WMU | Arts & Sciences | WMU Philosophy | Newman | Research
Please do not copy, distribute, or cite this paper. This is a work in progress.
David V. Newman
August 3, 1997
Copyright © 1997, David V. Newman
I argue in this paper that the mind-body problem can be solved using an emergence account of the mental. I review the concept of emergence as presented by C. D. Broad, and argue that mental properties should be understood as emergent properties rather than reductive or functional properties. I give a more precise definition of emergence, and I suggest that chaotic nonlinear dynamical systems provide a concrete example of emergent properties. I then argue that mental states are emergent properties of this kind. I conclude by arguing against Kim's recent suggestion that emergence theories are incompatible with physicalism, and by suggesting why the theory of emergence presented here is superior to O'Connor's theory of emergence.
Like a criminal with many aliases, the mind-body problem has come to be called by a number of evocative names. Levine calls it the problem of the "explanatory gap" (Levine 1983). Chalmers calls it the "hard problem of consciousness" (Chalmers 1995). Guzeldere calls it the "further how" problem (Guzeldere, 1995). An earlier discussion of the problem is summarized by Nagel's "what is it like to be a bat" (Nagel 1974). All these evocative phrases refer to the problem of explaining the existence of the mental (or more specifically, the subjective) in materialistic terms. The explanatory gap lies in the metaphorical distance between qualia and reductive or functional accounts of the mind. Consciousness is a hard problem because even the most sophisticated existing accounts of the mental do not include satisfactory explanations of how it arises from the activities of human brains. A full understanding of almost any current theory of the mind and brain still allows one to ask for a further account of how consciousness arises. A flurry of recent work on the mind-body problem reflects the fact that in spite of the difficulty that the problem has historically posed, many philosophers are confident that a solution is possible.
I believe that a solution to the mind-body problem lies in the doctrine of emergence.(2) In this paper, I argue that subjective mental states are emergent properties of the brain that arise because the brain is a chaotic nonlinear dynamical system. First, I discuss the concept of emergence. Second, I attempt to show on very general grounds that an emergence theory can solve the problem characterized so colorfully by Nagel, Levine, Chalmers, and Guzeldere. Third, I review portions of the theory of chaotic nonlinear dynamical systems and its application to the brain. Fourth, I argue that mental states can be analysed as emergent properties defined in terms of the basins of attraction of strange attractors of brain subsystems. Finally, I consider Kim's argument against emergence and O'Connor's competing theory of emergence, and I suggest why I think they are wrong. I believe that emergence is useful in accounting for consciousness because it explains how properties of a qualitatively different character from most physical properties can arise in purely physical systems.
In order to consider the claim that mental states are emergent, we must first understand the concept of emergence. Emergence is a relation between entities at different levels of organization in the world first discussed explicitly by C. D. Broad and other advocates of the doctrine of emergent evolution.(3) For Broad and the other emergent evolutionists, emergence was an alternative to the doctrines of vitalism and mechanism. The emergent evolutionists held that certain entites or properties should not be explained in terms of mysterious entelechies or in terms of strictly mechanical interactions. In more modern terms, the emergent evolutionists held a position different from either dualism or physicalism. As Broad put it, "an emergent quality is roughly a quality which belongs to a complex as a whole and not to its parts" (Broad 1925 p. 23). Thus, the doctrine of emergence is a kind of holism. A more specific explanation of the doctrine of emergence can be seen in the following quotation from Broad:
The emergent theory asserts that there are certain wholes, composed (say) of constituents A, B, and C in a relation R to each other; that all wholes composed of constituents of the same kind as A, B, and C in relations of the same kind as R have certain characteristic properties; that A, B, and C are capable of occurring in other kinds of complex where the relation is not of the same kind as R; and that the characteristic properties of the whole R(A,B,C) cannot, even in theory, be deduced from the most complete knowledge of the properties of A, B, and C in isolation or in other wholes which are not of the form R(A,B,C). (Broad 1925 p. 61)
In the first part of the definition, Broad rejects vitalism by requiring that emergent properties supervene on lower-level properties. This is partly accomplished by requiring that systems with emergent properties be mereological sums of lower-level entities. In the second part of the definition, Broad rejects mechanism by denying that emergent properties are nomologically related to lower-level properties. Thus, emergent properties of a system are not properties of some mysterious entelechy associated with that system; they are non-deducible properties of the system itself. Broad's emergence is a form of nonreductive supervenience,(4) since his emergent properties are completely determined by and dependent upon lower-level physical and relational properties, though the emergent properties cannot be defined in terms of the lower-level properties of the constitutive entities.(5)
On Broad's account, emergence theories also explain each emergent property. These explanations are provided by the structures of the wholes that give rise to emergent properties. Broad says that
[A non-vitalistic theory] "tries to explain the difference of behavior wholly in terms of difference of structure." On the first form of the theory the characteristic behavior of the whole could not, even in theory, be deduced from the most complete knowledge of the behavior of its components, taken separately or in other combinations, and of their proportions and arrangements in this whole. This alternative, which I have roughly outlined and shall soon discuss in detail, is what I understand by the "Theory of Emergence." (Broad 1925 p. 59)
Broad denies that such explanations lead to reduction on the grounds that emergent properties are those for which there are no general laws relating the properties of wholes to properties of their constituents (Broad 1925 pp. 64-65). Broad's view is that emergent properties can only be explained by examining the structures of the wholes in which they appear; nothing can be learned about emergent properties by studying the constituents of those wholes in isolation or by studying other wholes in which those constituents appear.
At this point some examples may help to explain the concept of emergence. In 1925, Broad felt that the property of being alive was a candidate for emergence since it was not apparently reducible to a mechanistic property of matter, and yet he thought that it could be explained in terms of the structure of particular living bodies without appeal to vital essences. Broad also suggested that liquidity and subjective mental properties (e.g. the characteristic odor of a chemical compound) were candidate emergent properties. Liquidity is now not a candidate emergent property because scientific advances have explained it in terms of mean molecular kinetic energy and the strength of inter-molecular attraction; I will argue that mental properties are emergent below. Some clearly non-emergent properties include the property of weighing 10 lbs, and having a particular physical shape. These properties can be easily derived from our knowledge of the weight and bonding properties of the atoms that make up any particular object. These two properties meet Broad's rough holistic criterion mentioned above, but they don't meet his more specific definition of emergence.
