two indiscernible T-states to have different outcomes. If these indiscernible T-states are the base for an emergent property then, given strong supervenience, they will subvene the very same emergent property. Therefore the

152

7 Emergence and Supervenience

emergent property will be unable to explain why there is divergence when you have T-indiscernible initial states. The lack of T-temporal supervenience is ‘brute’ (relative to U at least). If we want T-divergence to be explained by the emergent features then we cannot have strong supervenience. To make this argument slightly differently, the radical emergentists believed that the behaviour of complexes was in principle unpredictable from a knowledge— however complete—of the entities, relations and the laws governing the elementary submergent features. They nonetheless took it that the emergents were supposed to explain the divergence of the behaviour of the complex from the behaviour of the complex as it would be if it were determined solely by submergent laws and states alone. But as noted, if we have strong supervenience then the complexes would always subvene the same emergent feature (if any). If the behaviour of the complex was the same in all possible worlds then we would recover temporal supervenience and hence totality and we would have restored conservative emergence. The action of the complex would after all be predictable on the basis of the state of the elementary submergent features constituting the complex. Thus if the emergents are to explain the divergent behaviour of complexes, we cannot have strong supervenience. Although completely mysterious from the point of view of fundamental physics, the emergentists thought that the emergence of high-level features was nonetheless a part of the natural order. Once we know that, for example, a particular chemical property arises from the combination of certain basic physical entities, we can infer that this chemical property will arise whenever we have this physical combination. As Broad puts it: ‘No doubt the properties of silver-chloride are completely determined by those of silver and of chlorine; in the sense that whenever you have a whole composed of these two elements in certain proportions and relations you have something with the characteristic properties of silver-chloride’ (Broad 1925, p. 64). But this relation is ‘a law which could have been discovered only by studying samples of silver-chloride itself, and which can be extended inductively only to other samples of the same substance’ (Broad 1925, p. 65, my emphasis). Thus it seems that since the emergence is not a product of T-laws acting by themselves, there are T-worlds that differ with respect to the emergents attendant upon the same T-state (this is a variation that does not violate any T-laws). But at the same time, within these worlds we can generalize the emergence of emergent features across all intra-world instances of indiscernible T-states. This is an exact statement of a claim of weak supervenience of the emergent features upon T-features. The situation is illustrated in Fig. 7.5. And we can then define radical emergence as follows: D18. U is radically emergent from T = U weakly supervenes upon T, where T is a non-total theory. There is an interesting and perhaps rather attractive symmetry of causation in radical emergence that is lacking in doctrines that espouse totality, and endorse conservative emergence. The failure of totality under radical emergence is explicable in terms of a very strong form of ‘top-down’ causation. Totality will fail when complex systems fail to act in the ways they should act if their behaviour was entirely generated by the interactions of their constituents according to the fundamental laws governing

7.6 Varieties of Emergence

153

Fig. 7.5 Radical emergence and weak supervenience

those constituents and their interactions. Our label for such failure is divergence, so we can say, in short, that divergence is, or ought to be, explicable by top-down causation. Now, as noted above, the divergence of complex systems as described by high-level theories is commonplace, and such divergence is explicable by bottom-up causation; we expect that high-level generalizations will fail because of intrusions ‘from below’ of effects stemming from lower-level processes or structures, as when a computer program outputs erroneous data because of a cosmic ray hit on a memory chip or a person contracts cancer because an ultra-violet photon has damaged some DNA in one of their constituent cells. Radical emergence entails that there will be exactly analogous intrusions ‘from above’ as well: genuine as opposed to the merely apparent—or at least entirely explicable in low-level terms—top-down causation found in total theories. When there is radical emergence complex systems, described in terms of low-level theory, will suffer from effects stemming from higher-level processes or structures, effects which are simply not predictable solely from the low-level state of the systems and not fully determined by them (i.e. not determined with the strength of strong supervenience). Another interesting feature of radical emergence is that it tends to conspire to give an illusion of totality. That is, radical emergence of U from T entails weak T-temporal supervenience (up to intrinsic randomness of T). Thus, within a world, T-complexes that are indiscernible all act exactly the same (or, at least, generate the same behavioural statistics). Such a world could ‘look’ like it was T-total and encourage the search for a total T-theory. A rather bizarre further highly speculative possibility is that such a ‘total’ theory could perhaps, given sufficient ingenuity, be found despite the existence of genuine radical emergence. The theory would be false, but not testable. One warning sign of such a situation might be the multiplication beyond plausibility of potential energy fields (of the sort discussed above) required to handle multi-component interaction. More likely, the very complexity of those T-systems in which radical emergence might be found would rule out any test of

154

7 Emergence and Supervenience

emergence. The systems of interest would just be too far from the T-constituents for any calculation based solely upon fundamental T-laws of how they should behave to be feasible. That is, of course, the situation we are in and shall remain in. The issue of testability could become more contentious if it should turn out that the mathematical details of our best fundamental theory rule out not only analytic solutions of critical equations (a situation we are already in, as discussed in Chaps. 5 and 6) but also simulatability.28 It is worth remembering that the totality of physics is not practically testable for the simple reason that the instruments used in physical experimentation are themselves highly complex physical entities for which the hypothesis of radical emergence would, technically, have to be ruled out. The discovery and verification of properties of the most basic physical entities are the very ones that require the most complex instruments, such as particle accelerator complexes, as well as the extremely long historical chains of experimental inference which necessarily involve myriads of highly complex instruments. If it should turn out that certain complex and actual physical systems required for the testing of basic theory are in principle unpredictable because of certain mathematical limitations then it may be that the totality of physics is simply not a testable hypothesis at all.29 The contrast between conservative and radical emergence can also be expressed in a familiar theological metaphor. Imagine God creating a world. He decides it shall be made of, say, quarks and leptons and a few bosons that, in themselves, obey certain laws. But He has a choice about whether His new world shall be total (relative to these elementary constituents) or not. That is, He must decide whether or not to impose serious laws of emergence ‘on top of’ the properties of the basic entities. Either way, a world appears, but the worlds are different. Which world are we in? It is impossible to tell by casual inspection and perhaps impossible to tell by any experiment, no matter how idealized. Thus it may be that radical emergentism cannot be ruled out by any empirical test whatsoever, and thus it may be that we live in a world of radical emergence.

Chapter 8

Generalized Epiphenomenalism

8.1 Universal Physical Resolution I want now to turn to the question of how we should understand the status of conservatively emergent phenomena. This is important because, as I have argued, the natural understanding of both the structure of current physical science and its evident goals suggest that the only acceptable form of emergence will be conservative or epistemological. I aim to show that a common and plausible interpretation of what science tells us about the fundamental structure of the world—the ‘scientific picture of the world’ or SPW for short—leads to what I’ll call ‘generalized epiphenomenalism’, which is the view that the only features of the world that possess genuine causal efficacy are fundamental physical features. I think that generalized epiphenomenalism follows pretty straightforwardly from the SPW. At first, it might seem that generalized epiphenomenalism is fairly innocuous, since its threat is too diffuse to provoke traditional worries such as those about the putative epiphenomenal nature of mental states.1 If mental states are epiphenomenal only in the same sense that the supposed powers of hurricanes, psychedelic drugs or hydrogen bombs are epiphenomenal, then probably there is not much to worry about in the epiphenomenalism of the mental. I agree that the epiphenomenalism of hurricanes and the like is manageable, but it will turn out that ensuring this manageability requires that mental states have an ontological status fundamentally different from that of hurricanes, drugs and bombs, a status that is in fact inconsistent with the SPW. So I’ll argue that generalized epiphenomenalism does have some seriously worrying consequences after all. The SPW takes as its starting point the modern naturalistic conviction that the basic structure of the world can be discovered by scientific investigation and that there is no ground for positing a metaphysical understanding of the world distinct from a scientific understanding (a slogan: fundamental science is metaphysics with numbers). As discussed in Chap. 7, three interlocking features seem of central importance to the SPW: completeness, closure and resolution. Focusing on physical science, completeness is the doctrine that everything in the world is physical and as such abides by W. Seager, Natural Fabrications, The Frontiers Collection, DOI: 10.1007/978-3-642-29599-7_8, © Springer-Verlag Berlin Heidelberg 2012

155

156

8 Generalized Epiphenomenalism

closure and resolution. Closure entails that there are no ‘outside forces’—everything that happens, happens in accordance with fundamental physical laws so as to comply with resolution. Resolution requires that every process or object be resolvable into elementary constituents which are, by completeness, physical and whose abidance with laws governing these constituents leads to closure.2 Take anything you like: a galaxy, a person, a flounder, an atom, an economy. It seems that anything can be resolved into the fundamental physical constituents, processes and events which determine its activity. Indeed, our best theory of the creation of the universe, as outlined in Chap. 2, maintains that at very early times after the big bang the universe was quite literally resolved into its elementary constituents. At that time the universe consisted of an extremely hot, highly active ‘sea’ of quarks (that would later, after sufficient cooling, combine into more familiar composite particles such as protons and neutrons), leptons (including the electron needed for the later—after still more cooling—combination of protons and neutrons into elements, that is, chemical kinds) and elementary force exchange bosons. It is these which, to speak roughly and to use an evocative expression of Morgan, provide the ‘go’ of the universe, driving it from state to state.3 Completeness, closure and resolution and their inter-relations are concisely expressed in the startling thought that the universe is ‘running’ entirely and solely upon the interactions of these elementary constituents no less today than when it was 10−37 s old. It is crucial to emphasize that the SPW is a metaphysical, not an epistemological doctrine. It does not concern itself with how or whether we could understand everything in terms of full resolution. In fact, such understanding is quite impossible, for reasons of complexity (of various sorts) that the SPW itself can spell out. Innumerable immensely difficult questions arise at every stage of resolution,4 and there is no practical prospect whatsoever of knowing the full details of the physical resolution of anything much more complex than even such an apparently simple object as a proton. The metaphysical picture is nonetheless clear. And since the world has no need to know the details but just runs along because the details are the way they are, the problems we have understanding complex systems in terms of fundamental physics are quite irrelevant to the metaphysics of the SPW. We can extend the philosophical model of computational simulation used in Chap. 5 to better reveal how completeness, closure and resolution are supposed to work, and which might also serve as a test of one’s attitude towards the SPW.5 Call the model the superduper computer simulation thought experiment. It goes like this. Imagine the day when physics is complete. A theory is in place which unifies all the forces of nature in one self-consistent and empirically verified set of absolutely basic principles. Not too long ago there were some who saw this day as perhaps drawing near (e.g. Hawking 1988; Weinberg 1992). Optimism seems to have somewhat fallen off of late however. No matter. I am talking of a perhaps very distant possible future or world. Of course, the mere possession of this ‘theory of everything’ will not give us the ability to provide a complete explanation of everything: every event, process, occurrence and structure. Most things will be too remote from the basic theory to

8.1 Universal Physical Resolution

157

admit of explanation in its terms; even relatively small and simple systems will be far too complex to be intelligibly described in the final theory. But seeing as our imagined theory is fully developed and mathematically complete it will enable us to set up detailed computer simulations of physical systems. The range of practicable simulations will in fact be subject to the same constraints facing the explanatory use of the theory; the modeling of even very simple systems will require impossibly large amounts of computational resources. Nonetheless, possession of a computational implementation of our final theory would be immensely useful. Real versions of something very like my imaginary scenario now exist and are already fruitful. For example, there are computer models of quantum chromodynamics that can compute the theoretically predicted masses of various sub-atomic particles in terms of their constituent quarks (see Dürr et al. 2008; for a more popular presentation of an earlier calculation see Weingarten 1996). The looming problem of computational intractability is all too evident, for realizing these calculations required, in the 1996 attempt, the development of special mathematical techniques, the assembling of a dedicated parallel supercomputer specially designed for the necessary sorts of calculations (a computer capable of eleven billion arithmetical operations per second) and roughly a year of continuous computing. Weingarten reports that a special 2-year calculation revealed the existence of a previously unrecognized particle, whose existence could be verified by examining past records from particle accelerator experiments. The later effort reported in Dürr et al. (2008) had access to vastly more powerful computational resources; their computer could crank out two hundred thousand billion operations per second but since their modeling was much more detailed they still required a year’s worth of computer time. Modeling the interactions of particles would be a much more challenging task, suggesting to the imagination computational projects analogous to the construction of medieval cathedrals, involving thousands of workers for many decades.6

8.2 Superduper Simulation Now I want to introduce a thought experiment that flatly ignores the inevitably insuperable problems of computational reality. Imagine a computer model of the final physical theory which has no computational limits (we can deploy as much memory as we like and compute for as long, or as fast, as we like). Further, imagine that detailed specifications of the basic physical configuration of any system, at any time, in terms appropriate for the final theory, are available. If the configuration of any physical system is specified as input then the output configuration of the system, for any later time, can be calculated (and appropriately displayed). If the final theory should turn out to be non-deterministic (unlikely as that may seem, given that quantum mechanics, which seems to form the basis of any final physics we can now envisage, provides for the completely deterministic evolution of the wave function of any system7 ) then we can permit multiple simulations to run simultaneously, thus to duplicate the statistics to be found in the real world. Now, there is nothing

158

8 Generalized Epiphenomenalism

incoherent in the idea of an absolutely perfect simulation. In fact, we might have some of them in physics already. The Kerr equations for rotating black holes are (if the general theory of relativity is true), absolutely perfect models of these strange objects. Recall Chandrasekhar’s confession that ‘in my entire scientific life…the most shattering experience has been the realization that an exact solution of Einstein’s equations of general relativity, discovered by the New Zealand mathematician Roy Kerr, provides the absolutely exact representation of untold numbers of massive black holes that populate the universe’ (as quoted in Begelman and Rees 1996, p. 188). In fact, we know that Chandrasekhar should not have been quite so shattered, for general relativity does not provide the complete description of any actual black hole. There are known quantum effects, such as Hawking radiation, and presumably only the long hoped for über-theory which will unite the realms of relativity and the quantum will provide the ‘exact’ representation of the objects we call black holes. However, even if certain physical systems allow of perfect simulation, we are not so lucky with the rest of the world in general, and so even within our dream certain approximations in the input configurations will have to be allowed. We cannot input field values for every point of space-time and it is conceivable that some configurations require an infinite amount of information for their specification if, to give one example, certain parameters take on irrational values which never cancel out during calculation. Let us therefore imagine that we can input specifications of whatever precision we like, to allow for modeling the system for whatever time we like, to whatever level of accuracy we desire. Even though it is not physically realizable, I think the idea of such a computer program as a thought experiment is perfectly well defined.8 So, let us imagine a computer simulation of a part of the world. Step one is to restrict our attention to something we call ‘simple’—a bob on a spring on the moon say. The simulation covers a restricted region of space and time (though the programmer would have to set up ‘boundary conditions’ that represent the influence of the rest of the world), and must be defined solely in terms of the values of fundamental physical attributes over that region. The programmer is not allowed to work with gross parameters such as the mass of the bob or the stiffness of the spring, or the gravitational force of the moon, but must write her code in terms of the really basic physical entities involved. (It might help to imagine the code written in terms of the properties of the atoms of the pendulum, its support structure and moon, though these are themselves not really physically basic.) The SPW predicts that the output of this computer simulation, appropriately displayed, would reveal a bob bouncing up and down, suspended above the lunar surface. Step two is to up the ante. Now imagine a simulation of a more complex situation, for example a father and child washing their dog, in their backyard on a lovely sunny day. Do you think the simulation would mimic the actual events? I venture to maintain that our current understanding of the structure of the world very strongly suggests that such a simulation—if only it were possible—would ‘re-generate’ both the action of the pendulum and the behaviour of the father, child and dog (along with tub, water, water droplets, soap, sunlight, etc.).

8.2 Superduper Simulation

159

Although something of a digression, it is worth considering the details of such simulations a little more closely. The thought experiment is of course outrageously idealized. We assume unlimited (but finite) memory and allow unlimited (but finite) processing time (or, indefinitely, but finitely, fast computing machinery). Even relative to such generous constraints, there are questions about the ‘feasibility’ of such simulations in general. For example, many extant theories can allow mathematical singularities to arise (as in the formation of a black hole) or suffer from other mathematical pathologies. Technical problems also abound which are exacerbated by the need for computationally efficient algorithms (these are not a concern for the thought experiment of course). While some of the mathematical models of ideal physical systems have exact solutions (e.g. the frictionless pendulum) almost all such models lack exact analytic solutions and thus have to be tackled by numerical approximation. The field of applied mathematics that deals with numerical approximations of differential equations is exceedingly complex, far transcending the elementary discussion given above in Chap. 6; for an overview see Thijssen 1999). It is known that not all systems can be simulated if we require the simulation to obey certain otherwise apparently desirable mathematical constraints corresponding to physical traits such as energy conservation which govern the real world process being simulated (see Ge and Marsden 1988; Umeno 1997). The general question whether any given theory is simulatable either fully, partially or for all physically realistic models cannot be answered apart from a detailed study of the theory in question.9 As discussed in Chap. 5, it is perhaps also possible that nature transcends the ‘Turing Limit’—that is, can only be described in terms of uncomputable functions. This adds another layer of uncertainty about whether a digital computer (equivalent to a universal Turing machine) can simulate all the mathematical functions which correctly and completely describe nature. Nature might use, so to speak, uncomputable functions in getting the world to move from state to state. A very simple example of an uncomputable function is the function E(x, y) defined as E(x, y) = 1 if x = y and E(x, y) = 0 if x = y. Even if x and y range over computable real numbers, E is not computable (since you’d have to check an infinite number of digits to verify the identity of two real numbers). If nature ‘uses’ real numbers in ways that make the evolution of some system depend closely (e.g. chaotically) on the exact value of E then our simulation may turn out to be impossible. But notice that we don’t necessarily have to restrict ourselves to Turing machines (or the souped up equivalents we all have on our desks and in our laboratories) or purely digital simulations of continuous differential equations. If nature works ‘beyond the Turing limit’ then presumably we can build computers that exploit that region of physics. One example of such an ‘enhanced’ computer that has been studied by theoretically minded computer scientists is what Turing himself called Oracle machines (see above Chap. 5, n. 11). These are Turing machines that can, at well defined points during computation, call upon an ‘oracle’ to provide the output of an uncomputable function. No one knows for certain that oracle machines cannot be built, but if nature ‘uses’ uncomputable functions then nothing seems to prevent us from in principle incorporating such aspects of nature into our computing machinery. For the argument advanced in this chapter, the crucial constraint involves resolution.

160

8 Generalized Epiphenomenalism

The simulation must operate solely over the basic constituents of nature and only over the properties of those constituents. The importance of this restriction has to do with the notion of emergence, to be discussed immediately below. The simulation thought experiment can be used to provide a simple and clear characterization of emergence. An emergent is anything that is not coded into the simulation. Thus a thunderstorm is an emergent entity since, I take it, we would not need, in addition to coding in the quarks, leptons and bosons and their properties, to add thunderstorms as such to our simulation code. Temperature would be an example of an emergent property (thermodynamical properties in general would be emergent properties), as would be such features as ‘being hydrogen’ (chemical properties in general would be emergent properties), ‘being alive’ (biological properties would in general be emergent properties), etc. This idea is hardly new. Dennett has a famous example of a chess playing computer that ‘liked to get its queen out early’ (see Dennett 1978a). There is no ‘get-the-queen-out-early’ code in the chess program. Another example is Douglas Hofstadter’s amusing tale of desperate computer users requesting that the ‘thrashing number’ be raised (thrashing being when the user load on a server overwhelms its multitasking capacity and memory causing it to descend into seeming paralysis). There is of course no such value written anywhere in the machine’s code that is the ‘thrashing number’—this value just emerges (see Hofstadter 1986).

8.3 Towards Epiphenomenalism As we have seen, although the founders of the philosophical doctrine of emergentism (Mill, Lewes, Morgan, Alexander, Broad) would have agreed with my examples as examples of emergent features, they wanted more from their emergents. They wanted their emergents to have an active role in the world, not a merely passive or derivative role. That is, they believed in radical emergence in addition to the unexceptionable conservative emergence. The world, according to these emergentists, goes differently because of the presence of emergents. It does not behave simply in the way it would if the properties of the basic physical constituents of the world were solely efficacious and the emergents were simply ‘conglomerations’ of basic physical entities obeying the basic laws of physics. In deference to their desires, we have distinguished a conservative, or epistemological, from a radical, or ontological, emergence. In terms of our simulation thought experiment, radical emergence is the claim that, despite being an entirely accurate representation of the basic physical constituents of the world, the simulation will not render an accurate simulation of the development of the whole world as these basic constituents combine and interact to form ever more complex structures. That is, despite the accuracy of the simulation with regard to basic physical features, we will have to code into our simulation additional features that come into play only when certain combinations of the basic features appear.10 The emergentists’ favourite example of what they took to be a relatively straightforward, uncontroversial example of radical emergence was chemistry. They took

8.3 Towards Epiphenomenalism

161

it to be the case that the theory of the atom simply could not account for all the various chemical properties and chemical interactions found in the world. And they meant this to be a metaphysical claim. They were well aware of the difficulties of complexity which stood, and stand, in the way of fully understanding chemistry in terms of physics, or feasibly predicting chemical properties on the basis of basic physical properties. The emergentists were not merely noting these difficulties of complexity. They were denying that chemistry resolves into physics (or that chemical entities resolve into physical entities) so as to obey closure. As Broad put it, even a ‘mathematical archangel’ could not deduce chemistry from physics. I have put our simulation of basic physics in place of the angel, so the claim of radical emergentism is just that the simulation will not provide an accurate representation of distinctively chemical events. This is a kind of empirical claim. And though in a certain sense it is untestable, the development of quantum mechanics has severely undercut the case for chemistry being radically emergent. It now seems clear that chemical properties emerge from the properties of the basic physical constituents of atoms and molecules entirely in accord with closure and resolution. So-called ab initio methods in computational chemistry increasingly fund this confidence, as for example in this passage unwittingly directly opposed to Broad’s pronouncement: ab initio computations ‘are derived directly from theoretical principles with no inclusion of experimental data’ (Young 2001, p. 19). And that is the sign of conservative emergence. The claim that conservative emergence exhausts all the emergence there is in the world is simply the claim that all features not coded into the simulation are subject to resolution under closure. Of course, this is not to say that the way chemistry conservatively emerges from more basic physical features follows anything like Broad’s conception of the purely compositional, mechanistic growth of complexity. We know that the relation of emergence in this case (and many others) is much more interesting insofar as it depends on the peculiar features of quantum mechanics. It may be that chemistry provides an illustration of what Paul Humphreys calls ‘fusion emergence’ (see Humphreys 1997b and for a dissenting view with regard to chemistry see Manafu 2011). But, as discussed above in Chap. 6, fusion emergence—and quantum mechanics in general— does not seem to take us outside the boundaries of conservative emergence. Thus the SPW, with its endorsement of the totality of fundamental physics (that is, its completeness, closure and resolution) asserts that all emergence is conservative emergence. Now, does conservative emergence entail generalized epiphenomenalism? The threat is clear. Conservatively emergent features have no distinctive causal job to do; whatever they ‘do’ is done through the agency of the basic physical features that subvene them. But that does not directly entail epiphenomenalism. The SPW does not impugn the existence of conservatively emergent features so one might suspect that some notion of ‘supervenient causation’—perhaps the kind of weak efficacy granted in Chap. 7—will suffice to grant efficacy to these emergents. And let me emphasize yet again that conservatively emergent phenomena have indispensable explanatory jobs and there is no prospect of or desire for their elimination.

162

8 Generalized Epiphenomenalism

Nonetheless, I believe that generalized epiphenomenalism does follow from the SPW. There are at least three strong arguments for this conclusion, which I’ll label the economy argument, the screening-off argument and the abstraction argument.11

8.4 The Economy Argument I take it that, as mentioned above, causation is a metaphysical relation, and that in particular it is the relation that draws forth one event from another across time, or that determines one state as the outcome of a previous state, or that constrains nature to only certain sequences of states.12 This is no definition and I do not want to prejudge issues about the possibility of backwards causation or about whether the relata of the ‘causes’ relation are necessarily limited to events as opposed, for example, to states or objects. I want only to draw attention to the central idea that causation is the ‘go’ of the world (to recall the vivid expression of Morgan); it is causation that drives the world from state to state. The issue of concern to me is, so to speak, how much ‘go’ there is in the world or how widely it is distributed throughout the world. Morgan was clear about how widely he wanted to ‘spread the go’: ‘There are physico-chemical events, as such; there are vital or organic events, as such; there are conscious events, as such. All are integrated in the effective go of the system as a whole’ (Morgan 1923, p. 131). I think to the contrary that the SPW forces all the go of the world into a very confined space. The definite question I want to address is this: is there, from the point of view of the scientific metaphysics outlined above, any need to posit causal efficacy at any level above that of the fundamental physics, or is all of the ‘go’ lodged at the metaphysical root of the world? This question must be sharply distinguished from the question whether we need to deploy theories of (or descriptions of) levels of reality far higher than those described by fundamental physics in order to predict occurrences in the world, to explain what happens in the world and to understand or comprehend what is happening around us. I think it is obvious that we require high-level theories or descriptions for these essentially epistemic tasks. You are not going to understand why a square peg won’t fit in a round hole in terms of the fundamental physics governing the constituents and environs of peg and hole. But that by itself does not entail that we need to posit any causal efficacy to ‘square pegged-ness’ or ‘round holed-ness’. No less evident than the need for high-level descriptions to understand this relationship, is that the fundamental physics of the relevant constituents is all you need to ensure that, as a matter of fact, the square peg just won’t go into the round hole.13 The superduper computer simulation thought experiment is supposed to draw this to our attention. Imagine the fundamental physics simulation of peg approaching hole. There is no need to code into the simulation anything about squareness or roundness, or whether something is a peg and something else is a hole, or that the peg is moving towards the hole or anything else at a level of description above that of fundamental physics. Nonetheless the world of the simulation reveals that the peg won’t go through the hole. How can that be if there really is some kind of genuine

8.4 The Economy Argument

163

causal efficacy to the peg’s being square or the hole’s being round? It would seem reasonable to suppose that if you leave some genuine efficacy out of your simulation, it won’t manage to remain similar to the world where that missing efficacy resides and has its effects (recall our definition of ‘radical emergence’ versus ‘conservative emergence’ here). Leaving out of the simulation features that make a genuine causal contribution to the evolution of the world’s state ought to cause the simulation to drift out of synchronization with the real world. But, by our hypotheses of completeness, closure and resolution, no high-level features are ever needed to get our simulation to accurately duplicate the world. Of course, if you regard causation as a non-metaphysical relation and instead think of it as some kind of explanatory or fundamentally epistemic notion then I will happily grant you its existence in the high-level features. Further, I suppose, it is then actually missing from the low-level fundamental features that are too complex, particular and bound to specific contexts to explain things like why square pegs won’t go in round holes. It is not implausible to think that our commonsense notion of causation is rather unclear about the distinction between epistemology and metaphysics (and this confusion might account for much of the trouble we have making sense of such things as the ‘causal relevance’ of, for example, mental properties).14 But whatever the proper analysis of ‘causes’ may be, there remains the metaphysical question of where the ‘go’ of the world resides and how much of it has to be posited to get the world going the way it actually does go. What I mean can be clarified by a simple example which highlights the difference between the metaphysical and explanatory aspects of causation. We know that smoking cigarettes causes lung cancer. Although this is a correct statement of an explanatory causal relationship, it does not pinpoint the fundamental features of the world that bring about this relationship. In any particular case of cancer caused by smoking there will be some underlying events whose own causal relationships will account for the gross link we are interested in. In any particular case of cancer, there will have been a chemical change in the DNA in some cell which led to that cell becoming cancerous and where the change involved was itself conditioned by certain specific chemicals in cigarette smoke. This chemical process accounts for the observed causal link. Nor are such chemical processes immune to a similar explanation in terms of yet more fundamental causal relationships. Thus explanatory causation has an obviously hierarchical but also promiscuous structure or is a ‘multi-inter-level’ phenomenon. By calling explanatory causation ‘promiscuous’ I mean that causal explanations happily invoke relations between any levels in the hierarchy of natural structure.15 The hierarchical nature of causation is what underlies the hierarchical structure of nature which we discover through both the applicability of high level theories, via our conscious perception of the world and our commonsense categorizations. The existence of high level structure anything like we find in our world is not a metaphysical necessity. There seems nothing incoherent, for example, in the idea of a universe that consists of nothing but diffuse hydrogen gas. The existence of high level causal relations is crucial for the existence of high level structure in general, for merely random or arbitrary conjunctions of basic physical features are not likely to admit of a theoretical treatment of their own. Nor are they, in

164

8 Generalized Epiphenomenalism

general, perceptible as objects. Thus the set of all objects with mass between ten and twenty kg and positioned less than 10 km from me is not susceptible to a distinctive theoretical treatment as such (though of course the objects in this perfectly well defined set fall under a large number of distinct theories and common categories). Explanatory causal relations form their own high level theoretical structure, which explains why we can discover them within the domain of high level causation. The commonsense or ‘folk’ theory of causation involves a host of principles and rules of thumb, including maxims such as causes precede effects, association tends to imply causality or that, all else equal, a cause suffices to bring about its effect. These commonsense ideas can be refined without any need to inquire into the basic physical mechanisms underlying any particular level of causal relationship, especially via Mill’s famous methods (introduced in Mill 1843/1963, Bk. 3, Chap. 8). The methods are level independent, and while they are normally used to discover causal relations, we can look at them backwards, so to speak, to see a domain of causal relations from the point of view of some high level theory. For example, the theory of plate tectonics validates its ontology of crustal plates, mid-ocean rifts, subduction zones, etc. by showing how these things fit into a system of causal relations mappable via Mill’s methods.16 Naturally, we humans have invented our theories to enable explanation and prediction. A good theory is one which supports the discovery of explanatory causal relationships, by way of such methods as noted above. The reason we are interested in causal relations at all is that they provide a basis for understanding why events occur as they do, and offer the hope of our being able to predict and sometimes control events of particular interest to us. But of course we can ask whether, and to what extent, the vastly intricate systems of causal explanation we have developed and applied to the world with such evident success over the last few centuries reflect the ontologically basic causal ordering (or fundamental constraints) of the world. The answer cannot simply be ‘read off’ the success of our theorizing. We can identify two core features of the explanatory aspect of causation that will help to answer this question. Explanations and predictions are offered to satisfy human desires, both practical and more purely epistemic, and thus necessarily depend on our being able to comprehend the frameworks within which such explanations and predictions are offered. Causal explanation is also level-promiscuous, in the sense that any level of theory (or causal structure) can be appealed to in explanation, and inter-level connections can also be deployed. Thus, for example, in the explanation of atmospheric ozone depletion and its effects, many levels are invoked. These include atmospheric dynamics, human preferences for refrigeration, swimming and sun-bathing, the thermodynamics of compression, condensation and expansion of gases, fairly low level chemistry of the interaction of chlorine and oxygen, and still lower level physics in the account of why ultraviolet light appears in the solar spectrum along with its typical harmful effects.17 We can label these two features of explanatory causation (1) the comprehensibility condition and (2) the level promiscuity condition. True metaphysical causation is most certainly not subject to (1), but what about (2)? As we have already noted,

8.4 The Economy Argument

165

high level causal relations have at least some dependence upon lower level causal relations. A natural and important question to ask is whether this dependence is total or, in other words, whether high level causal structure supervenes upon low level structure. The characteristic claim here is that any two possible worlds that agree on their lowest level causal structure will also agree about all high level causal structure. As we have frequently emphasized, it is by no means obvious that such a supervenience claim is correct. Some philosophers have disputed it. Morgan’s (1923) emergent evolution and Broad’s (1925) emergentism both clearly denied the causal supervenience claim in favor of there being genuine inter-level causal relations. These emergentists postulated two classes of laws: intra- and inter-level laws (Broad called the latter ‘trans-ordinal laws’). The intra-level laws are more or less standard laws of nature which, from the explanatory point of view, would be confined to a single science. The inter-level laws, or at least some of these laws, are laws of emergence and dictate novel causal powers that come into being upon the creation of certain lower level structures. Both Broad and Morgan saw chemistry as a prime example of this kind of emergence and the confidence engendered by this supposedly uncontroversial example encouraged them to extend the domain of radical emergence. Unfortunately for them, the codification of quantum mechanics after 1925 provided strong evidence that chemical properties supervene on underlying physical features, and that chemistry does not require the postulation of any novel causal powers (see McLaughlin 1992). But as they saw it, chemical structures, such as H2 O, lawfully form from lower level entities but the properties of water are not merely the resultant of the causal powers of these atomic constituents and their low-level interactions. Rather, novel properties with distinctive causal powers emerge when hydrogen and oxygen come together. However, such ontological emergence is itself a lawful phenomenon expressing the ‘inbuilt’ inter-level structure of our universe. In terms of our characterization of ontological causation in terms of constraints upon allowable state sequences, the emergentists postulated that emergent features impose further or extra constraints beyond those imposed by the underlying features. Put in terms of our discussion in Chap. 7, radical emergents stand in a form of weak supervenience to their subvening features and processes. That is, the emergents can be different across possible worlds that are indiscernible with respect to the relevant underlying features. But while radical emergence is a coherent position, we have seen abundant evidence in Part I of this book why the SPW rejects it. Emergence is nonetheless an extremely important feature of the world, but it is limited to conservative or epistemological forms. Hierarchical and promiscuous causal explanation depends on a rich system of conservatively emergent features without which we would have no intelligible account of the structure and dynamics of the world. I cannot deny that the primary sense of the word ‘cause’ and its cognates may well be this epistemic or explanatory sense, but I also think that the metaphysical question about the ‘go’ of the world remains—and remains philosophically pressing. So, if you are inclined to think that causation is primarily an epistemological or explanatory relation, or that both metaphysical and epistemological notions jointly constitute our concept of causation, I won’t argue about the word. Define ‘kausation’

166

8 Generalized Epiphenomenalism

as the metaphysical relation between events (or whatever the appropriate relata may be) that drives the world forward or constrains state sequences. Our question then is whether high-level features have any kausal efficacy; the metaphysical question remains as pressing as ever. (In what follows, however, I generally keep to the ‘c’ spelling to spare the reader’s sensibilities.) We ought not to multiply entities beyond necessity. In the particular case of the metaphysical question of where causation works in the world and how much of it there is, we ought to posit the minimum amount, and the simplest nature, necessary to obtain the phenomena we seek to account for. The ‘phenomena’ are just the events that make up our world (at any level of description). The minimum amount of causation we need to posit is causation entirely restricted to the level of fundamental physics. This follows from closure, completeness and, especially, resolution. Fundamental physics (at the moment) suggests there are, currently active, four forces (weak, strong, electromagnetic, and gravitational) whose concerted exertions (within a backdrop of spacetime and quantum fields) generate all the variety we can observe in the world at large.18 Crudely speaking, our superduper simulation requires only the simulation of these forces (and fields) in spacetime to provide a total simulation of the world at every level. This is fairly obvious if we think of simulating the world when it was just a few nanoseconds old; completeness, closure and resolution entail that it is no less true of simulations of the current universe (or parts thereof). The SPW with conservative emergence sees the world as structured into roughly demarcated ‘levels’. It is enough that such demarcation be rough and imprecise because nothing of deep metaphysical import hangs on it. Broadly speaking, a level of reality exists when a family of properties forms a distinct system, which typically involves a set of lawful regularities, causal (not kausal) relations and the like and where the system is more or less autonomous, or relatively insulated against disturbances generated by properties outside the system. It is telling, however, that this autonomy is always subject to what I called in Chap. 7 intrusions from below. These levels are in essence what Dennett has called ‘patterns’ (see Dennett 1991). Patterns are structures, and relations amongst structures, that are visible from certain viewpoints. Only from a viewpoint which encompasses the nature, goal and rules of chess is a forced checkmate visible. It is from the viewpoint of chemical science that the systems of chemical kinds and affinities are apparent, and such patterns might not be visible from other viewpoints (even ones not too far distant, as for example Paracelsus’s vitalistic iatrochemistry which is ‘in the ballpark’ of our chemistry). But although they are familiar, patterns are somewhat odd. They inhabit a curious zone midway between, as it were, objectivity and subjectivity for patterns are there to be seen, but have no function if they are not seen.19 By the former, I mean that patterns are not just in the eye of the beholder; they are really in the world (it is not optional for us to decide that salt does or does not dissolve in water or that a checkmate is forced in two moves) and they provide us with an indispensably powerful explanatory and predictive grip upon the world. By the latter, I mean that the only role they have in the world is to help organize the experience of those conscious beings who invent concepts for them and then think in terms of these concepts. That is, although the world is rightly described as exemplifying a host of patterns, the world itself, so to

8.4 The Economy Argument

167

speak, has no use for them. In terms of our thought experiment again, high-level patterns do not need to be coded into the world-simulation in order to ensure the accuracy of the simulation, and this is just because it is the fundamental physical features of the world which organize the world into all the patterns it exemplifies and they do this all by themselves, with no help from ‘top-down’ causation. Some philosophers, such as Kim, attempt to ground the causal efficacy of reducible or functionally definable higher-order features via the claim that ‘the causal powers of an instance of a second-order property are identical with (or a subset of) the causal powers of the first-order realizer that is instantiated on that occasion’ (Kim 1998, p. 116). But this won’t work; first-order realizers are complexes of fundamental features and thus, according to my argument, have in themselves no causal efficacy. Everything they can do in the world is entirely the work of their own constituents. Realizers are not fundamental but are themselves patterns which are picked out by their relation to pre-existing high-level patterns (such as the elements of psychology, economics, geology or whatever). The economy argument shows that there is no need to suppose that realizers as such have any efficacy; if we imagine a world in which they lack efficacy the world proceeds just as well as an imagined world in which they do have efficacy, unless, of course, we enter the realm of radical emergence but we are here trying to understand the commitments of the SPW. Metaphysical economy then strongly suggests that we take our world to be the former world. In fact, realizers are in a sense worse off than the unrefined descriptions of high-level theory, for at least these latter have an explanatory role within their own theory whereas the realizers are epistemically inaccessible and explanatorily (as well as causally) impotent. We believe in them because we believe in completeness, closure and resolution but they are, in most cases, very remote from the features they are supposed to realize and can rarely take part in our explanatory projects. Doubtless there is a harmless sense of ‘top-down causation’ which is perfectly acceptable and appropriate for use within pattern-bound explanations. For example, researchers developing the scanning tunneling electron microscope were able to manipulate individual atoms to form an ultra-tiny billboard spelling out the name of their corporate sponsor, IBM (see Eigler and Schweizer 1990). Thus we can explain the location of a particular atom by reference to the intentions of the operator of a very macroscopic machine. But the SPW takes it that those very intentions are completely, albeit elusively, accommodated within a vastly intricate web of micro-states which, within their environment, ‘push’ the target atom to its final location. Intentions, like planets, animals and molecules, have no need to be specially written into the code of the world-simulation. Consider, for instance, the Coriolis force, which gunnery officers must take into account when computing the trajectory of long-range cannon shells. The general significance of the Coriolis force can be gathered from this resolutely realist passage from the 1998 edition of the Encyclopedia Britannica: The Coriolis effect has great significance in astrophysics and stellar dynamics, in which it is a controlling factor in the directions of rotation of sunspots. It is also significant in the earth sciences, especially meteorology, physical geology, and oceanography, in that the Earth is a rotating frame of reference, and motions over the surface of the Earth are subject to

168

8 Generalized Epiphenomenalism

acceleration from the force indicated. Thus, the Coriolis force figures prominently in studies of the dynamics of the atmosphere, in which it affects prevailing winds and the rotation of storms, and in the hydrosphere, in which it affects the rotation of the oceanic currents.

