Review of Matter and Consciousness with comments

Review of Matter and Consciousness with comments

Paul M. Churchland

MIT Press, 1984

ISBN 0-262-53074-0

After an introductory chapter, follow very useful chapters on ontological, semantic, epistemological, and methodological problems of consciousness. These are followed by simple chapters on artificial intelligence and neurology, and a final weak summary chapter.

Four chapters of the book summarize positions taken on four interrelated major questions of consciousness: the mind/body (or ontological) problem, the semantical problem, the epistemological problem, and the methodological problem. Two chapters provide brief overviews of contributions from artificial intelligence and neuroscience. A final chapter speculates on the long range prospect for consciousness in the universe.

The Mind/Body Problem

Ontology means the consideration of what exists. For consciousness, this question has traditionally been called the mind/body problem. The two major categories of positions are called dualism and materialism. Each of these categories has several varieties. It will become clear that the set of positions set out may be more nearly characterized as lying along a scale from extreme dualism to extreme materialism, that such is the order of presentation, and that some care needs to be taken near the boundary to distinguish a dualist position from a materialist position.

Dualism again falls into two branches, substance dualism and property dualism. Substance dualism maintains that there is a substance different from ordinary matter of which mind is made. Property dualism maintains that the mind has a set of properties shared by no other physical systems.

Cartesian dualism is the classical position established by Rene Descartes. The mind is a non-spatial thinking substance entirely distinct from the body. The mind is in causal interaction with the body. The reasons Descartes advanced to support this position were first that by introspection it was clear that he was a mind and not a body; furthermore, that he could not imagine how a physical system could use (second) language or (third) reason.

A problem with Cartesian dualism that even Descartes struggled with was how a substance with no mass, no shape, and no position could interact with a body at all. Also, with time, we have changed our notions of matter (so that an electron is currently described as particle of matter with no extension in space, for instance). Accordingly Descartes' notions of where to divide spirit from matter, if such a division is to be made, now seem founded on an outmoded notion of matter, and the position is of primarily historical interest.

Popular dualism supposes a thinking substance distinct from ordinary matter, but having at least a location, to wit, inside the body with which it is associated. The spatial overlap provides more scope for providing an interaction mechanism, as by exchanging some form of energy not currently known to physics. Since matter is now viewed as a form of energy, the spirit might also be some form of energy.

Epiphenomenalism is a form of property dualism holding that mental properties are caused by the brain but do not affect the brain. They are simply "above" (epi-) the phenomena of the brain. The mental properties are regarded as emergent properties of the brain. This view might be taken on considering that neuroscience shows a thoroughly material system, with no trace of another kind of substance than matter, which seems to account for all of the activity of the brain and body while at the same time wishing to acknowledge the reality of personal experience of beliefs, desires, and perceptions. Epiphenominalism thus strikes a bargain to admit the reality of mental properties but to demote them to phenomena which have no role in the scientific explanation of human or animal behavior.

Interactive property dualism also holds mental properties to be emergent properties of the brain, but maintains that the properties are irreducible - novel properties beyond prediction or explanation of physical science. Furthermore, the mental properties do have causal effects on the brain. The irreducibility of mental properties is what leads to calling this a dualist position. A problem with this position is the simultaneous claim of evolutionary emergence of mental phenomena and of their irreducibility.

Churchland considers four general arguments advanced for dualism. The first is that religions are more general theories than theories of consciousness, and that it is a valid argument to appeal for support from the more general to the specific. Second is the argument from introspection, that one is not aware introspectively of a neural network pulsing with electrochemical activity, but of beliefs, values, goals, and so forth. Third is the argument that various phenomena (language and reasoning as suggested by Descartes, and more recently perceptions and meaning) are not reducible to physical phenomena. Fourth is that only a dualist view can explain parapsychological phenomena.

Each of the these arguments for dualism has serious problems however. Religions have been rather spectacularly poor as theories of the structure of the universe, so appeal to them as a more general theory is appeal to a weak general theory.

The introspection argument simply assumes that what we perceive internally is what is real. But we now believe that what we perceive externally, such as solids and sounds, does not reflect what we would claim to exist - collections of atoms or compression waves in a gas. Thus why should we expect internal perception to reflect what we would claim to exist within the body?

Irreducibility arguments are appeals to negative results. But current inability to explain a mental phenomenon does not imply perpetual inability to explain it. In particular current research on reasoning, language, and perception rather strongly suggests that these can be reproduced by understandably material means.

The parapsycholgy claim is a specific instance of an irreducibility claim. Even if the phenomena are real, physical explanations can be suggested. But the claimed phenomena Churchland then sets forth the main arguments against dualism before going on to description of the materialist positions. Dualism violates Occam's Razor. Dualism lacks explanatory power compared to material explanations. Observations of damage to various portions of the brain show that the capacity for specific mental abilities is correspondingly altered (often removed), whereas, if dualism held, then one would expect the various mental states to be unaffected by changes in the brain. The evolutionary account of our origins shows that we have a nervous system in common with most of our multicell ancestors and their living representatives, so that our nervous system differs in degree but not kind from theirs.