Emergence can be characterized in terms of three features of Broad's account: nonreduction, supervenience, and structural or holistic explanation. A theory containing emergent properties or entities is not reducible to lower-level theories. At the same time, a theory of emergent properties does not postulate anything that is not ontologically justified by the lower-level theories. Finally, the theory allows us to explain all the properties or entities required by the theory in an acceptable way. However, Broad's account of emergence can be improved. A stronger account of emergence will avoid the charges of epistemological relativism that were made against Broad's theory.(6) A stronger account of emergence will also show how it is possible for structural explanations of emergent properties of wholes to exist while there are no general laws relating emergent properties of wholes to properties of the constituents of those wholes.(7) I argue below that a definition of emergence following Broad but using contemporary analyses of supervenience, reduction, and explanation is sufficient to solve the mind-body problem.
The mind-body problem arises when we attempt to reconcile two theses about the world. The first thesis is that of materialism: everything that exists in the universe is a material or physical thing, reduces to material or physical things, or supervenes on material or physical things. The second thesis is what I will call mentalism: there are things that have mental properties. In the second thesis, the word 'mental' is normally intended to apply to such things as emotions, sensations, and thoughts. For a materialist, reconciling these two theses consists in explaining mental properties in physical terms. Such explanations are difficult because mental properties have a special character that seems to foil any attempt at explaining them in physical terms. Davidson explains the special character of the mental by appeal to what he calls the anomalousness of the mental (Davidson 1970). For Davidson, the mental is anomalous because there are no laws relating the mental to the physical in spite of the fact that the mental is dependent upon the physical.(8) Nagel explains the special character of the mental by appeal to the subjectivity of the mental (Nagel 1974). For Nagel, the mental is intrinsically first-person or subjective and it cannot be explained in physical terms since such explanations must be made from a third-person or objective point of view. A materialistic solution to the mind-body problem must thus somehow explain the mental in physical terms while simultaneously respecting the special character of the mental.
Most theories of the mind fail to meet these desiderata. Dualistic theories, whether substance dualism or property dualism, fail to explain mental properties in physical terms. It is, in fact, impossible for dualism to provide such an explanation since dualism requires that the mental and the physical be fundamentally different. Behaviorism and type-identity theories also fail to provide an account of the special character of the mental. These theories try to reduce the mental to the physical in a way that does not allow the mental to have a special character. Reductive accounts must explain the mental as simply another physical property without any particular distinguishing characteristics, otherwise the reduction cannot go through. Thus, straightforward versions of materialism or mentalism cannot meet both of the desiderata above because they emphasize only one of them, or only one of the two theses that must be reconciled.(9)
Computational functionalism also fails to meet the desiderata proposed above. Ned Block has argued that computational functionalism is unacceptable because it does not accurately characterize the class of sentient creatures (Block 1978). For Block, some clearly non-sentient entities can meet the functional specification of a sentient entity without having any mental properties. For example, Block argues that the people of China, if suitably organized and instructed, could implement the computational-functional characterization of a mind. However, Block suggests that there will be no qualia in such a situation. Thus, some crucial mental properties are left out of the functional description. In short, while it is possible for something that meets the functional specification of a sentient being to have a mind, it isn't necessary. Thus, the functional account of the mind isn't sufficiently explanatory.
Pure supervenience theories are also insufficiently explanatory. Pure supervenience theories say that the mental must depend upon and be determined by the physical (Kim 1982). However, they add little or nothing to this basic claim. While this allows for special mental properties, it doesn't say how they arise, or give any account of why they are special. Thus, neither computational functionalism nor pure supervenience are acceptable solutions to the mind-body problem.
In contrast, an emergence theory can meet the desiderata due to the three properties characteristic of emergence. Because emergence is a kind of supervenience that supports explanation, emergence theories of the mental can explain the mental in physical terms. Because emergence is nonreductive, it allows for mental properties that have a special character different from standard physical properties (which are reductive). The nonreductive character of emergence is consistent with both Davidson's and Nagel's characterization of the special character of the mental -- it allows us to explain the subjectivity and anomalousness of the mental without contradiction. Since an emergence theory of the mental requires that mental properties be irreducible, there will be no laws relating mental properties and physical properties, and there will be no third-person objective descriptions of mental properties in physical terms. Thus, an emergence theorist holds that both Davidson and Nagel are correct in suggesting that the mental has a special character, but denies that this character puts the mental completely beyond the reach of any materialistic explanation. This is the strength of emergence theories: they can incorporate properties of a special character while requiring that those properties supervene on the material world in a way that allows for acceptable explanations.(10)
So far, I have presented only the most general kind of reason to think that an emergence theory can solve the mind-body problem. There are two main obstacles at this point. First, one may ask if the concept of emergence is coherent. That is, one may ask if it is possible to concoct an emergence theory without contradiction or other undesirable consequences. Second, one may wonder if there is any reason to think that emergent mental properties actually exist. In the remainder of the paper, I present a precise account of emergence, I suggest how such an account can be used to explain mental properties, I argue against Kim's suggestion that downward causation is a problem for any theory of emergence, and I argue that the theory of emergence presented below is superior to O'Connor's theory of emergence.