This force is a conservatively emergent property of the earth, or any other rotating system, of evident usefulness in a variety of high-level descriptions of the world. But in the context of assessing causal efficacy in terms of the fundamental physical features of the world, it is highly misleading to say that the Coriolis force causes diversions in, for example, a shell’s trajectory. At least, if we really thought there was such a force—hence with its own causal efficacy, the world would end up being a much stranger place than we had imagined. Just think of it: rotate a system and a brand new force magically appears out of nowhere, stop the rotation and the force instantly disappears. That is radical, brute emergence with a vengeance. Of course, there is no need to posit such a force (and it is called, by physicists if not engineers, a fictitious force, as is centrifugal force). The Coriolis phenomena are related to the underlying physical processes in a reasonably simple way—in fact simple enough for us to ‘directly’ comprehend, but, no matter the complexity, our imaginary computer model of any rotating system would naturally reveal the appearance of a Coriolis force. The Coriolis force can serve as a more general model for the relation between basic and high-level features of the world. The Coriolis force is an ‘artifact’ of the choice of a certain coordinate system. If you insist upon fixing your coordinates to the surface of the Earth, you will notice the Coriolis force. If you take a more natural, less geocentric, non-rotating, coordinate system as basic, the force will never appear (though artillery shells will of course still track a curved path across the surface of the Earth). In general, high-level features are ‘artifacts’ which arise from the selection of a particular mode of description. If, so to speak, we impose the ‘chemical coordinate system’ upon ourselves, we will find peculiar ‘chemical forces’ at work (or we will find chemical properties to be apparently efficacious). If we make the metaphysical choice of the more fundamental basic physical framework, these distinctively chemical efficacies will as it were disappear (though, of course, the phenomena chemists like to explain via ‘co-valent bonds’, ‘hydrogen bonds’, etc. will still occur). The fact that these phenomena result solely from fundamental physical activity suggests that no efficacy should be granted to chemical features as such (rather as the Coriolis force should be seen as an artifact arising from a certain point of view or choice of reference frame). Thus, the simplest and most natural interpretation of efficacy in the SPW restricts efficacy to the fundamental constituents of the world. They are by themselves able to generate all the variety to be found in the world. The high-level descriptions of the world we find so useful are just that: useful—indispensably useful even— epistemological aids to beings with a certain explanatory and predictive agenda who are smart enough to be able to pick out and/or formulate such descriptions. It would, for example, be insane to forbid gunnery officers or pilots from thinking in terms of the Coriolis force. But that cuts no metaphysical ice. The economy argument strikes me as a powerful reason to accept generalized epiphenomenalism. But there are other arguments as well.

8.5 The Screening Off Argument

169

8.5 The Screening Off Argument It is often difficult to tell the difference between a cause and ‘mere correlation’. Suppose we notice a correlation between, say, A and B, one in which B is a candidate for causing A. A probabilistic symptom of such a situation is where P(A | B) > P(A). That is, the presence of B increases the chances of obtaining A (as for example the presence of cigarette smoking increases the chances of lung cancer). One way to distinguish a mere correlate from a cause is the statistical property of ‘screening off’ (for a general discussion of the importance of this concept in issues of causality see Salmon 1984). Essentially, the idea is to see if one can discover a feature distinct from B that accounts for the correlation between A and B, and accounts for it in such a way as to undercut B’s claim to efficacy. We say that C screens off B from A whenever: (C1) P(A | C ∧ B) = P(A | C), and (C2) P(A | C ∧ B) = P(A | B) What the first condition asserts is that the ‘extra factor’ C destroys the statistical relevance of B on the occurrence of A. Once we take into account C, B no longer makes any difference to the probability of A happening. The second condition asserts that C makes a difference, relative to B, for the occurrence of A. We might say that, when (C1) and (C2) are met, C usurps the putative causal role of B. The test is at best good evidence for denying efficacy to B, for it seems possible that P(A | C ∧ B) might end up equal to P(A | C) just by accident. In such a case, B might well be non-efficacious but the screening off test could not reveal this (there would be a standoff with respect to the test). Nonetheless, the test is often effective. A classic example is an apparent link between cancer and consumption of coffee (this textbook example can be found all over but I have no idea whether anyone was ever fooled by this20 ). This spurious correlation resulted from not distinguishing coffee drinkers who smoke from those who do not smoke. Since coffee drinkers tend, more than non-drinkers, to be smokers a merely statistical correlation between coffee drinking and cancer arose. The screening off test reveals this since the statistics end up as follows: P(cancer | smoking ∧ coffee) = P(cancer | smoking) > P(cancer | coffee). The weakness of the test is nicely revealed in this example too, for it is evidently possible that absolutely every coffee drinker should be a smoker and vice versa which would falsify (C2) despite coffee’s harmlessness.21 The screening off test can be applied to the question of the causal efficacy of high-level features, with perhaps surprising results.22 Let’s begin with a toy example. Suppose I have a pair of dice. Let the low-level or micro features be the exact values of each die upon any throw (for example getting a 2 and a 6). A host of high-level or macro features can be defined as ‘patterns’ of low-level features, for example, if we take the sum of the individual outcomes, getting an even number (e.g. by rolling a 3 and a 3), getting a prime number, getting a result divisible by six, etc. Despite its simplicity the model has some suggestive attributes. There is obviously a relation

170

8 Generalized Epiphenomenalism

of supervenience of macro upon micro features, that is, micro features determine macro features. But at the same time there is multiple realizability (there are many micro-ways to roll a prime number for example). And there is a weak analogue of non-reducibility in the sense that there is only a disjunctive characterization of macro features in terms of micro features (it is the simplicity of the example that renders the disjunctions tractable and easily comprehensible—for contrast suppose we were rolling 1023 dice). The point of the model is that micro features can clearly screen off macro features from various outcomes. Consider the following probabilities. The probability of rolling a prime number is 5/12. The probability of rolling a prime given that one has rolled an odd number is 7/9. These are macro descriptions. Consider now one of the ways of rolling a prime number, say rolling a 3 and a 2. The probability of rolling a prime given a roll of a 3 and a 2 is, of course 1. Thus (C2) of the screening off relation is fulfilled. It is also evident that (C1) is met. Thus rolling a 3 and a 2 screens off rolling an odd number from obtaining a prime number. Since it is impossible to get an odd number except via some micro realization or other, and since each such micro realization will screen off the macro feature we should conclude that it is the micro features that carry ‘efficacy’. I need the preceding scare-quotes since there is, of course, no causation within my purely formal model, but that is not a serious defect. In fact, we could add genuine causation simply by imagining a ‘prime detector’ which scans the dice after they are rolled and outputs either a ‘1’ (for prime) or ‘0’ (for non-prime). If we assume for simplicity that the detector is perfectly reliable (probability of proper detection is 1) then the probabilities of the simple example carry over directly and we get genuine causal screening off of the macro features by the micro features. Notice that the probabilities of the outcomes contingent upon macro features are determined by some mathematical function of the probabilities of those outcomes contingent upon all the possible realizing micro features (in this case the function is simply the number of prime outcomes divided by the number of odd outcomes, 14/18). This function, as it were, destroys the detailed information available in the micro features, changing probabilities that are all 1 or 0 to the ‘smudged out’ values characteristic of the macro feature probabilities (in our example, a bunch of 1s and 0s transform into 7/9). We can apply the lessons of this simple model to more interesting cases. Consider whether basic physical features screen off chemical features. Take as an example, the chemical feature of acidity or ‘being an acid’. This is multiply realizable in quite distinct physical micro features (viz. the physical realization of HCl versus that of H2 SO4 ). It seems fairly clear that micro structure will screen off macro structure relative to the effects of acids. We might, for example, claim that substance S’s dissolving in X was caused by X’s acidity. But consider that there are acids too weak to dissolve S so the probability of dissolution is not 1, but each realization is an acid either strong enough to dissolve S or not, and each realization guarantees we have an acid, so screening off will occur in a way quite analogous to our simple dice example. It is important to remember in this discussion that we are concerned with the metaphysical facts about efficacy here, not the explanatory merits of the alternatives.

8.5 The Screening Off Argument

171

It is easy to imagine cases where the appeal to the macro features will be much more explanatorily useful and where in fact appeal to the micro features would be entirely opaque. That does not matter if the question is about the root of efficacy in the world. Of course, the most interesting case is that of mentality. The same lessons can be drawn. We might claim that Fred raised his arm because he believed he knew the answer. Of course, believing that one knows the answer does not guarantee that one will raise one’s arm. That is, the probability of any behaviour, given an explanatory mental state is not 1. But the micro structure in which beliefs are realized—the specific fundamental physical features of subject and (more or less local) environment—can plausibly be thought either to guarantee that Fred’s arm goes up or to guarantee that it does not (at the very least to provide probabilities different from those attendant upon the original psychological description). And since the realizing state by hypothesis subvenes the psychological state, we fulfill the two conditions of screening off. We expect that the probability of behaviour given a more or less gross psychological characterization is a mathematical function of the probabilities of that behaviour contingent upon the possible realizing states. These latter probabilities will, in general, be different as information is smudged out or lost in the macro psychological state relative to the determinate physical realization that will obtain in every instance of a psychological state. Yablo (1992) has pointed out that the debate on causation looks different from a point of view that takes the determinable and determinate distinction as its model of inter-level connection. To take a common example, if a bull gets mad because it sees a red cape, the causal efficacy of the redness of the cape does not seem—at first glance—to be undercut by noting that the cape must actually have been some determinate shade of red. Although I don’t think that it is very plausible to regard the physical realization states as determinates of a mentalistic determinable,23 the screening off argument seems to work against the causal efficacy of determinables versus their determinates. Regard the bull as a ‘red-detector’ (I hear that bulls are really colour blind, but let’s not spoil a traditional example with pedantry). Detectors are more or less efficient; that is the probability of detection varies with the shade presented. Thus the probability of the bull getting mad given that it is presented with red is not 1, and is in fact determined by the particular probabilities of getting mad given presentation of various shades. Thus perhaps the bull is very good at detecting crimson but weak at vermilion. Since both crimson and vermilion are shades of red, we will fulfill the two conditions for screening off. We ought to attribute efficacy to the determinates (the shades or red) over the determinable (red). The fact that bull, so to speak, doesn’t care about or even know anything about these determinates is irrelevant to assigning efficacy. The bull’s behaviour is governed by them, and correlations between colours and behaviours stem from the power of the determinates rather than the diffuse and derivative ‘power’ (as expressed in the probabilistic correlations) of the determinable. (Of course, we can go further, and demote the shades as well in favour of the ultimate, fundamental physical determinants of the bull’s behaviour.) As always, for the purposes of intelligible explanation the appeal to redness would usually trump appeal to the particular shade, which might in fact be positively misleading.

172

8 Generalized Epiphenomenalism

Another way to approach the screening off of macro by micro features is in terms of Dennett’s stances (Dennett 1971). Recall that Dennett outlines three stances, that is, viewpoints from which to make predictions or explanations of behaviour (broadly construed to include more than that of biological organisms). We have first the ‘intentional stance’ in which the ascription of, most basically, beliefs and desires according to dictates of rationality serves to predict and explain behaviour. The intentional stance is undoubtedly highly successful and widely applicable throughout, at least, the animal kingdom and certain subsets of the machine kingdom. But sometimes the application of the intentional stance to a system for which it is generally successful fails. If there is no intentional stance (i.e. psychological) re-interpretation that plausibly accounts for the failure we must descend to a lower stance, in the first instance to the ‘design stance’. The ‘design stance’ provides the sub-psychological details of the implementation of the system’s psychology, but in functional terms. (Dennett, along with most cognitive scientists, typically imagines the psychology is implemented by a system of ‘black boxes’ with a variety of sub-psychological functions, viz. short term memory, visual memory buffers, edge detectors, etc.) Certain features of the design of a system can account for why intentional explanations fail. For example, it may be that unavoidable resource constraints force the design to be less than fully rational. Any real world system will surely fail to deduce all the consequences of its current information, and the way it winnows out the wheat of useful information from the chaff of irrelevant noise will always be susceptible to failure or, frequently, exploitation by enemies in particular cases (as in the famous eye-spots found on the wings of certain moths). The details of how relevance is assigned may lead to bizarre, psychologically inexplicable behaviour in certain circumstances. A well known example is the digger wasp’s (Sphex ichneumoneus) seemingly purely mechanical, stereotypical and mandatory nest provisioning behaviour which cannot be explained in terms of rational psychology but, presumably, follows from features of its neurological ‘design’ (see Dennett 1984, pp. 10 ff.; for some kinder thoughts on Sphex see Radner and Radner 1989). But there will be failures of design stance predictions and explanations too. A random cosmic ray may chance to rewrite a bit in your computer’s memory leading to behaviour that is inexplicable from the design stance (even at the level of the individual memory chip there is a design stance, and the action of the cosmic ray forces us out of it). Intrusions of brute physical malfunction are by definition inexplicable from the design stance but can never be completely ‘designed away’. In such cases, we must descend to what Dennett calls the ‘physical stance’ to account for the failure of the system to abide by its design. The possibility of the failure of the intentional stance as well as the design stance will be reflected in the probabilities of behaviour contingent upon intentional states or design states. The physical stance—at the level of fundamental physical reality—will however provide the rock bottom probabilities of behaviour contingent upon particular physical realizations of either intentional or design states. In general, these probabilities will differ from those at the intentional or design level, for their maximal information content will be smudged out across the very large disjunctive fundamental physical characterizations of the intentional or

8.5 The Screening Off Argument

173

design level states. Thus the existence of and relationships between the three levels results in the screening off of high-level features by the low-level features in just the way I have outlined above. The low-level, physical stance features are what generate the correlations that show up at the higher levels. It is they that deserve to be granted efficacy. This screening off reveals that no efficacy needs to be assigned to the high-level features to obtain the correlations observed at those higher levels. Thus it seems that a good probabilistic test of efficacy further suggests that efficacy resides at the most basic level of physical reality. Yet another argument serves to reinforce this conclusion further.

8.6 The Abstraction Argument A good high-level description manages to explain and predict the behaviour of very complex physical systems while somehow ignoring most of their complexity. Perhaps it is not a necessary truth, not even a nomologically necessary truth, but only a lucky cosmic accident that fundamental physical features should conspire to constitute well behaved structures that obey laws which abstract away from the details of those very structures. But we live in a world where such behaviour is ubiquitous. That is what makes high-level theories (and theorizers like us) so much as possible. The way that high-level theory abstracts away from the physical details undercuts any claim that high-level features have causal efficacy, as opposed to explanatory or predictive usefulness. Begin with a simple and famous example. While developing his theory of universal gravitation, Newton postulated that every particle of matter attracts every other particle with a gravitational force that obeys an inverse-square relation. But it would be obviously impossible to calculate any behaviour resulting from gravitational forces if one faced an extreme version of the many-body problem before one could even begin. And, naively, it seems that one does face such a problem because the Earth, and all other celestial bodies for which we might be interested in computing orbits, are manifestly made of a very large number of material particles (evidently held together by gravity itself among other possible forces). However, the structure of the gravitational force law and the newly invented calculus enabled Newton to prove that a spherical body will have a gravitational field (outside of itself) exactly as if all the matter of the body were concentrated at a single point—the geometric centre of the body. This is a beautiful piece of mathematics and makes calculating the orbits of the planets at least possible if not easy. Newton could treat a planet (at least from the point of view of gross gravitational studies) as a single gravitating ‘point’, although it was not long before the non-spherical shape of the Earth had to be taken into account. He had developed a high-level description of the gravitational field of material bodies that abstracts away from the details of their material composition, to our great advantage. But would anyone want to deny that the efficacy of gravitation still resides entirely within the mass of each particle that makes up the body in question? Although Newton’s theorem would provide an irresistible shortcut for any real world simulation, our imaginary superduper simulation will reveal the effects of gravitation while blithely ignoring the fact that, so to speak, all

174

8 Generalized Epiphenomenalism

of a body’s gravity can be viewed as emanating from its geometric centre (even if, by chance, the body has a hollow centre and there is no matter at all at the ‘gravitational centre’). Such a pure mathematical abstraction cannot participate in the ‘go’ of the world however useful it might be in the business of organizing our view of the world. Another example to the same effect is that of the temperature and pressure of a gas. The reduction of thermodynamics to statistical mechanics is one of the glories of physical science and we can see it as the discovery of how a particular set of high-level features are grounded in lower level physical entities. We now know how temperature and pressure are realized in a gas. The temperature of a gas (at equilibrium) is the average kinetic energy of the particles which constitute the gas. The pressure of a gas is the average force per unit area which these particles exert on the gas’s container. But averages are mathematical abstractions that in themselves cannot cause anything at all. If one is inclined to think otherwise, consider this example: a demographer might say that wages will go up in the near future since the average family size fell twenty odd years ago (and so now relatively fewer new workers are available). There is not the slightest reason to think that ‘average family size’ can, let alone does, cause things although I think we easily understand the explanation to which such statistical shorthand points.24 By their very nature, pressure or temperature are not one whit less statistical ‘fictions’ than is average family size. The ascription of causal efficacy to, say, pressure is only a façon de parler, a useful shorthand for the genuine efficacy of the myriad of micro-events that constitute ‘pressure phenomena’. It is entirely correct to use the overworked phrase, and say that pressure is nothing but the concerted actions of the countless particles that make up a gas. Within our thought experimental simulation, pressure and temperature will emerge without having been coded into the simulation program. In this case, our high-level features have to be conservatively emergent by their very nature for they are simply—by definition— mathematical abstractions of certain properties of the low-level features which realize them. As such, no genuine efficacy can be granted to them.25 Now, we don’t know what sort of realization function will be appropriate for mental states. It may be that the way that thermodynamical properties emerge out of the molecular dance may have more relevance to psychology than mere metaphor. There are interesting, perhaps deep, analogies between thermodynamics and the dynamics of neural networks (see for example Churchland and Sejnowski 1992, Chap. 3; Rumelhart et al. 1986, Chap. 7), and if the latter underlie psychology then some psychological properties may be surprisingly closely analogous to thermodynamical properties. Be that as it may, whatever mathematical structure underlies the transition from neural to mental states, it will be a mathematical abstraction from the underlying properties. (And the underlying neural structures will in turn be an abstraction from the more fundamental physical structures that realize them.) Insofar as mental states are seen as such mathematical abstractions they cannot be granted causal efficacy. They will be mathematically convenient ways of thinking about the mass actions of the fundamental physical constituents, just as temperature is a mathematically convenient way to work with the mass action of the molecules that make up a gas.

Chapter 9

The Paradox of Consciousness

9.1 Mind Dependence The general structure of the SPW—most especially the properties of closure, completeness and resolution, plus the three arguments just advanced (which merely draw out or highlight certain consequences of the scientific picture)—makes a strong case for generalized epiphenomenalism. It remains to consider if the doctrine has any significant consequences for our view of the world and our place in it. I will argue that in the realm of consciousness there are dire consequences, ones that on the face of it will require a rethinking of the spw from the ground up. Given such a grandiose claim I need to make clear that my goal here is limited. The topic of consciousness is vast, already possessed of a staggeringly large literature ranging over a huge number of disciplines. My project falls within the purview of the philosophical problem of consciousness, the general outline of which encompasses two main difficulties: what is the nature of consciousness and how is it generated in the world. It would take us very far afield to survey and assess all the various theories of consciousness on offer, even if we never strayed beyond the bounds of philosophical accounts, and I will not attempt this here (for an effort in this direction see Seager 1999). The second question of how consciousness is generated does of course link to the topic of emergence, but my discussion is limited with respect to this issue as well. The ‘generation problem’ has received intensive and extensive investigation over the last fifty years and remains completely unresolved (for a now classic articulation of the problem see Chalmers 1996, especially Chaps. 1–4).1 I will not tackle the generation problem here but rather I want to focus on a particular implication of the spw for consciousness which yields a problem connected to but distinct from the more general problem of generation. This problem begins with generalized epiphenomenalism. The spectre of classical epiphenomenalism about consciousness is frightening because it turns us into mere spectators, utterly unable to affect the world no matter how much we might desire to. Classical epiphenomenalism makes our position in the world analogous to that of someone on a ride in Disneyland, able to watch what’s W. Seager, Natural Fabrications, The Frontiers Collection, DOI: 10.1007/978-3-642-29599-7_9, © Springer-Verlag Berlin Heidelberg 2012

175

176

9 The Paradox of Consciousness

happening and feel real fear or elation in accord with those events, but unable to get out of the vehicle in which they are inexorably carried along and grapple with the world outside. Generalized epiphenomenalism does not seem quite so frightening. It grants no more, but also no less, efficacy to our mental states than it does to any other high-level feature of the world. We can understand perfectly well the place of high-level phenomena in the scientific picture. Such phenomena are integrated into the causal structure of the world in virtue of their being realized by structures that are indisputably efficacious (structures that really are the ‘go’ of the world). So even if high-level features are, so to speak, metaphysically epiphenomenal, we can still coherently make a distinction between genuine and merely apparent causes at high levels of descriptions. We discover that coffee drinking does not cause cancer, but not because it, like every other high-level feature, is epiphenomenal, but because the realizers of coffee drinking do not drive the world into a state which can be given the high-level description: cancer. We find the opposite story for smoking. Try to imagine the chances of success of Big Tobacco executives arguing that generalized epiphenomenalism shows that smoking cannot cause cancer. The SPW is austerely beautiful, and seems to find everything it needs to construct the entire structure of the world, in all its complexity, in a handful of elementary features that constitute those structures and drive their interactions entirely ‘from below’. And it seems that just as high-level theories provide useful and in fact indispensable ‘abbreviations’ for certain well disciplined complexes of elementary aspects of the world, so too we could speak of ‘high-level causation’ as a useful abbreviation for the twin stories of realization and basic causation; high level efficacy would be more or less as efficacy is defined in Chap. 7. That would be a happy ending to my story, if it was the end. Unfortunately it is not. It is impossible to regard all mental states as high-level features of the world in the same way that chemical, geological, etc. states are high-level features. Thus it is impossible to regard the epiphenomenalism of the mental as unproblematic in the way I have conceded that the epiphenomenalism of high-level features in general is unproblematic. The reason, baldly stated, is that the high-level features for which generalized epiphenomenalism can be regarded as a harmless metaphysical curiosity are, in a certain sense, mind-dependent.2 High-level features are elements of patterns, and patterns are, so to speak, invisible to the world. Their relationships, including those we like to call relations of high-level cause and effect, are only visible, sometimes literally visible, to conscious creatures possessed of certain ways of looking at the world, and they serve only as aids to explanation, prediction and understanding for those beings, rather than the metaphysical ‘go’ of the world. Can that be right? And, if it is right, what does that tell us about the world and the place of consciousnesses within it? Mind dependence comes in grades. We might define the absolute mind dependence of certain entities as requiring that if there were no minds there would be no such features. For example, money is absolutely mind dependent—without minds there could be no money. However, most of the high level features of reality which the SPW admits as conservatively emergent are not absolutely mind dependent. It is

9.1 Mind Dependence

177

not the case that a world devoid of mind would necessarily fail to manifest features which could be recognized as high level features to appropriately prepared minds. According to the SPW, it is very likely that there was a time when there were no minds. Though it is impossible to say when the first minds appeared in the universe, the early universe of the SPW (maybe the first half-billion years or so) seems pretty clearly to preclude the kinds of organization necessary to underpin thought and experience. It would be extremely odd to say that chemical processes were not occurring at that time. While true, this does not entirely undercut the mind-dependency of chemical processes as high-level features. For to say that chemical processes were occurring before the advent of mind is just to say that the world was running in ways such that the application of chemical concepts would have been explanatorily fruitful (had there been minds) before minds had in fact appeared on the scene. The world had, of course, no sense of chemistry and no need of it just because chemistry is, by its high-level nature, epiphenomenal. To put the point another way, it is impossible to understand the nature of high-level features without a prior understanding of the epistemic needs and goals of conscious thinkers. They are the sort of thing partly constituted out of those epistemic needs and goals. Consider an example. The stars visible from the Earth (say from the northern hemisphere) form literal patterns which are apparent to anyone (a little preparation or guidance helps one to see the patterns favoured by one’s own culture). These constellations are useful in a variety of ways, of which the most basic, and ancient, is probably navigation. It is, I hope, abundantly clear that these patterns are minddependent, even though the configurations of the stars might have been the same as it has been throughout recent human history at any time in the past, even long before any consciousness of any kind was around to notice the stars (as a matter of fact, on a long time scale constellations are relatively ephemeral but that’s irrelevant to the example). One might object that constellations aren’t ‘real’. But why not? In support of their unreality, the most natural insult to heap upon the constellations is that they don’t do anything except via the mediation of mind. This will not distinguish chemistry from constellations; the point of generalized epiphenomenalism is to show that highlevel features are uniformly without causal efficacy as such. Chemistry only looks efficacious from the point of view of a mind prepared to notice and exploit ‘chemical patterns’. In the loose sense countenanced above, in which efficacy is granted to a high-level feature via the efficacy of its low-level realizers, the constellations again come out on a par with chemistry. As in chemistry, the low-level realizers of the constellations have lots of effects, including effects on (the realizers of) human senses. One might complain that the patterns of the constellations are purely arbitrary. But that is just not true. There are natural groupings visible to all and recognized by diverse cultures (albeit with very different interpretations and divergence about exact boundaries). Ancient Europeans and the Iroquois of North America both noted the ‘big dipper’ and pictured it as a celestial bear. The sort of mind dependence that such conservatively emergent high level features possess is what might be called epistemic mind dependence. No one disputes that high level features are indispensable to our understanding of the world. The issue is

178

9 The Paradox of Consciousness

whether high level features are, so to speak, part of nature’s way of structuring the dynamics of the world. So far as the SPW can discern, the drivers of the world are all, and only, the fundamental features. So while it would be wrong simply to say that high level features are absolutely mind dependent it is correct to say that high level features play no role in the world except as apprehended by minds. Until and unless some high level feature is taken up in conscious perception and thought by some cognitive agent, that high level feature stands as a merely potential epistemological resource. To paraphrase Kronecker’s famous remark: God created the particles, all else is the work of man. To use another popular philosophical metaphor, the SPW allows that nature has joints to carve, but there are few of them and they are all down at the fundamental level. Once basic science has done the carving, we can make whatever sandwiches strike our fancy, aesthetic or utilitarian as the case may be, constrained always but only by the nature of the fundamental properties and relations.3 Consider for example the sad story of Pluto, recently demoted from planethood to become a mere ‘dwarf planet’. It is entirely up to us what counts as a planet, though we are obviously constrained by a host of real considerations of utility, elegance and coherence in our decision making. There is no Platonic Form of Planet against which we seek to assess Pluto independent of such desiderata.

9.2 Abstractive Explanatory Aids What is distinctive about high level features is their explanatory salience. High level features are those combinations of low level features that happen to fit into fruitful theoretical schemes. Fruitful, that is, to some intelligence which aims to predict, explain and understand the world. Generally and abstractly, the best of the high level features fall under systems of concepts to which Mill’s methods and their more sophisticated statistical descendants can be applied to outline high level causal relationships. Another important feature of the best of the high level features is that they conform to the strictures laid out by Ian Hacking which delineate a ‘natural kind’. These demand that kind ascription will with high reliability indicate possession of a range of other properties or suitability for various human purposes, both quotidian and more strictly scientific. A caveat is that the reliability of natural kind ascription is not conditional on the recognition of the ‘social role’ of the kind itself, as in the example of money, which in general respects admirably meets the reliability condition (see Hacking 1991, especially pp. 116 ff.). It is possible that some people somehow fail to appreciate the nature of social kinds such as money and may then come to regard monetary value as intrinsic or natural. We might even imagine a group that begins to exchange tokens as proxy for ‘directly’ valued goods and does so via something like a conditioned reflex or mere habit so that, for them, in some sense, money would be akin to a natural kind. Interestingly, chimpanzees can be trained to use plastic tokens

9.2 Abstractive Explanatory Aids

179

in something like this way without, presumably, fully understanding what is going on (see Matsuzawa 2001). Crucially, they do not try to eat the tokens. The SPW denies that there is any deep distinction between sorts of high level features save ones that serve the epistemic purposes of classification and manipulation that underlies the whole system of high level concept application. It has no problem with their being meta-high level features. As a fascinating aside to our concerns, Hacking also points out Peirce’s criticism of Mill’s definition of ‘real kinds’ which depends on the idea that the real kinds share innumerable novel properties unpredictable from what is known about the kind. Peirce, in his entry on kinds in Baldwin’s Dictionary of Philosophy and Psychology (1901) notes that the whole enterprise of science is contrary to Mill’s notion: ‘Mill says that if the common properties of a class thus follow from a small number of primary characters, “which, as the phrase is, account for all the rest”, it is not a real kind. He does not remark, that the man of science is bent upon ultimately thus accounting for each and every property he studies’ (as quoted in Hacking 1991, p. 119). The SPW sides with Peirce in denying the existence of any of Mill’s real kinds. With respect to Peirce’s criticism of Mill however, I think Peirce may be missing Mill’s radical emergentism. For Mill the inexhaustible set of properties associated with a kind could be linked to the primary properties, or fundamental properties, of the kind by what he called ‘heteropathic laws’, which are essentially the rules by which radical emergence, as defined in Chap. 7, is supposed to operate (see Mill 1843/1963, Bk. 3, Chap. 6). As we have emphasized, high level features emerge because of the propensity of fundamental features to aggregate into ‘self-sustaining’ systems.4 Crucial to aggregation are the fundamental laws of nature and the conditions under which these laws are operating. Thus very shortly after the big bang, although the laws of nature permitted aggregation, the conditions then extant prevented it (essentially, the temperature was so high that no stable aggregation could persist). In our world, these conditions ameliorated over time so that ordinary matter, chemical structures, stars and galaxies, planets and eventually life could emerge. But it is easy to imagine conditions that forever preclude the creation of many high level features (see Barrow and Tipler (1988) for a host of ‘cosmological’ constraints upon the emergence of high level features relevant to our own existence). There is a huge variety in the sorts of high level features available for thinking, working and theorizing. One class is important for my purposes because its relation to the fundamental features of the world is much more abstract and perhaps in a way deeper. The preconditions of the emergence of these features are not tied so closely to the details of either the laws or the initial conditions of the worlds in which they appear, though to be sure each has necessary conditions. Because of this, these high level features possess a high degree of multiple realizability. The high level features I am thinking of are those associated with theories—or folk theories—of thermodynamics, information theory, evolutionary biology, and folk psychology (the first two, and perhaps also the last two, are extremely closely related). Consideration of these theories will both reveal some core features of emergence, but will also provide yet more, and quite fundamental, evidence that within the SPW such emergence is always and only conservative emergence.