Churchland begins the list of materialist positions with philosophical behaviorism. Here, it is asserted that statements about phenomena such as sensations and emotions are really shorthand expressions for longer descriptions of externally observable behavior. The position give no account of the subjective experience of these phenomena, a problem seriously addressed by the behaviorists and discussed at a later point by Churchland. A second problem with behaviorism was the extensive, and apparently infinite, descriptions required to specify what external observables characterized an internal state. With the development of other materialist positions, behaviorism was rapidly abandoned.

Reductive materialism, or the identity theory asserts that mental states are states of the brain. As an analogy, a one to one mapping between the pitch of a tone and the frequency of the vibrations in air.

As to why one might expect to find a one to one mapping for mental phenomena, four reasons are advanced. First is the material development of an adult human from a single cell. Second is a coherence with an evolutionary view of how the brain developed. Third is the experimentally observed neural dependence of all known mental phenomena. Fourth is the success in totally unraveling the neurological structures of a few simple creatures.

Several arguments against the identity theory are found to be weak. Introspective experience is in fact easily explained rather than impossible to explain. Claims that identity theory make a category error do not stand up to examination.

Functionalism is the heir to the behaviorist position. It is asserted that essential characteristics of any mental state are the set of causal relations with bodily behavior, other mental states, and external effects on the body. The difference between this and the behaviorist position is the insistence that the set of relations between mental states cannot be reduced or eliminated.

A functionalist critique of the identity theory highlights their difference. An alien being can easily be imagined who would have the same functional mental states as a human, but an entirely different physical basis. If there are many ways of realizing a set of mental states, then it seems unlikely to find an identity.

The functionalist position, like the behaviorist position, does not include notions of the inner perception of mental states. Various objections to the functionalist position have been raised and debated on this basis. They do not seem to eliminate the position from consideration.

Eliminative materialism expects that rather than a one to one mapping between the currently conceived set of inner mental states and the brain it will be found that an entirely different set of states will be suggested by neurological research. It is argued that the currently conceived set of inner states constitutes a theory of the mind, our first such theory. Since our first theories in every other field were wildly off the mark, why should we expect to have hit it just right in this field? Development of a more accurate theory can be expected to change our perceptions (just as the perception of a literal sphere of the heavens was changed by astronomical theories) and in such a way as to allow much more powerful predictions, it is asserted.

Arguments for eliminative materialism include first that the current theory has many important features, such as sleep, learning, and mental illness which are not understood. Second, is the historical wrongness of our first theories cited above. Third, it is argued that reduction, as required for an identity theory, is requires a two sided construction and is thus intrinsically more difficult than a one sided construction from the ground up.

Only a weak argument against eliminative materialism stands scrutiny, but it makes an important point. Identity theory and eliminative materialism can be seen as end points on a continuum in which some of the current theory will be discarded and some will survive.

A major strength of eliminative materialism even if it is not correct, in my view, is the identification of our current perceptions of our mental states as a theory. It thus places the entire subject in a framework within which we have provided ourselves with a number of effective tools.

The Semantical Problem

Churchland describes three notions of the basis for the meaning of terms used in discussing the mind.

An ostensive definition is one given by pointing. It is suggested by some that only by an "inner ostention", pointing to inner experiences, can such terms as pain, itch, and sensation-of-red be defined. Inner ostention is clearly inadequate for giving the meaning of propositional beliefs, and is thus incomplete at best. Even in the clearest cases, such as "pain", the terms cover a range of non-uniform phenomena (such as headache, burns, blinding light or deafening noise) which makes the contention that any specific mental phenomenon can be pointed questionable. Also, since the ostention is inner, it cannot be clear that the phenomena nominally pointed to are identical in different people.

Behaviorists claim that the meaning of a term describing a mental state is to be found by associating the term with externally observable states. Some problems with this were noted before. An argument given by Wittgenstein in favor of behaviorism is shown to draw too strong a conclusion.

While presenting the two previous alternatives, Churchland effectively dismisses them, effectively asserting that only one notion of the source of meaning for mental terms is now considered viable. This is the network theory of meaning: the terms of a theory have meaning as a part of a network of concepts. The terms we use in describing mind are terms in what is literally a theory of mind, which Churchland terms Folk Psychology. The correspondences between the apparent Folk Psychology and other theories are strong. According to the network theory of meaning, the terms for states of mind gain their meaning from a rather substantial set of general statements connecting mental states with each other, external circumstances, and overt behavior. [The influence between this semantical notion and the functional and eliminative materialistic positions on the mind/body problem look strong.]