The following definition of emergence is a modification and modernization of Broad's less precise account. Like Broad's definition, this definition expresses the idea that emergence is a variety of nonreductive supervenience, and it requires an explanation-supporting relation between emergent properties and their supervenience base. The explanation-supporting relation must be weak enough that reduction is not required. For this reason, nomological-deductive models of explanation are ruled out since they are so closely related to traditional accounts of reduction. The definition also avoids the chief problems with Broad's account. The first problem was that Broad's concept was interpreted to be epistemically relative. This is avoided by requiring that any theories discussed in the definition be ideal theories (i.e. theories that get everything right -- the best possible theories). Following Broad, the definition is cast in epistemological terms, but this does not imply that emergence is a strictly epistemological relation. Like many epistemological approaches to reduction, the definition is intended to characterize a metaphysical relationship. Requiring ideal theories in the definition helps here also, since ideal theories are clearly a definitional tool rather than a characterization of any particular epistemic situation. The second problem was that Broad's theory allowed for structural explanations of emergent properties, which some theories of explanation might suggest are incompatible with the non-reductive nature of these properties. Thus, the definition requires that emergent properties be nonstructural or not completely structural. The definition expresses each of the main concepts of supervenience, nonreduction, and explanation in a separate clause:
A property designated by a predicate P in an ideal theory T is emergent if and only if the following conditions are met:
1. T describes a class of systems SC which are structured aggregates of entities described by T'; T' is an ideal theory of those entities, and the entities described by T strongly supervene on those described by T'.
2. It is epistemically impossible to identify occurrences of the property designated by P with any occurrence of a property finitely definable in T'.
3. Each occurrence of the property designated by P is an occurrence of one of a definable set of properties PC, which is modeled by T'. Each member of PC is epistemically indistinguishable in T' from some other member of PC. (Newman 1995, 1996)
Some further explanation of the definition may be helpful. The first clause of the definition requires the kind of structured wholes that Broad discusses in the quotation provided earlier. It also requires strong supervenience as defined by Kim (Kim 1987). However, the particular form of supervenience is not crucial to what follows. I believe that Horgan's regional supervenience would work just as well (Horgan 1993). The second clause guarantees nonreducibility by preventing the definition of the emergent properties in terms of a lower-level theory. The idea here is that the emergent properties have some metaphysically special character that prevents their definition. This should also prevent emergent properties from being purely structural since structural properties are relational properties that can presumably be defined in physical terms. I have phrased the restriction in terms of epistemic impossibility because it must not contradict the first clause. However, the epistemic impossibility of the identities must be universal in order to avoid the charges of epistemic relativism levelled against Broad: it must be impossible to formulate the identities in any epistemic situation. I have used the notion of finite definability because I believe that any physical property can be characterized using an infinite disjunction describing each of its instances, and I want to rule out this kind of characterization as a possible reductive definition. The third clause guarantees the existence of some form of model-based explanation. If emergent properties are members of a larger class of properties which can be defined even when the particular members of the class cannot be defined, then explanations of those properties can proceed in terms of the features that all members of the class share. In other words, an emergent property will be explicable by virtue of some general characterization of that property that is not sufficiently precise to qualify as a reductive definition. The definition of emergence given above will be applied to a concrete example below.
There are two ways that the proposed definition might be incoherent: it might be internally contradictory, or it might not be distinct from pure supervenience, reduction, or dualism. First, consider the possibility that the definition is incoherent. If explanation or supervenience require reduction, then the definition would be internally inconsistent. Since contemporary models of explanation don't require reduction, it seems safe to assume that explanation and nonreduction are mutually consistent. Since supervenience is a more general concept than reduction (i.e. instances of reduction are instances of supervenience) they must be mutually consistent. Finally, since supervenience says nothing about explanation, and is a very general thesis, it too seems consistent with the existence of an explanation for supervenient properties. Thus, it seems a reasonable working hypothesis that the definition is internally consistent. Now consider the possibility that the definition is not distinct from pure supervenience, reductive materialism, or dualism. First, the relation defined is distinct from supervenience because the third clause is strictly stronger than supervenience. Second, if clause (2) successfully rules out reduction, emergence as defined here must be distinct from reduction.(11) It seems clear that reduction is ruled out since reduction relies on bridge principles between higher and lower-level theories, and the second clause effectively prohibits the formulation of such principles by ruling out any instances of such principles. If supervenience rules out dualism, then the inclusion of the first clause guarantees that the definition is distinct from dualism. Thus, the definition appears to be suitable subject to the worry that it might not have any instances in the universe and that it might not apply to mental states; these worries will be dealt with in the following two sections.
To prepare the way for the arguments to follow, a brief discussion of chaotic nonlinear dynamical systems (hereafter CNDSs) is in order.(12) I will briefly review here some important features of the class of CNDSs that are also dissipative systems. Roughly speaking, dissipative systems are systems that consume energy. There are chaotic systems that are not dissipative, but these do not play a role in my argument. I will also briefly review an argument I have given elsewhere that CNDSs have emergent properties. Finally, I will briefly present the evidence that human brains are chaotic systems.
CNDSs cannot be analyzed using the mathematical models and techniques that have dominated most scientific research. They exhibit apparently random behavior, which is part of the reason they are so difficult to model. However, they can be analyzed using a set of techniques sometimes called dynamical systems theory. Any dynamical system can be described by some number of variables; the abstract space constructed from the possible values of these variables is called the phase space of the system (this is also sometimes called a state space). The behavior of a system can be described as a set of trajectories within its phase space in the sense that when the system is started in any particular state, its dynamical evolution will result in a sequence of states that form a trajectory in the phase space. The trajectories in the phase space of a dynamical system typically converge on one or more regions in the phase space; these regions are called attractors, and the region of the phase space in which all trajectories converge on a particular attractor is called the basin of attraction of that attractor. Dynamical systems can be classified according to the topological properties of their attractors. The dynamical systems approach allows us to classify all dynamical systems using four patterns of behavior.