180

9 The Paradox of Consciousness

Consider how thermodynamical laws emerge out of statistical mechanics, especially the second law. Almost any system of interacting particles will obey the second law simply because of the probability distribution of system states corresponding to particular entropy values. This holds for water molecules in a bathtub or colored balls being mixed in an urn. But nonetheless it is perfectly possible to define systems that violate the second law. In our bathtub, if hot water is added at one end, we find that, after a while, the whole bathtub is slightly warmer (and the entropy of the system has increased). But if we imagine a bathtub in which all the velocities of the molecules are exactly opposite to those of the warm tub, we will observe the hot water separating itself out and moving to one end of the tub (thus decreasing entropy). There is nothing physically impossible about the imagined state (as noted in Chap. 7 this idea goes back to Josef Loschmidt, who used it to argue against Boltzmann’s attempted mathematical derivation of the second law; see Sklar 1993). Thermodynamical evolution is ultimately no less governed by the fundamental features and laws of the world than any other process. There are no magic high level thermodynamical properties that act on the low level features, disciplining them to conform to abstract laws. Thermodynamics is one more conservatively emergent phenomenon. What is especially interesting about thermodynamics however is that it is a high level feature which abstracts away from the details of the fundamental features to a very great extent. That represents a tremendous ‘mental victory’—it is a testimony to the immense cleverness of human scientists who managed to find a way to group together a vast array of physical systems under a common conceptual scheme which happens to be instantiated in our universe. I am not denigrating the second law. Its being a conservative emergent is compatible with it being an absolutely inescapable fact about our world. The dependence of such abstract features as thermodynamical properties upon the fundamental features of the world can be revealed via an examination of the perennial dream of the perpetual motion machine (pmm). We can define a pmm simply as a device from which unlimited work can be extracted. Since work essentially involves a system moving, or being moved, from a higher to a lower entropy state, the second law forbids a pmm. Thus anyone who claims to have a pmm can be refuted simply by appeal to the second law. However, analysis of any putative pmm will show why it violates physical principles other than brute thermodynamics. Every candidate pmm will fail because of a conflict with some or other lower level physical constraint rather than simply conflict with thermodynamics (there is an analogy here with the screening off argument discussed in Chap. 8—in a way, low level features of the putative pmm screen off the 2nd law5 ). This is a sign of conservative emergence. For example, here’s a schematic design for a ‘capillary action’ pmm6 (see Fig. 9.1). Place a narrow tube in a water reservoir. Capillary action will force water up the tube. Cut a hole in the side below the level to which the water rises and let water that drips out turn a small water wheel. The water then returns to the reservoir. We have thus extracted energy from a system that returns to its initial state and thus is apparently capable of perpetually generating energy. The second law of thermodynamics requires that this device fail, but that law is not what makes the device fail. The device fails because of the physics of atmospheric

9.2 Abstractive Explanatory Aids

181

Fig. 9.1 Capillary action pmm

pressure, surface tension and adhesion between water and the particular material of the tube. That is, low level features cause the pmm to fail (of course, adhesion, pressure etc. are only relatively low level and are themselves not fundamental; they stem from lower level features, right down to the truly fundamental). Why are there no candidate pmms that fail only because of their violation of the second law? Because thermodynamics is causally impotent, even though few principles in physics have greater explanatory potency. Information theory is very closely related to thermodynamics. The analogy between the two theories is fascinating although less than perfectly clear. But the ‘information law’ analogous to the second law of thermodynamics is that information degrades over time as entropy increases. It is important to recall here that ‘information’ simply means information capacity; there is no attempt to provide any theory of the content or semantical value of this information. We might liken the process of information decay to the fading of a printed page, where the parallel between entropy and information is quite clear. And what makes a physical process an ‘information channel’ is just as abstract as thermodynamical properties. Information, and information channels, are multiply realizable; they represent an extreme abstraction from the fundamental physical processes which ultimately constitute them. The information equivalent of a pmm would be a device that generated unlimited amounts of information capacity. Such a device will fall afoul of the second law however, since information capacity is directly linked with relative low entropy. However, just as in the case of thermodynamics, no information channel will fail to generate ‘extra information’ for free simply because of a violation of information theory. Rather, every information channel’s particular characteristics will devolve from the fundamental physical processes constituting it.7 At what is perhaps a still higher level of abstraction lies evolutionary theory, whose applicability depends upon only three exceptionally general principles: heritable reproduction, variation and selection. Obviously these can be implemented in a huge number of diverse ways, and need not be restricted to the realm of biology. Hence

182

9 The Paradox of Consciousness

we ourselves can and do produce a variety of implementations of ‘artificial life’, that is, systems of computer code that metamorphose according to the above principles (for a philosophically oriented discussion see Boden 1996). Gerald Edelman (1987) has advanced the theory that neural structure is created via a Darwinian mechanism of competition and selection of groups of neurons within the working brain. Richard Dawkins (1989), Daniel Dennett (1995) and Susan Blackmore (2000) have pointed out and explored how abstract mechanisms of evolution can be applied in the realm of psychology, sociology and culture in the theory of the spread of ideas and other cultural artifacts collectively known as ‘memes’. If the speculations of Lee Smolin (1999) should turn out to be correct, perhaps the universe as a whole is a vast collection of more or less causally isolated sub-universes which embodies a kind of evolutionary system of which our ‘local’ sub-universe is a product. Smolin hypothesizes that every black hole is, somehow, the source of a new universe so that over vast cosmic time (however that would be defined) there will be a kind of selection for universes which produce black holes more copiously. Smolin also speculates that black hole fecundity is correlated with general features which favor the existence of life and intelligence, so his theory would be an important part of the explanation of the emergence of high level structure. The applicability of the basic concepts of evolution is so extensive because they impose such small demands on their implementation. Thus many such implementations can easily spring into existence. And while it is of course a testament to our intelligence that we are able to spot these implementations and put them to use in our explanatory schemes, they remain nothing more than the kind of high level feature, or epistemological resource, we have been discussing. The notion of relative fitness (defined in terms of a particular system of reproducing entities within a given environment), for example, helps to explain innumerable features of the biological world, but nature makes no use of it in the running of the world. Just as in the case of the pmm, every time an organism does better than some other because of its fitness in a certain environment, the causal determination of that event depends solely and entirely on underlying mechanisms and never on fitness as such. Again, we see the telltale signature of conservative emergence. My final example of an extremely abstract high level feature is psychology, by which I mean not particular scientific theories, but rather what philosophers call ‘folk psychology’ or commonsense belief-desire psychology. I suspect that the general applicability of folk psychology is closely related to the similarly general applicability of evolution in biology (see Dennett 1978b; Seager 1990, 2000b). This is because an organism’s desires are primarily focused on what is good for it from the biological point of view and its belief system focuses on information that is likely to be important for the satisfaction of these desires. Once organisms evolve sensory systems and mobility we will find that folk psychology inevitably becomes applicable to them to at least some degree. Thus I can predict that if you spill some sugar on the floor, the ants will come and get it because, as we are prone to say, they like sugar and, via information delivered by their sensory systems, they will come to believe that sugar is available to them whereas if you spill some motor oil on the floor you are safe, at least from ants. Of course, whether ants should be credited with genuine beliefs and

9.2 Abstractive Explanatory Aids

183

desires is, to say the least, a moot point, but it is undeniable that folk psychology is pragmatically effective at a basic level even within the insect realm. Within the realm of high level features we can see a host of structures ranging from, at the abstract end of the scale, the thermodynamical, informational, evolutionary and psychological to more concrete structures such as tectonic plates and chemical kinds. But from the viewpoint of the SPW it is evident that all these features are nothing more than patterns of events which are explanatorily salient to us. High level features are thus ‘for us’, conscious subjects. Only through being consciously conceptualized can they take any role in the world. High level features are, we might say, only ‘visible’ from certain consciously adopted explanatory standpoints. In this respect, they are rather like the clouds that appear in the shape of animals as we watch them drift overhead. Some clouds really do look like animals (to us, from where we lie) but that does not matter to what happens in the sky. The fact that some other high level features (such as animals) not only look like but act like animals does not change their status as epiphenomena of the low level features’ unutterably complex dynamical interactions. But now, if patterns are mind-dependent and all high-level features are patterns then, given that minds are high level features, they will themselves be minddependent. There is an obvious trivial sense in which this is true: if there were no minds there would be no minds. The serious problem is that the SPW requires minds in order to integrate high-level features into a world that obeys completeness, closure and resolution. But minds appear to be high-level features themselves. Therefore, there is one high-level feature (at least) that cannot be integrated into the SPW in the usual way. If we try to regard minds as a ‘normal’ conservatively emergent high-level feature of the world then instead of integration with the SPW, we simply are left with ‘another’ mind which serves as the viewpoint from which the target mind appears as a pattern. This is either a vicious regress or an unacceptable circularity. There is a real paradox lurking here. It is a paradox of consciousness. Looked upon as a behaviour modifying and adaptive ability, conceptualization is just another high level feature of certain rather special emergents. Conceptualization, and its precursor, categorization, are psychological features which are crucial for the applicability of belief-desire psychology to a system (this is trivial: no beliefs without content and no content without, at least, categorization). There is no problem here until we consider beings which are conscious of themselves and other high level features of the world. Within the SPW, conscious awareness most certainly does not appear to be a basic feature of the world but rather takes its natural place as another high level feature, one that involves the mass action of at least millions, if not hundreds of millions or even billions, of neurons. But if we accept that consciousness is another high level feature, we must conclude that it, like all other high level features, is causally impotent. Consciousness does not participate in the ‘go’ of the world; it does not add any constraints upon state evolution that are not already present because of the fundamental physical features. That is, we must conclude that the SPW entails that consciousness is epiphenomenal.

184

9 The Paradox of Consciousness

9.3 Paradox But that is not the paradox. After all, we could bite the bullet of mentalistic epiphenomenalism which is a coherent, if bizarre and implausible, doctrine. It has, in one form or another, been defended by serious thinkers (for example, Thomas Huxley 1974; Frank Jackson 1982; William Robinson 2004). Generalized epiphenomenalism reveals, however, that the SPW, and in particular, the scientific picture of the metaphysics of high-level features, is incoherent. It seeks to regard mind as a highlevel feature of reality, but then it finds a conscious mind is needed to make sense of high-level features. It does not seem possible to evade this problem, for the only place for mind within the current SPW is as a high-level feature. But in order for it take such a place, there must be a viewpoint from which the ‘patterns of mind’ are apparent amidst the complex turmoil of the elementary features of the world. Such a viewpoint is itself an aspect of mind, and so we have failed to understand mentality in general as a high-level feature.8 For example, if, say, my consciousness is just another high level feature of the world, it too stands as a mere potential epistemological resource, except insofar as it is taken up into some actually and consciously deployed explanatory system. But the conscious deployment of an explanatory system is one more high level feature which itself is a mere epistemic potentiality unless it is taken up into a further consciously deployed explanatory system. This seems to be a vicious regress which leaves consciousness as nothing but merely potential, never ‘actualized’. Each standpoint, as a high level feature, requires a conscious appreciation of it to transcend mere potentiality, but if consciousness itself is a high level feature then each consciousness requires a further consciousness to appreciate it. So the worry is that consciousness could never get off the ontological ground in the first place. This undercuts the SPW’s understanding of consciousness, which it sees as emerging slowly and in initially primitive forms, attaining the status of reflective, theorizing consciousness only very late in its development. But—I take this as self evidently true—my current consciousness, or indeed the consciousness of any conscious being, is a fact about the world that is not at all merely potential and could persist as a fact even if all other consciousnesses were, this instant, obliterated. Furthermore, it seems completely implausible to suppose that my consciousness depends in any sense on the explanatory activities of any other beings (whereas I think there is little difficulty in seeing the sense in which objects such as trees, planets or galaxies do have a kind of dependence upon the activity of beings who think of the world in terms of trees, planets and galaxies). The worry can be sharpened if we consider the Earth of some millions of years ago, before there were any beings capable of taking up an explanatory standpoint from which consciousness would appear to be at work in the world. Nonetheless, the world was full of conscious animals, suffering all the torments of mortal existence, and all the pleasures as well. It is hard to deny that their pains caused them to avoid certain things, their pleasures led to the avid pursuit of other things. These conscious states were not mere potential epistemic resources, awaiting recognition by more powerful

9.3 Paradox

185

minds bent on explanation and prediction. That is not a standpoint dependent fact. It seems absurd to suppose that, absent beings who consciously explained animal behavior in terms of those animals’ experiences, animal consciousness was a mere epistemic potentiality awaiting the distant future birth of human beings. The paradox developed here can perhaps be better understood if it is contrasted with a less serious difficulty that besets a number of philosophical accounts of high level properties, most especially mental properties. This latter difficulty is a failure in the project of naturalization. This project is the attempt to show that all high level features are conservatively or epistemologically emergent. This is not the same as showing that everything reduces to the fundamental physical features, although it is compatible with strict reductionism. It will suffice for naturalization of some domain if we can show how its features conservatively emerge from more fundamental features of the world, or at least provide good evidence that there is such an account available in principle. We have seen abundant evidence for the wide ranging if not universal power of the naturalistic outlook in the first three chapters of this book. Successful naturalization of any domain must abide by a few simple rules. Roughly speaking we can characterize these as follows. X has been naturalized if and only if (1) The emergence of X has been explained in terms Y. (2) Y is properly natural. (3) Y does not essentially involve X. This notion of naturalization has several virtues. It is reasonably clear and is directly and quite properly aimed at the scientific integration of the naturalization’s targets. After all, it would seem very strange if not perverse first to embrace the SPW and then boast about how there are some phenomena that defy all attempts to give an account of how they fit into that world-view. In terms of the idea of supervenience, the need for an explication of the supervenience relation between target and base domains is obvious (otherwise, to generalize a remark of Simon Blackburn’s supervenience is part of the problem rather than part of the solution; see Blackburn 1985). To illustrate how naturalization can fail let’s examine a contemporary example: Daniel Dennett’s instrumentalist theory of intentional mental states (as outlined originally in Dennett 1971). It is somewhat curious that Dennett, who by and large writes from a perspective that clearly endorses the scientific view of the world and which is supposed to be non-eliminativist, espouses a theory of intentionality which blocks the naturalization of the mind. Although Dennett’s theory of the intentional stance is by now intricate and subtle, it remains essential to it that the mental states of a subject, S, be understood as states (no doubt physical states, probably of S’s brain) whose mentality resides in their underpinning an intentional interpretation of S. Now, the commonsense view is that mental states have the job of generating behaviour which can be interpreted from the intentional stance, but the mentalistic interpretation is parasitic upon the mental properties of these states. It is because they are mental that we can successfully interpret their subject as having a mind. Fundamentally, Dennett sees things the other way around: it is because we can interpret subjects as behaving

186

9 The Paradox of Consciousness

in accord with a more or less rational ascription of intentional mental states that they count as having mental states at all. But, of course, the notions of ‘interpretation’, ‘intentional stance’, ‘predictive purposes’, etc. are one and all notions which are themselves fundamentally mentalistic. This is formally obvious, but let’s be clear how the problem arises. It is not that, as a matter of fact so to speak, mental state ascriptions are parasitic upon behaviour which can be interpreted mentalistically; the problem of naturalization is that we cannot explain what mental states are without appeal to notions shot through with their own mentalistic implications. You can’t understand what a mind is unless you already know what a mind is, since you can’t understand mentality without understanding the intentional stance, which requires you to already understand a host of essentially mentalistic concepts. Another approach to this is via the comparison of the case of the mind with that of chemistry; the two are entirely dissimilar. One can imagine learning chemistry by learning its naturalization along with a host of defined terms—at the end one would really know what chemistry was. Of course, after this beginning, because of typical problems of complexity, one would have to learn to ‘think chemically’ to really get anywhere in chemical studies. According to Dennett, you can’t do this for the mind, since you’d already have to know what a mind was to ‘get’ the intentional stance.9 Dennett’s approach in ‘Real Patterns’ (Dennett 1991) suggests that interpretability can stand as the objectively existing, mind independent reality of minds. It is just a fact that certain systems can be interpreted via the intentional stance and this is true whether or not any creature takes up the intentional stance itself and engages in any real time interpreting. But the paradox of consciousness that worries me stems from the fact that high level features are mere epistemic resources the play no role in nature save insofar as they are conceptualized by some conscious being. This too leads to a failure of naturalization which threatens the SPW but it also leads to the much worse consequence that consciousness cannot appear in the world save as recognized by a conscious being. This consequence makes the SPW incoherent if it insists that consciousness is a standard high level feature of the world.

9.4 Reflexive Consciousness Could the problem be evaded if consciousness somehow served as its own epistemic standpoint, looping back on itself and thus generating itself without necessarily violating its own merely conservative emergence? The idea that consciousness is reflexive or in some way necessarily self-referential goes back a long way, at least to Aristotle who formulated an argument based on the fact that we are conscious that we are conscious. In Book 3 of De Anima, Aristotle writes

9.4 Reflexive Consciousness

187

Since we perceive that we see and hear, it is necessarily either by means of the seeing that one perceives that one sees or by another [perception]. But the same [perception] will be both of the seeing and of the colour that underlies it, with the result that either two [perceptions] will be of the same thing, or it [sc. the perception] will be of itself. Further, if the perception of seeing is a different [perception], either this will proceed to infinity or some [perception] will be of itself; so that we ought to posit this in the first instance. (Translation from Caston 2002)

The argument is somewhat elusive, but Aristotle evidently thinks that if conscious states are not reflexive or self-representational then a vicious regress will ensue. This follows only if every mental state is a conscious state in the strong sense that its subject is aware of it. If some mental states are not conscious in this sense then no regress is generated.10 Although the idea that mental states are essentially conscious is also a very old idea with illustrious defenders, such as Descartes, we do not nowadays balk at the claim that there are unconscious mental states. Nonetheless, there is some intuitive pull to the idea that consciousness is essentially reflexive. After all, whenever we notice that we are conscious we have an awareness of our mental state and when we do not notice our own consciousness we of course have no sense that we are conscious. But for all of that, we could be—I believe we are—mostly conscious without having any awareness of our own consciousness. I think this is also the perpetual state of most animals many of whom are conscious beings but very few have any sense that they are conscious. Still, it is worth thinking about whether the reflexivity of consciousness would have any mitigating effect on the paradox of consciousness. It seems to me this is very unlikely because of a number of considerations. First, the kind of reflexivity which conscious states are supposed to possess enables, or embodies, a special, indeed metaphysically unique, relation between the subject and the subject’s own consciousness, whereas the relation between subjects and their explanatory posits is entirely neutral between self and other ascription. Thus the reflexivity of consciousness and the system of conservative emergence are entirely distinct. In other words, even if consciousness is essentially reflexive that will not show how it is conservatively emergent from neural processes. The situation actually threatens to get worse, because this mysterious feature of reflexivity puts an added explanatory burden on the system of conservative emergence. One can easily imagine theories of mental or neural representation which involve self-representing representations, and it does not seem too difficult to imagine these self-representers being conservatively emergent. But these won’t dissolve the paradox since they are themselves stance dependent explanatory aids to conscious beings bent upon understanding cognition. To solve the paradox we need some kind of ‘magic’ reflexivity which generates consciousness. Perhaps it exists, but it will not fit into the system of conservative emergence characteristic of the spw. Second, and related, the reflexivity of consciousness is supposed to be a primitive phenomenological fact about subjectivity. The explanatory stance of emergence has no particular relation with the phenomenology of consciousness and there is no reason to suppose that simply taking up such a stance via the hypothesis that consciousness is essentially reflexive could generate any kind of phenomenology.

188

9 The Paradox of Consciousness

Third, the only way reflexivity could evade the paradox of consciousnesses is if the reflexive qualities of conscious states included the full machinery of an explanatory framework in which conservative emergence makes sense. But this is extremely implausible. The concepts of this framework are sophisticated, multifaceted and numerous. To make their possession a necessary condition for consciousness would put such a cognitive burden on conscious beings as to be quite incredible. That is, even if we accept that consciousness is in some way reflexive this must be understood so not to rule out rudimentary forms of consciousness, such as presumably exist in many animals and human infants. It is important to note here a subtlety in the argument. The paradox arises because of the way the SPW integrates conservatively emergent features into its overall view of the world. This integration requires there exist a standpoint of a conscious ‘explainer’ in order for emergents to have a role in the world. However, this argument puts no constraints on the nature of consciousness itself. It does not, for example, entail that all consciousness is sophisticated, curiosity driven theoretical thought or that all conscious beings must possess concepts enabling such thought. Finally, and in any case, from the point of view of the SPW the reflexive feature of consciousness is simply another high level feature. As such, it stands as merely a potential epistemic resource awaiting the intellectual gaze of some conscious being bent upon understanding the world to have any role in the world. So even if reflexivity could hold consciousness up by its own bootstraps, the reflexive aspect of consciousness itself would remain unrealized until taken up by an explanation seeking consciousness. That is, the reflexive loop which is supposed to solve the paradox has no being save as an epistemological resource. This engenders a vicious regress. And, unlike in the case of consciousness, there is no reason at all to suppose that the reflexive loop itself is some kind of self-representational state and no evidence whatsoever for this from our own experience of consciousness. I think that ultimate reason for the paradox comes down to the apparent fact that consciousness is an intrinsic, non-relational property whose instances are real time existents which stand as ontologically independent (though not causally independent).11 In contrast, the properties which are conservatively emergent are in a certain sense relational. What I mean is that conservatively emergent properties are essentially structure dependent as befits their status as patterns which can stand as epistemological resources in various epistemic projects. Consciousness has, or is, an intrinsic nature which no conservative emergent possesses as such. Recall Galileo’s words when he first observed the Milky Way by telescope: ‘the Galaxy is nothing else than a congeries of innumerable stars’. In general, conservative emergents dissolve into their fundamental constituents remaining only as patterns ready to serve the epistemological and explanatory purposes of conscious beings bent upon understanding the world. It seems obvious that consciousness itself cannot be similarly dissolved without destroying its real time, phenomenal, intrinsic and immediately accessible reality.

Chapter 10

Embracing the Mystery

10.1 Watchful Waiting Broadly speaking, there are only a few possible responses to the paradox of consciousness. I cannot say I have much confidence in any of them but permit myself some hope that the list of options considered in this chapter is complete, so that one of them will provide the escape route. Three of the four routes are familiar, the fourth, perhaps the most radical of them all, is more novel. The first option is simply to conservatively insist that consciousness is a standard conservatively emergent phenomenon that will eventually be fitted into the SPW in the usual way. On this view, it is far too early to junk the scientific picture of the world. If the problem persists and grows worse more drastic steps can be taken but the history of science suggests that eventually the problem of consciousness will resolve itself. Call this the option of Watchful Waiting. Our discussion has focused on the paradox of consciousness but of course the phenomenon of consciousness raises a host of additional worries. The general problem of consciousness is that there seems to be no way to explain exactly how matter generates or embodies phenomenal consciousness given our present, and likely future, understanding of the nature of the physical world. That is, the conservative emergence of consciousness is entirely opaque to us. This issue has attracted sustained philosophical attention for the last fifty years or so,1 under various labels, most recently that of the problem of the ‘explanatory gap’ (Levine 1983, 2001) and the ‘hard problem’ (Chalmers 1996, 1997). An enormous literature has ensued but there is no sign of any consensus on a solution, on the exact specification of the problem or even whether it is not some sort of philosophically induced pseudo-problem. But one clear response in the face of epistemic opacity is to plead that our ignorance is excusable and/or explicable. The most straightforward plea is simply that more time is necessary for the study of the brain mechanisms responsible for consciousness, as well as for the advancement of various requisite attendant disciplines—whatever these might be. The depth of our ignorance is perhaps indicated by the fact that candidate disciplines range all the way from physics to philosophy. W. Seager, Natural Fabrications, The Frontiers Collection, DOI: 10.1007/978-3-642-29599-7_10, © Springer-Verlag Berlin Heidelberg 2012

189

190

10 Embracing the Mystery

A more radical line is to endorse essential ignorance. This position has been defended by Colin McGinn (1989) who claims that while the emergence of consciousness is conservative it is impossible for us (humans) to understand the particular mechanism of emergence at work in this case. We are, in McGinn’s phrase, cognitively closed with respect to the problem of consciousness in much the same way that a chimpanzee is cognitively closed with respect to calculus. McGinn stresses that there is nothing metaphysically mysterious about the emergence of consciousness. There might be creatures for which it is intelligible how consciousness appears as a standard emergent from fundamental physical processes. Unfortunately, we lack the intellectual chops to join that club. A third way to endorse Watchful Waiting is to envisage some fundamental change in our understanding of the physical world, or consciousness, which will permit new, but recognizably legitimate, mechanisms of conservative emergence. There can be no doubt that the explanatory structures within science have changed, sometimes radically, over the course of its development. A salient example is the transmutation in the doctrine of mechanism brought about by the acceptance of Newton’s theory of gravitation in the 17th century. The introduction of a non-contact force which acted instantaneously over any distance was regarded with extreme skepticism. Newton could never bring himself to believe in it and wrote in a famous passage: ‘that one body may act upon another at a distance through a vacuum, without the mediation of anything else, by and through which their action and force may be conveyed from one to another, is to me so great an absurdity that I believe no man who has in philosophical matters a competent faculty of thinking can ever fall into it’ (Janiak 2004, p. 102). Thus one might think that the problem of consciousness awaits some new, revolutionary development in science which will unveil the unmysterious mechanisms of conservative emergence. Perhaps this is what Noam Chomsky intends to suggest with this analogy: Suppose that a nineteenth century philosopher had insisted that ‘chemical accounts of molecules, interactions, properties of elements, states of matter, etc. must in the end be continuous with, and harmonious with, the natural sciences,’ meaning physics as then understood. They were not, because the physics of the day was inadequate. By the 1930s, physics had radically changed, and the accounts (themselves modified) were ‘continuous’ and ‘harmonious’ with the new quantum physics. (Chomsky 2000, p. 82)

Perhaps some similarly radical transformation in physics will allow for consciousness to take its place as a standard conservative emergent. Although Chomsky is in the end noncommittal about this with regard to consciousness, and the mind in general, he does note that ‘[c]ommonly the “fundamental’’ science has to undergo radical revision for unification to proceed’ (p. 106). The philosopher John Searle sees the problem of consciousness as akin to the difficulties encountered when the fundamental science lacks the necessary resources to explicate the mechanisms of standard conservative emergence. He says ‘the “mystery’’ of consciousness today is in roughly the same shape that the mystery of life was before the development of molecular biology or the mystery of electromagnetism was before Clerk–Maxwell’s equations’ (Searle 1992, pp. 101–2). This seems to

10.1 Watchful Waiting

191

endorse the idea that we need a revolution in our existing sciences to enable us to understand the emergence and nature of consciousness. In a similar vein, Thomas Nagel writes that ‘the status of physicalism is similar to that which the hypothesis that matter is energy would have had if uttered by a Presocratic philosopher. We do not have the beginnings of a conception of how it might be true’ (Nagel 1974, p. 177). Famously, Nagel goes on to express grave doubts that any advance in science will ever make it possible to understand how consciousness could be a conservative emergent. Although these thinkers have considered the idea that revolutionary advances in science are needed to solve the problem of consciousness, one scientist has positively endorsed the view. In The Emperor’s New Mind, Roger Penrose argues that consciousness can only be scientifically understood via a radical change in physics which will do no less than introduce some features that transcend Turing computability (see Penrose 1989). The reason Penrose offers for this is not the usual one which regards the generation of phenomenal or qualitative aspects of consciousness as scientifically inexplicable but rather, and notoriously, that Gödel’s incompleteness result entails that the intellectual power of the conscious mind somehow goes beyond the power of algorithmic computation. The connection to consciousness arises because only in the case of conscious thoughts in which the Gödel problem is appreciated does the problem appear. Although the position is far from clear, writings with Stuart Hameroff sometimes suggest that something like a quantum mechanical entanglement based emergentism or, as they sometimes write, panpsychism is the correct way to link consciousness with scientific theory (see e.g. Hameroff and Penrose 1996). Note that it is not particularly strange to link emergentism and panpsychism. The introduction of fundamental mentality opens the door to the conservative emergence of more complex forms while avoiding the problem and paradox of consciousness as we shall see below. As Niels Bohr is reputed to have said, making predictions is difficult, especially about the future. The option of Watchful Waiting is a hostage to the fortunes of future science. Will physics undergo a revolution that will introduce a consciousness-like feature in its foundations? No one can say, but it seems far from likely. Nor do I think defenders of the idea that consciousness is a standard conservative emergent would be satisfied if physical science had to undergo that sort of reformation. Will some conceptual revolution reveal what has been missed for so long—the key to understanding that consciousness is a bog standard conservative emergent? This cannot be ruled out, but of course we have no inkling of how this could work at present. What is interesting is the sense that science as we know it is at a loss to explain how consciousness could be a conservative emergent of purely physical processes. But there is something of a red herring quality to this issue. Although the hard problem and explanatory gap are, in my view, deadly serious and very far from being resolved I do not want to add to their discussion here. This is because I think that the paradox of consciousnesses stemming from the account of conservative emergence within the SPW creates a distinct problem for the defenders of Watchful Waiting. No matter what new science or conceptual revolution occurs, the nature

192

10 Embracing the Mystery

and role of conservative emergents remains clear. They are causally (i.e. kausally) inefficacious and exist solely as potential epistemic resources for conscious beings intent upon understanding, explaining and predicting the world. As discussed in the previous chapter, this means that the claim that consciousness itself is simply another conservative emergent leads to incoherence. Only the total rejection of the SPW can evade this problem. The nature of conservative emergence ensures this since it requires that all higher level features depend upon fundamental features which completely determine their natures.

10.2 Embrace Emergence If Watchful Waiting is a non-starter, we should turn to options that in one way or another reject the SPW itself. The obvious first choice is to Embrace Emergence, that is, to reconsider the possibility of radical emergence.2 We can leverage the discussion in Chap. 7 to provide a succinct characterization of the difference between conservative and radical emergence. Recall that the doctrine of supervenience requires that all high level features be determined by low level or fundamental features. The basic form of a supervenience claim is expressed in the following formula (D5 from Chap. 7) in which the property family U is supervenient upon the family T : (∀σ )(∀F ∈ U )(Fσ → (∃G ∈ T )(Gσ ∧ (∀π )(Gπ → Fπ ))) What is crucial here is the second necessity operator () which expresses the strength of the necessity relation which holds between the low level, subvenient domain and the high level domain (the initial operator serves only to express the modal status of the entire claim). For present exposition we can simplify this formula. Consider a single representative property, F, and neglect the outer necessity operator (plus let us take for granted that G is an element of the appropriate subvening family of properties). The simplified form is then: Fa → (∃G)(Ga ∧ (∀π )(Gπ → Fπ )) Now I need to re-complicate the formula to express explicitly the way emergence is supposed to be a function of the constituents of the supervening feature. That is, the canonical form of emergence is that a system depends on the nature and interrelationships of some more fundamental features or entities which together generate the system’s structure and properties. Bear in mind that the notion of ‘constitution’ here is very loose and is not restricted to so-called part-whole or mereological relations. For example, the property of being an uncle presumably supervenes on fundamental physical features but these features are not restricted to those fundamental entities which are, in any ordinary sense of the term, constituents of the particular uncle in question. The needed complication is simply to bring in reference to the relevant constituents, as follows:

10.2 Embrace Emergence

193

Fa → (∃)(∃z 1 , z 2 , . . . , z n )(C z 1 , z 2 , . . . , z n , a∧z 1 , z 2 , . . . , z n ∧ x (∀x1 , x2 , . . . , xn , y) (C z 1 , z 2 , . . . , z n y ∧ z 1 , z 2 , . . . , z n → F y))

What does this say? In its unfortunately but unavoidably prolix fashion it says that there exists a bunch of constituents of the system a (these are the z i ). C stands for the appropriate relation of constituency and  is the relation in which the z i stand which subvenes F (we assume that, taken together, C and  express the laws of the low level entities involved, that is, the z i ). The supervenience claim is recast in the latter half of the formula to say that any system with constituents arranged according to the relation  will have the higher level property F. Again, the crucial point is the strength of the necessity operator, whose openness is suggested in the formula by use of x . There are two ‘settings’ which matter: straight on full logical necessity versus causal or nomological necessity. Conservative emergence aligns with the former, radical emergence with the latter. Why is that? Conservative emergence requires that the underlying fundamental base totally determine the high level features in such a way that no modification of the low level laws is required to generate all the behaviour in which the high level features can participate. For example, if we fix all the atomic facts then the chemical facts are thereby also fixed—there is no need to add any new distinctively chemicallevel principles, laws or activities in order for all the characteristically chemical-level behaviour to be determined. That is what it means for the chemical to be a conservative emergent from the underlying physical basis. Those who Embrace Emergence tell a starkly different story. They agree that the low level features have a set of laws corresponding to that domain. But they add that under certain conditions, systems made up of low level entities will bring about the emergence of entirely new features that will influence the behaviour of the total system in ways not accountable purely from the low level. Recall the thought experiment of the superduper computer simulation. Radical emergentists deny that the simulation would perfectly duplicate any real world system that possesses radical emergence, no matter how well the simulation mimics the lower level. The novel emergent features will have their own causal efficacy which will ‘interfere’ with the low level processes, forcing them to drift away from the simulated version whose behaviour is, by design, entirely governed by low level processes and low level processes only. Therefore, radical emergentism maintains there are possible worlds that differ in their ‘laws of emergence’ without supposing that there is any difference in the subvenient level laws governing these worlds. In terms of our formula this is expressed entirely in the grade of the necessity operator so that two natural forms appear, one where we use l , standing for full logical necessity and the other which features n , standing for nomological necessity. The former, corresponding to conservative emergence, entails that there are absolutely no possible worlds in which a set of constituents stand in some operative subvening relation but fail to have the associated supervening high level property. The latter, corresponding to radical emergence, allows for worlds in which the various laws of emergence are different, or absent altogether. The emergentist who believes in only conservative emergence thinks the

194

10 Embracing the Mystery

actual world is in the class of worlds that lack any laws of emergence—they are completely unnecessary. The radical emergentist thinks that a world lacking laws of emergence will not support all the high level features we actually find, notably including consciousness. Their world is in essence like the time dependent Life world we examined in Chap. 5 where the fundamental laws of Life will not be able to generate all the phenomena that occur and they believe the behaviour of the actual world similarly outruns what the fundamental laws of physics (plus of course the fundamental physical state of the world) can generate strictly on their own. Radical emergence seems to be a coherent doctrine (see McLaughlin 1992), although there have been attempts to show otherwise. For example, Thomas Nagel’s (1979) argument in favour of panpsychism depends on the denial of the possibility of radical emergence. Nagel’s argument is very succinct. The first premise is that ‘the properties of a complex system must derive from the properties of its constituents, plus they way they are combined’ (Nagel 1979, p. 185, my emphasis). All emergentists can agree with this (modulo the caveat about the liberal interpretation of ‘constituent’). The second premise is that ‘true causes do necessitate their effects: they make them happen or make them the case’ (p. 186). Here Nagel has to assume that the kind of necessitation involved is full logical necessity. But this is highly implausible. What can cause what depends on the laws of nature and various contingent factors. The value of the fine structure constant in in quantum electrodynamics does not appear to be necessitated by any (other) law of nature and it has a powerful influence on the causal powers of atomic structure and hence chemical kinds. As noted by Barrow and Tipler ‘it transpires that the gross properties of all atomic and molecular systems are controlled by only two dimensionless physical parameters—the fine structure constant…and the electron to proton mass ratio’ (1988, pp. 295–6). It seems pretty obviously true that, one, there are causal processes in the actual world that involve these parameters and, two, there are other possible worlds where the causal processes are different because of variation in the values of these parameters. Nagel is right that once we fix the laws of nature and the state of the world then it is a matter of pure logical necessity what is going to happen and what high level features will appear (if the laws are intrinsically indeterministic then the range of outcomes is logically determined). But this does not show that all emergence is conservative emergence unless we assume that the laws of physics exhaust the fundamental laws of nature.3 Since the radical emergentist explicitly denies this Nagel is simply begging the question against this form of emergence. And, in principle, there does not seem to be anything incoherent in the idea that there are irreducible laws of emergence that go beyond the laws of fundamental physics. Although the heyday of radical emergentism centered on the early 20th century, the idea that there are irreducible laws of emergence has not entirely disappeared. A very interesting recent example is Anthony Leggett’s suggestion that complexity of material constitution can modify or override the laws of quantum mechanics. Leggett writes that ‘it is quite conceivable that at the level of complex, macroscopic objects the quantum superposition principle simply fails to give a correct account of the behavior of the system…any concrete proposal of this type would require us to

10.2 Embrace Emergence

195

introduce new physical laws which are not contained in the quantum formalism itself’ (Leggett 1987, p. 98). This can be read as endorsing a kind of radical emergentism, and Leggett’s further anti-reductionist remarks reinforce this interpretation. Note also that, in line with our general discussion of radical emergence, Leggett’s idea would be empirically testable. In fact, Leggett has engaged in and spurred highly interesting experimental efforts to realize macroscopic, or almost macroscopic, devices that can be put in superpositional states. Thus far, emergentism has not fared well. In the year 2000, a superconducting ring about the size of a human hair was put in a state with superposed bi-directional electric currents in accordance with quantum mechanics (see Friedman et al. 2000). Leggett admitted to the New York Times that the experiment provided ‘reasonably foolproof evidence you do have a superposition of macroscopic quantum states’ (Chang 2000). Galen Strawson has also argued against the coherence of radical emergentism (Strawson 2006). His argument depends on the claim that ‘it is built into the heart of the notion of emergence that emergence cannot be brute in the sense of there being absolutely no reason in the nature of things why the emerging thing is as it is’ (p. 18). This is true if the ‘nature of things’ includes the laws of emergence which the radical emergentist posits. If the ‘nature of things’ is however restricted to the fundamental physical laws then emergence will be brute but not in any way that would or should disturb a radical emergentist. Is there something incoherent about a brute law of nature? If there is then there are no fundamental laws at all, but there seems to be nothing unintelligible about basic, non-derivative relationships between things. Strawson also argues against the conservative emergence of phenomenal consciousness on the basis that the experiential and non-experiential are so disparate that there is no explanatory relationship between them which could intelligibly reveal exactly how consciousness is generated by matter. Although such an argument is not relevant to our present concerns, one might wonder whether Strawson’s tactic could be adapted in an argument against radical emergence. Could one argue that even irreducible laws of emergence must relate features which are, in some sense, not too different from each other? There is then nothing wrong with brute relations between certain elements of physical reality but, one might argue, there cannot be irreducible relations of emergence across ontological categories. Or, to put it another way, there would be no inter-level brute emergence; intra-level primitive determination relations could then be regarded not as emergence but causation. Bear in mind, however, that in light of the discussion of Chap. 7 there is little difference between causation regarded as temporal supervenient determination and radical emergence. Perhaps there is some interest in this idea, but absent a clear account of the nature of ‘ontological categories’ and an independent argument that ontological borders are necessarily closed, the suggestion once again begs the question against radical emergentism. However, there is a very great distance from the claim that radical emergence is coherent to the claim that the actual world exemplifies it. The first three chapters of this book stand as powerful evidence that there are no empirical signs of radical emergence. One can regard the history of science as a failed search for such emergence or, equally, as a triumphant progress towards ontological unity coupled with

196

10 Embracing the Mystery

ubiquitous conservative emergence. Chap. 6 then argued that there is no theoretical reason to think the world exhibits any radical emergence. These lines of evidence are not, of course, anywhere near conclusive. Our ignorance of the natural world still completely dwarfs our knowledge. Nonetheless, the scale of our knowledge is staggeringly large and comprehensive. So it’s passing strange that radical emergence should exist but be completely invisible to us despite the depth and range of our scientific investigation of the world. Although emergence has become something of a buzzword, it does not take much study to see that its current scientific defenders one and all mean to champion the conservative form (for a sample see Anderson 1972; Schweber 1993; Laughlin 2005; Holland 1998; Morowitz 2002). There are more fundamental reasons to doubt the existence of radical emergence. In Chap. 7 we saw that if a theory is total then it can only allow conservative emergence. Recall that the hallmarks of a total theory, T, were the trio of features called completeness, closure and resolution which were defined as follows: Completeness is the doctrine that everything in the world has a non-trivial T-description and as such abides by closure and resolution. Closure entails that there are no ‘outside forces’— everything that happens, happens in accordance with fundamental T-laws so as to comply with resolution. Resolution requires that every process or object be resolvable into elementary constituents which are, by completeness, T-entities and whose abidance with T-laws governing these constituents leads to closure. It seems clear that physical theory at least aims to be a total theory. It has been constructed so as to home in on a system of fundamental entities and forces which certainly appear to provide completeness, closure and resolution. It has had persistent success at this endeavour even in the face of regular setbacks, some of which required revolutionary reconstruction of its foundations. The outcome of these episodes of rebuilding has always been renewed strength and a strengthened sense of completeness, closure and resolution. It is true that the one time dream of an accessible epistemic reduction of all phenomena to basic physics—if it was ever even half-heartedly entertained—has long since been relegated to the dustbin of the history of philosophy. But that is not the issue on which radical emergentism stands or falls. What matters is the possibility that in (the future of) physics we can attain to a total theory. Radical emergence faces a series of basic problems insofar as it conflicts with the totality of physics. Physics says that energy is conserved in closed systems but radical emergence requires that systems complex enough to initiate emergence deviate from the behaviour the system would exhibit if governed solely by physical law applied to the system’s constituents and their interactions. This alteration in behaviour is going to require some redistribution, creation or destruction of energy that will appear at odds with the fundamental physics of the situation. Again, this is not a problem of logical coherence. Even if energy conservation is sacrosanct4 this would not imply that radical emergence is false. For it is possible to imagine that the laws of radical emergence add and subtract energy here and there so as to retain the overall balance (see McLaughlin 1992 for a discussion of this issue). But there is no reason for the emergence generated changes in energy distribution to be subtle or to be hard to measure, yet there have been no traces of such phenomena. The same problem

10.2 Embrace Emergence

197

could arise for any conserved quantity that affects the motion of physical matter (momentum, angular momentum, charge and so on). Radical emergentists are placed in the uncomfortable position of either denying that these conservation laws hold or rather gratuitously hypothesizing that radical emergence operates so as to miraculously patch up any discrepancies with the distribution of conserved quantities as determined by the fundamental physical processes alone. Both horns of this dilemma are extremely unattractive. Yet another difficulty facing radical emergence is that it appears as an ad hoc hypothesis invoked solely to explain the presence of consciousness in the world. This is not how the progenitors of emergentism saw things. Emergentists from Mill onwards saw radical emergence as a pervasive feature of the world operating at the very first level, so to speak, above physics. To the extent that it is now very difficult to argue that chemistry or biology exemplify radical emergence, the only phenomenon which remains a viable candidate for emergence is consciousness. But why should the world have waited so long to exploit its power of radical emergence? Consciousness, with its subjective, first-person aspect, is special. However, this specialness does not in any way suggest that radical emergence is the unique or best way to account for consciousness. It seems that by its nature, radical emergence should appear throughout nature but so far as we can tell it appears nowhere except in an ever shrinking domain where we feel at a total explanatory loss. This makes radical emergence look like a dodge invoked only to plug a gap in understanding with no independent evidence for it whatsoever.5 Despite all the foregoing, the option of Embracing Emergence does have the advantage that it can solve the problem of consciousness and avoid the paradox of consciousness. The basic emergentist story is that when certain physical systems attain sufficient and appropriate complexity, as in the human brain, there emerges a novel feature, consciousness, with all its attendant phenomenal qualities. This new property of the brain is lawfully related to its subvening brain state but is not determined solely by the underlying physical state and purely physical law. It has its own causal powers which can act back on the physical world. This is genuine downward causation, not the anodyne simulacrum allowed by conservative emergence. There is no sense in which this new feature is a mere epistemic potential or pattern—it has its own stance-independent being. Thus it seems to me that the option of Embracing Emergence is preferable to that of Watchful Waiting whose allegiance to the SPW precludes any real solution to the paradox of consciousness.