In light of the network theory of meaning, we see an old theory, Folk Psychology, coming to be confronted with a new theory which will be based on a matured neuroscience. Several of the positions on the mind/body problem can be seen as different expectations on how this confrontation will be resolved. The identity theory proposes that the old theory will be uniformly reduced to neurological terms. The dualism theories maintain that the old theory will not be reduced to the new one on the grounds that behavior has nonmaterial sources. Functionalism also maintains that the old theory will not be reduced to the new one, but because there will be too many possible reductions to single out any particular one as the unique reduction. Eliminative materialism maintains that the old theory will fail to reduce, but on the grounds that the old theory is too confused and inaccurate for a successful reduction.

Epistemological Problems

Epistemological problems are questions of how one knows something. In the case of consciousness, Churchland addresses how one can know something about other minds and how anything can be conscious of itself. To both, he shows that understanding the semantical problem leads to understanding the epistemological problem.

The basic epistemological problem of other minds is how one can justify belief in other minds and their states when connection between the object of belief and behaviors cannot be observed since other minds cannot be observed. A classical answer was by induction - from observation of one's own mind and behavior, one could induce that others had minds. There are three problems with this classical answer. First it is induction from exactly one case, the weakest possible induction. Second, it would deny that a color-blind person could attribute color sensations to others, etc. Third, the "observations" of one's own mind are apparently learned, and to grant their existence is to beg the question of the grounds for belief in any mind.

Behaviorism attempted to argue that the generalisations made about other minds were not generalizations at all, but definitions. The inability to show that finite definitions were possible has led to rejection of this thesis.

When folk psychology is seen as a theory, then it is easy to compare it with other theories which often have unobservable entities. Reinforcing the basis of seeing folk psychology as a theory is that it allows us to make useful predictions of other peoples behavior. The "mind" with which we attribute another person is a theoretical construct, justified not by direct observation but by its predictive and explanatory power.

In discussing self consciousness, Churchland begins with a contemporary view rather than a historical view. Considering the development of ability to discriminate internal states with age, and the likely better ability of novelists or psychologists to discriminate internal states than many other professions, it appears that self-consciousness is both learned and has different degrees. The learned aspect of inner perception accords with observation of the learned aspect of other perception. As with other perception, there needs to be some means of perception, but in this case it can be identified with a general ability to make judgments. Self-consciousness, then, has it roots in signals of inner physical events, or useful introspective discriminations, and in ability to make predictively useful judgments.

The traditional view has been that inner perception is distinct from external perception, in that there is no distinction between appearance and reality. Aside from a long history, it does seem that there is merit to the argument that there cannot be an infinite regress of sensation of sensation of . . . of an internal phenomenon.

Against the traditional view, it is argued that there is no reason to suppose evolution would supply two mechanisms for perception when one would do. The argument that there is no difference between appearance and reality can hold only if internal perception is free of error. But it is not. Additional information can change a judgment. Neurology suggests that there are (error prone) intermediaries between internal phenomena and judgment. And short times degrade all kinds of perceptions. The thesis of lack of errors for internal phenomena just might hold for perception, but there is no reason to expect it to hold for propositional beliefs, since the propositions are certainly subject to errors.

In short, all perception is theory laden. {The filling in of colors outside the small forward area in which they are demonstrably perceived is a classic example.} The brain is a theorizer (learner) from the start.

Methodological Problems

Underlying questions of methodology (what will "scientific method" mean for study of mind), is a long standing discussion of realism and idealism. {The pure realist asserts not only the existence of a reality separate from our minds, but also the ability directly to know that reality.} The pure idealist asserts that there is no reality apart from our constructs. Neither of these extreme forms survives. Kant asserted a position which has had a lasting influence. This position is that material objects are constructs which are real to us, but may lack a transcendental reality. Scientific realism is a modem position more towards realism still, that our theories will make closer and closer approximations to a transcendental reality. {However, it is not possible to rule out convergence to some local maximum in predictive ability rather than convergence to a global maximum, because the transcendental reality is simply not available for our inspection.}

Phenomenology is a word covering many varieties, but all are rooted in Kantian idealism, and all assert that methods for the study of mind must differ radically from methods for the study of external phenomena. Standard scientific method constructs realities, and construction is not applicable to the constructor. The traditional epistemological view is part of this methodological view, and was previously shown to have several weaknesses.

Behaviorist methodology insisted on explaining observed behavior without reference to folk psychology at all. The position was found to be extreme, for instance, in that most theories have terms which do not have operational definitions. The rejection of folk psychology has been a lasting influence even if the position was found to be extreme.

The most popular research methods are currently based in cognitive psychology andartificial intelligence. These disciplines see and address separately a wide variety of mental phenomena. They both seek computational explanations for mental phenomena. The approaches differ in the criteria considered relevant to showing a computational method explains a mental phenomenon. The artificial intelligence approach considers but the single criterion: will the proposed computation do a task that a creature is observed to do? The cognitive psychology approach adds two other criteria. Its second criterion is whether secondary characteristics in the observation of the behavior simulated, such as timing and error patterns, are also reproduced. Its third criterion, less applied because of difficulty in studying neural systems, is that the neural system in question should be able to carry out the suggested computation.

Last Revised 21 January 1999