The first three kinds of behavior are associated with systems that can be analysed using traditional mathematical methods. If all the trajectories of a system converge on a single point, the system is said to have a point attractor. An example of such a system would be a simple pendulum with friction. Regardless of the speed or angle at which the pendulum is swung, it will always slow down and stop, hanging vertically. If all the trajectories of a system converge on a loop in the phase space, the system is said to have a limit cycle. An example of this kind of system is a driven pendulum with friction. Regardless of the speed or angle at which the pendulum is swung, it will eventually come to swing at a fixed frequency determined by the length of the pendulum, the weight of the bob, and the magnitude and frequency of the driving force. If all the trajectories of a system converge on a torus shape in such a way that there is no limit cycle, then the system is said to be quasiperiodic. An example of this kind of system is a pair of driven pendulums with incommensurable frequencies (i.e. the frequencies make an irrational ratio). The frequency of one pendulum controls the interval in which the system proceeds around the hole in the torus, and the frequency of the second pendulum controls the interval in which the system proceeds around the smaller cross-section of the limb of the torus. Because the two frequencies are incommensurable, the attractor of such a system wraps densely around the torus, but is not periodic. Systems with one of these three kinds of attractor are easily understood using traditional mathematical models and analytic techniques.
CNDSs exhibit behavior different from each of the three classes of systems introduced above. The trajectories of a CNDS converge on a highly complicated shape that is typically a fractal (i.e. it has non-integer dimension). This shape is a topologically closed set of non-periodic trajectories, and it is characteristically a set of closely packed highly similar trajectories something like a filo pastry. An example of a CNDS is a driven hinged pendulum. The behavior of this kind of system is highly sensitive to the initial speed and angle at which it is swung, and such a system will behave in a seemingly random fashion. The apparent randomness in the behavior of a CNDS arises because CNDSs have sensitive dependence on initial conditions. This is sometimes called the 'butterfly effect'. This means that very small perturbations of the state of a CNDS will have relatively large effects on its behavior (typically the effects are an exponential function of the perturbation, while non-chaotic systems may respond to perturbations in a linear or non-exponential way). The attractors that describe the behavior of CNDSs are known as 'strange attractors'.
Using the property of sensitive dependence on initial conditions, it is possible to show that CNDSs have emergent properties. In some CNDSs, perturbations so small as to be immeasurable can have effects on the behavior of the system that are so large as to invalidate any possible prediction of the course of a CNDS' behavior (Kellert 1993). Because we can only measure the state of a system with finite precision, and because any error in our measurement is magnified in an extreme way by a CNDS' sensitive dependence on initial conditions, there is a sense in which it is impossible to define the states of some CNDSs. For the same reason, it is impossible to define (or measure) the strange attractor of such a CNDS. However, if we understand the dynamics of a CNDS, we can determine an equivalence class of strange attractors within which its behavior will fall.(13) For this reason, being in the basin of attraction of a strange attractor is an emergent property of a CNDS. This property supervenes on the physical state of the CNDS, so clause (1) of the definition above is met. Because the states and attractors of the system cannot be defined, it is impossible to reduce the basin of attraction to physical terms, and clause (2) of the definition is met. Finally, because the behavior of the system can be explained in virtue of the equivalence class of strange attractors, clause (3) of the definition is met. Thus, there are physical systems that have emergent properties, and one of the remaining worries regarding emergence is eliminated.(14)
The arguments below regarding the emergence of mental properties depend on the fact that the brain is a CNDS. There are two basic reasons to think that this is the case: the first is empirical, the second is theoretical. Empirically speaking, there are two sets of results that suggest the brain is a CNDS. First, there are a number of studies that find strange attractors in human EEG recordings (Basar 1990). Second, Skarda and Freeman have found that rabbit olfactory systems exhibit different strange attractors when presented with different odors (Skarda and Freeman 1990). This suggests that all mammalian neural systems are CNDSs, and thus that human brains are CNDSs. Theoretically speaking, the brain is just the kind of nonlinear dynamical system that typically has chaotic dynamics. The constituent neurons and synapses of the brain exhibit nonlinear behavior, and they are connected together in such a way that sensitive dependence on initial conditions will result at a higher level (Foss 1992). Thus, although further empirical work remains to be done, it seems very likely that human brains are CNDSs.
It is now possible to consider the possibility that mental properties are emergent. If being in a mental state is analysed in terms of being in the basin of attraction of a strange attractor of the brain, then mental states are emergent because the basins of attraction of the strange attractors of any CNDS are emergent, as discussed above. The mental states I am concerned with here are the occurrent mental states. These are the states that are most closely associated with consciousness, subjectivity, and the mind-body problem as characterized by Nagel et. al. They are the states that are most difficult to explain, and they are the states that the emergence account is intended to handle. They include sensations, emotions, and having a thought, and they are typically defined in terms of their content. I argue here that we should understand these states in terms of strange attractors because our folk psychological understanding of these states requires that they be analyzed in terms of the basins of attraction of strange attractors rather than the basins of attraction of point attractors, limit cycle attractors, or torus attractors.(15)
There are a number of features of occurrent mental states that are easily noted, and which can be used to show that mental states should be associated with strange attractors. The first thing to notice about occurrent mental states is that they are part of a dynamical system, and that they are extended in time (Morton 1988). Mental states are part of an 'ecology' of the mind in which states come and go, and in which states and their relations to each other change as time passes. For example, if one is angry, one can also be hungry, and the hunger can come and go while the anger remains. Further, the state of anger itself may evolve over time, becoming (e.g.) less extreme as the source recedes in time, or changing as one considers different methods of exacting one's revenge. Thus, mental states have duration and are not completely uniform through that duration. In addition, mental states are not noticeably rhythmic: they do not have reoccurring features that occur at regular time intervals. Finally, mental states are neither predictable nor strictly random; no one can precisely predict the sequence, duration, or quality of our mental states, yet our mental states are not normally so random that they are surprising or shocking. Because the mind is plausibly constituted of a number of dynamical systems (e.g. the vision subsystem and the language subsystem), it is reasonable to ask about the attractors of these systems. By doing this, we will be able to show that mental states are plausibly considered to be emergent properties.