10.3 Favour Fundamentality Still, there might be better alternatives equally able to solve the problem without suffering from the rather large number of substantial difficulties which face radical emergence. A venerable strand of thought about the mind opposes emergence without endorsing the SPW by making consciousness fundamental. To adopt this option is, I will say, to Favour Fundamentality. A number of familiar metaphysical approaches to

198

10 Embracing the Mystery

the mind-body problem can be grouped under this umbrella option: neutral monism, dual aspect theory and traditional panpsychism. Aficionados will balk at such a crude assimilation and indeed there are important differences among these views as well as many possible sub-varieties of each (see Stubenberg 2008) but such details will not matter to our discussion here. Panpsychism is the most straightforward theory. It maintains that every basic entity has some mental properties and these properties are ontologically fundamental. Generally speaking, panpsychism accords the same status to the physical, leaving mind and matter as equally fundamental. Panpsychism is the natural foil of radical emergentist views. As we shall see, panpsychism is in some ways the minimal modification of the SPW which provides a solution to the problem and paradox of consciousness.6 Contrary to the strict dictates of panpsychism, neutral monism holds there is an underlying neutral foundation of reality to which both mind and matter reduce in some sense but which is itself neither mental nor physical. The doctrine is well expressed by one of its early champions, Bertrand Russell: The stuff of which the world of our experience is composed is, in my belief, neither mind nor matter, but something more primitive than either. Both mind and matter seem to be composite, and the stuff of which they are compounded lies in a sense between the two, in a sense above them both, like a common ancestor. (Russell 1921, p. 7)

The neutral remains ineffable and utterly mysterious, for we have no concepts save those which apply to the way the neutral appears to us and our instruments. According to this view neither mind nor matter are truly fundamental but they are so to speak relatively co-fundamental (nothing we have access to explains them and neither is ‘nearer’ to the neutral than the other). In the face of the need to provide some further theoretical articulation of the view and some characterization of the neutral itself, neutral monism faces the constant danger of slipping into either materialism or panpsychism. For example, the self proclaimed neutral monist William James eventually remarked in a notebook ‘the constitution of reality which I am making for is of the psychic type’ (see Cooper 1990 for provenance and discussion). And Russell declares, to my mind mystifyingly, that sensations are examples of the neutral.7 Dual aspect theory regards mind and matter as two equally co-fundamental aspects of an otherwise ineffable and inaccessible reality. Although this sounds much like neutral monism the theories should not be confounded. There is no canonical formulation of either view and they are frequently lumped together but despite being obviously closely related, we can, following Stubenberg, usefully distinguish them. Dual aspect accounts maintain that every element of fundamental reality is represented or expressed in both the mental and physical aspect, and deny that there is a reduction of the mental or the physical to this more basic ur-reality. It seems to me that dual aspect theory ‘solves’ the problem of consciousness by merely sidestepping around it. The pure form of dual aspect theory is a parallelism akin to that of Spinoza in which mind and matter are serenely and utterly independent of each other.8 No ‘problem’ of consciousness can arise because all linkages between the physical and mental realms have been severed. On this view, we necessarily have

10.3 Favour Fundamentality

199

access only to the mental realm and this in turn suggests that the whole physical side of the equation is an idle hypothesis. It is then hard to stop short of some form of idealism in which the physical realm is re-integrated with the mental as a mere construction out of mental building blocks. I do not mean to suggest that idealism is not a legitimate metaphysical position but it fails to pay sufficient attention both to commonsense belief in a robustly non-mental realm of being and to the incredibly beautiful, vastly expansive and explanatorily powerful scientific picture of the world. A solution to the problem of consciousness that disengages mind from the world that science presumes to study fails to get to grips with what makes that problem interesting in the first place. Neutral monism and panpsychism hold out the hope of more honest solutions, but in my opinion the latter more so than the former. Neutral monism posits an unknown something-or-other out of which emerges both matter and consciousness. Some versions, often labeled Russellian in honour of their inspirational source, posit some unknown intrinsic properties of the world as revealed by physical science which account both for the structural relations to which our physical science is limited and from which consciousness can also emerge (see Lockwood 1989 and Stoljar 2006 for modern discussions). As mentioned above, such views have a tendency to either privilege the physical over the mental insofar as the unknowable intrinsic feature can be regarded as a kind of physical property, or to slide towards panpsychism insofar as the neutral is taken to be more akin to mentality. Panpsychism abjures the hypothesis of the unknowable background of the neutral. Reality is instead exhausted by the two known co-fundamental categories of being: matter and consciousness. Panpsychism agrees with traditional neutral monism that neither mind nor matter is more fundamental than the other. But it avoids positing the mysterious relation of reduction by which both consciousness and the physical realm somehow emerge from the neutral. There is a straightforward argument in favour of panpsychism whose form can I think be traced back at least to the Presocratic philosophers who faced the general problem of emergence in the 5th century BCE (see Mourelatos 1986). More recently, Thomas Nagel has presented a succinct and elegant version which I adapt here (Nagel 1979; the argument is also endorsed in Strawson 2006). Nagel’s argument is naturally integrated into our discussion of emergence. It asserts, first, that the only sort of emergence which is possible is conservative emergence, but, second, that it is impossible that consciousness should be a conservatively emergent feature of the world. In support of the second premise, Nagel’s basic strategy is to appeal to the explanatory gap that lies between matter and consciousness. Conservative emergence implies that there is, in principle, an explanation of the mechanisms of emergence, but there is no such explanation in the case of the transition from brute matter to consciousness. With respect to the first premise, why does Nagel also believe that radical emergence is impossible? He does not provide any argument for this and, as we have seen above, it is hard to see why radical emergence is incoherent or metaphysically impossible (even if we cannot point to a single non-controversial example).

200

10 Embracing the Mystery

If we grant for the sake of the argument that the only coherent doctrine of emergence is restricted to conservative emergence and if we follow Nagel in the claim that consciousness does not or cannot conservatively emerge from physical structures and processes then the conscious mind cannot be an emergent from the physical world. Yet, Nagel proceeds, mental properties are properties of organisms which are physical entities. Thus, ‘if the mental properties of an organism are not implied by any physical properties but must derive from properties of the organism’s constituents, then those constituents must have nonphysical properties from which the appearance of mental properties follows when the combination is of the right kind’ (Nagel 1979, p. 182). If these properties—the properties of the constituents—are not themselves mental then the problem of emergence will simply reappear for them. Thus the constituents’ properties must be mental. Furthermore, Nagel claims, since we can ‘build’ an organism out of any kind of matter, all matter must possess these elementary mental properties.9 That is, panpsychism is true. This vision of panpsychism accepts the SPW almost to the letter, with the one caveat that science has missed some of the basic properties which characterize its own fundamental entities. It accepts the general story of emergence which the SPW propounds with, again, the single reservation that some emergents are the result of the elementary mental features of matter via, in Nagel’s words, ‘a kind of mental chemistry’ (1979, p. 182). Panpsychism thus pays its respects to the SPW by making the minimal change needed to accommodate consciousness. It is odd for a view that denies physicalism, but in essence both the methodology and the general world view of the SPW is accepted by panpsychism. The overall coherence, power, elegance and motivation of the SPW can be accepted with only a small, albeit crucial, alteration by the panpsychist. Despite these virtues, panpsychism gets little respect from philosophers. For example, John Searle describes panpsychism as an ‘absurd view’ and (question beggingly) asserts of examples such as thermostats that they do not have ‘enough structure even to be a remote candidate for consciousness’ (Searle 1997, p. 48). Colin McGinn labels panpsychism either ‘ludicrous’, for a strong version which asserts that everything has full fledged consciousness, or ‘empty’, for a weak form which asserts only that everything is made of physical stuff which has the capacity to be a constituent of a system with mental properties (McGinn 1999, p. 95 ff.). Leaving aside the fact that given the view that complex consciousness emerges from elemental consciousness there is no need to hold that everything possesses consciousness (and very few panpsychists have asserted this), why is panpsychism the object of such scornful, low-content, ridicule? While there are many objections to panpsychism, which is of course a highly speculative doctrine, I think the principal reason that philosophers make fun of it is methodological. Panpsychism has, in the cutting phrase of Thomas Nagel, ‘the faintly sickening odor of something put together in the metaphysical laboratory’ (Nagel 1986, p. 49). Once the SPW is reasonably well articulated there is a natural philosophical project of completing it and deploying it in the overall project of metaphysics, which is, roughly ‘to understand how things in the broadest possible sense of the term hang together in the broadest possible sense of the term’ (Sellars 1963b, p. 1).

10.3 Favour Fundamentality

201

Given the cultural prominence of science, philosophers have strong motivations to devise a world view fully consonant with the SPW. To endorse panpsychism is to admit the failure of this project, somewhat like a climber giving up and taking a helicopter to the summit. The project of completing the SPW is an undeniably worthy one and no one can say with certainty that it is unattainable. But I have argued that the option of Watchful Waiting is not going to be able to solve the paradox of consciousness. This at least suggests that any effort to save the SPW will require a complete overhaul of its metaphysics which will, first, somehow undercut the argument that high level phenomena are epistemic or explanatory potentia without any independent being and, second, find a way to understand high level phenomena as genuinely efficacious without undermining the totality of physics (that is, its commitment to completeness, closure and resolution). I think achieving all this will be a very tall order. To take one obvious difficulty, most if not all high level features are vague. It is hard to see how vagueness could be an objective property out there in the world, especially if the underlying ontology, given by fundamental physics in whatever form it will eventually take, is fully determinate.10 On the other hand, it is equally difficult to think about high level objects without imputing vagueness to them. Consider the property of ‘being a mountain’. How could it be that there is some critical piece of information which settles whether some object at the borderline between mountain and hill is really one or the other? It seems intuitively obvious that we could know everything about the object’s height, the heights of surrounding objects, all social and linguistic facts relevant to the word ‘mountain’ and still be faced with a borderline case of mountainhood.11 It is easier to believe, in accord with the standard SPW, that our concepts are indeterminate, a natural inheritance from their source as mere explanatory aids to be applied as needed and strengthened or weakened at our whim (as in the example of Pluto’s demotion to dwarf planet). But even supposing there were vague objects we also need to provide them with genuine causal efficacy to avoid generalized epiphenomenalism. It is very hard to believe that there is some actual causal power which mountains possess but which borderline non-mountains do not. What could it be? On the other hand, if there were such a litmus test for category inclusion we could use it to definitively answer the question whether some large lump of matter was a mountain or not thus presumably handing victory to the epistemicists (see note 11 above) about vagueness after all. This possibility strikes me as ludicrous. The central point is that without the standard SPW this issue and a myriad of others turn into bizarre scientific questions. One is at a complete loss as to exactly how they could be tackled by science. Not only does there seem to be no way to articulate any scientific project of determining what is a genuine mountain, it is a project which is obviously scientifically completely worthless. Questions of this nature may well be genuine questions. There is no doubt that the proper understanding of vagueness is an extremely difficult problem. But on the standard SPW, they are properly left to philosophy with the tacit understanding that the philosophical solution will fit smoothly into the SPW and may draw inspiration and crucial empirical data that bears on the issue from ordinary science. It is not a scientific question, for example, whether

202

10 Embracing the Mystery

ordinary objects such as chairs, mountains and tectonic plates exist or not. Notice that I include a scientific concept in this list. There is, of course, a scientific question whether tectonic plates exist which has already been answered in the affirmative. These are questions that come up once the epistemic resources of modern geology have been marshaled. The philosophical question is of a different order, one rather more directed at the status of the conceptual system as a whole. If the Watchful Waiting option thus requires a foray into radical metaphysics then, in a way, Favouring Fundamentality actually holds truer to the spirit of the SPW than Watchful Waiting. But can it solve the problem and paradox of consciousness? Aside from panpsychism’s supposedly obvious implausibility it faces some serious objections of which only one need concern us here (for a more comprehensive discussion see Seager and Allen-Hermanson 2008). According to this objection panpsychism is simply emergentism in disguise. The problem goes back to William James in his criticism of a form of panpsychism which, with characteristic aptness, he called the mind dust theory. The worry is that even if we suppose that consciousness is a fundamental feature of the world we still have to explain how complex minds—the only sort we have any acquaintance with— come into being and we have to explain their relationship with the elemental mental features. This looks like a job for emergence, but how? As James complains: ‘Take a sentence of a dozen words, and take twelve men and tell to each one word. Then stand the men in a row or jam them in a bunch, and let each think of his word as intently as he will; nowhere will there be a consciousness of the whole sentence.’ (James 1890, p. 160). It is worth quoting at some length James’s specific worries about consciousness: Where the elemental units are supposed to be feelings, the case is in no wise altered. Take a hundred of them, shuffle them and pack them as close together as you can (whatever that may mean); still each remains the same feeling it always was, shut in its own skin, windowless, ignorant of what the other feelings are and mean. There would be a hundred-and-first feeling there, if, when a group or series of such feelings were set up, a consciousness belonging to the group as such should emerge. And this 101st feeling would be a totally new fact; the 100 original feelings might, by a curious physical law, be a signal for its creation, when they came together; but they would have no substantial identity with it, nor it with them, and one could never deduce the one from the others, or (in any intelligible sense) say that they evolved it. (p. 160, my emphasis)

Evidently, James has doubts that Nagel’s ‘mental chemistry’ even makes sense in the case of consciousness. But notice he also considers the option of radical emergence wherein the new emergent feeling is directly caused to occur by the conglomeration of the simple feelings. James does not say this is impossible but that it will not provide any way to ‘deduce’ the one from the other (this is one of the hallmarks of radical emergence). That is, there is no way to see this process as a case of conservative emergence. This is the objection: the only way to put together a panpsychist theory on the model of the SPW requires the admission of radical emergence, which is directly contrary to that model. The way that panpsychism is supposed to mimic the operation of the SPW turns out to be illusory. Furthermore, if panpsychism requires the

10.3 Favour Fundamentality

203

admission of radical emergence then why not simplify the theory and let complex consciousness radically emerge directly from the physical? It may well be possible to counter this argument. James has a very narrow conception of conservative emergence which is very distant from what is recognized in modern science and especially quantum mechanics (as discussed above in Chap. 6, pp. 150 ff.). James’s view of conservative emergence stems for a picture of the physical world more akin to C. D. Broad’s conception of extremely austere mechanism, as discussed in Chap. 5, than that of the SPW. I have argued elsewhere that the role of information in quantum entanglement might serve as a model for a kind of conservative emergence of complex mental states out of elemental forms (see Seager 1995). But even supposing we accept that there is a viable form of conservative emergence available to the panpsychist, this is just to jump from the frying pan into the fire. For conservative emergence is all that is required to rob complex consciousness of its efficacy and to generate the paradox of consciousness. The panpsychist hypothesizes that there are elemental mental properties which belong to the fundamental physical entities of the world. If these elemental features have their own causal powers (that is, are not themselves epiphenomenal) then by the logic of conservative emergence they will usurp efficacy from the complex conscious states which they subvene. Furthermore, if complex consciousness conservatively emerges in anything like the standard way endorsed by the SPW, albeit via mental chemistry, then the paradox looms, since the natural way to regard conservative emergents is as stance dependent conceptual/epistemic resources. There is an elegant way around this problem available to the panpsychist. Recall in our discussion of conservative emergence we allowed for the possibility of the emergence of new ‘large simples’—entities which are the lawful consequence of the interaction of initial constituents but which ‘absorb’ or supersede these to become self standing entities in their own right (see Chap. 7, n. 23).12 This possibility does not violate the strictures of conservative emergence nor those of the totality of the underlying theory (completeness, closure and resolution). Such ‘large simples’ would show up in a simulation restricted to the state and laws of the fundamental account. Perhaps within the realm of classical general relativity black holes are examples. They arise via physical law from a set of elementary progenitors but take on their own identity as fundamental entities characterized by only three basic physical properties: mass, charge and angular momentum. Now, this is nothing more than an example because general relativity is not the final physics and it seems very likely that whatever the final physics is it will assign constituent structure to black holes (strings on the black hole boundary, or something). Nonetheless the example is highly illuminating insofar as it shows there is nothing incoherent about the conservative emergence of large or macro simples. The panpsychist can then argue, very speculatively indeed, that the elementary mental features associated with the elementary physical features of the world can, when combined in the appropriate way, as in for example human brains, generate a new large simple mental entity. It is such entities which are the foundation of the sort of complex consciousness with which we are familiar in introspection. As a new

204

10 Embracing the Mystery

simple, with no mental constituents (though with a history of formation from a set of elemental mental features), this entity can have its own efficacy without violating the strictures of conservative emergence.13 But note that Watchful Waiting cannot avail itself of this argument to save the SPW’s requirement that all emergence be conservative emergence. It is true that large simples can conservatively emerge but without any elemental mental features to draw on no mentalistic large simple can emerge save by radical emergence. Of course, it remains true that the behaviour of the world envisaged by the panpsychist will be in principle empirically distinguishable from that of a purely physical world. In terms of our simulation thought experiment, the purely physical simulation would fail to duplicate the actual world’s evolution if this sort of panpsychist account is correct. On the principle that one might as well be hung for a goat as a sheep, I regard this as a virtue of the account and in fact it seems to be virtually a requirement to avoid the paradox of consciousness. Without empirical distinguishability the arguments for generalized epiphenomenalism will reappear in the mentalistic realm posited by the panpsychist. This is not to say that there is any practical prospect of an experiment which would test panpsychism. We should not expect the elemental mental features to necessarily have any measurable effects at the micro-level, any more than we expect to be able to detect the gravitational field of a single electron. At the macro-level, there is no way to definitively tell whether the human brain, for example, ever violates any physical laws such that its state evolution diverges from that which would result from the operation of a purely physical model of it. There is all the abundant indirect evidence that the brain is a physical system like any other, albeit of immense complexity, but it would be hard to imagine any experimental test which could directly verify this short of demonstrated non-physical parapsychological effects.14 Both the option of Favouring Fundamentality and Embracing Emergence have the consequence that, in addition to any perceived metaphysical defects of the SPW, physical science provides an empirically false account of the world which, in principle at least, could be revealed via standard experimentation. This is to their credit as genuine hypotheses about the nature of the world. But they face up against the long and surprisingly smooth history of explanatory success which the SPW has enjoyed. After some five centuries of concentrated effort our science has come to encompass virtually all of reality with which we can empirically grapple. There is no sign whatsoever that this pattern of success will not continue and deepen as we unravel how the cosmos began under the dictates of a fairly small set of basic physical entities, laws and processes. Even more impressive is the hierarchy of conservatively emergent higher level features: properties, entities and laws. Each element of this hierarchy appears to connect to the basic level in intelligible ways with multiple lines of explanatory interconnection. Although far from complete and facing impossible problems of sheer complexity, the overall picture of the world which science has developed and which was outlined in Part I gives every appearance of a seamless whole. One would have to be very brave or perhaps foolhardy to bet against the continued smooth growth of the SPW.

10.4 Modify Metaphysics

205

10.4 Modify Metaphysics It is thus especially interesting that there is a way to accept the explanatory power of the standard SPW without endorsing Watchful Waiting. This way avoids the paradox of consciousness and sidesteps the problem of phenomenal consciousness. Label this approach: Modifying Metaphysics. The modification in question stems from the observation that there is an implicit assumption lurking within all the other options and the SPW itself. The assumption is scientific realism, which is, roughly, the thesis that science aims at and is providing the truth about the ‘deep structure’ of the world. It is obvious that the option of Watchful Waiting endorses scientific realism. It goes so far as to identify the scientific project of the SPW with the metaphysical quest to discover the ultimate nature of reality. According to it, is only a matter of time before all features of reality will take their rightful place within the system of conservative emergence founded on the ultimate reality which is the basic physical structure of the world. But why do I say that Embracing Emergence and Favouring Fundamentality also endorse scientific realism? Because both these views take it as a given that consciousness has to be integrated with the SPW with as little disturbance as possible. It is admitted that both of these options deny that science tells us the whole truth about the world. According to them science has either missed the existence of radical emergence or the existence of an extra set of fundamental mental features of the world. But they both nonetheless set themselves up as ways to integrate consciousness into a minimally tweaked SPW. At bottom, both options agree with the core of the philosophical response to the SPW, which essentially involves, in the words of Bas van Fraassen a ‘strong deference to science in matters of opinion about what there is’ (van Fraassen 2002, p. 48). It is possible instead to interpret the difficulties the SPW has with consciousness as indicating not the need for some extra parameter we can bolt onto it but rather as hinting that the core assumption of scientific realism is dangerously overstressed and cannot bear the theoretical weight the problem of consciousness loads upon it. Obviously, I cannot lay out a fully developed anti-realist or arealist account of science here. But drawing on the substantial work of Bas van Fraassen (1980; 2002), with more indirect help from John Dupré (1993) and Nancy Cartwright (1999), we can outline how the problem of consciousness is transformed when we reject scientific realism.15 The most fully worked out anti-realist position is the constructive empiricism of van Fraassen. Fundamentally, his account of science replaces the quest for truth with the search for empirical adequacy which is the epistemic virtue of ‘saving the phenomena’. Empirically adequate theories make correct claims about observable reality, notably the readouts of our instruments, but also about the whole range of phenomena that underpin modern technology. Typically, scientific theories also make claims about unobservable entities but the anti-realist puts no credence in these, or at least does not hold that the empirical success of the theory in the observable domain should necessarily lead one to accept claims about the unobservable.

206

10 Embracing the Mystery

Such an account of science does not rule out belief in unobservables, nor does it deny that the empirical success of science might provide evidence in favour of its claims about the unobservable. In fact, it is obvious that science does provide such evidence. Within philosophy there is a long standing debate about the upshot of this evidence. To my mind there is a pre-existing intuition that unobservable entities exist and this means that the empirical success of science which postulates these entities deserves considerable—but limited—epistemic respect. I don’t think however that this forces us to embrace the SPW. Instead, I think one should regard the scientific enterprise as one of model building with the object of the model being to generate correct descriptions and predictions of observable and measurable phenomena. These models are usually idealized and simplified theoretical representations of some portion of reality. The scale of such models is breathtakingly wide, ranging from ultra-microscopic vibrating string-like entities to the entire universe. The SPW lays claim to the idea that there is what might be called a total model— one in which, as its defenders see things, every facet of reality is included. Of course, this is not the claim that there is any practical prospect of our deploying such a model to represent any significant region of space and time. We know that the overwhelming complexity of even very tiny systems absolutely precludes any such thing. But that is not the point of the exercise, which is philosophical and not scientific. The idea is that the concept of a total theory is coherent and this underwrites the ‘metaphysical’ possibility of a complete model.16 The option of Modifying Metaphysics as I see it need not and does not deny the coherence of the idea of a total model in this philosophical sense, that is, a model which in principle is entirely empirically adequate. What it denies is the step from the model to the claim that it accurately represents every feature of reality or captures the metaphysics of genuine efficacy. The paradox of consciousness shows that it does not and cannot. The SPW seems to be very similar to what Nancy Cartwright calls ‘fundamentalism’ about physics. I’m not completely sure about this because, as we have observed with other thinkers, it is actually hard to tell whether she is making an epistemological or ontological point. Only the latter could threaten the SPW. Cartwright characterizes fundamentalism as the view that ‘all facts must belong to one grand scheme and moreover that this is a scheme in which the facts in the first category have a special and privileged status. They are exemplary of the way nature is supposed to work. The others must be made to conform to them’ (Cartwright 1999, p. 25). Although Cartwright describes the first category of facts rather broadly, as ‘those that are legitimately regimented into theoretical schemes’ (p. 24) and does not spell out exactly what she requires for one fact to conform to the set of first category facts, there is a clear affinity with the SPW. What is exemplary of how the world works is the way the world works according to fundamental physics and all other facts must conform to the fundamental physical facts in the sense that the former are completely determined by the latter via the completeness, closure and resolution of the total theory of final physics.

10.4 Modify Metaphysics

207

The option of Modifying Metaphysics emphatically agrees that this is the core mistake which both engenders the problem and paradox of consciousness and makes them completely intractable. But denying this mistaken understanding of the nature of science is compatible with the possibility of an in principle model which is empirically adequate. It is thus somewhat curious that Cartwright sometimes appears to deny the in principle existence of the total model which the SPW endorses. She uses an example of a thousand dollar bill dropped out of a high window. Obviously, there is no practical, usable model in any science, or any combination of sciences for that matter, which will predict where the bill will land with precision. She says that the fundamentalist will insist that ‘there is in principle (in God’s completed theory?) a model in mechanics for that action of the wind, albeit probably a very complicated one that we may never succeed in constructing’ (p. 27). The defender of the SPW certainly agrees with the fundamentalist here. We can use our usual metaphor here: is there an in principle possible computational simulation (under relaxed computational constraints) of the physical system in question which will accurately track the trajectory of that thousand dollar bill? Does Cartwright actually deny the possibility of the ‘total model’? Her discussion is maddeningly elusive. Of fluid mechanics (in relation to the thousand dollar bill problem) she says ‘it does not have enough of the right concepts to model the full set of causes, or even all the dominant ones’ (p. 27). But the SPW uses God’s ultimate model, that is, fundamental physics, from which fluid mechanics conservatively emerges and is quite distant from the basic level. Is it possible to read Cartwright as endorsing either the Favouring Fundamentality option or that of Embracing Emergence? Both entail that fundamental physics is not a total theory. It either fails to include primitive causally efficacious features that need to be added to the fundamental level of reality or fails to include causally efficacious radically emergent high level features that would not appear as consequences of the pure model generated by basic physics. Cartwright explicitly accepts the existence of unobservable entities postulated by scientific theory. She has written that her book The Dappled World (Cartwright 1999) ‘defends scientific realism; well-supported claims of science have as good a claim to truth as any’ (Cartwright 2001, p. 495). But she goes on to add that while there are ways to ‘reconcile how successful science is practiced with the metaphysics of the single unifying theory…[t]he one great theory is not incompatible with the evidence; it is just badly supported by it’ (p. 495). The practice she has in mind is the way scientists deploy a grab bag of disparate theories and technologies in the design, setup and explanation of their experiments. I think that the reconciliation strategies alluded to involve the commitment to conservative emergence of the high level features to which these disparate theories appeal coupled to the uncontroversial appeal to the staggeringly vast complexity of the high level phenomena when considered from a low level viewpoint. If so, Cartwright is implying that the evidence for conservative emergence is weak. Much of this book disputes this. The evidence for conservative emergence is indirect but I don’t think it is either weak or insignificant. Save for the problem and paradox of consciousness, where is there any evidence against the metaphysical

208

10 Embracing the Mystery

position which holds that the entities described by basic physics form a complete, causally closed system into which all other features can be resolved? Once Cartwright admits that we have good evidence for the existence (and nature) of the entities postulated by fundamental physics it will be hard to resist the slide towards the totality of physics. The only way to stop the slide is to deny that the structure of the world posited by basic physics possesses completeness, closure and resolution. This, in turn, seems to require adopting some form of our two radical options of Favouring Fundamentality or Embracing Emergence. It may be that in the work of John Dupré (e.g. Dupré 1993) we find something like this dialectic worked out to its natural conclusion. Like Cartwright, Dupré is keen to deny that any level in the ontological hierarchy is to be privileged and he launches a detailed and even passionate attack on reductionism. However, the reductionism he attacks is one which focuses on the epistemic issue of the in principle possibility of identifying the entities of one level with those of other levels, whereas the core issue here is whether physics describes (or at least aims to describe) that level of reality which determines every other level according to the constraints of completeness, closure and resolution. Physics does not need to maintain that there is some way to identify collections of sub-atomic constituents with, for example, particular mountains; it need only make the claim that things like Mt. Everest depend entirely upon low level physical entities for their existence and properties. The upshot is that, as in Cartwright’s case, I am not completely sure what Dupré’s attitude is towards the SPW in its pure form. But he does make some highly suggestive claims. With respect to the condition of closure he says that ‘the central purpose of the ontological pluralism I have been defending is to imply that there are genuinely causal entities at many different levels of organization. And this is enough to show that causal completeness at any one level is wholly incredible’ (Dupré 1993, p. 101). Of course, many initially incredible things turn out to be true—especially in the domain of modern science, so this is a somewhat odd, or at least very careful, choice of phrase. Nonetheless, I think the natural reading of this is the denial that all the causal powers of high level entities are completely determined by those of the lower level entities and processes. Since Dupré, like Cartwright, professes to be a scientific realist, it appears that his position will have to endorse one of our radical options and, it would seem, the favoured option for both of them would be that of Embracing Emergence. We can see that it is the acceptance of scientific realism that leads Cartwright and Dupré down the garden path towards radical emergentism. The denial of realism does not say anything about how close to empirical adequacy the model suggested by basic physics can get. However, it does suggest that interest in this question should largely evaporate. Obviously, there is absolutely zero prospect of ever developing a working version of the model which could make any predictions. Thus, without the sense (which many physicists almost instinctively possess) that one is laying bare the ultimate nature of reality the significance and appeal of the total model quickly fades away. Without the underlying commitment to scientific realism there is no way to leverage the indirect support for conservative emergence into support for the existence of the total model. Instead, this evidence simply forms a web of

10.4 Modify Metaphysics

209

interconnected theories and models which serve to generate theoretical explanations, empirical predictions and, of course, new sources of ideas for novel technologies. Furthermore, the anti-realist explains the interconnectedness of all our models in terms of the need for science to generate empirical content. We can borrow the evocative metaphors of Cartwright: the anti-realist sees a ‘dappled’ world of diverse sorts of entities which can be somewhat organized under a ‘patchwork’ of laws of nature. And we can follow Dupré in denying that there is a complete hierarchical structure to reality in which some levels are ontologically privileged. Despite the coherence of a stance which rejects scientific realism, it is hard to articulate in any way that seems plausible. This is in part due to the undeniably immense power and elegant attractiveness of the metaphysical picture suggested by the SPW. In addition, there is a kind of cultural inculcation of the idea that science is in the business of searching for ultimate reality.17 After all, is it not simply obvious that all material objects are constituted out of components which determine their properties? And if that is false then must it not be the case that certain assemblages of matter have radically emergent properties? The option of Modifying Metaphysics requires that if these questions are understood as aimed at metaphysical goals then they must be rejected. Ultimately, existence is a mystery. Pursuing the dream of the SPW in the hope of solving this mystery is a hopeless quest. ‘The most perfect philosophy of the natural kind only staves off our ignorance a little longer’ wrote David Hume in 1748 (2000, p. 28) and in some areas we must be content to remain very close to this irresolvable mystery. This does not preclude researchers in neuroscience from investigating and perhaps discovering the so-called neural correlates of consciousness nor the potential development of devices to read minds or gauge a subject’s conscious state. But there will be no model of the conservative emergence of consciousness. The success of science in staving off mystery does not get us one step closer to integrating consciousness into the SPW, for the very methodology which funds its success leaves it unable to grapple with consciousness. Perhaps the option of Modifying Metaphysics enjoins a kind of quietism about consciousness: take it as given, a part of the natural world and, in a way, metaphysically primitive. It is open to standard scientific investigation; we can find the neurological conditions which underpin it. But we cannot fit it into the SPW as a conservative emergent and if we abandon scientific realism we are under no obligation to do so. I cannot bring myself to endorse any of the options we have studied. It seems obvious that the mainstream response is and will be to go with Watchful Waiting and insofar as this primarily enjoins extending science in the normal way it has some obvious virtues. Perhaps with enough accumulated knowledge of the brain and its environment and measurable links to a host of mental states, the scales will fall from our eyes and consciousness will take its place alongside chemistry and life as a standard conservative emergent. But I think the problem of consciousness and most especially the paradox of consciousness makes it very hard to see how this could happen.