The folk psychological properties of mental states examined above make it plausible that mental states should not be modeled by point attractors, limit cycles or torii.(16) Because mental states evolve through time, and because they show no tendency to stay in a particular state for any extended period of time, it seems clear that the mind does not have point attractors within its phase space. Since mental states show no evidence of periodicity, we can also eliminate limit cycles from the possible features of the phase space of the mind. For the same reason, we can eliminate torii since the mind would exhibit periodicity with respect to a particular set of features (though not with respect to its total state) if there were torii in its phase space. Thus, by elimination, mental states should be modeled by strange attractors. There are also positive reasons to think that mental states are like strange attractors. For example, the seeming randomness of behavior characterized by a strange attractor may be likened to the unpredictability of human nature, and the sensitivity of human mental states to very small stimuli may be indicative of sensitive dependence on initial conditions. Finally, Skarda and Freeman's result concerning rabbit olfactory systems is very suggestive (Skarda and Freeman 1990). Since rabbits' brains exhibit different strange attractors when presented with different odors, and since experiencing a smell is a typical mental state of the kind that we seek to explain, it seems reasonable to conclude that rabbit mental states are associated with strange attractors. From this conclusion, it is a small leap to suppose that other mammals, including humans, have mental states similar to rabbits' mental states. Thus, when the mind is treated as a dynamical system, mental states should be modeled using the basins of attraction of strange attractors, and they are thus emergent by the argument above that being in the basin of attraction of a strange attractor is an emergent property of any chaotic system.
On this account, mental states are defined in terms of classes of similar strange attractors of brain subsystems. More precisely, being in a mental state is being in the basin of attraction of a strange attractor of some subsystem of the brain. Mental states are thus constituted by the basins of attraction of all members of a class of similar strange attractors in the phase space of the brain. This is not a reductive account since our inability to measure the brain with precision sufficient to overcome its sensitive dependence on initial conditions prevents us from defining any instance of a mental state in physical terms. In other words, we cannot identify mental state occurrences with the basin of attraction of particular strange attractors. Because we regard mental states as the basins of attraction of classes of strange attractors, this view is a kind of functionalism (broadly construed). Using a class of similar strange attractors to define a mental state is a functional definition of a mental state, since any physical state that instantiates the causal dynamics of the strange attractors in the class is a mental state of the corresponding kind. Unlike computational functionalism, the emergence account has a story to tell about the special character of mental states. Like computational functionalism, the emergence account suggests that mental states are multiply realizable. Further, because mental states are described by a class of strange attractors that all have similar features, we can explain the general features of those mental states even though we cannot identify them precisely. The level of precision available here seems to be about the level of precision afforded by folk psychology. If an account of mental states in terms of classes of strange attractors can explain the predictive successes and failures of folk psychology, then this is another reason to think that it is plausible.(17)
The key question, however, is whether or not the theory advanced here can solve Chalmers' hard problem. Can this theory bridge the explanatory gap? For two reasons, I believe that it can. First, the emergence theory postulates the existence of special physical properties fundamentally different from other physical properties. Since the emergent properties with which I analyze mental properties are supervenient on the physical, they are firmly grounded in accepted metaphysical entities. Since they are not reducible to the physical, emergent mental properties are fundamentally different from traditional physical properties like those considered in reductive or functional approaches to the mind-body problem. These two features suggests the foundations or footings of the desired bridge between the physical and the mental. The bridge itself is the holistic explanation of groups of emergent properties in terms of their supervenience base. The difficulty of constructing the bridge lies in the fact that individual emergent properties cannot be explained. This failure of individual explanation comes about because mental properties are fundamentally non-reductive and hence anomalous: mental properties are not finitely definable and they consequently cannot participate in any finitely expressible laws or explanations. Thus, if emergent properties are plausibly identifiable with mental properties the emergence theory relates the mental and the physical in a way that explains why the mind-body problem is so difficult. The emergence theory thus provides at least a plausible solution to the mind body problem. However, some will still be unsatisfied for the reason that the explanation given so far does not address the question of qualia.
The second reason I think that the emergence theory can bridge the explanatory gap is that I believe the emergence theory can explain some of the interesting features of qualia. Emergent properties are plausibly identifiable with mental properties if we consider the properties shared by all qualia rather than those of particular qualia. This procedure is required because the theory presented here suggests that no individual mental state can be explained. This comes about because the individual mental states cannot be adequately identified since they are states of chaotic systems, which have the property of sensitive dependence on initial conditions.(18) In particular, if mental states are not finitely describable we can explain Nagel's 'what it is like' and Guzeldere's further-how question. Mental states that are not finitely definable would be inherently private since it would be impossible to make measurements precisely enough to distinguish one mental state from another similar mental state. Similarly, any definition proposed for something that is not finitely definable would be incomplete, allowing a further-how question to be asked. Thus, the emergence theory agrees with the other investigators who have suggested that the subjective quality of mental states is inexplicable. However, the emergence theory offers an explanation for why this is the case, and still provides a materialistic foundation for those mental states and their unusual properties.
The emergence theory also allows for some explanation of other qualities of experience as described by William James using the metaphor of the stream of consciousness (James 1976). James' metaphor suggests that occurrent mental states have a dynamic kind of flow in which the boundaries between particular mental states are unclear, and in which the distinctions between different mental states are very subtle. If being in a mental state is being in the basin of attraction of a strange attractor of a brain subsystem, the subtlety of the boundaries between mental states and the subtlety of their differences is explained by the fact that the basins of attraction of strange attractors can have fractal dimension, meaning that their boundaries are so complicated that they cannot be easily characterized. Furthermore, several such basins of attraction can be tangled together in a complex way, making the differences between them very small. James' metaphor also suggests that occurrent mental states are inchoate and non-linguistic. While this can only be true of some mental states, it too is explicable in the theory presented here. Low-level physical states of the brain, it seems, must be the foundation on which linguistic states are built. For if qualitative mental states were linguistic, then the problem of expressing what they are like would seem to be less severe than it is. Thus, qualia are probably non-linguistic, which is explained in the current theory by fact that strange attractors are non-linguistic. So James' classic description of consciousness and qualia is largely explained by the emergence theory.
The theory presented here has not really explained the features of qualia for which many seek an explanation. Much work remains to be done, both theoretical and experimental, to make the proposed relation between qualia and strange attractors more plausible, and to further elucidate the nature of our mental states. However, if the theory presented here is correct, there will always be unanswered questions about mental states. Indeed, if the theory presented here is correct, some of these questions will be unanswerable. Thus, it is not a criticism of this account to suggest merely that it does not provide a complete explanation of mental states. It is my hope that further investigation will provide more explanation than is given here, but I believe that the foundation laid here is sufficient to allow a view across the explanatory gap even if no bridge is possible.
Having given a positive argument for the thesis that mental properties should be understood as emergent properties of the brain. I turn now to Jaegwon Kim's argument that emergence is not a plausible account of the mental. This argument suggests that any emergence account must accept the existence of downward causation and that this violates the causal closure of the physical world (Kim 1993). Further, Kim suggests that violation of the causal closure of the physical world is not to be countenanced by any physicalist. Kim's argument is that unless they are considered epiphenomenal, mental states must have causal powers. It is certainly plausible that mental states cause further mental states. However, Kim argues that mental states must also cause physical states since otherwise, there is causal overdetermination of the subsequent mental states (i.e. from the physical realization of the second mental state and from the first mental state). Kim suggests that the solution to this problem is to allow that each mental state causes the physical realization of subsequent mental states to occur, thus also causing the subsequent mental state. This is why Kim thinks that emergence theorists must accept downward causation. Kim then claims that this violates the causal closure of the physical since on this account the physical realization of a mental state cannot cause the physical realization of a subsequent mental state on pain of epiphenomenalism. Thus, says Kim, emergentists cannot be physicalists since physicalism requires the causal closure of the physical.
The causal closure of the physical is not really violated in this case. The mental state that Kim is worried about is supervenient on a physical state. All the causal powers of the mental state must supervene on the causal powers of the physical state. While the causal powers of the mental state are not reducible to those of the physical state, they do not originate outside the physical world. I would suggest that the causal powers of the mental state are emergent, making them acceptable to the physicalist in the same way that the mental state itself is acceptable to the physicalist.(19) Another reason to think that Kim's argument is incorrect is that it applies to functionalism as well as emergence since it nowhere relies on features of an emergence theory that are not also shared by a functionalist theory. Yet it is highly plausible that one can consistently be both a functionalist and a physicalist. Thus, contrary to Kim's suggestion, the causal closure of the physical world is maintained, and emergence theorists can be physicalists.
Before concluding, I would also like to consider a competing account of emergence by Timothy O'Connor (O'Connor 1994). This account of emergence is formulated in terms of supervenience, non-structuralism, and novel causal effectiveness. O'Connor understands supervenience in much the same way that I do, so any differences between my view and O'Connor's will involve differences between his non-structuralism and novel causal effectiveness conditions on the one hand, and my non-reduction and explanatory conditions on the other. For O'Connor, nonstructuralism rules out structural properties and properties inherited by a whole from one or more of its parts. I believe O'Connor thinks non-structural properties of a whole are neither dependent upon any physical relation that holds between the constituents of a whole, nor a kind of summation of a property over all constituents of a whole. Nonstructuralism and nonreduction both play similar roles in their respective accounts of emergence, but the two conditions are quite different. Nonstructuralism seems to be strictly stronger than nonreduction since nonreduction allows structural properties so long as they cannot be defined, yet it seems impossible for nonstructural properties to be definable. For O'Connor, novel causal effectiveness is a form of downward causation from the emergent level to a lower level that brings about effects that would not have occurred if the emergent property had not been instantiated. This is at least a denial of epiphenomenalism with respect to emergent properties. Novel causal effectiveness is a very different kind of constraint than the explanatory constraint in the account presented above. Due to the interactions between the various components of the two accounts, a conceptual comparison is not straightforward; I will follow O'Connor in comparing the extensions of the defined concepts.
O'Connor criticizes accounts of emergence similar to Broad's on the grounds that they classify some non-emergent structural properties as emergent. His claim is that any account of emergence based on the idea that emergent properties cannot be deduced must include within the class of emergent properties causal properties that depend on the macro-structure of the wholes that have them. He further claims that these causal properties are not emergent because they are structural properties, and that nondeducibility definitions of emergence are therefore invalid. In order for this argument to apply to the account presented above, it must be the case that my definition of emergence cannot rule out all causal properties that depend on the macro-structure of the wholes that have them, and it must be the case that all such causal properties are not emergent. To show that the account of emergence given above is not subject to O'Connor's criticism, I must show that one or the other of these claims is false.
Consider the claim that there are causal properties that depend on the macro-structure of the wholes that have them which meet the definition of emergence presented above. That definition suggests that properties that can be finitely defined in terms of the properties of the constituents of a whole are not emergent. It thus rules out those causal properties that depend on the macro-structure of the wholes that have them in a way that is finitely definable. However, the definition above does not rule out all of O'Connor's worrisome properties since it seems possible that there are properties that depend on undefinable macro-structural properties. Hence we must consider O'Connor's second claim. For O'Connor, emergent properties cannot depend on the macro-structure of the wholes that have them. I suspect that the reason for this is either that such dependence seems to allow us to explain the emergent property in a reductive fashion, or that such dependence amounts to downward causation. Since we have rejected downward causation in the discussion of Kim's criticism above, this leaves the possibility of reductive explanation. However, as we have seem above, dependence may only allow us to make a very general kind of explanation, particularly when the emergent property is undefinable. Therefore, if emergent properties can be dependent on the macro-structure of a whole in a way that does not allow detailed explanations, O'Connor's second claim is false, and his criticism of Broad's account will not apply to the account advanced here.
I believe that the account presented here is superior to O'Connor's for several reasons. First, defining emergence in terms of indefinability has the useful consequence that we can allow emergent properties that are weakly dependent on the wholes that have them. This seems right since otherwise we have trouble explaining the emergent properties, and the supervenience of the emergent properties on their supervenience base becomes mysterious. Second, properties that meet the third clause of the definition presented above will have novel causal effectiveness (when this is taken in a way not dependent on downward causation) since these properties will have at least the causal powers of the larger class of properties of which they are members. Third, O'Connor's definition of emergence depends on the notion of downward causation, which I believe is troublesome and fraught with difficulty. Fourth, O'Connor's concept of emergence allows for emergent properties that are inexplicable in a strong sense, and this means that there will be little practical difference between these properties and truly mysterious properties for which there is no scientific justification.(20) Fifth, and finally, I believe that definability is a clearer concept than nonstructuralism, and that it is more useful due to its connections to computability and other concepts. Thus, the choice between O'Connor's theory of emergence and that presented above rests on the idea that emergent properties can be weakly dependent on structural (but undefinable) properties of a whole, but I believe that the theory presented here more closely represents our intuitive notion of emergence.