210

10 Embracing the Mystery

What of the two radical options of Embracing Emergence and Favouring Fundamentality? These seem to me to suffer from the urge to minimize the disturbance consciousness introduces into the SPW yet both end up adding elements to the scientific picture for which there is no empirical evidence. The final, most radical, option avoids saddling science with radical emergence or new elementary features, but its cost is to demote science from a metaphysical guide to the nature of reality to an epistemic project of better dealing with the observable world. The scientific project has a distinctive methodology which is that of building a hierarchy of conservative emergence. The Modifying Metaphysics option asserts this methodology cannot successfully grapple with consciousness. So it enjoins us not to attempt the impossible, to accept consciousness as a part of the natural world which is a metaphysically primitive mystery leaving us free to link consciousness with our rich system of scientific models, regarded as nothing more than models. It seems that, broadly speaking, the four options outlined here exhaust the possible responses to the problem of consciousness. I wish I could show that one of these options was definitively correct. Failing that, I can only leave to the reader the job of picking the winner.

Notes

Chapter 2 1

Because of the precession of the Earth’s axis of rotation, Polaris has been located reasonably close to the celestial pole only for the last 2000 years or so. In ancient Egypt, Thuban was the north star but it now resides some 25 from the celestial north pole. 2 Our imaginary astronomers would have another method at their disposal: the transits of the sun by the inner planets, Mercury and Venus. By measuring the apparent position of, say, Venus against the backdrop of the Sun from different locations on Earth, it is possible to use parallax to determine the Earth–Venus distance. Kepler’s laws then allow one to deduce the distances of all planets from the sun. But such measurements are difficult to make and require a very accurate determination of the orbital periods of the planets. How could this be determined in the absence of a fixed backdrop of stars? Incidentally, a search of the Internet will discover a fascinating ‘movie’ of the 1882 transit of Venus which has been assembled from still photos taken at the Lick Observatory. 3 The hypothesis was anticipated by Newton in his first letter to the theologian Richard Bentley (see Janiak 2004, Chap. 4; Newton’s correspondence with Bentley can also be found at The Newton Project (http://www.newtonproject.sussex.ac.uk); see Larson 2003 for an overview of current accounts of star formation). 4 But as of 2009 Pluto has been officially demoted to mere dwarf planet, initiating a novel method of solar system exploration: complete the project of visiting all planets by pure semantics! So much cheaper than building spacecraft after all. 5 Protoplanetary disks are now routinely studied. See Lagage et al. (2006) or visit the Subaru telescope web page: http://www.naoj.org/Pressrelease/2004/04/ 18/index.htm 6 Over seven hundred as I write this, but more are discovered almost daily. See the Internet Extrasolar Planets Encyclopedia (http://exoplanet.eu). 7 Our knowledge of stellar ages results from a wonderfully clever combination of the astrophysics of star composition and the statistical analysis of the

W. Seager, Natural Fabrications, The Frontiers Collection, DOI: 10.1007/978-3-642-29599-7,  Springer-Verlag Berlin Heidelberg 2012

211

212

Notes

correlations between the luminosity and color of stars, as codified in Hertzsprung– Russell diagrams. Such diagrams of globular clusters, which can be assumed to contain stars of roughly the same age (whatever it might be) are highly distinctive and suggest all by themselves that the clusters are very old. 8 See http://www.aip.de/People/MSteinmetz/E/movies.html for some spectacular simulation runs. 9 Measurement of the distances to galaxies is another fascinating aspect of astronomy. Galaxies are much too far away for the parallax technique. But it is possible to measure the luminosity of some individual stars in nearby galaxies and compare them to similar stars in our own galaxy where we have a better handle on distance measurements. Astronomers have constructed an intricate ‘ladder’ of overlapping distance indicators which let them assign distances to a host of galaxies, from which they can establish the Hubble law and then use it to assign distances to galaxies yet further away. 10 In a curious, if faint, echo of the Lonely Earth thought experiment, our ability to detect the background radiation may be a relatively parochial feature of our present temporal location in the universe. Given the accelerating expansion of the universe imposed by the so-called ‘dark energy’ (see Glanz 1998; Carroll 2004) there will come a time when the CMB will be so ‘stretched’ out that it will be undetectable and this time is, on a cosmic scale, not too far in the future. Perhaps in a mere 150 billion years or so the denizens of our galaxy will see only the galaxy drifting in an otherwise empty universe (see Krauss and Scherrer 2008). A truly optimistic spirit would point out that future cosmologists would have the history of cosmology available to them. The contention of Krauss and Scherrer also suggests a new use of Brandon Carter’s infamous doomsday argument (Carter 1983). Suppose, first, that we represent a random selection from all the creatures capable of undertaking scientific cosmology. For reductio, suppose second that such intelligent creatures will exist throughout the life of the universe (when, of course, conditions allow for the existence of such observers at all). If, following Krauss and Scherrer, during almost all of the life of the universe there will be no evidence available to cosmologists about the overall state of the universe, in particular that it is expanding, then we would expect that we—as random observers within that history—would not be in a state where universal expansion is detectable. Since we can detect this expansion it ‘follows’ that cosmology capable observers will not survive for very long (in cosmic terms). For a thorough philosophical discussion of Carter’s argument see John Leslie (1998). 11 See http://lambda.gsfc.nasa.gov/product/cobe/firas_image.cf. The authors state ‘The FIRAS data match the curve so exactly, with error uncertainties less than the width of the blackbody curve, that it is impossible to distinguish the data from the theoretical curve’. 12 Arp is no less infamous for the controversial and disturbing account of his professional ostracism occasioned by apostasy from Hubble orthodoxy. See Arp (1987) or the later (but equally inflammatory) Arp (1998). 13 See http://www.lightandmatter.com/html_books/7cp/ch04/ch04.html for a nice example.

Notes 14

213

Incidentally, the value of this ratio puts constraints on the overall structure of the universe. The relative amount of deuterium is a function of the density of baryonic or ‘ordinary’ matter in the very early universe, and the observed ratio implies that the density of the universe is only 10% of the ‘critical value’—the value that would geometrically close the universe. Recent observations suggest that the universe is not closed, but that the density is something like one third the critical value. Hence there must be a lot of ‘missing mass’, an inference that fits in with much other data suggesting that the universe is mostly unseen ‘dark matter’ and the so-called ‘dark energy’ responsible for the apparent acceleration in the rate of cosmic expansion (see Glanz 1998; Seife 2005). 15 It is important to bear in mind, however, that this apparent simplicity could be an artifact of our inability to closely observe the details of the early universe. The further we move away from rich sources of data the more dependent we become on the models we devise of the target phenomenon. It may well be that it is our models of the early universe which are (relatively) simple rather than the universe itself. 16 If the half-life of a proton is, say, 1031 years then a tank containing that many protons should reveal about one decay every year. That’s not a hugely big tank of water (approximately 100 cubic metres). The actual observations are conducted in vast chambers far underground and in them one would expect to see proton decays every few days. No unequivocal events of proton decay have ever been observed. 17 Essentially, inflation ‘steals’ this energy from the gravitational field energy of the universe which is negative. Hence the overall energy balance of the universe remains constant. Curiously, it is quite possible that the total energy of the universe is exactly zero! See Guth (1997), Appendix A, for a nice qualitative argument that the energy of a gravitational field is negative. 18 The experiments at Brookhaven involve smashing gold nuclei together at very high energies and have the curious distinction of threatening the existence of the entire universe. At least, a few people suggested before the heavy ion collider began its runs that there was a chance that the experiment could produce ‘strangelets’—a form of matter composed of quarks with the quantum mechanical property of ‘strangeness’—which could set off a chain reaction converting all matter in the universe to strange matter, destroying all life etc. in the process. This hypothesis—worthy of the Hitchhiker’s Guide to the Galaxy—was stated at the time to have an extremely low probability, a reply which, if the probability is nonzero, leads to interesting speculations about the expected utility of the experiments. The worry was apparently set to rest in 2000 by the calculation that Brookhaven could, at worst, produce positively charged strangelets while only negative charged ones could pose a danger (see Madsen 2000). 19 Peter Higgs (among others) worked out the mathematics to show how particles could acquire mass via interaction with a special kind of field in the 1960s. No one yet knows whether nature really uses this mechanism, but the particle quantum mechanically associated with the field, called the Higgs boson, ought to be detectable by next generation accelerators such as the large hadron collider (LHC) at CERN is now in operation. There were some hints in 2000 that CERN’s large electron positron collider (LEP) had detected the Higgs boson, but the claim, based

214

Notes

on the statistics of certain quite limited experiment outcomes, could not be verified before the LEP was shutdown to make way for construction of the LHC (see Abbott 2000). In December 2011, CERN announced preliminary data that, if confirmed, would represent the discovery of the Higgs boson (see http://press.web. cern.ch/press/PressReleases/Releases2011/PR25.11E.html). In July 2012, further data have made the discovery of the Higgs boson virtually certain. 20 There have been conjectures that certain physical constants are not really constant. If these turn out to be correct then the basic physical properties of protons and other quite fundamental particles may well have changed over time. Perhaps, that is, the mass, or the charge, of the proton in not quite the same as it was shortly after the big bang. But obviously such conjectures do not support in any way the idea that individual protons have acquired biological properties sometime in the last 12 billion years.

Chapter 3 1

Now available in a CDROM edition, Hooke (1665/1998). 2 One can’t help but note the analogy between this puzzle and the more fundamental asymmetry of matter over anti-matter in the universe even though physical processes almost universally do not favor production of the one over the other. 3 For a review of the relevant brain mechanisms and an interesting philosophical application of them in the understanding of pleasure and desire, see Schroeder (2004). 4 Some day the machines may not have to be so ungainly. Recent work on imaging with very weak magnetic fields (on the order of 30 milli-Teslas for the main magnet) where the measurement is effected by ultra sensitive superconducting quantum interference devices (SQUIDs) has already produced rudimentary medical images (see Zotev et al. 2007). 5 There are two distinct relaxation times in fact, labeled T1 and T2, which reflect different physical processes (basically interaction of proton spins with each other and with the atoms in the subject). Image extraction exploits the differences between T1 and T2 and how they vary with tissue type and imposed magnetic fields. This is extremely complex applied physics, but it is nothing more than physics. 6 A much older imaging method, positron emission tomography (PET), also tracks metabolic processes (see Brownell 1999). It is another anchor point, if you will. PET works by injecting a radioactive ‘tracer’ into a subject which insinuates itself into the chemical processes that constitute cell metabolism. The tracer element decays by emitting positrons-the anti-matter version of the electron-which are almost instantly annihilated when they encounter nearby electrons. The destruction of the positron/electron pair yields a pair of gamma ray photons that travel in exactly opposite directions (so the net momentum is zero as it must be). The detector need only watch for coincidental pairs of gamma rays. An image can then be constructed from information about where the coincident pairs of gamma

Notes

215

rays are detected (essentially, the locus of activity will be where the trajectories of the gamma rays intersect). Various tracers are best suited for studying different processes. Brain PET scans use a radioactive isotope of oxygen and indirectly measure brain activity by measuring oxygenation. In this, they are similar to fMRI. PET scans are of relatively low resolution, somewhat invasive and very expensive (a cyclotron is required at the facility to produce the short lived radioactive tracers). MRI is now the much more common technology. 7 This is severely oversimplified. There are a number of interacting parameters at work here that have to be disentangled during image acquisition. For a comprehensive overview see Noll (2001). 8 There is a veritable flood of new findings in this area, all of which point surprisingly directly to the conclusion that mental processes can actually be discerned in the activity of the brain. Some notable recent results include the partial reconstruction of visual perceptual imagery from fMRI data (Nishimoto et al. 2011) and the reconstruction of heard words from fMRI imaging of human auditory cortex (Pasley et al. 2012). If, as seems likely, internal cognition and imagination uses some of the same brain mechanisms that underlie visual and auditory perception such studies suggest the eventual possibility of ‘listening in’ on thought and imaging via prospective highly advanced brain imaging technology. 9 See also Buckner (2003).

Chapter 4 1

In Figures 4.1, 4.2 and 4.3 brain imagery courtesy of Mark Dow and can be found at http://lcni.uoregon.edu/*dow. 2 Extremely high level and abstract cognitive modules have been postulated, the most notable of which is perhaps the so-called ‘theory of mind’ module, the absence of which, or defects within, is conjectured to underlie autism (see Leslie 1992, Baron-Cohen 1995). A fairly recent critical discussion can be found in Nichols and Stich 2003, pp. 117ff.). Another high-level cognitive module involving social cognition is the ‘cheater detection’ module which is supposed to aid us specifically in situations of more or less logically complex ‘social exchange’ (see Cosmides 1989; a general philosophical attack on a number of core theses of evolutionary psychology, and the cheater detection module hypothesis in particular, can be found in Buller 2006). 3 More precisely, sensory input on the left (right) generally is processed in the right (left) hemisphere and motor signals generated in the right (left) hemisphere actuate in the left (right) side of the body—the anatomical property of nerve crossover is called ‘decussation’. Why this is so remains mysterious, but for some reason lost in the depths of our evolutionary history, perhaps a mere accident or spandrel forced by other constraints, perhaps a crucial if obscure evolutionary advance, the vertebrate nervous system is generally organized in this crosswired form. The pioneering neuroscientist Ramön y Cajal hypothesized that decussation

216

Notes

was dictated by the need to rectify the inverted retinal image. Nowadays it is suggested that the crosswiring in primitive vertebrates allowed for a quicker response to predators via a direct activation of the muscles on the side opposite the threat (see Sarnat and Netsky 1981). 4 It may be worth pointing out that something a little bit similar happens in ordinary subjects who are subjected to two sound streams, one to each ear (an experimental paradigm called dichotic listening). Though subjects are conscious of only one stream—the one they are attending to—the interpretation of ambiguous phrases or words in the attended stream is influenced by disambiguating information from the non-attended stream (see Hugdahl 1988 for an overview of dichotic listening experimentation and theory). Note however that the dichotic listening case is a phenomenon of divided attention, a feature totally lacking in blindsight where subjects are in fact directed to attend to potential visual stimuli. 5 I cannot resist noting that Alan Cowey and Petra Stoerig (Cowey and Stoerig 1995) report on a fascinating experiment that opens a small window into the visual consciousness of monkeys. It gives a hint that the ‘kind of nothing’ blindsighted monkeys experience is quite similar to that of human beings. Monkeys which have had the visual cortex of just one hemisphere of the brain removed exhibit blindsight for the opposite side visual field. But Cowey and Stoerig ingeniously trained the blindsighted monkeys to respond to the absence of any visual stimuli in their good visual field—the monkeys learned to push a special button when they saw a blank presentation screen, and other buttons for more normal targets (such as light spots on the screen). What happens when this experiment is performed on the monkey’s blind visual field? Despite the fact that in other experiments the monkeys can and do respond to light spots in their blind visual fields, under the conditions of this experiment the monkeys press the key corresponding to ‘no light spot present’ (normally sighted control monkeys have no trouble indicating when they see a light and when they do not see a light). It is hard to avoid the conclusion that the monkeys are, just as humans are, not visually conscious of any light spots. 6 Damasio uses Gage’s accident and its aftermath to illustrate and bolster his own theory about the significance of emotion in cognition and may sometimes go further than the evidence strictly warrants. For an extremely detailed and more nuanced account of the Gage incident see Macmillan (2000). 7 A more recent study of brain response to the ultimatum game involved accomplished Buddhist ‘mindfulness’ meditators. Its result is rather peculiar: while confirming the general outlines of earlier research it found that meditators will accept ‘unfair’ offers at a much higher rate than non-meditators (they are, in a certain sense, more rational); see Kirk et al. (2011). 8 A remarkable study (Sahin et al. 2009) recently carried out on epilepsy patients with a number of electrodes implanted within Broca’s area (among other regions of the brain) reveals several basic processing stages of speech production, localized in both space and time. The presence of the electrodes allows for spatial and temporal resolution far superior to that of MRI techniques and supports the idea that lexical, grammatical and phonological features are processed in sequence in a process typically taking about one second.

Notes 9

217

At the moment I will only mention in a footnote that the philosophical big game in this area would be a link from neural organization to the subjective character of conscious experience. Of course there is a huge difference between the phenomenology of vision and hearing and anything that could shed light on how differential neural structure bears on sensory phenomenology is of the greatest interest. Certain phenomenological traits of some aspects of vision have been investigated by the philosopher Paul Churchland (see Churchland 1986), but his approach is limited to the relational features of color experience (e.g. such features as that orange is ‘closer’ to red than blue) rather than the origin of subjectivity from neural activity (a difficulty I call the ‘generation problem’). For a discussion of Churchland’s approach in relation to the generation problem, see Seager (1999), Chap. 2. 10 This experiment may provide some small measure of empirical support for David Chalmers’s principle of organizational invariance by which the subjective qualities of conscious experience supervene upon the functional architecture rather than more basically physical properties of whatever system happens to be implementing or generating the relevant states of consciousness (see Chalmers 1996, Chap. 7). 11 The area is called the right fusiform gyrus (a gyrus is one of the characteristic mounds created by the folding of the cortex; the valley between two gyri is called a sulcus); the region is sometimes labeled the FFA (fusiform face area). Its function is somewhat controversial, at least if one was inclined towards the perhaps naive view that this region of the brain is specifically and uniquely geared towards recognizing human facial features. In fact, the FFA is recruited by, no doubt among many, many others, expert bird-watchers and car lovers to help discriminate birds and automobiles (see Gauthier et al. 2000). Its function might better be described in terms of subtlety of categorization based upon very fine differences. Thus, conditions analogous to prosopagnosia have been observed impairing recognition of birds and makes of cars as well as faces. 12 The rapid pace of innovation in this area puts such remarks in jeopardy. Real time fMRI is under current development and promises to open brand new windows on the brain and, most definitely, the mind (see Bagarinao et al. 2006). For still more recent work see Monti et al. (2010). 13 See Menon et al. (1998) for discussion of a case of observed activation, via positron emission tomography (PET), of the face area in a patient in the so-called ‘vegetative state’. The face area responded appropriately when the patient was exposed to pictures of people familiar to her. It would be interesting to discover if the rivalry effects observed by Tong et al. (1998) also occurred in such patients (but I know of no such work). With regard to the warning given in the text above, we should also bear in mind that we don’t know with much certainty that such patients are actually unconscious; perhaps some are and some are not depending upon the nature of their illness (to see how disturbingly inaccurate diagnoses of persistent vegetative state may be, see Andrews et al. 1996). In fact, it now seems quite clear that a significant number of vegetative state patients have a high level of awareness. Recent work using fMRI has revealed that such patients can respond to instructions to imagine certain activities (for example playing tennis and walking about one’s house) which can be clearly distinguished in the resulting

218

Notes

brain scans (see Owen 2008). It has now been shown that this methodology can be adapted to open up a slow and difficult line of communication with people diagnosed as being in a vegetative state (see Monti et al. 2010). 14 The presence of Marc Hauser in the list of authors of this article casts an unfortunate but unavoidable shadow on its credibility, but I can find no evidence that the work here was in any way academically suspect. 15 For a video demonstration of the effect of TMS on speech, see http://www. newscientist.com/blogs/nstv/2011/04/how-a-magnet-can-turn-off-speech. 16 The possibility of manipulating the brain also generates a number of ethical issues. Many of these are obvious, having to do with state intrusion into the brains of their citizens. Perhaps a set of more interesting ethical questions arises from the voluntary use of TMS and other neuroscientific interventions for medical, cognitive enhancement or entertainment purposes. As such technology advances its price declines and before too long some of it will be readily available outside of the laboratory (in fact there is already a small community of DIY neural enhancers exploring some of these techniques). Issues of safety and unfair advantage are the most obvious here (for some discussion see Kadosh et al. 2012; Hamilton et al. 2011). 17 Perhaps this is not universally true. Some would argue that severely negative experiences can be, maybe even regularly are, ‘repressed’ as a kind of defense mechanism. It is not clear that this really occurs although long term severe stress can lead, because of an oversupply of stress hormones, to damage of the hippocampus (see Bremner 1999). 18 Our knowledge in this area depends largely on the surgical tragedy of Henry Molaison, who had sections of his brain, including both hippocampi, removed to cure severe epilepsy in 1953. Known in the medical literature only by the initials H. M. he died in 2008, leaving his brain to science. Project H. M. aims to preserve his brain by sectioning it into thin slices which are digitally photographed at very high resolution. The project intends to open the resulting database to the world via the internet (see http://thebrainobservatory.ucsd.edu/hmblog/). 19 There is a surprisingly large number of syndromes that involve more or less wild and free confabulation which resists rational correction, even in patients who are apparently fairly rational in general. For an extensive investigation of the range and causes of such afflictions, and a rich philosophical discussion of them, see Hirstein (2005). 20 Very curiously, however, their performance is actually mathematically more correct than that of normal subjects, better following the dictates of Bayes theorem which is a fairly recondite mathematical statement of how the chance of a hypothesis being true changes as evidence accumulates (for discussion see Stone and Young 1997, p. 342). 21 There is some data that suggest that a suspicious cast of mind is associated with Capgras syndrome, whereas a closely related delusional syndrome, Cotard’s delusion, in which a subject believes that he or she is dead, is associated with depression (see Stone and Young 1997, pp. 343ff.). 22 For some research on culture and schizophrenic manifestations see Tateyama et al. (1998) or, for a review, Stompe et al. (2003).

Notes

219

23

Moniz, already in his sixties with a very distinguished career, began operating in 1935 but did not perform many lobotomies. His major study encompassed twenty patients with poor follow up. The real driving force behind this kind of psychosurgery was the American neurologist Walter Freeman (see El-Hai 2005), who invented the ‘ice-pick’ method. This procedure—the trans-orbital lobotomy— involved literally hammering an ice-pick like device into a patient’s skull, through an eye socket, and then rapidly swiping the instrument from side to side to grossly lesion the brain. Freeman personally performed some 3500 operations and with his encouragement many tens of thousands of people were lobotomized with at best very marginal and equivocal outcomes and at worst the horrible destruction of many human minds. 24 The precise nature of the executive function of the anterior cingulate is not very clear. For a brief sampling of some hypotheses in the area see Awh and Gehring (1999). 25 As usual, it is not so clear whether the insula specifically underlies feelings of disgust themselves (though it seems certainly involved in them); for diverging experimental results see Schienle et al. (2002) and Wright et al. (2004). 26 In fact, in rats and presumably in humans as well, certain sorts of stimulation of the insular cortex can produce lethal cardiac arrhythmia, and this may help explain why some stroke victims suffer heart attacks (see King et al. 1999). 27 There is indeed a representational theory of consciousness which is the subject of some current excitement among philosophers. For an introduction see Seager and Bourget (2007); for influential theory development see Dretske (1995) and Tye (1995).

Chapter 5 1

For other discussion of the philosophical significance of cellular automata in the issue of emergence see Bedau (1997), Dennett (1991), Rosenberg (2004). 2 There are any number of Life implementations available as downloadable software (such as gtklife for Linux or Life32 for windows) or as web pages. A good one can be found at http://www.ibiblio.org/lifepatterns. 3 Zuse is a very interesting figure. He was a pioneer in the construction of digital computers, starting in Germany in the early 1930s and continuing through the war. He developed one of the first programming languages and built a calculating machine that could be programmed from memory and was ‘Turing complete’ (his adventures racing around Germany at the very end of the war to preserve his incomplete Z4 computer are like something out of Gravity’s Rainbow, see Zuse 1993). 4 Additional recent work which attempts to reconcile some aspects of quantum physics with cellular automata models can be found in ’t Hooft (2009, 2003). 5 A very interesting approach to the nature of computation taken by Rolf Landauer leads to a view ultimately rather similar to Fredkin’s. Landauer postulates that in some way physical law itself will have to respect the limits of

220

Notes

physical computation: ‘I am proposing that the ultimate form of the implementable laws of physics requires only operations available (in principle) in our actual universe’ (Landauer 1991, p. 28). Landauer sees the laws of nature and computational limitations as somehow (one has to add, mysteriously) codependent. Thus, given the apparently finite actual computational possibilities the laws themselves will describe a discrete universe that gives rise to those very computational possibilities with those very limitations. 6 I do not mean to imply here that only the CA approach can intelligibly make space and time emergent features of a radically non-spatial and non-temporal substrate. This is one of the hopes of many of those working on the presumed successor theory which will unify quantum and relativistic physics. 7 It is also important to bear in mind that the physical theory which will incorporate all phenomena in one unified description of nature, gravitational and quantum alike, will very likely dispense with continuous space and time. This is because the successor theory is almost bound to be some kind of quantum field theory which will incorporate space and time, or spacetime, in some quantized form, perhaps as emergent features of some still more fundamental aspect or structure of nature. The scale at which quantized spacetime would become apparent is extremely small; the so-called Planck length is about 10-35 meters. It is thus somewhat surprising that current technology may be able to provide evidence of the graininess of spacetime but that is the goal of the holometer being constructed at Fermilab (see http://holometer.fnal.gov/index.html) which may be able to verify the existence of ‘Planckian holographic noise’ (Hogan 2010) which is a kind of ‘jitter’ induced in position measurements by the presumed quantum nature of spacetime and which might just be detectable by highly sensitive linked pairs of interferometers. Of course, discrete spacetime does not imply that nature is describable as a CA but it is consonant with that hypothesis. 8 The literature on what Einstein called ‘spooky action at a distance’ is immense (see Aczel 2002; Hughes 1992, especially Chap. 6). Entanglement will be discussed as a possible form of radical emergence in Chap. 6 below. 9 That cellular automata can be Turing universal had been shown long before by von Neumann (his universal machine dates back to 1952 but the proof that it is universal was not presented until after his death, and was completed by Arthur Burks; see von Neumann 1966). 10 One caveat. If the generation rules of a real world finite CA should somehow essentially depend upon uncomputable numbers then they could not be simulated by a Turing machine (see below for how this might work). 11 The halting function, H (x,y), takes two numbers as input. First, the identifying number of some Turing machine as indexed according to some cataloging system (there are lots of ways to do this, the point being that the Turing machines match up one to one with the integers) and, second, an integer. The function returns a 1 or 0 depending on whether the indexed Turing machine halts with that input or not. That is, H (x, y) = 0 if Turing machine number x halts when given the number y as input and H (x, y) = 1 if TMx does not halt when given y as its input. Alan Turing famously showed that no Turing machine could compute the

Notes

221

halting function (strictly speaking, Turing provided the resources for this proof but did not present the halting problem as such; see Copeland 2004, which includes a reprint of Turing’s article and illuminating introductory remarks). The halting function is perfectly well defined but uncomputable by any Turing machine, or computationally equivalent system, such as cellular automata. 12 Is there any hope of actually building some kind of accelerating Turing machine? Work by Itamar Pitowsky (see Pitowsky 1990) and Mark Hogarth (Hogarth 1992) suggests that there are mathematical models consistent with general relativity that might permit a Turing machine to perform an infinite number of operations in a finite time. The world line of the machine would encompass infinite proper time but would be entirely in the past lightcone of an appropriately situated outside observer. It seems very unlikely that these models could be physically realized in our world (see Earman and Norton 1993). 13 There is a long history of logical discomfort arising from assumptions of continuity, going back at least to Zeno’s paradoxes of motion. Even apparently commonsense notions can lead to severe trouble. Consider the natural idea that matter is continuous, solid and impenetrable plus the idea that matter interacts by collision. Are such collisions possible? Consider two spheres approaching each other. Arguably, they cannot contact each other, for if they are in contact they share a point (which is impossible since they are each impenetrable). But if they do not share a point then they are not in contact (an infinity of points exist between their two surfaces). For a nice discussion of such oddities see Lange (2002), Chap. 1. 14 Such a system would be something like Turing’s ‘oracle machines’ in which a standard Turing machine is augmented with a device which can, on demand, supply answers to what may be called uncomputable questions, such as the answer to the halting problem for any given Turing machine plus input pair. The mode of operation of the oracle is unspecified although Turing somewhat cryptically says that it ‘cannot be a machine’ (Turing 1939/2004, p. 156). According to Max Newman (see Newman 1955, p. 259) Turing conceived of the oracle’s operation as a kind of representation of the mathematical intuition required for creative theorem formulation and proof. This is especially interesting since Turing endorses quite a different view in his famous paper on machine intelligence (Turing 1950). There he argues that machines computationally equivalent to Turing machines will be capable of creativity and at least give all the appearances of intuitive thought. In a curious echo of the oracle concept, Turing explains a fallacious argument against machine creativity thus: ‘One could say that a man can ‘inject’ an idea into the machine, and that it will respond to a certain extent and then drop into quiescence’ (p. 454). 15 This means that there is a standard Life configuration whose evolution can be interpreted as simulating the CA world corresponding to my new rules since, as mentioned above, Life is itself Turing universal (which raises an important issue about emergence). Of course the Life configuration which emulates my new rules will not look like a glider plowing through any pattern of cells it comes across. There are some niceties to be taken into account as well. What happens if two invulnerable gliders meet? Whatever we like of course, but I suggest that both be annihilated (or they could just pass through each other).

222 16

Notes

The idea of temporal supervenience is explored in greater depth in Chap. 7 below. A sympathetic examination of Morgan’s views can be found in Blitz 1992; for an excellent general discussion of ‘classical British’ emergentism see McLaughlin 1992; for an interesting exploration of the relation between supervenience and emergentism see Kim 1993, Chap. 8. 18 That is, there is no Turing computable way to efficiently predict the future state of a CA. If we could harness some of the speculative hypercomputational powers discussed above, perhaps we could transcend this barrier. In this respect, predictability remains a relative notion (closely linked here to the notion of ‘relative computability’—see Copeland and Sylvan 1999). In the absence of any real hypercomputational machine, we are limited to human computing power, as emulated and amplified but not essentially transformed by the digital computer. Incidentally, the much vaunted quantum computer promises a radical speed up of our computations but, it seems, no hypercomputational powers (see the stark pronouncement in the ‘bible’ of quantum computation, Nielsen and Chuang 2000, p. 126: ‘quantum computers also obey the Church-Turing thesis’). 19 We can only say ‘probably’ here because, first, there could be hypercomputation methods to which we could conceivably attain access and, second, there is no hard proof yet there is not some clever but otherwise normal algorithm that will make these hard problems efficiently solvable. This is the P=NP question. Roughly speaking, P problems are solvable in a time that is merely a polynomial function of the input size, so if n measures the size of the input, an algorithm that solves a problem in time proportional to, for example, n2 is in P. NP problems are ones that have solutions that can be verified efficiently (i.e. in polynomial time). Thus factoring an integer into its prime factors is relatively difficult compared to verifying that a given set of possible factors are indeed correct and it may be that factoring has no efficient algorithm (the evident difficulty in factoring large integers is the current basis of Internet security). No proof of this exists (in the computational complexity business in general, proofs are distressingly scarce). Furthermore, it is known that an efficient quantum computational algorithm (Shor’s algorithm) exists for prime factorization. This does not quite imply that factoring is in P since the quantum algorithm only provides an answer with some probability (which can be made as small as we like). In any event, it strongly appears that NP includes problems that are fundamentally harder than those in P, whose algorithms can do no better than a time proportional to an exponential function of n, as say 2n which quickly dwarfs n2 as n increases (see Harel 2000 for a gentle introduction to computational complexity). Although this is an open mathematical problem and indeed perhaps the most significant issue in theoretical computer science, almost all mathematicians and computer scientists think that P is not the same as NP. For what it’s worth, a particularly bizarre (and still unpublished in any journal so far as I know) ‘proof’ that P=NP is based upon Fredkin’s digital physics—see Bringsjord and Taylor (2004). The proof moves from the idea that since some natural processes ‘compute’ NP problems efficiently (as the soap film that forms over a set of pins solves the so-called Steiner Tree Problem of finding the shortest set of links between a set of points) and the digital 17

Notes

223

physics idea that the universe is running on a Turing machine equivalent cellular automaton, then there must be an efficient algorithm for solving NP problems, hence P = NP. Offhand, it seems the proof fails since the physical process at issue in the Steiner Tree Problem case only produces an approximation to the optimal solution, which gets worse as the number of points increases. Also, the proof that nature solves the Steiner Tree Problem assumes that nature is continuous and thus, from the digital physics point of view, has a false premise and hence is unsound. 20 There are much smarter ways to tackle the problem than an exhaustive look through every possible route, but they gain efficiency by trading generality: either they are not guaranteed to discover the optimal solution or have more or less limited applicability. One approach of the latter sort, called the cutting plane method, has found an exact solution for a particular set of almost 25,000 cities (all 24,978 cities, towns, and villages in Sweden to be precise), using somewhat fewer resources than the total universe (only the equivalent of about 92 CPU years on a single Intel Xeon 2.8 GHz processor). For more information visit the TSP web page at http://www.tsp.gatech.edu/. 21 There is a very lively current debate about the exact nature of this sort of explanation, focused on the issue of a priori entailment. The question is whether or not it is possible (as always, in principle) to deduce the emergent given only the ideal description of the submergent base and mere possession of the concepts descriptive of the emergent domain (this latter restriction explains the choice of the a priori label). The debate is especially significant for that most difficult of emergents: consciousness, and leads into pretty dense thickets of technical philosophy in the area of two dimensional modal logic. See Jackson (1998), Chalmers and Jackson (2001) for a defense of the a priori entailment thesis; see Block and Stalnaker (1999) for an attack upon it. We can safely ignore the intricacies of this debate in our present discussion of emergence. 22 The significance and nature of conservative emergence can itself be debated. While Batterman takes the very strong line that physical understanding positively requires reference to explanatorily emergent features, Gordon Belot (Belot 2005) regards such reference as basically heuristic in nature—it does not by itself render the world inexplicable in the terms of fundamental theory. Belot’s argument, in part, seems to be that there must be ‘in principle’ deducibility of the less fundamental theory from the more fundamental theory and that this will provide— again, ‘in principle’—understanding of the former from the presumed understanding of the latter. To the extent that we can understand how emergents appear, Belot must be right; but it is not given that that such understanding is attainable. It is not given that we could come to understand such phenomena as turbulence in the absence of the concepts of the explanatorily emergent features of the world. However, it is very interesting that in his reply to Belot, Batterman (Batterman 2005) states that in many cases the explanatorily emergent features (to use my term) simply do not exist. They are nothing but explanatory aids but ones which are indispensable for our understanding (perhaps something in the way that one could not understand Christmas in North America without ‘reference’ to Santa Claus). So this points to a real difference between what I mean by conservative

224

Notes

(i.e. explanatory or epistemological) emergence and Batterman’s ideas. I am thinking: tornadoes (they are real); Batterman is thinking of certain mathematical structures in continuous fluid dynamics (there aren’t any instantiated in nature, since the atmosphere is not a continuous fluid). The two ideas come back together when we recall the fact that the mathematical object serves as an excellent model for the real-world object. It seems Belot and Batterman are debating about Broad’s mathematical archangel. Batterman thinks the angel would need some concepts from the emergent domain in order to understand what was going on, even in a world entirely governed and determined by its underlying fundamental features. 23 It is easier to engender confusion about this than you would think. Silvan Schweber writes that ‘the reductionist approach that has been the hallmark of theoretical physics in the 20th century is being superseded by the investigation of emergent phenomena’ (Schweber 1993, p. 34) and that fields such as condensed matter physics investigate ‘genuine novelties in the universe’ (p. 36) which depend upon ‘emergent laws’. One could be forgiven for entertaining the exciting hope that here we have a forthright endorsement of radical or ontological emergence. But no. Schweber means only that the theories which describe the emergent phenomena are not simple consequences of underlying theory, that ‘condensed matter physics is not just ‘‘applied elementary-particle physics,’’ nor is chemistry applied many-body physics’ (p. 36). And as for the ‘genuine novelties’, Schweber is being completely literal: modern science produces things that have not existed in the universe before, just like Paris Hilton produces phenomena, e.g. television shows and the like, that are a genuine novelties in the world. The metaphysical import of both observations is about the same. Schweber actually goes somewhat out of his way to make it clear that ontological emergence is not the issue, emphasizing that ‘one may array the sciences in a roughly linear hierarchy according to the notion that the elementary entities of science X obey the laws of science Y one step lower’ (p. 36). Schweber goes on to write that recently developed ‘methods have changed Anderson’s remark ‘‘the more the elementary particle physicists tell us about the nature of the fundamental laws, the less relevance they seem to have to the very real problems of the rest of science,’’ from a folk theorem into an almost rigorously proved assertion’ (p. 36). Curiously, Noam Chomsky reports this passage as stating the refutation of reductionism, eliding the difference of interest between radical and conservative emergence entirely (Chomsky 2000, p. 145).

Chapter 6 1

The concepts of realization and multiple realization have been the subject of much philosophical discussion since at least the birth of the functionalist theory of mind with Hilary Putnam’s early papers (see for example Putnam 1960) and have generated a huge literature. For a good overview see Bickle 2008. Some recent important works on the topic include Melnyk 2003, Kim 2005 and Shoemaker 2007.