If the arguments I have presented above are correct, there are a number of important things to notice. First, emergence is a variety of metaphysical realism with respect to the mental. Emergence accounts do not eliminate the mental, and yet they suggest that mental properties are metaphysically different from standard reductive physical properties. Second, emergence is not a form of epiphenomenalism since qualia are information-bearing states of physical brains that in some sense constitute human mental life. As suggested in the response to Kim, emergent mental properties have causal powers, and they come by them honestly. Third, emergence provides an opportunity for a partial explanation of mental states, and thus is superior to functionalism and dualism in this respect. Finally, the hypothesis is empirically testable. We can perform experiments to determine if human brains are chaotic systems, and whether or not mental states are correlated with strange attractors in the state space of the brain. All four of these features are advantages for a theory of the mind.
The emergence account also provides an explanation for the various aliases under which the mind-body problem has come to be understood. The explanatory gap exists because the accounts we have examined to date do not provide the kind of positive explanatory structure that emergence does, and because the nature of mental states foils other explanatory structures. The problem seems hard because we have been looking for reductive or functional accounts rather than an emergence account. We can ask Guzeldere's further-how question because we have focused on reductive explanations, blinding ourselves to other possibilities. Thus, since the emergence account has a number of theoretically advantageous properties, and because it explains the difficulty we have encountered in trying to solve the problem, there are good reasons to investigate emergence further, both theoretically and empirically.
Basar, Erol. 1990. Chaotic Dynamics and Resonance Phenomena in Brain Function: Progress, Perspectives, and Thoughts. In E. Basar (Ed.). Chaos in Brain Function. 1-30. Berlin: Springer-Verlag.
Beckermann, Ansgar, Hans Flohr, and Jaegwon Kim (Ed.). 1992. Emergence or Reduction?: Prospects for Nonreductive Materialism. Edited by R. Posner and G. Meggle. Foundations of Communication and Cognition. Berlin: Walter de Gruyter.
Block, Ned. 1978. Troubles With Functionalism. In C. W. Savage (Ed.). Perception and Cognition: Issues in the Foundations of Psychology. 261-325. Minneapolis: University of Minnesota Press.
Broad, C. D. 1925. The Mind and its Place in Nature. Edited by C. K. Ogden. International Library of Psychology, Philosophy and Scientific Method. New York: Harcourt, Brace and Company.
Causey, Robert L. 1977. Unity of Science. Edited by J. Hintikka, R. S. Cohen, D. Davidson, G. Nuchelmans and W. C. Salmon. Synthese Library. Dordrecht: D. Reidel Publishing Company.
Chalmers, David. 1995. Facing up to the Problem of Consciousness. Journal of Consciousness Studies 2 (3) : 200-219.
Davidson, Donald. 1970. Mental Events. In L. Foster and J. W. Swandon (Ed.). Experience and Theory. 79-101. Amherst: University of Massachusetts Press.
Devaney, Robert L. 1989. An Introduction to Chaotic Dynamical Systems. Second Edition. Edited by R. L. Devaney. Addison Wesley Studies in Nonlinearity. Redwood City, California: Addison-Wesley Publishing Company, Inc.
Foss, Jeffrey. 1992. Introduction to the Epistemology of the Brain: Indeterminacy, Micro-Specificity, Chaos and Openness. Topoi 11 : 45-57.
Gleick, James. 1987. Chaos: Making a New Science. New York: Penguin Books.
Guzeldere, Guven. 1995. Consciousness: What it is, How to Study it, What to Learn from its History. Journal of Consciousness Studies 2 (1) : 30-51.
Hilborn, Robert C. 1994. Chaos and Nonlinear Dynamics: An Introduction for Scientists and Engineers. Oxford: Oxford University Press.
Horgan, Terrence. 1993. From Supervenience to Superdupervenience: Meeting the Demands of a Material World. Mind 102 (408) : 555-586.
Horgan, Terrence, and John Tienson. 1988. Settling into a New Paradigm. The Southern Journal of Philosophy 26 (Supplement) : 97-113.
Jackson, Frank. 1982. Epiphenomenal Qualia. Philosophical Quarterly. 32 : 127-136.
Jackson, F. 1986. What Mary Didn't Know. Journal of Philosophy 83 : 291-295.
James, William. 1976. Essays in Radical Empiricism. Edited by F. H. Burkhardt. The Works of William James. Cambridge, Massachusetts: Harvard University Press.
Kellert, Stephen H. 1993. In the Wake of Chaos: Unpredictable Order in Dynamical Systems. Edited by D. L. Hull. Science and Its Conceptual Foundations. Chicago: The University of Chicago Press.
Kim, Jaegwon. 1982. Psychophysical Supervenience as a Mind-Body Theory. Cognition and Brain Theory 5 (2) : 129-147.
Kim, Jaegwon. 1984. Concepts of Supervenience. Philosophy and Phenomenological Research 45 : 153-176.
Kim, Jaegwon. 1987. 'Strong' and 'Global' Supervenience Revisited. Philosophy and Phenomenological Research 48 : 315-326.
Kim, Jaegwon. 1993. The Non-Reductivist's Troubles with Mental Causation. In J. Heil and A. Mele (Ed.). Mental Causation. 189-210. Oxford: Clarendon Press.
Klee, Robert L. 1984. Micro-Determinism and Concepts of Emergence. Philosophy of Science 51 (1) : 44-63.
Levine, Joseph. 1983. Materialism and Qualia: The Explanatory Gap. Pacific Philosophical Quarterly 64 : 354-361.
McLaughlin, Brian P. 1992. The Rise and Fall of British Emergentism. In A. Beckermann, H. Flohr and J. Kim (Ed.). Emergence or Reduction? Prospects for Nonreductive Materialism. 49-93. Berlin: De Grutyer.
Morton, Adam. 1988. The Chaology of Mind. Analysis 48 (June) : 135-142.
Nagel, Thomas. 1974. What is it Like to be a Bat? Philosophical Review 4 : 435-450.
Newman, David. 1995. Chaos and Consciousness. Ph.D. Dissertation, University of Texas.
Newman, David V. 1996. Emergence and Strange Attractors. Philosophy of Science 63 (2) : 244-260.
Nicolis, G., and I. Prigogine. 1989. Exploring Complexity: An Introduction. New York: W.H. Freeman and Company.
O'Connor, Timothy. 1994. Emergent Properties. American Philosophical Quarterly 31 (2) : 91-105.
Port, Robert F., and Timothy van Gelder (Ed.). 1995. Mind as Motion: Explorations in the Dynamics of Cognition. Cambridge, Mass.: MIT/Bradford.
Putnam, Hillary. 1975. The Nature of Mental States. In Mind Language and Reality. 429-440. Cambridge: Cambridge University Press.
Skarda, Christine, A., and Walter J. Freeman. 1990. Chaos and the New Science of the Brain. Concepts in Neuroscience 1 (2) : 275-285.
Smolensky, Paul. 1987. The Constituent Structure of Connectionist Mental States: A Reply to Fodor and Pylyshyn. Southern Journal of Philosophy 26 (Supplement) : 137-160.
Stephan, Achim. 1992. Emergence - A Systematic View on its Historical Facets. In A. Beckermann, H. Flohr and J. Kim (Ed.). Emergence or Reduction? Prospects for Nonreductive Materialism. 25-48. Berlin: De Grutyer.
Stewart, Ian. 1989. Does God Play Dice?: The Mathematics of Chaos. Cambridge, Massachusetts: Basil Blackwell Inc.
Van Cleve, James. 1990. Mind-Dust or Magic? Panpsychism Versus Emergence. Philosophical Perspectives 4 : 215-226.
(1) My thanks to Nick Asher, Rob Koons, Johanna Seibt, Fred Kronz, Cory Juhl, James Garson, Robert Causey, Jeff Foss, Pat Manfredi, and the philosophy faculty of Western Michigan University for fruitful discussion of these issues, or for comments on various versions of this research.
(2) For other recent discussions of the concept of emergence, see (Beckermann, Flohr, & Kim 1992).
(3) I use Broad as the paradigmatic example of an emergence theory because I find Broad's work most similar to my own way of thinking about emergence. However, Klee (1984) compares different theories of emergence and argues that those like Broad's are more plausible than other versions of emergence. Klee divides emergence theories into four groups based on the fundamental characteristic of emergent properties: unpredictability, novelty, variability, and downward causal effectiveness. Broad's theory and mine characterize emergence in terms of unpredictability or undefinability.
(4) I mean supervenience in the general sense of (Kim 1984).
(5) In this paper, I rely on a standard understanding of reduction (Causey 1977). Cory Juhl has suggested to me that it is possible to interpret emergence as a weak kind of reduction. Even if this is possible, I think it confuses things rather than clarifying them.
(6) See (Nagel 1961 pp. 366-379) and (Hempel 1948 pp. 245-290).
(7) If a structural explanation of a property involves general laws relating that property to the properties of the entities that constitute the structure, then Broad's program fails. It is thus important to be clear on how it is possible for emergent properties to be explained in ways that do not require reduction. I hope to discuss this issue at another time.
(8) Note the similarity between Broad's view and Davidson's. Both suggest that mental properties are not related to physical ones by any general laws.
(9) Panpsychism meets the desiderata, but (Van Cleve 1990) argues that emergence is a better hypothesis than panpsychism if the argument for panpsychism is Nagel's. Mental properties are also unexplained under panpsychism.
(10) Of course, models of explanation that do not require objective laws and reductions to physics are required for the emergence approach to work.
(11) Searle's version of emergence doesn't seem to be distinct from reduction (Searle 1992). Searle himself calls his view causal reductionism but denies that ontological reduction of the mental is possible.
(12) Nontechnical discussions of chaotic nonlinear dynamical systems can be found in (Gleick 1987) and (Stewart 1989). More technical material can be found in (Devaney 1989) and (Hilborn 1994).
(13) The equivalence class of strange attractors will be defined by topological conjugacy of strange attractors. This is a kind of topological isomorphism between regions of the state space.
(14) This argument was originally presented more fully in (Newman 1995) and (Newman 1996).
(15) For further discussion of the approach in which one treats the mind and brain as a dynamical system, see (Horgan 1988) and (Port and Van Gelder 1995).
(16) For the purposes of this analysis, I have made the simplifying assumption that environmental influences do not significantly alter the nature of mental states or strange attractors. This deserves further investigation in the light of recent externalist accounts of the mental. On the other hand, Nicolis and Prigogine suggest that environmental influences like those that human beings are under are typical of chaotic systems. This influence, which they call a 'nonequilibrium constraint' may be an identifying feature of chaotic systems (Nicolis & Prigogine 1989).
(17) For a similar idea applied to connectionism, see Section 3 of (Smolensky 1987).
(18) A number of other arguments also suggest that it is impossible to explain individual mental states in a functional or reductive way, including Putnam's arguments for multiple realizability (Putnam, 1975), Block's argument against functionalism (Block, 1978) and Jackson's argument against physicalism (Jackson, 1982, 1986).
(19) Fahey & Zenzen made a similar defense of nonreductive physicalism in their 1995 presentation to the APA Eastern Division conference entitled "Kim on Emergentism: Dead Again?", as does (O'Connor 1994).
(20) I believe Broad would call a theory with such properties vitalistic.
WMU | Arts & Sciences | Philosophy Dept. | Newman | Research
Last Modified: April 30, 1999
By: David Newman