Notes 2

225

Written in dimensional form, this expression appears as T2LM/T2LM thus canceling all the dimensions as promised. 3 It is not clear from Poe’s words whether or not the flow is irrotational. Poe writes: ‘She [the trapped vessel] was quite upon an even keel—that is to say, her deck lay in a plane parallel with that of the water—but this latter sloped at an angle of more than forty-five degrees, so that we seemed to be lying upon our beamends. I could not help observing, nevertheless, that I had scarcely more difficulty in maintaining my hold and footing in this situation, than if we had been upon a dead level ; and this, I suppose, was owing to the speed at which we revolved.’ Does this perhaps suggest that the observer is always facing the centre of the vortex? If so, it is a rotational flow (one revolution of the observer per one rotation around the centre). Some small support for this interpretation can be found in a brief remark in Milne-Thomson (1957) who likens the maelstrom to the Rankine ‘combined vortex’ which is rotational in its ‘core’, irrotational outside the core. It turns out that if the velocity of the water around the centre of the vortex varies as the inverse of the distance from the centre, there will be irrotational flow. This condition is closely approximated in many real world vortices. 4 Vortices figure in a fascinating footnote to the history of science. In the late nineteenth century William Thomson, aka Lord Kelvin, championed the theory that the ultimate nature of matter was knotted vortices in some fundamental continuous, homogeneous and inviscid fluid (see Thomson 1867). Thomson was impressed with Helmholtz’s demonstration of the stability and vibratory properties of vortices. For example, he thought the then recently discovered atomic spectra could be explained in terms of characteristic oscillations of the knotted vortex which constituted each species of matter. Although the theory has not survived, it inspired George Tait (who became somewhat famous for his parlour room demonstrations of smoke rings), in collaboration with Thomson, to investigate the fundamental properties of knots which led to whole new branch of topology, knot theory. 5 Eugene Wigner provided a now classic discussion of this mystery: why mathematics is so successfully applicable to the natural world (Wigner 1960). An extended philosophical clarification and discussion of Wigner’s puzzle can be found in Mark Steiner (1998). There are at least two sides to this question. Why is mathematics applicable to the world? Here one might follow Leibniz who noted that any data will be describable by some mathematical function. In the Discourse on Metaphysics (Leibniz 1686/1989) Leibniz noted that ‘in whatever manner God might have created the world, it would have been regular and in accordance with a certain order’ because all structures or sequences of data are subject to mathematical description. As he puts it, using an example: ‘… let us assume… that someone jots down a number of points at random on a piece of paper… I maintain that it is possible to find a geometric line whose motion is constant and uniform, following a certain rule, such that this line passes through all the points in the same order in which the hand jotted them down’ (p. 39, note I am following Gregory Chaitin’s tiny emendation of the translation with ‘motion’ replacing ‘notion’; see Chaitin 2004). Of course, Leibniz makes it easy on himself, taking a

226

Notes

finite set of points, but I think his point is well taken. But the other aspect of the question is why humanly contrivable mathematics is so well able to describe the world (on this point Steiner tends to think that the universe is somehow ‘friendly’ to us). It is tempting to try out an anthropic explanation. Perhaps any world too complex or too weird for us to grapple with mathematically (and bear in mind our grapplings are only partially successful) would not be a world in which could evolve and persist. I have no idea how one would go about proving such a thesis. 6 It is worth stressing that this issue looks completely different from the point of view of any remotely feasible scheme of real world simulation of dynamical systems. There we find that the intrinsic limitations on our knowledge of the initial states of real world systems as well as their necessarily incomplete models severely limit predictability. One very interesting issue that arises here is the natural consideration that, in the face of our epistemic limitations, we should and would be content with gaining a merely qualitative understanding of how a certain system will tend to evolve. The simple approximation algorithms we have examined have a bad fault in this respect; they fall into the class of ‘nonsymplectic’ integrators, which means that they do not respect the conservation laws which will inevitably stem from various symmetries embodied in the systems under study. This inevitability is purely mathematical, as demonstrated in 1918 in the amazing theorem of Emmy Noether (for a brief history and discussion of her results see Byers 1999) which links continuous symmetries with conservation laws. Thus the symmetry of the laws of physics with respect to time implies and is implied by the conservation of energy. The failure to respect conservation principles in a method of numerical simulation means that as the evolution of a model drifts further from the evolution of the system it is intended to represent it will tend to enter regions of phase space which are simply inaccessible to the target system. Symplectic integrators can be constructed that will respect conservation laws despite accumulating error (one can envisage this as a restriction of the evolution of the model to an abstract ‘surface’ in phase space). This can be much more revealing about the qualitative behaviour of the system under study. Of course, such considerations are of less significance in the airy realm of relaxed computational constraints. 7 There are deep philosophical (and scientific) issues in this area. For an excellent philosophical discussion see Sklar (1993). 8 For a taste of the intricacies of climate modeling see Peixoto and Ort (1992). 9 Here is the calculation. Multiplying our length scale, 3000 metres by e4 we get a value which measures acceptable predictability. It happens to be about 163,795. We need then to solve this equation: 3  1010 ex ¼ 163; 795: Thus, x ¼  163;795  ln 310 which is about 34 days. 10 10 Recall the story of how Columbus, trapped on Jamaica in 1504 and threatened by the natives, used his knowledge of a lunar eclipse predicted to the minute (using a geocentric model) long before by Johannes Müller, who also went by the Latin nickname of Regiomontanus, in his Ephemerides of 1475. Columbus’s display of power over the heavens prompted a change of heart in the locals, and he and his

Notes

227

crew were rescued some months later. You can never tell when some piece of seemingly esoteric knowledge might come in handy. 11 In Newton’s own words: ‘... blind Fate could never make all the Planets move one and the same way in Orbs concentrick, some inconsiderable Irregularities excepted, which may have risen from the mutual Actions of Comets and Planets upon one another, and which will be apt to increase, till this System wants a Reformation’ (Newton 1730/1979, p. 402). Famously, Leibniz pounced on this seeming admission that God almighty Himself was incapable of building a ‘watch’ that kept good time. 12 However, for a less sanguine view of the long term stability of the inner Solar System see Laskar and Gastineau (2009). 13 An extremely interesting possibility is that quantum mechanics could impose an emergence wall which would be hit much sooner than we might expect. Wojciech Zurek (Zurek 1998) calculates that the application of the uncertainty principle to the motions of the planets ought to lead to predictive breakdown at a timescale of just a few million years. This outrageous conclusion suggests to Zurek the need for some extraneous factor which ‘classicizes’ the solar system: decoherence via interaction with the dust making up the background galactic environment. In any event, the lesson is clear. The underlying structure of the system imposes an emergence wall. This example will be discussed further below. 14 Proponents of the dynamical systems approach to cognition like to see themselves as radical revolutionaries battling the stultifying hegemony of the computationalist paradigm. For an interesting assessment of how radical (or not) the approach is see Grush (1997). 15 The calculation is based on this formula: t ¼ ac ½sinh vc; where c is the speed of light, v is the final velocity and a is the acceleration (both latter given in the problem); sinh is the hyperbolic sine function. 16 This remains somewhat controversial. David Bohm’s version of quantum mechanics remains a contender in the race for an acceptable interpretation of quantum mechanics and it restores full determinacy of position and momentum and provides an account of why our knowledge is restricted in accord with the Heisenberg relations (see Bohm and Hiley 1993). This possibility makes no difference to the discussion here. 17 It is interesting to compare Sperry’s thoughts on emergence with those of his student, Michael Gazzaniga, who some have also seen as endorsing radical emergence with downward causation. Gazzaniga (in 2011) recognizes a difference between what he calls ‘weak’ and ‘strong’ emergence but his discussion leaves it unclear whether his strong emergence is really radical, as defined here. The problem is that Gazzaniga pays insufficient attention to the crucial distinction between accessible and inaccessible explanatory structures and thus tends to see inexplicability or unpredictability as directly entailing something like radical emergence whereas, as we have seen, inaccessible explanations and unpredictability are compatible with conservative emergence.

228 18

Notes

Thus we treat the quantum uncertainty engendered by Heisenberg’s principle as fixing a minimum range of initial conditions from which error will inevitably grow, and grow very quickly in chaotic systems. An amusing example of this is the problem of balancing a pencil on its point. The uncertainty principle interpreted as placing a limit on the initial condition of the pencil forbids getting the pencil into a state with zero motion and perfectly balanced through its centre of gravity. Assuming nothing else prevents balancing the pencil (e.g. an errant breeze, imperfections of the pencil’s point, inhomogeneities in the Earth’s gravitational field, etc.), how long could the pencil remain standing under the most ideal conditions? Seemingly, only about four seconds! (For the problem and calculations, see Morin 2004). 19 The associated video can be viewed at http://www.hitachi.com/rd/research/ em/doubleslit.htm. 20 The term was first used to describe this property of quantum systems in 1935 by Erwin Schrödinger where he opines that ‘I would not call that one but rather the characteristic trait of quantum mechanics, the one that enforces its entire departure from classical lines of thought’ (Schrödinger 1935, p. 555). 21 It seems the mere registration of information about the particles in the detector, whether or not it affects the particles themselves, suffices to destroy the interference. This suggests the following bizarre question: what if we erase the information from the detectors? Quantum mechanics says that the interference then returns! The way this works is somewhat complex and so counterintuitive as to drive one physicist to complain that quantum mechanics has ‘more the character of medieval necromancy than of science’ (Jaynes 1980, p. 42). For an analysis of the ‘quantum eraser’ see Scully and Drühl (1982); for a brief philosophical commentary see Seager (1996). 22 There have been some efforts to understand the role of decoherence in brain processes. Max Tegmark has argued in an influential article that neural processes are subject to severe rapid environmental decoherence (see Tegmark 2000); for a rejoinder from the camp of ‘quantum consciousness’ theories see Hagan et al. (2002). Some recent work suggests a surprising degree of coherence even at room temperature in the central biological process of photosynthesis (see Collini et al. 2010). However, the time scale at issue is not very different from Tegmark’s theoretical value and seemingly much too short to be of significance in the neural processes underlying mentality. But you never know. 23 The decoherent histories approach thus provides a promising answer to one long standing problem with the many worlds interpretation of quantum mechanics, the so-called preferred basis problem. In quantum mechanics there are many equally valid choices of attributes which can be used to describe a system’s state only some of which will meet the condition of generating a more or less classical looking world history. Environmental decoherence may be able to explain why one choice is preferred. There are many other problem with the many worlds interpretation, most especially the difficulty of recovering a robust sense in which events have definite probabilities when, in a sense, the interpretation holds that every event that can happen does happen. Suffice it to say that the many worlds theory remains controversial and unorthodox (for an overview see Vaidman 2008).

Notes

229

24

It may be that Teller is only endorsing the weaker claim that quantum systems exhibit a kind of irreducible ontological holism. I think that discussions of emergence in quantum mechanics tend to miss the distinction between holism and radical emergence. This is often coupled with the assimilation of conservative or epistemological emergence with ‘part whole reductionism’. But while conservative emergence is compatible with mereological reductionism it is not equivalent to it. 25 The minus sign in singlet is irrelevant to our concerns. There is another joint state with a plus sign, called the ‘m ¼ 0 triplet’ that has subtly different properties than singlet which emphasize the holistic character of these sorts of states (see Maudlin 2007, pp. 53 ff.).

Chapter 7 1

‘Non-trivial’ is added here and above to prevent properties like ‘having charge pffiffiffi ?1 or not’ rendering anything and everything a physical entity ð 2 has this property). 2 As discussed above in Chap. 5, a very clear and austere characterization of mechanism is given in Broad (1925, Chap. 2). Modern views which espouse the idea that physical reality’s fundamental description is some kind of cellular automaton provide a different characterization of mechanism (see Wolfram 2002; Fredkin 1990). 3 This is an extremely lax notion of efficacy. For example, it completely ignores obvious problems that arise from overdetermination, finkish dispositions (see Martin 1994) or other philosophical chicanery. But it will serve my purposes here. 4 For my purposes we can generally understand counterfactual dependency in commonsense terms. A counterfactual conditional is true if its consequent is true in the possible world most like the actual world except that in that world the antecedent is true (this minimally different world of evaluation is frequently, if loosely, called the possible world ‘nearest’ to the actual world). For example, we evaluate the counterfactual ‘if the Supreme Court of the United States had not, in effect, ruled that Bush won the 2000 election, then Gore would have become President’ by considering the situation which includes the Supreme Court ruling against Bush but which is otherwise as similar as possible to the actual world. We judge this counterfactual to be true if Gore turns out to be President in that world. If, say, we think that in that case the Florida recounts would have left the Bush victory intact, then we think the counterfactual is false. Philosophers have formalized this understanding of counterfactuals with the machinery of modal logic but we need not delve into such technicalities here (the pioneering philosophical work can be found in Stalnaker 1968 and Lewis 1973). 5 An obvious imperfection glossed over is the existence of indexical terms. With a little suppression of one’s critical, or is it pedantical, faculties the point of the example should be clear.

230 6

Notes

A nice way to define physicalism stems from considering physically indistinguishable possible worlds. Call a minimal physical duplicate of a world, w, to be a world physically indistinguishable from w but which contains nothing else in addition to the physical (to use a theological metaphor, the minimal physical duplicate of w is created if God copies all the physical features of w into a new world and then stops). Physicalism can then be defined simply as the claim that any minimal physical duplicate of the actual world is a total or complete duplicate of the actual world (for details see Lewis 1983 and Jackson 1998). 7 A ‘weak’ version of local supervenience can be expressed in terms of worlds as: ð8wÞð8rÞð8pÞð8F 2 UÞððð8G 2 TÞðGrw  GpwÞ ^ Frw ! ðFpwÞ: It is a trivial consequence of local supervenience. Also, one can regard the entire world as one system thus encompassing global supervenience within this definition if it should turn out that unrestricted global supervenience is the appropriate relation needed to tackle certain properties. 8 A direct translation of strong supervenience in possible worlds form is easy to produce, but is of very little interest: ð8wÞð8rÞð8F 2 UÞðFrw ! ð9G 2 TÞ ðGrw ^ ð8wÞð8pÞðGpw ! FpwÞÞÞ: 9 The definition of ‘efficacy’ given above won’t necessarily capture such details. That depends on how certain counterfactuals turn out. Suppose, for example, that someone bends over to pick up a piece of paper on the road because it’s a twenty dollar bill. Would they have done so if that paper had not been real money? If we suppose that in the nearest world where this piece of paper is not money it is a good counterfeit (physically very similar to its actual counterpart) then we get the result that the subject still picks it up. So the property of being money is not efficacious according to our definition, as seems intuitively correct. But it is possible to disagree about how to evaluate such counterfactuals. 10 Obviously, the restriction to a single system could be relaxed but I want to focus on the case of the evolution of one system for the moment (in any case, there is no reason we could not treat the case of two systems as that of a single composite system). 11 A possible illustrative example of de-randomization in the micro to macro relationship is given by Ehrenfest’s equations (as briefly discussed above in Chap. 6), which assert that the expectation value of an observable such as position or momentum will evolve in accordance with classical laws of mechanics. In a macroscopic system made of huge numbers of microsystems we might expect (or hope) that such statistical features will exhibit a stability sufficient to allow us to identify the expectation value with the values obtained by particular observations, thus resulting in de-randomization and providing a reason to expect top-down discipline. But note that in general, the issue of the ‘retrieval’ of classical physics from quantum physics is extremely complex, incompletely researched and still poorly understood (see Ford 1989, Belot and Earman 1997). 12 There are other grounds for suspicion that such disjunctions of subvening states can support any robust sense of reduction, for which see Owens (1989), Seager (1991), Kim (1993, Chap. 16).

Notes 13

231

This thought experiment goes back to the inception of statistical mechanics in the controversy between Ludwig Boltzmann and Josef Loschmidt about the possibility of deriving the second law of thermodynamics from the mathematics of statistical mechanics (see Sklar 1993, Chap. 2). 14 See Sellars’s discussion of the postulation of two kinds of gold (Sellars 1963a, p. 122) and van Fraassen’s commentary (van Fraassen 1980, pp. 32ff.). 15 One good reason for this lack of concern is the recognition of distinctively lower-level ‘intrusions’ into the high-level dynamics which are simply not within the purview of the high-level theory. See Dennett (1971) for a classic discussion of this. 16 An interesting discussion of this constraint on theorizing which Hans Radder calls the ‘generalized correspondence principle’ can be found in Radder (1991). 17 Technically, the supervenience condition is included in the definition of topdown discipline, but it is clearer to emphasize the role supervenience plays as a separate part of the proof. Also I did not specify the grade of supervenience in the definition of TDD but left it loose to implicitly form a family of relations of topdown discipline. 18 Note also that this argument does not lead to ‘efficacy inflation’ in the sense that the U-state in question helps to bring about every T-state in any system. My dream last night is not efficacious in producing an earthquake in Turkey even assuming that earthquake has a unique physical predecessor. On the assumption of full temporal supervenience, the nearest possible world in which I don’t have my dream is different from the actual world right back to the beginning of time, but even so there is no reason to think it is different with regard to the earthquake’s occurrence. In testing for efficacy, we can pick any outcome state we wish so we can find one for which my dream is efficacious. This does not lead to my dream’s having efficacy ‘everywhere’ because the counterfactual change in my dream does not necessarily lead to the non-existence of every possible realizer of the earthquake (of course we can’t absolutely rule out this possibility either). 19 Note we must assume strong T-temporal supervenience to get this result, since in considering strong supervenience we have to consider other physically possible worlds. 20 For an independent argument in favour of this assumption see Seager (1988). 21 For ease of exposition I am being somewhat sloppy in my notation here and in what immediately follows. It would better follow our practice to write Fr to designate the U-state in question (where F is a property of the system r). 22 Or, more strictly speaking, a set of possible T-realizers fs1 ; s2 ; . . .; sn g The argument is not affected by this detail, which is thus omitted for simplicity of presentation. 23 Here I assume that if there is a T-description of a system then there is a description in T-elementary terms. This is an innocuous assumption since, by itself, it does not imply that every T-state has a constituent structure formed out of T-elementary features, for maybe some ‘large’ T-states are themselves elementary (call such things ‘large simples’). It is hard to think of genuine examples, but here is a possibility. Classical black holes can have but three physical properties that fully characterize them: mass, charge and angular momentum. These properties are

232

Notes

a function of the properties of the elementary constituents that have formed the black hole. But, once formed, there is no sense in which the black hole is composed of little bits and pieces that individually have various masses, charges or angular momenta (string theory may alter our perspective on this, but, of course and interestingly, in a way that makes black holes resolvable into a new—but still physical of course—kind of elementary constituent structure). Thus the black hole cannot be resolved into sub-components. This is no violation of the totality of physics however, since charge, mass and angular momentum are themselves allowable elementary features. A black hole is, so to speak, a kind of elementary ‘particle’ (and one that can, of course, take a constituent place within larger physical assemblies such as multi-star systems, galaxies, etc.). 24 Notice we do not need to assume that U possesses top-down discipline for this argument to work. The single case of r’s divergence violates T-temporal supervenience. 25 However, this at least suggests that there may be novel emergentist doctrines that derive from global or local supervenience relations. Perhaps we can imagine emergent properties that depend upon total world states for their existence. These are emergent properties dependent upon the total state of the ‘whole universe’ even though they might be properties of individual things. I can’t think of any examples of such properties however, although there are clear cases of non-local emergents. ‘Being money’ is such a non-local (but hardly fully global) emergent, but because of its lack of efficacy and our possession of some idea of how we might explicate the existence of money in terms of non-monetary properties, we tend and ought to regard this as a form of conservative emergence. Another example of a very non-local but far from fully global emergent property might be the value of the gravitational field at any point; it may well be that the state of the entire universe figures in determining this value (though perhaps not, depending on whether there are regions of the universe that are not in causal contact with each other, which currently seems very likely). The important point made by these examples is that even in non-local emergence, the emergent property depends upon quite definite, if ‘spread out’ features of the submergent domain. 26 This is why, I think, Morrison inserts into my ellipses in the above quote the otherwise puzzling claim that the emergents ‘cannot be explained in terms of microphysics’. There are lurking here large issues about the nature of explanation. The discussion above in Chap. 5 of the views of Philip Anderson is relevant here as well. 27 A shadow of a doubt about this might arise from noting that such predictions are in principle possible only if it is in principle possible to mathematically deduce such predictions from a given state. As discussed in Chap. 5 some properties of a system cannot be mathematically deduced, as for example whether a cellular automata will ever get into a certain configuration, but it remains true that the evolution of such systems is mathematically describable via simulation. Chap. 5 also delved into the question of whether it is perhaps conceivable that there are fundamental mathematical impediments to such predictions (e.g. non-computability). Of course, in this eventuality it would still be true that the emergents were completely determined by the subvening domains’ structures and laws.

Notes

233

The epistemological framework which emphasizes explanation in principle of the determination relation is thus ultimately unnecessary but adds explanatory vivacity to the account of conservative emergence. 28 As discussed in Chap. 5, simulatability is the feature of a theory that it is possible to calculate, in principle, the state transitions of any system in terms of the fundamental description of an initial state. Simulatability does not require that this calculation be mathematically exact; approximations are allowable so long as we can mathematically guarantee that the error of the approximation can be made as small as we like. For example, while the equations governing an isolated pendulum can be simulated by a mathematically exact representation of the system, the problem of simulating even a three-body gravitationally bound system is mathematically unsolvable. But the many-body problem can be approximated to whatever degree of accuracy we like (given arbitrarily large computing resources). There may be systems which cannot even be approximated in this sense however. 29 This failure of the ‘purity’ of empirical testing of complex physical theory is emphasized by Nancy Cartwright, who argues that we can neither understand nor build experimental apparatus without appeal to theories which are incompatible with each other. For example, she maintains that any aspect of nature, most especially scientific experiments, requires application of diverse laws from separate and perhaps incompatible theories: ‘neither quantum nor classical theories are sufficient on their own for providing accurate descriptions of the phenomena in their domain. Some situations require quantum descriptions, some classical and some a mix of both.’ (Cartwright 2005, p. 194). In order to argue that this is merely a practical necessity, one woud need a proof that the realm of classical physics really does conservatively emerge from more fundamental quantum theory. Cartwright’s views will be discussed further in Chap. 10 below.

Chapter 8 1

Epiphenomenalism, the doctrine that mental states are causally impotent products of physical processes, was first articulated and defended by Thomas Huxley (Huxley 1874). The worry that epiphenomenalism is a consequence of modern physicalist accounts of the mind has been reinvigorated by Jaegwon Kim (see for example Kim 1998). The idea that Kim’s argument generalizes beyond the case of the mind has also been explored (see Block 2003; Bontly 2002). Kim’s argument depends on his ‘exclusion principle’ which, roughly speaking, states that no event has more than one total cause. My argument in this chapter makes no use of the exclusion principle, though it could be seen as something of a defense of some form of such a principle. 2 It may be worth reemphasizing here that this is not an endorsement of so-called ‘part-whole reductionism’, though it is consistent with it. For example, we know from quantum mechanics that the states of ‘wholes’ are not simple functions of the states of their parts but this does not tell against the characterization given in the

234

Notes

text. As discussed in Chap. 6, quantum mechanics is a celebration of how the interactions of things can be understood—rigorously understood—to yield new features. It is, if you like, the mathematical theory of emergence, but one that obeys the strictures of resolution and abides by the strictures of conservative emergence. 3 This is expressed from a ‘particle perspective’. Perhaps it would be more accurate to say that all these particles are quanta of underlying quantum fields which have a better claim to be the truly fundamental physical structure of the world. 4 For example, something as simple as the spin of a proton turns out to be the product of an almost inconceivably complex interaction between the three constituting (or ‘valence’) quarks, a sea of virtual particles within the proton as well as additional components of orbital spin. The so-called proton (or nucleon) spin puzzle, the problem of explaining where the proton spin comes from given that measurement reveals that only about 30% of the spin can be attributed to the constituent quarks, has bedeviled physics for over 20 years. New lattice QCD calculations, recent observations and a deeper theoretical understanding of the role of the quantum mechanical vacuum may point to its resolution (see Bass 2007; Thomas 2008). Our ignorance about this is sobering and somewhat disturbing as is the awesome idea that nature embodies such staggering complexity within every minute portion of the world. 5 What follows is, so to speak, a purely metaphysical exercise, not intended to provide any insight into the use or significance of computer simulation in science. The topic of computer simulation and modeling has raised important questions in the philosophy of science concerning the epistemological status of simulations and the dangers of their interpretation, especially in light of the extra-theoretical adjustments required to get them running and producing sensible output (some of which were touched on in Chaps. 5 and 6 above). An excellent introduction to the philosophy of scientific simulation can be found in Winsberg (2010). 6 Would it ever make sense to start such a project? Not if computer technology progresses sufficiently quickly. Suppose the original length of the computation is n years and technology advances so quickly that after d years have passed the computation would take less than n  d years. If this level of technological progress persisted, it would never make sense to start the computation! Of course there are non-computer technical constraints on the time required for such computations and presumably the pace of progress in computer technology must eventually slow down rather than continue its heretofore exponential acceleration. For some n, the computations make sense, as evidenced by the real world examples given above, but the problem of this note is equally well illustrated (see Weingarten 1996). Computers of the 1980s would have taken about one hundred years to perform the reported computations. It was just not worth starting. 7 I am of course being somewhat playful here. There are very large issues lurking here, most especially the so-called measurement problem. If one believes that there are non-deterministic processes that drastically and uncontrollably alter the wave function then we can resort to the multiple simultaneous simulation model outlined immediately below.

Notes 8

235

Reflection upon the superduper computer simulation thought experiment suggests a purely philosophical question. Could we define physicalism in terms of this imaginary computer implementation of final physics? We might try something like this: physicalism is the doctrine that everything that occurs/exists in the actual world would have its exact counterpart in a final physics computer simulation of the world, or that the simulation would be, in some appropriate sense, indistinguishable from the actual world. Such a formulation has the advantage of automatically including what Hellman and Thompson (1975) call the principle of physical exhaustion. But it obviously requires a clearer specification. However, I think a more direct characterization of physicalism in terms of minimal physical duplicates of the actual world as pioneered by David Lewis (Lewis 1983; see also Jackson 1998) is preferable. The simulation model would fit seamlessly into such an approach without pretending to usurp its basic role in the definition of physicalism. 9 Some interesting preliminary work on this has been done by Warren Smith for the case of ideal Newtonian mechanics and basic quantum mechanics; see Smith 1999. The idea that quantum computers (of various types) can be used to simulate a range of, possibly all, quantum mechanical systems goes back to work of Richard Feynman (see e.g. Feynman 1982). Further work on the use of quantum computers in physics simulation has been done by David Deutsch and Seth Lloyd (see Deutsch 1985; Lloyd 1996). 10 There is an obvious epistemological problem here. How would one distinguish a case of radical emergence from a theory of the basic constituents which was merely false? One can imagine various ways such difficulties could be addressed. For example, suppose—what we already know to be false—that our best theory of the elementary features could not explicate even the simplest chemical properties of atoms. After enough failure, we might have sufficient reason to come to believe that chemical properties were brute and radically emergent features of certain complex structures. We have seen enough to appreciate how radical a step that would be however, putting great pressure on our theoretically informed views of nature. 11 It is also worth noting that supervenience and lack of genuine efficacy are compatible. It is not hard to think of candidate examples of such epiphenomenal supervenients. In addition to more or less bizarre properties, like the property of existing in a universe with a prime number of goats, there are many ordinary properties that appear to supervene on the physical but which lack genuine causal efficacy: moral properties, aesthetic properties, monetary properties, etc. It strikes me as obvious that monetary properties, such as the property of being a one dollar coin, have no causal powers (it might be different for the non-monetary mental property of believing that X is a one dollar coin of course). What could a one dollar coin do that a perfect duplicate counterfeit could not? Perhaps, actually purchase something legally? But a legal purchase is not a purely causal transaction. 12 I do not want to be committed to any particular view of the nature of time here and certainly not to the claim that the correct description of the world is one in which states ‘unfold’ in time. It is hard, and I won’t try, to avoid talk of nature as temporally dynamic but this way of talking can always, I think, be recast in terms

236

Notes

of the constraints imposed by nature on the set of states which form allowable sequences. Such sequences can be viewed as 4-dimensional entities stretched across the temporal as well as spatial dimensions. 13 This example is of course from Hilary Putnam (Putnam 1975, pp. 295 ff.). It is important to remember that Putnam is explicitly concerned with the issue of explanation and never questions that the fundamental physical features serve to fully determine the possible motions of the peg and hole. 14 It is also possible that this confusion is part of the reason why philosophers have failed after over 250 years of serious effort to come up with an acceptable philosophical account of causation and why the idea that science makes no reference to nor needs any appeal to causation is so persistent (this claim was famously made by Bertrand Russell (Russell 1917/1981, Chap. 9) for an updated defense see Norton 2003). 15 For an amusing but not altogether uninstructive example of this kind of explanatory promiscuity, as well as others, I commend to the reader the country music song ‘Third Rock from the Sun’ by Joe Diffie whose music video can be readily found on youtube. 16 While Mill’s methods are rather crude and modern statistical analyses provide a much more powerful and extensive tool kit, these newer methods stand as refinements rather than replacements of Mill’s methods. A sophisticated philosophical extension of Mill’s methods has been advanced as the ‘manipulability’ or ‘interventionist’ theory of causation (for an overview see Woodward 2008). This theory has been used in an attempt to show that there is no ‘exclusion problem’ about mixing macro and micro causation, with particular reference to the problem of mental causation and the threat of epiphenomenalism (Shapiro and Sober 2007). I think the interventionist account is clearly set at the epistemic level and nicely shows, and exploits, the natural inter-level promiscuity we should expect to find there. For an interesting review of current psychological theories of ‘inferring causes’ that provides insight into the folk theory of causation, as well as a defense of her own account, see Cheng (1997). 17 Taken together these factors allow the (controversial) calculation of the effects on human health of ozone depletion which the American Environmental Protection Agency estimates at about 70,000 deaths by 2075 (Reilly 1992). 18 For an engaging defense of a very strong form of this claim of scientific completeness, see the physicist Sean Carroll’s post at the blog Cosmic Variance (Carroll 2010). 19 Note that Dennett says ‘These patterns are objective—they are there to be detected—but from our point of view they are not out there independent of us, since they are patterns composed partly of our ‘subjective’ reactions to what is out there, they are the patterns made to order for our narcissistic concerns’ (Dennett 1987, p. 39). 20 A cursory search of the journal Lung Cancer turns up recent studies that go both ways on the coffee—lung cancer link. All the researchers are well aware of smoking as a possible confounder of course.

Notes

237

21

In light of this example, one might wonder why we have an inequality in (C2) rather than requiring that PðAjC ^ BÞ [ PðAjBÞ: We seek only to determine causal relevance amongst competing factors. Relative to A, C’s relevance ‘absorbs’ B’s putative relevance. But although (C2) guarantees that C makes a difference, it is possible that B could modify C’s causal influence, either strengthening or weakening it. Supposing that coffee either adds or subtracts power to tobacco smoke’s carcinogenic influence we can get (C2) to go either way. 22 My results are contrary to an earlier attempt by Robert Brandon to use the screening off test to show that high-level features can take efficacy away from lowlevel features in the context of evolutionary biology (see Brandon 1982; Brandon’s approach was criticized in Sober 1992). 23 For a critique of Yablo’s approach see Cox (2008). 24 One possible snare: the conscious apprehension of ‘average family size’ appears able to cause things but examples like these are—if they are examples of efficacy of any kind—examples of the efficacy of representational states of mind, not of the efficacy of what is represented. Thoughts about unicorns have their effects, but admitting this does not concede any causal powers to unicorns. 25 I must note these remarks of Richard Lewontin. Though he made them in a debate about the nature of IQ and IQ testing, I think the point ought to be generalized: ‘It is important to point out that the distinction between mental constructs and natural attributes is more than a philosophical quibble, even when those constructs are based on physical measurements. Averages are not inherited; they are not subject to natural selection; they are not physical causes of any events’ (Lewontin 1982).

Chapter 9 1

What I am calling the generation problem has very close connections to both the problem of the ‘explanatory gap’ (see Levine 1983) or what is called the ‘hard problem’ of consciousness (see Chalmers 1996). 2 The core idea of the view I will develop here can be traced back at least to Leibniz who, speaking of what he called ‘aggregates’ (which in my treatment are the conservative emergents), holds that it is ‘appropriate to conceive of them as a single thing. . . but then all these unities are made complete only by thoughts and appearances’ (Leibniz 1967, p. 126). Leibniz himself cites Democritus as holding a similar view. More recent philosophical work which explores somewhat similar, if perhaps rather more extreme, ideas include Merricks (2001) and van Inwagen (1990). The philosophical debate on the nature and status of ‘ordinary objects’ remains in flux. 3 Other conscious creatures may live in entirely different worlds insofar as they conceptualize things differently. Such creatures need not be particularly sophisticated but they do need to be conscious or else there is no sense-beyond our own projection-in which distinctive high level patterns play any role for them.

238

Notes

For a fascinating attempt to get inside the mind of a ‘simple’ creature whose way of looking at the world is quite different from ours see the discussion of the life of a tropical jumping spider in Harland and Jackson (2004). 4 A very rich conception of how this might work can be found in Richard Boyd’s discussion of the general structure of natural kinds (Boyd 1999). Although Boyd focuses on biological kinds his treatment is intended to extend to high level structure in general. One might hope that Boyd’s approach could be integrated with our discussion of the emergence of the classical world in Chap. 6 perhaps to serve as some kind of overall general superstructure of conservative emergence. Boyd himself sees his approach as part of a much broader scientific realism which would resist the austerity the SPW as I am presenting it. 5 Once again, it is important to bear in mind the distinction between the explanatory and metaphysical domains of causation, or, as I labeled it, the difference between causation and kausation. There is an interesting confluence of the two in this case. There is no doubt that the 2nd law provides a great deal of explanatory power and intelligibility across a wide range of applications. But it is also true that the appreciation of how the lower level features ‘conspire’ to verify the 2nd law deepens our understanding of the nature of thermodynamics. It nonetheless remains true that the drivers of all thermodynamical systems are the fundamental low level kausal processes. 6 There is a host of benighted designs that aim to exploit capillary action in various ways to achieve perpetual motion. One of the earliest was a system of sponges and weights invented by William Congreve around 1827 (see Ord-Hume 1977, Chap. 6 for a survey of capillary action and sponge wheel based attempts). 7 It’s worth noting that entropy increase and information degradation are themselves both reflections of non-fundamentality. At least, insofar as basic theory is what we called ‘total’ in Chap. 7 above there is perfect information preservation as an isolated system evolves. Since it does seem that our basic physics aims at totality, how can this be? Because systems of interest to us are not isolated (only the whole universe—or maximal causally interacting parts of it—are isolated) and the information spreads out from the system at issue into the larger system plus environment (see p. 71 above). 8 The paradox of consciousness is vaguely analogous to an aspect of a problem, the measurement problem, that arose early in the struggle to interpret quantum mechanics. The theory implies that systems will form superpositions of all their possible outcome states. For example, a detector monitoring the spin of a particle whose spin is a superposition of spin states should, if regarded as a quantum system, itself enter a superposition of detector states. But if conscious beings are also simply quantum systems, then they too should go into superpositions of distinct observations when they consult the detector. But we know that consciousness never appears to itself in such an indeterminate state. Thus consciousness appears to have a distinctive and irreducible place in the quantum world (see Wigner 1962 for the classical presentation). Of course, there are many way to respond to or evade this problem. I only wish to point out the structural similarity to the paradox developed here.

Notes

239

9

I discuss this case in further detail as well as what I regard as similar failures of naturalization in both Ruth Millikan’s and Jerry Fodor’s accounts of meaning in Seager 2000b. 10 Aristotle’s argument was endorsed and extended by Franz Brentano (1874/ 1973, pp. 130 ff.) and remains a prominent doctrine of the phenomenological school of philosophy. Currently, there is something of a renaissance of interest in reflexive accounts of consciousness in analytic philosophy, with several books out or on the way (e.g. Janzen 2008; Kriegel 2009; Brook and Raymont, forthcoming). At least one modern theory of consciousness denies the essential reflexivity of consciousness but follows Aristotle in requiring that conscious states be such that their subjects must be aware of them. This is the so-called higher order thought theory of consciousness espoused by, for example, David Rosenthal and—in a somewhat different form—Peter Carruthers (see Rosenthal 2005; Carruthers 2000). HOT theory halts the regress by accepting that there are some nonconscious mental states. In particular, the thoughts which confer consciousness on certain mental states in virtue of being about them are not themselves conscious so there need be no conscious awareness of every conscious mental state, contrary to Aristotle and Brentano. 11 Recognition of the importance of the fact that consciousness is an intrinsic property goes back at least to Leibniz and still funds some radical ideas about the nature of consciousness and its place in nature (see Strawson 2006; see also Seager 2006).

Chapter 10 1

Which is not to say that it was not noticed in one way or another long ago; for an historical survey of the problem see Seager (2007). 2 See Timothy O’Connor’s ‘Emergent Properties’ (1994) for an endorsement and philosophical development of a radical emergentism. 3 We have to add the caveat about fundamental laws since laws of nature can be conservatively emergent in the usual sense that they are determined by the fundamental laws. For example, the Weidemann-Franz law which states a positive correlative relationship between thermal and electrical conductivity of a metal is an emergent or derived law, which depends on the fact that both heat and electricity conduction depend on the presence of free electrons in metals. The radical emergentist of course posits the existence of primitive, irreducible laws of emergence. 4 The principle of the conservation of energy has been doubted by highly reputable scientists. In the 1930s Neils Bohr briefly advocated rejecting the principle in the face of anomalous evidence from radioactive decay, but the postulation of the almost undetectable neutrino by Wolfgang Pauli offered an escape hatch (the name ‘neutrino’ was Fermi’s however). The neutrino was finally detected in 1956 by Frederick Reines and Clyde Cowan.

240 5

Notes

There is a way for the radical emergentist to avoid many of the difficulties advanced above which is to take the unorthodox (for traditional emergentists) route of epiphenomenalism. Standard epiphenomenalism is a kind of radical emergence which says that the brain causes conscious states as non-physical attendants which have no causal effect back on the physical realm. This avoids the danger of violating conservation laws but the price is high. The supposed benefit of emergentism is that it does not displace consciousness from the physical world. Plus, of course, epiphenomenalism has difficulties of its own which are quite formidable such as the charge that epiphenomenalism entails that our conscious states are not the source of our (supposed) knowledge of them or even our knowledge that we are conscious at all (for a detailed discussion from a proepiphenomenalism stance see Robinson 2004). It is also worth pointing out here that Cartesian dualism can also be regarded as a form of radical emergentism but one in which the emergent is not a property of a physical entity but a novel substance brought into existence under certain conditions. Of course, for Descartes it is no law of nature that governs the emergence of mental substance but divine decree. Obviously, substance dualism has all the difficulties of radical emergence (conflicts with conservation laws, no sign of mental action, etc.) plus difficulties of its own (for example the problem of how two utterly disparate substances can be causally linked, a problem that was raised by Princess Elizabeth in correspondence with Descartes as far back as 1643—their correspondence can be found in Shapiro 2007—and very forcefully raised again recently by Jaegwon Kim with what he calls the ‘pairing problem’, see Kim 2005). 6 Panpsychism is an ancient doctrine with a lengthy philosophical pedigree ranging from thinkers of the ancient world to contemporary philosophers. See Skrbina (2005) for a detailed historical study; Seager and Allen-Hermanson (2008) for a theoretical overview. There has been a recent revival of interest in panpsychism amongst analytic philosophers, occasioned by the persistently intractable difficulty of integrating consciousness into the SPW. The first spark of panpsychism’s renewal goes back to Thomas Nagel’s argument for panpsychism (Nagel 1979). See also Cleve (1990); Chalmers (1996), Chap. 8; Seager (1995); Strawson (2006) and the recent collections Freeman (2006); Skrbina (2009); Blamauer (2011). 7 Russell is part of a tradition in which perception is founded on basic elements often referred to as sense-data. These were not generally regarded as mental in nature. After all, sense-data have spatial extent and qualities like redness which seem foreign to mental states (how could a mental state be red). But the mental states in question are states such as appearing red and these are certainly mental and do not require that anything actually be red in order to occur in some mind. 8 In the words of Spinoza himself: ‘The modes of each attribute have God for their cause only insofar as he is considered under the attribute of which they are modes, and not insofar as he is considered under any other attribute’ (Spinoza 1677/1985, Bk. 2, Prop. 6, p. 450). 9 This inference seems decidedly too swift. It is not easy to see how to build an organism out of nothing but bosons, or nothing but electrons. So it seems there is

Notes

241

logical room for some of the physically fundamental entities to lack mentality even if we grant the overall cogency of Nagel’s argument. The view that at least some but perhaps not all of the physically fundamental constituents of the world possess some sort of mental properties has been called ‘micropsychism’ by Galen Strawson (see Strawson 2006). On the other hand, if we are willing to grant that some physically basic entities possess mental properties, there does not seem to be much reason to balk at their universal distribution. At the very least this avoids the question of exactly why certain fundamental physical entities possess while others lack mentality. 10 It is worth forestalling a possible objection. Quantum superpositions are not vague. Williams and Barnes (2009), following on the famous argument of Gareth Evans (Evans 1978), provide an interesting argument that an underlying determinate ontology is incompatible with the existence of vague objects. 11 The ‘essential indeterminacy’ of mountainhood (and all other supposedly vague properties) is something which so-called epistemicists about vagueness deny. Philosophers willing to bite this bullet must insist that there is a critical fact which determines whether or not X is a mountain, but this fact is for various reasons more or less completely inaccessible to us (see Williamson 1994 for an extended defense of epistemicism). 12 The notion of large simples is not essentially connected to emergence. There could be large simples that are part of the fundamental furniture of the world from its inception or a part of its eternal structure. Newton’s conception of absolute space might count as an example. At least, certain philosophers have held the view that space is an extended simple. Absolute space is not the causal result of any interaction of elementary physical features but instead stands by itself. Yet it arguably does not have parts, save in a some purely notional sense of having regions within it-it is not composed of these parts and it cannot be divided into them. This is a complex issue and I only put it forward as an illustrative example; for discussion see Holden (2004). 13 I attempt to flesh out this suggestion in more detail in Seager (2010). The viability of this general strategy is examined and defended in Jonathan Powell’s PhD dissertation (Reading University); some of his work was presented at Tucson 2010 Towards a Science of Consciousness conference. See also Hameroff and Powell (2009). The suggestion is also highly reminiscent of the fusion operation of Paul Humphreys, as discussed in Chap. 6 above (see Humphreys 1997a, b). The crucial difference is that the current suggestion does not assume that consciousness arises from the fusion of purely physical precursors. 14 The prospect of refuting physicalism has always been one of the driving forces behind ESP research. Despite more than a century of at least quasi-scientific efforts to isolate a demonstrable paranormal effect there has been no decisive, clear and uncontroversial experimental evidence for its existence, still less any successful technology dependent upon it. Although it is hard to prove a negative this does not seem to be a thriving, progressive research program. It is worth remembering that a thinker of stature no less than Alan Turing regarded ESP as a strong, though not irrefutable, argument against the idea that a computer could exhibit intelligent

242

Notes

thought and also makes the remarkable claim that ‘the statistical evidence, at least for telepathy, is overwhelming’ (Turing 1950, p. 453). Sadly, Turing, along with C. D. Broad, were largely persuaded by the then highly influential card guessing experiments of Graham Soal whose results were later shown to be fraudulent (see Markwick 1978). In Broad’s case there was an underlying and amusing reason for his interest in the paranormal. He professed not to really care very much about the question of post-mortem survival but eagerly if forlornly hoped that ESP research would show that the scientific vision of the world ‘may prove to be as inadequate as it certainly is arrogant and ill-informed’ (Broad 1962, p. x). 15 Though these three thinkers provide important tools for developing a antirealist view of science which can be applied to the problem of consciousness, I do not mean to imply that they are all anti-realists themselves. In fact, both Dupré and Cartwright are happy to believe in the existence of unobservable entities discovered by science. Nonetheless in their opposition to the unity of science (Dupré) and ‘physics fundamentalism’ (Cartwright) they provide grist for the antirealist mill, especially with regard to questions about the nature of what the SPW regards as high level, conservatively emergent features of the world. 16 For an interesting and highly negative assessment of what he calls the ‘perfect model model’ of science see Teller (2001), which emphasizes the basic disconnect between the epistemic and explanatory role of models and the unimaginable complexity of any hypothetical model which completely captures some portion of reality. This attitude is nicely expressed in the report of the United States LHC Communication Task Force. The aim of the Task Force is to sustain and build support for particle physics in the USA by advertising the involvement of American scientists and science funding agencies in the LHC. The first strategy towards this goal listed in their report is to ‘promote recognition by key audiences of the value to the nation of particle physics, because of… its unique role in discovery of the fundamental nature of the universe’ (Banegas et al. 2007, p. 5).

References

Abbott, A.: CERN considers chasing up hints of Higgs boson. Nature 407(6800), 8 (2000) Aczel, A: Entanglement: The Greatest Mystery in Physics. Raincoast Books, Vancouver (2002) Alexander, S.: Space, Time and Deity. Macmillan & Co, London (1920) Anderson, P.W.: More is different. Science 177(4047), 393–396 (1972) Andrews, K., Murphy, L., et al.: Misdiagnosis of the vegetative state: retrospective study in a rehabilitation unit. Br. Med. J. 313(7048), 13–16 (1996) Ariew, A., Cummins, R., et al. (eds.): Functions: New Essays in the Philosophy of Psychology and Biology. Oxford University Press, Oxford (2002) Arp, H.: Quasars, Redshifts and Controversies. Interstellar Media, Berkeley (1987) Arp, H.: Seeing Red: Redshifts, Cosmology and Academic Science. Apeiron, Montreal (1998) Arp, H.: Catalogue of Discordant Redshift Associations. Apeiron, Montreal (2003) Ashby, N.: Relativity in the global positioning system. Living Rev. Relativ. 6(1). (2003). URL http://www.livingreviews.org/lrr-2003-1 Avrutin, S.: Linguistics and agrammatism. GLOT Int. 5(3), 87–97 (2001) Awh, E., Gehring, W.: The anterior cingulate cortex lends a hand in response selection. Nat. Neurosci. 2(10), 853–854 (1999) Bacciagaluppi, G.: The role of decoherence in quantum mechanics. In: Zalta, E. N. (ed.) The Stanford Encyclopedia of Philosophy. The Metaphysics Research Lab (2007). URL http://plato.stanford.edu/archives/fall2008/entries/qm-decoherence Bagarinao, E., Nakai, T., et al.: Real time function MRI: development and emerging applications. Magn. Reson. Med. Sci. 5(3), 157–165 (2006) Baldwin, J. (ed.): Dictionary of Philosophy and Psychology. The Macmillan Company, New York (1901) Banegas, D. et al.: At discovery’s horizon: report of the task force on US LHC communication. Technical report, National Science Foundation. (2007). URL http://wwwlhcus.org/ communication/documents/US_ILC_Report_101807_v3.pdf Barbour, J.: The End of Time: The Next Revolution in Physics. Oxford University Press, Oxford (2000) Barnes, J.: The Presocratic Philosophers. Arguments of the Philosophers, Routledge, London (1983) Baron-Cohen, S.: Mindblindness: An Essay on Autism and Theory of Mind. MIT Press, Cambridge (1995) Barrow, J., Tipler, F.: The Anthropic Cosmological Principle. Oxford University Press, Oxford (1988) Bartels, A., Zeki, S.: The neural basis of romantic love. NeuroReport 11(17), 3829–3834 (2000) Bartels, A., Zeki, S.: The neural correlates of maternal and romantic love. Neuroimage 21(3), 1155–1166 (2004)

W. Seager, Natural Fabrications, The Frontiers Collection, DOI: 10.1007/978-3-642-29599-7,  Springer-Verlag Berlin Heidelberg 2012

243

244

References

Bass, S.: How does the proton spin? Science 315, 1672–1673 (2007) Batterman, R.:The Devil in the Details: Asymptotic Reasoning in Explanation, Reduction and Emergence. Oxford University Press, Oxford (2002) Batterman, R.: Response to Belot. Philos. Sci. 72, 154–163 (2005) Bauer, R.: Autonomic recognition of names and faces in prosopagnosia: a neuropsychological application of the guilty knowledge test. Neuropsychologia 22, 457–469 (1984) Bedau, M.: Weak Emergence. In: Tomberlin, J. (ed.) Philosophical Perspectives 11: Mind, Causation, and World, pp. 375–399. Blackwell, Oxford (1997) Bedau, M., Humphreys, P.: Emergence: Contemporary Readings in Philosophy and Science. MIT Press, Cambridge (2008) Begelman, M., Rees, M.: Gravity’s Fatal Attraction: Black Holes in the Universe. W. H. Freeman, New York (1996) Belot, G.: Chaos and fundamentalism. In: Howard, D. (ed.) PSA 98, pp.S454–S465. Newark, DE: Philosophy of Science Association. (16th Biennial Meetings of the Philosophy of Science Association, Part II: Symposium Papers.) (2000) Belot, G.: Whose devil? which details? Philos. Sci. 72, 128–153 (2005) Belot, G., Earman, J.: Chaos out of order: quantum mechanics, the corrspondence principle and chaos. Stud. Hist. Philos. Mod. Phys. 28, 147–182 (1997) Berlekamp, E., Conway, J. et al.: What is life? In: Winning Ways for Your Mathematical Plays, vol. 2, Chap. 25. Academic Press, New York, (1982) Bickle, J.: Multiple realizability. In: Zalta, E. N. (ed.) The Stanford Encyclopedia of Philosophy. The Metaphysics Research Lab, fall 2008 ed. (2008) Blackburn, S.: Supervenience Revisited. In: Hacking, Ian (ed.) Exercises in Analysis: Essays by Students of Casimir Lewy. Cambridge University Press, Cambridge (1985) Blackmore, S.: Alien abduction. New Sci. 144(1952), 29–31 (1994) Blackmore, S.: The Meme Machine. Oxford University Press, Oxford (2000) Blamauer, M. (ed.): The Mental as Fundamental: New Perspectives on Panpsychism. Ontos Verlag, Frankfurt (2011) Blitz, D.: Emergent Evolution: Qualitative Novelty and the Levels of Reality. Kluwer, Dordrecht (1992) Bloch, F.: Nuclear induction. Phys. Rev. 70, 460–474 (1946) Block, N.: Do causal powers drain away? Philos. Phenomenol. Res. 67, 133–150 (2003) Block, N., Stalnaker, R.: Conceptual analysis, dualism, and the explanatory gap. Philos. Rev. 108, 1–46 (1999) Boden, M. (ed.): The Philosophy of Artificial Life. Oxford University Press, Oxford (1996) Bogen, J.: Physiological consequences of complete or partial commissural section. In: Apuzzo, M. (ed.) Surgery of the Third Ventricle, 2nd ed. pp. 167–86. Willams and Wilkins, Philadelphia (1998) Bohm, D., Hiley, B.: The Undivided Universe: An Ontological Interpretation of Quantum Mechanics. Routledge, London (1993) Bontly, T.: The supervenience argument generalizes. Philos. Stud. 109, 75–96 (2002) Boyd, R.:Homeostasis, species, and higher taxa. In: Wilson, R. (ed.) Species: New Interdisciplinary Studies, pp. 141–185. MIT Press, Cambridge (1999) Bozarth, M.: Pleasure systems in the brain. In: Warburton, D. M. (ed.) Pleasure: The Politics and the Reality, pp. 5–14. Wiley, New York (1994) Brandon, R.: The levels of selection. In: Asquith, P., Nickles, T. (eds.) PSA 1980, vol. 1, pp. 315– 323. Philosophy of Science Association, East Lansing (1982) Braun, A., Balkin, T., et al.: Regional cerebral blood flow throughout the sleepwake cycle: An H2(15)O PET study. Brain 120, 1173–1197 (1997) Bremner, J.D.: Does stress damage the brain? Biol. Psychiatry 45(7), 797–805 (1999)

References

245

Brentano, F.: Psychology from an empirical standpoint. international library of philosophy and scientific method. Routledge and Kegan Paul, London. (Original edition edited by O. Kraus; English edition edited by L. McAlister. Translated by A. Rancurello, D. Terrell and L. McAlister) (1874/1973) Bringsjord, S.: Taylor, J.: P=NP. (2004). URL http://arxiv.org/abs/cs/0406056 Broad, C.D.: Mind and its Place in Nature. Routledge and Kegan Paul, London (1925) Broad, C.D.: Lectures on Psychical Research: Incorporating the Perrott Lectures Given in Cambridge University in 1959 and 1960. Routledge and Kegan Paul, London (1962) Broadfield, D., Holloway, R., et al.: Endocast of sambungmacan 3 (Sm 3): A new homo erectus from Indonesia. Anat. Rec. 262, 369–379 (2001) Brock, W.: The Norton History of Chemistry. W. H. Norton & Co., New York (1993) Brook, A., Raymont, P.: The unity of consciousness. In: Zalta, E. N. (eds.) The Stanford Encyclopedia of Philosophy. Stanford University: The Metaphysics Research Lab, winter 2009 ed. (2010). URL http://plato.stanford.edu/archives/win2009/entries/consciousness-unity Brook, A., Raymont, P.: A Unified Theory of Consciousness. MIT Press, Cambridge, (forthcoming) Brownell, G.: A history of positron imaging. Presentation given in celebration of the 50h year of services by the author to the Massachusetts General Hospital on 15 Oct 1999. URL http://www.petdiagnostik.ch/PET_History/alb.html Brush, S.: History of the Lenz-Ising model. Rev. Mod. Phys. 39(4), 883–893 (1967) Buckingham, E.: On physically similar systems: illustrations of the use of dimensional analysis. Phys. Rev. 4(4), 345–376 (1914) Buckner, R.: The hemodynamic inverse problem: making inferences about neural activity from measured MRI signals. Proc. Nat. Acad. Sci. 100(5), 2177–2179 (2003) Buller, D.: Adapting Minds: Evolutionary Psychology and the Persistent Quest for Human Nature. MIT Press (Bradford Books), Cambridge, (2006) Buss, D. (ed.): The Handbook of Evolutionary Psychology. Wiley, Hoboken (2005) Byers, N.: E. Noether’s discovery of the deep connection between symmetries and conservation laws. The Heritage of Emmy Noether in Algebra, Geometry and Physics, vol. 12 of Israel Mathematical Conference (1999). URL http://www.physics.ucla.edu/*cwp/articles/noether. asg/noether.html Capgras, J., Reboul-Lachaux, J.: L’illusion des ‘‘sosies’’ dans un délire systématisé chronique. Bulletin de la Société de Médecine Mentale, 11, 6–16. (English translation available in History of Psychiatry (1994) 5, 119–130) (1923) Carroll, S.: Why is the universe accelerating? In: Freedman, W. (ed.) Measuring and Modeling the Universe, pp. 235–55. Cambridge University Press, Cambridge. (2004). URL http://xxx.lanl.gov/abs/astro-ph/0310342 Carroll, S.: The laws underlying the physics of everyday life are completely understood. Cosmic Variance, 23 Sept (2010). URL http://blogs.discovermagazine.com/cosmicvariance/2010/09/ 23/the-laws-underlyingthe-physics-of-everyday-life-are-completely-understood/ Carruthers, P.: Phenomenal Consciousness: A Naturalistic Theory. Cambridge University Press, Cambridge (2000) Carter, B.: The anthropic principle and its implications for biological evolution. Philos. Trans. R. Soc. Lond. A310(1512), 347–363 (1983) Cartwright, N.: The Dappled World. Cambridge University Press, Cambridge (1999) Cartwright, N.: Reply to Anderson. Stud. Hist. Philos. Mod. Phys. 32(3), 495–497 (2001) Cartwright, N.: Another philosopher looks at quantum mechanics, or what quantum theory is not. In: Ben-Menahem, Y. (ed.) Hilary Putnam, Contemporary Philosophy in Focus, pp. 188–202. Cambridge University Press, Cambridge (2005) Caston, V.: Aristotle on consciousness. Mind 111(444), 751–815 (2002)

246

References

Chaitin, G.: Leibniz, information, math and physics. In: Winfried Herausgegeben von Löffler and Paul Weingartner (eds.), Wissen und Glauben/Knowledge and Belief, Internationalen Wittgentstein-Symposiums, pp. 277–286. Vienna: ÖBV & HPT. (2004). URL http://www. umcs.maine.edu/*chaitin Chalmers, D.: The Conscious Mind: In Search of a Fundamental Theory. Oxford University Press, Oxford (1996) Chalmers, D.: Facing up to the problem of consciousness. In: Shear, J. (ed.) Explaining Consciousness, pp. 9–32. MIT Press, Cambridge (1997) Chalmers, D., Jackson, F.: Conceptual analysis and reductive explanation. Philos. Rev. 110, 315– 361 (2001) Chandrasekhar, S.: (1987). Shakespeare, Newton, and Beethoven. Truth and Beauty: Aesthetics and Motivations in Science, pp. 29–58. University of Chicago Press, Chicago Chang, K.: Here, There and Everywhere: A Quantum State of Mind. New York Times (July 11) (2000) Cheng, P.: From covariation to causation: a causal power theory. Psychol. Rev. 104(2), 367–405 (1997) Choisser, B.: Face blind. (2007). URL http://choisser.com/faceblind Chomsky, N.: New Horizons in the Study of Language and Mind. Cambridge University Press, Cambridge (2000) Churchland, P., Sejnowski, T.: The Computational Brain. MIT Press, Cambridge, MA (1992) Churchland, P.: Some reductive strategies in cognitive neurobiology. Mind 95, 279–309. Reprinted in Churchland’s A Neurocomputational Perspective: The Nature of Mind and the Structure of Science, MIT Press, Cambridge. 1991 (1986) Clayton, P., Davies, P. (eds.): The Re-Emergence of Emergence: The Emergentist Hypothesis From Science to Religion. Oxford University Press, Oxford (2006) Van Cleve, J.: Mind-dust or magic? Panpsychism versus emergence. Philos. Perspect. 4, 215–226 (1990) Collini, E., Wong, C.Y., et al.: Coherently wired light-harvesting in photosynthetic marine algae at ambient temperature. Nature 463(7281), 644–647 (2010) Connolly, J., Anderson, R., et al.: FMRI evidence for a ‘‘parietal reach region’’ in the human brain. Exp. Brain Res. 153(2), 140–145 (2003) Cooper, W.: William James’s theory of mind. J. Hist. Philos. 28(4), 571–593 (1990) Copeland, J.: Hypercomputation. Mind. Mach. 12, 461–502 (2002) Copeland, J. (ed.): The Essential Turing. Oxford University Press, Oxford (2004) Copeland, J., Sylvan, R.: Beyond the universal Turing machine. Aust. J. Philos. 77(1), 46–66 (1999) Corradini, A., O’Connor, T. (eds.): Emergence in Science and Philosophy. Routledge, London (2010) Cosmides, L.: The logic of social exchange: Has natural selection shaped how humans reason? Studies with the Wason selection task. Cognition 31, 187–276 (1989) Cover, J.A.: Non-basic time and reductive strategies: Leibniz’s theory of time. Stud. Hist. Philos. Sci. 28(2), 289–318 (1997) Cowey, A., Stoerig, P.: Blindsight in monkeys. Nature 373, 247–249 (1995) Cox, E.: Crimson brain, red mind: Yablo on mental causation. Dialectica 62(1), 77–99 (2008) Cronin, J., Pizzarello, S.: Enantiomeric excesses in meteoritic amino acids. Science 275(5302), 951–955 (1997) Damasio, A.: Descartes’ Error: Emotion, Reason and the Human Brain. G. P. Putnam’s Sons, New York (1994) Damasio, A.: The Feeling of What Happens: Body and Emotion in the Making of Consciousness. Harcourt Brace and Co., New York (1999) Darwin, C.: On the Origin of Species by Natural Selection. Murray, London (1859) Davis, M.: Computability, computation, and the real world. In:Termini, S. (ed.),Imagination and Rigor: Essays on Eduardo R. Caianiello’s Scientific Heritage. Springer, Milan (2006)

References

247

Dawkins, R.: The Selfish Gene. Oxford University Press, Oxford (1989) Dennett, D.: Intentional Systems. J. Philos. 68(4):87–106. (Reprinted in Dennett’s Brainstorms, Bradford Books, Montgomery (1978).) (1971) Dennett, D.: A cure for the common code. Brainstorms: Philosophical Essays on Mind and Psychology, pp. 90–108. Bradford Books, Montgomery. This first appeared as Dennett’s review of Jerry Fodor’s The Language of Thought in the journal Mind, April 1977 (1978a) Dennett, D.: Why the law of effect will not go away. in Brainstorms: Philosophical Essays on Mind and Psychology. Bradford Books, Montgomery. Originally published in The Journal of Theory of Social Behavior 2, 169–187 (1978b) Dennett, D.: Elbow Room: The Varieties of Free Will Worth Wanting. MIT Press (Bradford Books), Cambridge (1984) Dennett, D.: The Intentional Stance. MIT Press, Cambridge (1987) Dennett, D: Real Patterns. J. Philos. 88:27–51.(Reprinted in Dennett’s Brainchildren: Essays on Designing Minds, MIT Press, Cambridge 1998) (1991) Dennett, D.: Darwin’s Dangerous Idea: Evolution and the Meanings of Life. Simon and Schuster, New York (1995) Deutsch, D.: Quantum theory, the Church-Turing principle and the universal quantum computer. Proc. Royal Soc. A 400(1818), 97–117 (1985) DiSalvo, F.J.: Thermoelectric cooling and power generation. Science 285(5428), 703–706 (1999) Dostoevsky, F.: The Idiot. Harmondsworth: Penguin. (Translated by David Magarshack). (1869/ 1955) Dretske, F.: Naturalizing the Mind. MIT Press, Cambridge (1995) Dupré, J.: The Disorder of Things. Harvard University Press, Cambridge (1993) Dürr, S., et al.: Ab initio determination of light hadron masses. Science 322(5905), 1224–1227 (2008) Earman, J.: A Primer on Determinism. Dordrecht: Reidel (1986) Earman, J., Norton J.: Forever is a day: Supertasks in Pitowsky and Malament-Hogarth Spacetimes. Philos Sci 60(1), 22–42 (1993) Edelman, G.: Neural Darwinism: The Theory of Neuronal Group Selection. Basic Books, New York (1987) Eigler, D.M., Schweizer, E.K.: Positioning single atoms with a scanning tunnelling microscope. Nature 344, 524–526 (1990) El-Hai, J.: The Lobotomist: A Maverick Medical Genius and His Tragic Quest to Rid the World of Mental Illness. Wiley, Hoboken (2005) Ellis, H., Young, A.: Accounting for delusional misidentifications. Br. J. Psychiatry 157, 239–248 (1990) Evans, G.: Can there be vague objects. Analysis 38:208 (1978) Everett, H.: ‘‘Relative state’’ formulation of quantum mechanics. Rev. Mod. Phys. 29, 454–462 (1957) Feynman, R.: Simulating physics with computers. Int. J. Theor. Phys. 21, 467–488 (1982) Feynman, R., Leighton, R., et al.: The Feynman Lectures on Physics. Addison-Wesley, Reading (1963) Fodor, J.: The Modularity of Mind. MIT Press, Cambridge (1983) Ford, J.: What is chaos, that we should be mindful of it? In: P. Davies (ed.) The New Physics, pp. 348–372. Cambridge University Press, Cambridge (1989) Fredkin, E.: Digital mechanics: an informational process based on reversible universal CA. Physica D 45, 254–270 (1990) Fredkin, E.: An introduction to digital philosophy. Int. J. Theor. Phys. 42(2), 189–247 (2003) Fredkin, E.: Five big questions with pretty simple answers. IBM J. Res. Dev. 48(1), 31–45 (2004) Freeman, A. (ed.): Consciousness and Its Place in Nature. Imprint Academic, Exeter (2006) Friedman, J., Patel, V., et al.: Quantum superposition of distinct macroscopic states. Nature 406(6791), 43–46 (2000)

248

References

Frisch, U., Hasslacher, B., et al.: Lattice gas automata for the Navier-Stokes equation. Phys. Rev. Lett. 56(14), 1505–1508 (1986) Galilei, G.: Sidereus Nuncius (The Starry Messenger). The University of Chicago Press, Chicago (1610/1989) Gardner, M.: The fantastic combinations of John Conway’s new solitaire game ‘‘life’’. Sci. Am. 223, 120–123 (1970) Gardner, M.: Wheels, Life, and Other Mathematical Amusements. W. H. Freeman, New York (1983) Gauthier, I., Skudlarski, P., et al.: Expertise for cars and birds recruits brain areas involved in face recognition. Nat. Neurosci. 3(2), 191–197 (2000) Gazzaniga, M.: The Bisected Brain. Appleton-Century-Crofts, New York (1970) Gazzaniga, M.: Who’s in Charge? Free Will and the Science of the Brain. Harper Collins, New York (2011) Ge, Z., Marsden, J.: Lie-Poison Hamilton-Jacobi theory and Lie-Poisson integrators. Phys. Lett. A 133(3), 134–139 (1988) Gençay, R., Dechert, W.: The identification of spurious Lyapunov exponents in Jacobian algorithms. Stud. Nonlinear Dyn. Econ. 1(3), 145–154 (1996) Ghirardi, G.: Collapse theories. In: Zalta E. N. (ed.), The Stanford Encyclopedia of Philosophy. The Metaphysics Research Lab, Fall 2008 ed. (2008). URL http://plato.stanford.edu/ archives/fall2008/entries/qm-collapse Glanz, J.: Cosmic motion revealed. Science 282(5397), 2156–2157 (1998) Gleick, J.: Chaos: Making a New Science. Viking, New York (1987) Glimcher, P.: Decisions, Uncertainty, and the Brain: The Science of Neuroeconomics. MIT Press, Cambridge (2004) Goodale, M., Milner, D.: Sight Unseen. Oxford University Press, Oxford (2004) Granqvist, P., Fredrikson, M., et al.: Sensed presence and mystical experiences are predicted by suggestibility, not by the application of transcranial weak complex magnetic fields. Neurosci. Lett. 379(1), 1–6 (2005) Greene, B.: The Elegant Universe: Superstrings, Hidden Dimensions, and the Quest for the Ultimate Theory. W. W. Norton and Co., New York (1999) Gross, C.: Brain, Vision, Memory: Tales in the History of Neuroscience. MIT Press, Cambridge (1999) Grush, R.: Review of Port and van Gelder’s Mind as Motion. Philos. Psychol. 10(2), 233–242 (1997) Guth, A.: The Inflationary Universe: The Quest for a New Theory of Cosmic Origins. AddisonWesley (Helix Books), New York (1997) Guth, A.: Inflation and the New Era of High-Precision Cosmology. MIT Physics Annual, pp. 28– 39 (2002) Hacking, I.: Representing and Intervening: Introductory Topics in the Philosophy of Natural Science. Cambridge University Press, Cambridge (1983) Hacking, I.: A tradition of natural kinds. Philos. Stud. 61(1/2), 109–126 (1991) Hagan, S., Hameroff, S.R., et al.: Quantum computation in brain microtubules: decoherence and biological feasibility. Phys. Rev. E 65(6), 061901 (2002) Hameroff, S., Penrose, R.: Conscious events as orchestrated spacetime selections. J. Conscious. Stud. 3, 36–53 (1996) Hameroff, S., Powell, J.: The conscious connection: a psychophysical bridge between brain and pan-experiential quantum geometry. In: Skrbina, David (ed.) Mind that Abides: Panpsychism in the New Millennium, pp. 109–127. Benjamins, Amsterdam (2009) Hamilton, R., Messing, S., et al.: Rethinking the thinking cap: Ethics of neural enhancement using noninvasive brain stimulation. Neurology 76(2), 187–193 (2011) Harel, D.: Computers, Ltd.: What They Really Can’t Do. Oxford University Press,. Oxford (2000)

References

249

Harland, D., Jackson, R.: Portia perceptions: the umwelt of an araneophagic spider. In: Prete, F. (ed.) Complex Worlds From Simpler Nervous Systems. MIT Press, Cambridge (2004) Haugeland, J.: Weak supervenience. Am. Philos. Q. 19, 93–103 (1982) Hawking, S.: A Brief History of Time. Bantam, New York (1988) Heath, R.: Electrical self-stimulation of the brain in man. Am. J. Psychiatr 120, 571–577 (1963) Heath, R.: Pleasure response of human subjects to direct stimulation of the brain: Physiologic and psychodynamic considerations. In: Heath, R. (ed.) The Role of Pleasure in Human Behavior, pp. 219–243. Hoeber, New York (1964) Hellman, G., Thompson, F.: Physicalist materialism. Nous 11, 302–345 (1975) Hellman, G., Thompson, F.: Physicalism: ontology, determination and reduction. J. Philos. 72, 551–564 (1977) Henson, S., Constantino, R., et al.: Lattice effects observed in chaotic dynamics of experimental populations. Science 294, 602–605 (2001) Hercz, R.: The God helmet. Saturday Night 117(5), (2002) Hirstein, W.: Brain Fiction: Self Deception and the Riddle of Confabulation. MIT Press, Cambridge (2005) Hofstadter, D.: Waking up from the Boolean dream, or, subcognition as computation. Metamagical Themas, pp. 631–65. Bantam Books, New York. Reprint of Hofstadter’s article in the July 1982 issue of Scientific American (1986) Hogan, C.: Interferometers as probes of Planckian quantum geometry. (2010). URL http://arxiv.org/abs/1002.4880v26 Hogarth, M.: Does general relativity allow an observer to view an eternity in a finite time? Found. Phys. Lett. 5, 173–181 (1992) Holden, T.: The Architecture of Matter: Galileo to Kant. Oxford University Press, Oxford (2004) Holland, J.: Emergence: From Chaos to Order. Addison-Wesley (Helix Books), Reading (1998) Hooke, R.: Micrographia. Oakland, Octavo (CDROM). This is an electronic photo-reproduction of the original 1665 edition (1665/1998) Horgan, T., Tienson, J.: Connectionism and the Philosophy of Psychology. MIT Press (Bradford Books), Cambridge (1996). Hugdahl, K. (ed.): Handbook of Dichotic Listening: Theory, Methods, and Research. Wiley, New York (1988) Hughes, R.I.G.: The Structure and Interpretation of Quantum Mechanics. Harvard University Press, Cambridge (1992) Hughes, R.I.G.: The Ising model, computer simulation and universal physics. In: Morgan, M., Morrison, M. (eds.) Models as Mediators: Perspectives on Natural and Social Science, pp. 97– 145. Cambridge University Press, Cambridge (2000) Hume, D.: An Enquiry Concerning Human Understanding. Oxford University Press (Clarendon), Oxford (2000) Humphrey, N.: Consciousness Regained: Chapters in the Development of Mind. Oxford University Press, Oxford (1984) Humphreys, P.: Emergence, not supervenience. Philos. Sci. 64, S337–S345 (1997a) Humphreys, P.: How properties emerge. Philos. Sci. 64, 1–17 (1997b) Hurley, S.: Action, the unity of consciousness and vehicle externalism. In: Cleeremans, A. (ed.) The Unity of Consciousness: Binding, Integration and Dissociation, pp. 78–91. Oxford University Press, Oxford (2003) Huxley, T.: On the hypothesis that animals are automata, and its history. Fortnightly Rev. 95, 555–580 (1874). URL http://aleph0.clarku.edu/huxley/CE1/AnAuto.html Jackson, F.: Epiphenomenal qualia. Philos. Q. 32, 127–136 (1982) Jackson, F.: From Metaphysics to Ethics. Oxford University Press (Clarendon), Oxford (1998) James, W.: The Principles of Psychology, vol. 1. Henry Holt and Co., New York (1890) Janiak, A. (ed.): Newton: Philosophical Writings. Cambridge University Press, Cambridge (2004) Janzen, G.: The Reflexive Nature of Consciousness. Advances in Consciousness Research. Amsterdam, John Benjamins (2008)

250

References

Jaynes, E.T.: Quantum beats. In: Barut, A. (ed.) Foundations of Radiation Theory and Quantum Electrodynamics, pp. 37–43. Plenum Press, New York (1980) Jensen, K., Call, J., et al.: Chimpanzees are rational maximizers in an ultimatum game. Science 318(5847), 107–109 (2007) Joos, E., Zeh, D. H., et al.: Decoherence and the Appearance of a Classical World in Quantum Theory. Springer, Berlin (2003) Kadosh, R.C., et al.: The neuroethics of non-invasive brain stimulation. Curr. Biol. 22(4), R1–R4. (2012). URL http://download.cell.com/images/edimages/CurrentBiology/homepage/curbio9329.pdf Kanwisher, N.: Neural events and perceptual awareness. Cognition 79, 89–113 (2001) Kay, K.N., Naselaris, T., et al.: Identifying natural images from human brain activity. Nature 452, 352–355 (2008) Kelley, D., Milone, E.: Exploring Ancient Skies: An Encyclopedic Survey of Archaeoastronomy. Springer, New York (2005) Kim, J.: ‘‘Downward Causation’’ in emergentism and non-reductive materialism. In: Beckermann, A., Flohr, H., Kim, J. (eds.) Emergence or Reduction? De Gruyter, Berlin (1992) Kim, J.: Supervenience and Mind. Cambridge University Press, Cambridge (1993) Kim, J.: Mind in a Physical World: An Essay on the Mind-Body Problem and Mental Causation. MIT Press, Cambridge (1998) Kim, J.: Physicalism, or Something Near Enough. Princeton University Press, Princeton (2005) King, A., Menon, R., et al.: Human forebrain activation by visceral stimuli. J. Comp. Neurol. 413, 572–582 (1999) Kirk, U., Downar, J., et al.: Interoception drives increased rational decision-making in meditators playing the Ultimatum Game. Frontiers Neurosci 5(0) (2011). URL http://www.frontiersin. org/decision_neuroscience/10.3389/fnins.2011.00049/abstract Koch, C.: The Quest for Consciousness: A Neurobiological Approach. Roberts and Company, Englewood (2004) Krauss, L.M., Scherrer, R. J.: The return of a static universe and the end of cosmology. Int. J. Mod. Phys. D 17, 685 (2008). URL http://dx.doi.org/10.1007/s10714-007-0472-9 Kriegel, U.: Subjective Consciousness: A Self-Representational Theory. Oxford University Press, Oxford (2009) Lagage, P., Doucet, C., et al.: Anatomy of a flaring proto-planetary disk around a young intermediate-mass star. Science 314(5799), 621–623 (2006) Landauer, R.: Information is physical. Phys. Today 44(5), 23–29 (1991) Lange, M.: An Introduction to the Philosophy of Physics: Locality, Fields, Energy and Mass. Blackwell, Oxford (2002) Larson, R.: The physics of star formation. Rep. Prog. Phys. 66(10), 1651–1697 (2003) Laskar, J., Gastineau, M.: Existence of collisional trajectories of Mercury, Mars and Venus with the Earth. Nature 459, 817–819 (2009) Lassonde, M., et al.: Effects of early and late transection of the corpus callosum in children. Brain 109, 953–967 (1986) Laughlin, R.: A Different Universe. Basic Books, New York (2005) Lecar, M., Franklin, F., et al.: Chaos in the Solar System. Ann. Rev. Astron. Astrophys. 39, 581– 631 (2001) Leggett, A.: Reflections on the quantum measurement paradox. In: Hiley, B.J., David Peat, F. (eds.) Quantum Implications: Essays in Honor of David Bohm, pp. 85–104. Routledge and Kegan Paul, London (1987) Leibniz, G.W.: Discourse on Metaphysics. In: Ariew, R., Garber, D., (eds.) G. W. Leibniz: Philosophical Essays, pp. 35–68. Hackett, Indianapolis (1686/1989) Leibniz, G.W.: The Leibniz-Arnaud Correspondence. Manchester University Press. Edited and translated by H. Mason, Manchester (1967) Leibundgut, B.: Time dilation in the light curve of the distant type Ia supernova SN1995k. Astrophys. J. Lett 466, L21–L24 (1996)

References

251

Leslie, A.: Theory of mind impairment in autism. In: Whiten, A. (ed.) Natural Theories of Mind: Evolution, Development, and Simulation of Everyday Mindreading. Blackwell, Oxford (1992) Leslie, J.: The End of the World: The Science and Ethics of Human Extinction. Routledge, London (1998) Levine, J.: Materialism and qualia: the explanatory gap. Pac. Philos. Q. 64, 354–361 (1983) Levine, J.: Purple Haze: The Puzzle of Consciousness. Oxford University Press, Oxford (2001) Lewes, G.: Problems of Life and Mind, vol. 2. Kegan Paul, Trench, Turbner & Co., London (1875) Lewis, D.: Counterfactuals. Blackwell, Oxford (1973) Lewis, D.: New work for a theory of universals. Aust. J. Philos. 61, 343–377 (1983) Lewis, D.: On the Plurality of Worlds. Blackwell, Oxford (1986) Lewontin, R.: Reply to Martin, Orzack and Tomkow. New York Review of Books, 29(1). This is Lewontin’s reply in the Letters section to criticisms of his review of Stephen Jay Gould’s The Mismeasure of Man, (1982) Liddle, A.: An introduction to cosmological inflation. In: Masiero, A., Senjanovic, G., Smirnov, A. (eds.) 1998 Summer School in High Energy Physics and Cosmology: ICPT, Trieste, Italy, 29 June, 17 July 1998. World Scientific Publishing Company, London (1999) Lloyd, S.: Universal quantum simulators. Science 273, 1073–1078 (1996) Lloyd, S.: Computational capacity of the universe. Phys. Rev. Lett. 88(23) (2002) Lockwood, M.: Mind, Brain and the Quantum. Blackwell, Oxford (1989) Logothetis, N., Pauls, J., et al.: Neurophysiological investigation of the basis of the fMRI signal. Nature 412(6843), 150–157 (2001) Lorenz, E.: Deterministic nonperiodic flow. J. Atmos Sci. 20(2), 130–141 (1963) Lovelock, J.: A physical basis for life detection experiments. Nature 407(4997), 568–570 (1965) Macdonald, G., Macdonald, C. (eds.): Emergence in Mind. Oxford University Press, Oxford (2010) Macmillan, M.: An Odd Kind of Fame: Stories of Phineas Gage. MIT Press (Bradford Books), Cambridge, (2000) Madsen, J.: Intermediate mass strangelets are positively charged. Phys. Rev. Lett. 85, 4687–4690 (2000) Manafu, A.: The prospects for fusion emergence in chemistry. Paper delivered at 2011 Canadian Philosophical Association meeting (2011) Marcel, A.: Blindsight and shape perception: deficit of visual consciousness or of visual function? Brain 121, 1565–1588 (1998) Marks, C.: Commissurotomy, Consciousness, and Unity of Mind. MIT Press (Bradford Books), Cambridge (1980) Markwick, B.: The Soal-Goldney experiments with Basil Shackleton: new evidence of data manipulation. Proc. Soc. Psych. Res. 56, 250–281 (1978) Martin, C.B.: Dispositions and conditionals. Philos. Q. 44, 1–8 (1994) Matsumoto, M., Saito, S., et al.: Molecular dynamics simulation of the ice nucleation and growth process leading to water freezing. Nature 416(6879), 409–413 (2002) Maudlin, T.: The Metaphysics Within Physics. Oxford University Press, Oxford (2007) McGinn, C.: Can we solve the mind-body problem? Mind 98(391), 349–366 (1989) McGinn, C.: The Mysterious Flame: Conscious Minds in a Material World. Basic Books, New York (1999) McGivern, P., Rueger, A.: Emergence in physics. In: Corradini, A., O’Connor, T. (eds.) Emergence in Science and Philosophy, pp. 213–232. Routledge, London (2010) McLaughlin, B.: The rise and fall of British emergentism. In: Beckermann, A., Flohr, H., Kim, J. (eds.) Emergence or Reduction, pp. 49–93. De Gruyter, Berlin (1992) McLaughlin, B., Bennett, K.: Supervenience. In: Zalta, E. N. (ed.) The Stanford Encyclopedia of Philosophy. Stanford University: The Metaphysics Research Lab, fall 2008 ed. (2008). URL http://plato.stanford.edu/archives/fall2008/entries/supervenience

252

References

Melnyk, A.: A Physicalist Manifesto: Thoroughly Modern Materialism. Cambridge University Press, Cambridge (2003) Menon, D., Owen, A., et al.: Cortical processing in persistent vegetative state. The Lancet 352(9123), 200 (1998) Merricks, T.: Objects and Persons. Oxford University Press, Oxford (2001) Mill, J. S.: A System of Logic, vol. 7–8 of The Collected Works of John Stuart Mill. University of Toronto Press, Toronto (1843/1963) Milne-Thomson, L.: Some hydrodynamical methods. Bull. Am. Math. Soc. 63(3), 167–186 (1957) Milner, D., Goodale, M.: The Visual Brain in Action. Oxford Psychology Series. Oxford University Press, New York (2006) Monti, M., Vanhaudenhuyse, A., et al.: Willful modulation of brain activity in disorders of consciousness. New Engl. J. Med. (2010) URL http://content.nejm.org/cgi/content/abstract/ NEJMoa0905370v1 Morgan, C.L.: Emergent Evolution. Williams and Norgate, London (1923) Morin, D.: Balancing a pencil. (2004) URL http://www.physics.harvard.edu/academics/ undergrad/problems.html Morowitz, H.: The Emergence of Everything: How the World Became Complex. Oxford University Press, Oxford (2002) Morrison, M.: Emergence, reduction, and theoretical principles: rethinking fundamentalism. Philos. Sci. 73, 876–887 (2006) Motl, L.: Hawking and unitarity. (2005). URL http://motls.blogspot.com/2005/07/hawkingand-unitarity.html Mourelatos, A.: Quality, structure, and emergence in later pre-socratic philosophy. In: Proceedings of the Boston Colloquium in Ancient Philosophy, vol. 2, pp. 127–94 (1986) Müller, J.: Müller’s Elements of Physiology. Thoemmes, Bristol. This edition, based upon the 1838–1842 English translation of William Baly, is edited by Nicholas Wade who provides an extensive introduction (2003) Murray, N., Holman, M.: The role of chaotic resonances in the Solar System. Nature 410(6830), 773–779 (2001) Musallam, S., Cornell, B., et al.: Cognitive control signals for neural prosthetics. Science 305(5681), 258–262 (2004) Nagel, T.: Brain bisection and the unity of consciousness. Synthese 22, 396–413. (Reprinted in Nagel’s Mortal Questions, Cambridge University Press, Cambridge 1979) (1971) Nagel, T.: What is it like to be a bat? Philos. Rev. 83(4), 435–50. (This article is reprinted in many places, notably in Nagel’s Mortal Questions, Cambridge: Cambridge University Press, 1979.) (1974) Nagel, T: Panpsychism. In: Mortal Questions, pp. 181–95. Cambridge University Press, Cambridge. (Reprinted in D. Clarke Panpsychism: Past and Recent Selected Readings, Albany: SUNY Press, 2004) (1979) Nagel, T.: The View from Nowhere. Oxford University Press, Oxford (1986) Narlikar, J., Vishwakarma, R., et al.: Inhomogeneities in the microwave background radiation interpreted within the framework of the quasi-steady state cosmology. Astrophys. J. 585, 593– 598 (2003) Newman, M.: Alan Matheson Turing: 1912–1954. Biograph. Mem. Fellows Roy Soc 1, 253–263 (1955) Newton, I.: Opticks, or, A Treatise of the Reflections, Refractions, Inflections and Colours of Light. Dover, New York (1730/1979) Nicholls, J., Martin, A. et al.: From Neuron to Brain: A Cellular and Molecular Approach to the Function of the Nervous System. Sinnauer Associates, Sunderland (2001) Nichols, S., Stich, S.: Mindreading. Oxford University Press, Oxford (2003) Nielsen, M., Chuang, I.: Quantum Computation and Quantum Information. Cambridge University Press, Cambridge (2000)

References

253

Nisbett, R., Peng, K., et al.: Culture and systems of thought: holistic vs. analytic cognition. Psychol. Rev. 108, 291–310 (2001) Nishimoto, S., Vu, A.T., et al.: Reconstructing visual experiences from brain activity evoked by natural movies. Curr. Biol. 21(9), 1641–1646 (2011) Noll, D.: A primer on MRI and functional MRI (2001). URL http://www.eecs.umich.edu/ *dnoll/primer2.pdf (Univ. of Michigan Depts. of Biomedical Engineering and Radiology) Norton, J.: Causation as folk science. Philosopher’s Imprint, 3 (4) (2003). URL:http://hdl. handle.net/2027/spo.3521354.0003.004 O’Connor, T.: Emergent properties. Am. Philos. Q. 31, 91–104 (1994) O’Connor, T., Wong, H. Y.: Emergent properties. In: Zalta, E. N. (ed.) The Stanford Encyclopedia of Philosophy. Stanford University: The Metaphysics Research Lab, spring 2009 ed. (2009). URL http://plato.stanford.edu/archives/spr2009/entries/properties-emergent Ord, T.: The many forms of hypercomputation. Appl. Math. Comput. 178(1), 143–153 (2006) Ord-Hume, A.: Perpetual Motion: The History of an Obsession. Adventures Unlimited Press, Kempton (1977) Owen, A.M.: Functional neuroimaging of the vegetative state. Nat. Rev. Neurosci. 9(3), 235–243 (2008) Owens, D.: Disjunctive laws. Analysis 49, 197–202 (1989) Pais, A.: Inward Bound. Oxford University Press, New York (1986) Pasley, B., David, S., et al.: Reconstructing speech from human auditory cortex. PLoS Biol. 10(1), e1001251 (2012). URL http://dx.doi.org/10.1371/journal.pbio.1001251 Paull, R., Sider, T.: In defense of global supervenience. Philos. Phenomenol. Res. 52(4), 833–854 (1992) Peixoto, J., Ort, A.: Physics of Climate. American Institute of Physics, New York (1992) Penfield, W.: The Excitable Cortex in Conscious Man. Liverpool University Press, Liverpool (1958) Penrose, R.: The Emperor’s New Mind: Concerning Computers, Minds and the Laws of Physics. Oxford University Press, Oxford (1989) Persinger, M.: Religious and mystical experiences as artifacts of temporal lobe function: a general hypothesis. Percept. Mot. Skills 57, 1255–1262 (1983) Persinger, M.: Electromagnetic remote control of every human brain. Percept. Motor skills 80, 791–799. (1995) Petrie, B.: Global supervenience and reduction. Philos. Phenomenol. Res. 48, 119–130 (1987) Phillips, M., Young, A., et al.: A specific neural substrate for perceiving facial expressions of disgust. Nature 389(6650), 495–498 (1997) Pinker, S.: How the Mind Works. W. W. Norton and Co., New York (1999) Pitowsky, I.: The physical Church thesis and physical computational complexity. Iyyun 39, 81– 99 (1990) Poincaré, H.: Science and Method. Dover, New York.(Reprint of F. Maitland’s English translation of Science et méthode, Paris: Flammarion, 1908) (1908/1960) Popper, K., Eccles, J.: The Self and Its Brain. Springer International, New York (1977) Port, R., van Gelder, T. (eds.): Mind as Motion. MIT Press, Cambridge (1995) Pour-El, M., Richards, I.: The wave equation with computable initial data such that its unique solution is not computable. Adv. Math. 39(3), 215–239 (1981) Puccetti, R.: Brian bisection and personal identity. Br. J. Philos. Sci. 24(4), 339–355 (1973) Pullman, B.: The Atom in the History of Human Thought. Oxford University Press, New York (1998) Putnam, H.: Minds and machines. In: Sidney Hook (ed.) Dimensions of Mind, pp. 148–80. New York University Press, New York. (Reprinted in Putnam’s Mind, Language and Reality, Cambridge: Cambridge University Press, 1975.) (1960) Putnam, H.: Philosophy and our mental life. In: Mind, Language and Reality: Philosophical Papers, vol. 2, pp. 291–303. Cambridge University Press, Cambridge (1975) Radder, H.: Heuristics and the generalized correspondence principle. Br. J. Philos. Sci. 42(2), 195–226 (1991)

254

References

Radner, D., Radner, M.: Animal Consciousness. Prometheus Books, Buffalo (1989) Rainville, P., Duncan, G., et al.: Pain affect encoded in human anterior cingulate but not somatosensory cortex. Science 277(5328), 968–971 (1997) Reilly, W. K.: Remarks at the fourth meeting of the parties to the Montreal protocol (1992). URL http://www.epa.gov/history/topics/montreal/05.htm Robinson, W.: Understanding Phenomenal Consciousness. Cambridge University Press, Cambridge (2004) Rosenberg, G.:A Place for Consciousness: Probing the Deep Structure of the Natural World. Oxford University Press, Oxford (2004) Rosenthal, D.: Consciousness and Mind. Oxford University Press, Oxford (2005) Roy, C. S. Sherrington, C. S.: On the regulation of the blood-supply of the brain. J. Physiol. 11(1–2), 85–108; 158ff (1890) Rumelhart, D., McClelland, J., et al.: Parallel Distributed Processing: Explorations in the Microstructure of Cognition, vol. 1. MIT Press, Cambridge (1986) Russell, B.: Mysticism and Logic, and Other Essays. Totowa, Barnes & Noble Books, New York (1917/1981) Russell, B.: The Analysis of Mind. George Allen & Unwin, London (1921) Sahin, N.T., Pinker, S., et al.: Sequential processing of lexical, grammatical, and phonological information within Broca’s area. Science, 326(5951), 445–449 (2009). URL http://www. sciencemag.org/cgi/content/abstract/326/5951/445 Salam, A.: The role of chirality in the origin of life. J. Mol. Evol. 33, 105–113 (1991) Salmon, W.: Scientific Explanation and the Causal Structure of the World. Princeton University Press, Princeton (1984) Sandage, A.: Edwin Hubble 1889–1953. J. Royal Astron. Soc. Can. 83(6) (1989) Sanfey, A., Rilling, J., et al.: The neural basis of economic decision-making in the ultimatum game. Science 300(5626), 1755–1758 (2003) Sarnat, H., Netsky, M.: Evolution of the Nervous System. Oxford University Press, New York (1981) Schienle, A., Stark, R., et al.: The insula is not specifically involved in disgust processing: an fMRI study. NeuroReport 13(16), 2023–2026 (2002) Schrödinger, E.: Discussion of probability relations between separated systems. Proc. Camb. Philos. Soc. 31: pp. 555–63 (1935) Schroeder, T.: Three Faces of Desire. Oxford University Press, Oxford (2004) Schweber, S.: Physics, community and the crisis in physical theory. Phys. Today 46, 34–40 (1993) Scully, M., Drühl, K.: Quantum eraser: a proposed photon correlation experiment concerning observation and ‘‘delayed choice’’ in quantum mechanics. Phys. Rev. A 25, 2208–2213 (1982) Seager, W.: Weak supervenience and materialism. Philos. Phenomenol. Res. 48, 697–709 (1988) Seager, W.: Instrumentalism in psychology. Int. Stud. Philos. Sci. 4(2), 191–203 (1990) Seager, W.: Disjunctive laws and supervenience. Analysis 51, 93–98 (1991) Seager, W.: Consciousness, information and panpsychism. J. Conscious. Stud. 2(3), 272–88. (Reprinted in J. Shear (ed.) Explaining Consciousness, Cambridge, MA: MIT Press, 1997) (1995) Seager, W.: A note on the quantum eraser. Philos. Sci. 63, 81–90 (1996) Seager, W.: Theories of Consciousness. Routledge, London (1999) Seager, W.: Introspection and the elementary acts of mind. Dialogue 39(1), 53–76 (2000a) Seager, W.: Real patterns and surface metaphysics. In: Ross, D., Brook, A., Thompson, D. (eds.) Dennett’s Philosophy: A Comprehensive Assessment, pp. 95–130. MIT Press, Cambridge (2000b) Seager, W.: Emotional introspection. Conscious. Cogn. 11(4), 666–687 (2002) Seager, W.: The ‘intrinsic nature’ argument for panpsychism. J. Conscious. Stud., 13(10–11), 129–145. (Reprinted in A. Freeman (ed.) Consciousness and Its Place in Nature, Exeter: Imprint Academic, 2006.) (2006)

References

255

Seager, W.: A brief history of the philosophical problem of consciousness. In: Zelazo, P., Moscovitch, M., Thompson, E. (eds.) The Cambridge Handbook of Consciousness, pp. 9–34. Cambridge University Press, Cambridge (2007) Seager, W.: Panpsychism, aggregation and combinatorial infusion. Mind and Matter 8(2), 167– 184 (2010) Seager, W., Allen-Hermanson, S.: Panpsychism. In: Zalta, E. N. (ed.) The Stanford Encyclopedia of Philosophy. The Metaphysics Research Lab, fall 2008 ed. (2008). URL http://plato. stanford.edu/archives/fall2008/entries/panpsychism Seager, W., Bourget, D.: Representationalism about consciousness. In: Velmans, M., Schneider, S. (eds.) The Blackwell Companion to Consciousness, pp. 261–276. Blackwell, Oxford (2007) Searle, J.: The Rediscovery of the Mind. MIT Press, Cambridge (1992) Searle, J.: Consciousness and the philosophers. NY Rev. Books 44(4), 43–50 (1997) Seife, C.: What Is the universe made of? Science 309(5731), 78 (2005) Sellars,W.: The language of theories. In: Science, Perception and Reality, pp. 106–126. Routledge and Kegan Paul (1963a) Sellars, W.: Philosophy and the scientific Image of Man. In: Science, Perception and Reality, pp. 1–40. Routledge and Kegan Paul, London (1963b) Shapiro, L., Sober, E.: Epiphenomenalism: the dos and the don’ts. In: Wolters, G., Machamer, P. (eds.) Thinking About Causes: From Greek Philosophy to Modern Physics. University of Pittsburgh Press, Pittsburgh (2007) Shapiro, L. (ed.): The Correspondence Between Princess Elisabeth of Bohemia and René Descartes. University of Chicago Press, Chicago (2007) Shepherd, G.: Foundations of the Neuron Doctrine. Oxford University Press, Oxford (1991) Shoemaker, S.: Physical Realization. Oxford University Press, Oxford (2007) Siegelmann, H., Sontag, E.: Analog computation via neural nets. Theor. Comp. Sci. 131 (1994) Silberstein, M., McGeever, J.: The search for ontological emergence. Philos. Q. 49, 182–200 (1999) Sklar, L.: Physics and Chance: Philosophical Issues in the Foundations of Statistical Mechanics. Cambridge University Press, Cambridge (1993) Skrbina, D.: Panpsychism in the West. MIT Press, Cambridge (2005) Skrbina, D. (ed.): Mind that Abides: Panpsychism in the New Millennium. John Benjamins, Amsterdam (2009) Smith, P.: Explaining Chaos. Cambridge University Press, Cambridge (1998) Smith, W.: Church’s thesis meets quantum mechanics. In: NEC Research Technical Report (1999). URL http://www.math.temple.edu/*wds/homepage/churchq.ps Smith, W.: Church’s thesis meets the N-body problem. Appl. Math. Comput. 178(1), 154–183 (2006) Smolin, L.: The Life of The Cosmos. Oxford University Press, Oxford (1999) Smolin, L.: Three Roads to Quantum Gravity. Basic Books, New York (2001) Sober, E.: Screening-off and the units of selection. Philos. Sci. 59, 142–152 (1992) Sousa, C., Matsuzawa, T.: The use of tokens as rewards and tools by chimpanzees. Anim. Cogn. 4(3–4), 213–21 (2001) Sperry, R.: In defense of mentalism and emergent interaction. J. Mind Behav. 12(2), 221–246 (1991) Spinoza, B.: Ethics. In: Curry, E. (ed.) The Collected Works of Spinoza, pp. 401–617. Princeton University Press, Princeton (1677/1985) Stalnaker, R.: A Theory of conditionals. In: Rescher, N. (ed.) Studies in Logical Theory, pp. 98– 112. Blackwell, Oxford (1968) Stannett, M.: The case for hypercomputation. Appl. Math. Comput. 178(1), 8–24 (2006) Steiner, M.: The Applicability of Mathematics as a Philosophical Problem. Harvard University Press, Cambridge (1998) Stewart, L., Walsh, V., et al.: TMS produces two dissociable types of speech disruption. Neuroimage 13(6), S45 (2001) Stoerig, P., Cowey, A.: Blindsight in man and monkey. Brain 120, 535–559 (1997)

256

References

Stoljar, D.: Ignorance and Imagination. Oxford University Press, Oxford (2006) Stompe, T., Ortwein-Swoboda, G., et al.: Old wine in new bottles? Stability and plasticity of the contents of schizophrenic delusions. Psychopathology 36(1), 6–12 (2003) Stone, T., Young, A.: Delusions and brain injury: the philosophy and psychology of belief. Mind Lang. 12(3/4), 327–364 (1997) Strawson, G.: Realistic monism: why physicalism entails panpsychism. J. Conscious. Stud. 13 (10–11), 3–31. (Reprinted in A. Freeman (ed.) Consciousness and Its Place in Nature, Exeter: Imprint Academic, 2006) (2006) Strutt, J.: The principle of similitude. Nature 95(2368), 66–68 (1915) Stubenberg, L.: Neutral Monism. In: Zalta, E. N. (ed.) The Stanford Encyclopedia of Philosophy. The Metaphysics Research Lab, fall 2008 ed. (2008) URL http://plato.stanford.edu/ archives/fall2008/entries/neutral-monism Sur, M., Garraghty, P., et al.: Experimentally induced visual projections into auditory thalamus and cortex. Science 242(4884), 1437–1441 (2000) ’t Hooft, G.: Can quantum mechanics be reconciled with cellular automata? Int. J. Theor. Phys. 42(2), 349–354 (2003) ’t Hooft, G.: Entangled quantum states in a local deterministic theory (2009). URL http://arxiv. org/abs/0908.3408 Tateyama, M., Asai, M., et al.: Transcultural study of schizophrenic delusions: Tokyo versus Vienna versus Tuebingen (Germany). Psychopathology 31(2), 59–68 (1998) Tegmark, M.: Importance of quantum decoherence in brain processes. Phys. Rev. E 61(4), 4194– 4206 (2000) Tegmark, M.: Many lives in many worlds. Nature 448(7149), 23–24 (2007) Tegmark, M.: The mathematical universe. Found. Phys. 38 (2), 101–150. (2007b). URL http://arxiv.org/abs/0704.0646 Teller, P.: Relational holism and quantum mechanics. Br. J. Philos. Sci. 37, 71–81 (1986) Teller, P.: Twilight of the perfect model model. Erkenntnis 55, 393–415 (2001) Thijssen, J.: Computational Physics. Cambridge University Press, Cambridge (1999) Thomas, A.: Interplay of spin and orbital angular momentum in the proton. Phys. Rev. Lett. 101 (10) (2008) Thomson, W.: On vortex atoms. Proc. Royal Soc. Edinb. 6, 94–105 (1867) Tong, F., Nakayama, K., et al.: Binocular rivalry and visual awareness in human extrastriate cortex. Neuron 21, 753–759 (1998) Tonomura, A.: Demonstration of single-electron buildup of an interference pattern. Am. J. Phys. 57(2), 117–120 (1989) Turing, A.: On computable numbers, with an application to the entscheidungsproblem. Proc. Lond. Math. Soc. 2(42), 230–267. (Reprinted in Copeland (2004)) (1936) Turing, A.: Systems of logic based on ordinals. In: Copeland, J. (ed.) The Essential Turing, pp. 146–204. Oxford University Press (Clarendon), Oxford (1939/2004) Turing, A.: Computing machinery and intelligence. Mind 59(236), 433–460. (Reprinted in Copeland (2004)) (1950) Tye, M.: Ten Problems of Consciousness: A Representational Theory of the Phenomenal Mind. MIT Press, Cambridge (1995) Tye, M.: Consciousness and Persons. MIT Press, Cambridge (2003) Umeno, K.: Integrability and computability in simulating quantum systems. In: Hirota, O., Holevo, A.S., Caves, C.M. (eds.) Quantum Communication, Computing and Measurement, the Language of Science. Plenum Press, New York (1997) Vaidman, L.: Many-worlds interpretation of quantum mechanics. In: Zalta, E. N. (ed.) The Stanford Encyclopedia of Philosophy. The Metaphysics Research Lab (2008). URL http://plato.stanford.edu/archives/fall2008/entries/qm-manyworlds van Fraassen, B.: The Scientific Image. Oxford University Press (Clarendon), Oxford (1980) van Fraassen, B.: The Empirical Stance. Yale University Press, New Haven (2002) van Inwagen, P.: Material Beings. Cornell University Press, Ithaca (1990)

References

257

von Neumann, J.: Theory of Self-Reproducing Automata. University of Illinois Press, Urbana (1966) Watkins, P.: Story of the W and Z. Cambridge University Press, Cambridge (1986) Weinberg, S.: Dreams of a Final Theory: The Search for the Fundamental Laws of Nature. Pantheon, New York (1992) Weingarten, D.: Quarks by computer. Sci. Am. 274(2), 104–108 (1996) Weiskrantz, L.: Consciousness Lost and Found. Oxford University Press, Oxford (1997) Wigner, E.: The unreasonable effectiveness of mathematics in the natural sciences. Commun. Pure Appl. Math. 13, 1–14. (Reprinted in Wigner’s Symmetries and Reflections, Bloomington: Indiana University Press, 1967) (1960) Wigner, E.: Remarks on the mind-body problem. In: Good, I. (ed.) The Scientist Speculates, pp. 284–302. Heinemann, London (Reprinted in J. Wheeler and W. Zurek (eds.) Quantum Theory and Measurement, Princeton: Princeton University Press, 1984.) (1962) Williams, R., Barnes, E.: Vague parts and vague identity. Pac. Philos. Q. 90(2), 176–187 (2009) Williamson, T.: Vagueness. Routledge, London (1994) Wimsatt, W.: The ontology of complex systems. In: Matthen, M., Ware, R. (eds.) Biology and Society: Reflections on Methodology, pp. 207–274. University of Calgary Press, Calgary. Can. J. Philos. (supplementary volume) (Reprinted in Wimsatt’s Re-Engineering Philosophy for Limited Beings, Harvard University Press, Cambridge 2007) (1994) Winsberg, E.: Science in the Age of Computer Simulation. University of Chicago Press, Chicago (2010) Wittgenstein, L.: Philosophical Investigations. The Macmillan Company, New York (1953) Wolfram, S.: A New Kind of Science. Wolfram Media, Champaign (2002) Wong, H.Y.: Emergents from fusion. Philos. Sci. 73, 345–367 (2006) Woodward, J.: Causation and manipulability. In: Zalta, E. N. (ed.) The Stanford Encyclopedia of Philosophy. The Metaphysics Research Lab, winter 2008 ed. (2008). URL http://plato. stanford.edu/archives/win2008/entries/causation-mani Wright, P., He, G., et al.: Disgust and the insula: fMRI responses to pictures of mutilation and contamination. NeuroReport 15(15), 2347–2351 (2004) Yablo, S.: Mental causation. Philos. Rev. 101, 245–280 (1992) Yang, E., Blake, R., et al.: Fearful expressions gain preferential access to awareness during continuous flash suppression. Emotion 7(4), 882–886 (2007) Young, D.: Computational chemistry: a practical guide for applying techniques to real world problems. Wiley, New York (2001) Young, L., Camprodon, J., et al.: Disruption of the right temporoparietal junction with transcranial magnetic stimulation reduces the role of beliefs in moral judgments. Proc. Nat. Acad. Sci. 107(15), 6753–6758 (2010) Zeman, A.: Consciousness: A User’s Guide. Yale University Press, New Haven (2002) Zimmer, C.: Soul Made Flesh. New York: Free Press, (2004) Zotev, V., Volegov, P. et al.: Microtesla MRI of the human brain with simultaneous MEG. (2007). URL http://arxiv.org/abs/0711.0222 Zurek, W.: Decoherence and the transition from quantum to classical. Phys. Today 44(10), 36–44 (1991). URL http://arxiv.org/abs/quant-ph/0306072. Updated version (2003) Zurek, W.: Decoherence, chaos, quantum-classical correspondence, and the algorithmic arrow of time. Phys. Scr. T76, 186–206 (1998) Zurek, W.: Decoherence and the transition from quantum to classical-revisited. Los Alamos Sci 27, 86–109 (2002). URL http://www.fas.org/sgp/othergov/doe/lanl/pubs/number27.htm Zurek, W., Paz, J.: Decoherence, chaos and the second law. Phys. Rev. Lett. 72(16), 2508–2512 (1994) Zuse, K.: Rechnender Raum. Friedrich Vieweg & Sohn, Braunschweig (Schriften zur Datenverarbeitung, Band 1) (1969) An English translation is available at ftp://ftp.idsia.ch/pub/ juergen/zuserechnenderraum.pdf Zuse, Konrad: The Computer—My Life. Springer, New York (1993)

Titles in this Series

Quantum Mechanics and Gravity By Mendel Sachs Quantum-Classical Correspondence Dynamical Quantization and the Classical Limit By Josef Bolitschek Knowledge and the World: Challenges Beyond the Science Wars Ed. by M. Carrier, J. Roggenhofer, G. Ku¨ppers and P. Blanchard Quantum-Classical Analogies By Daniela Dragoman and Mircea Dragoman Life—As a Matter of Fat The Emerging Science of Lipidomics By Ole G. Mouritsen Quo Vadis Quantum Mechanics? Ed. by Avshalom C. Elitzur, Shahar Dolev and Nancy Kolenda Information and Its Role in Nature By Juan G. Roederer Extreme Events in Nature and Society Ed. by Sergio Albeverio, Volker Jentsch and Holger Kantz The Thermodynamic Machinery of Life By Michal Kurzynski

W. Seager, Natural Fabrications, The Frontiers Collection, DOI: 10.1007/978-3-642-29599-7,  Springer-Verlag Berlin Heidelberg 2012

259

260

Titles in this Series

Weak Links The Universal Key to the Stability of Networks and Complex Systems By Csermely Peter The Emerging Physics of Consciousness Ed. by Jack A. Tuszynski Quantum Mechanics at the Crossroads New Perspectives from History, Philosophy and Physics Ed. by James Evans and Alan S. Thorndike Mind, Matter and the Implicate Order By Paavo T. I. Pylkkänen Particle Metaphysics A Critical Account of Subatomic Reality By Brigitte Falkenburg The Physical Basis of The Direction of Time By H. Dieter Zeh Asymmetry: The Foundation of Information By Scott J. Muller Decoherence and the Quantum-To-Classical Transition By Maximilian A. Schlosshauer The Nonlinear Universe Chaos, Emergence, Life By Alwyn C. Scott Quantum Superposition Counterintuitive Consequences of Coherence, Entanglement, and Interference By Mark P. Silverman Symmetry Rules How Science and Nature Are Founded on Symmetry By Joseph Rosen Mind, Matter and Quantum Mechanics By Henry P. Stapp Entanglement, Information, and the Interpretation of Quantum Mechanics By Gregg Jaeger

Titles in this Series

Relativity and the Nature of Spacetime By Vesselin Petkov The Biological Evolution of Religious Mind and Behavior Ed. by Eckart Voland and Wulf Schiefenhövel Homo Novus—A Human Without Illusions Ed. by Ulrich J. Frey, Charlotte Störmer and Kai P. Willführ Brain–Computer Interfaces Revolutionizing Human-Computer Interaction Ed. by Bernhard Graimann, Brendan Allison and Gert Pfurtscheller Extreme States of Matter On Earth and in the Cosmos By Vladimir E. Fortov Searching for Extraterrestrial Intelligence SETI Past, Present, and Future Ed. by H. Paul Shuch Essential Building Blocks of Human Nature Ed. by Ulrich J. Frey, Charlotte Störmer and Kai P. Willführ Mindful Universe Quantum Mechanics and the Participating Observer By Henry P. Stapp Principles of Evolution From the Planck Epoch to Complex Multicellular Life Ed. by Hildegard Meyer-Ortmanns and Stefan Thurner The Second Law of Economics Energy, Entropy, and the Origins of Wealth By Reiner Kümmel States of Consciousness Experimental Insights into Meditation, Waking, Sleep and Dreams Ed. by Dean Cvetkovic and Irena Cosic Elegance and Enigma The Quantum Interviews Ed. by Maximilian Schlosshauer

261

262

Titles in this Series

Humans on Earth From Origins to Possible Futures By Filipe Duarte Santos Evolution 2.0 Implications of Darwinism in Philosophy and the Social and Natural Sciences Ed. by Martin Brinkworth and Friedel Weinert Probability in Physics Ed. by Yemima Ben-Menahem and Meir Hemmo Chips 2020 A Guide to the Future of Nanoelectronics Ed. by Bernd Hoefflinger From the Web to the Grid and Beyond Computing Paradigms Driven by High-Energy Physics Ed. by René Brun, Federico Carminati and Giuliana Galli Carminati The Dual Nature of Life Interplay of the Individual and the Genome By Gennadiy Zhegunov