The Mind-Body Problem
JERRY A. FODOR
Jerry A. Fodor is professor of philosophy at Rutgers University. He is renowned for defending the view that we
think in an inner language, akin to the symbolic codes used in ordinary computers. Fodor's highly influential
works include The Language of Thought, The Modularity of Mind, and Psychosemantics.
Copyright @ 1981 by Scientific American, Inc. Reprinted by permission of the publisher. All rights reserved.
Modem philosophy of science has been devoted largely to the formal and systematic description of the
successful practices of working scientists. The philosopher does not try to dictate how scientific inquiry and
argument ought to be conducted. Instead he tries to enumerate the principles and practices that have
contributed to good science. The philosopher has devoted the most attention to analyzing the methodological
peculiarities of the physical sciences. The analysis has helped to clarify the nature of confirmation, the logical
structure of scientific theories, the formal properties of statements that express laws and the question of
whether theoretical entities actually exist.
It is only rather recently that philosophers have become seriously interested in the methodological tenets of
psychology. Psychological explanations of behavior refer liberally to the mind and to states, operations and
processes of the mind. The philosophical difficulty comes in stating in unambiguous language what such
Traditional philosophies of mind can be divided into two broad categories: dualist theories and materialist
theories. In the dualist approach the mind is a nonphysical substance. In materialist theories the mental is not
distinct from the physical; indeed, all mental states, properties, processes and operations are in principle
identical with physical states, properties, processes and operations. Some materialists, known as behaviorists,
maintain that all talk of mental causes can be eliminated from the language of psychology in favor of talk of
environmental stimuli and behavioral responses. Other materialists, the identity theorists, contend that there
are mental causes and that they are identical with neurophysiological events in the brain.
In the past fifteen years a philosophy of mind called functionalism that is neither dualist nor materialist has
emerged from philosophical reflection on developments in artificial intelligence, computational theory,
linguistics, cybernetics and psychology. All these fields, which are collectively known as the cognitive sciences,
have in common a certain level of abstraction and a concern with systems that process information.
Functionalism, which seeks to provide a philosophical account of this level of abstraction, recognizes the
possibility that systems as diverse as human beings, calculating machines and disembodied spirits could all
have mental states. In the functionalist view the psychology of a system depends not on the stuff it is made of
(living cells, metal or spiritual energy) but on how the stuff is put together. Functionalism is a difficult concept,
and one way of coming to grips with it is to review the deficiencies of the dualist and materialist philosophies of
mind it aims to displace.
The chief drawback of dualism is its failure to account adequately for mental causation. If the mind is
nonphysical, it has no position in physical space. How, then, can a mental cause give rise to a behavioral
effect that has a position in space? To put it another way, how can the nonphysical give rise to the physical
without violating the laws of the conservation of mass, of energy and of momentum?
The dualist might respond that the problem of how an immaterial substance can cause physical events is not
much obscurer than the problem of how one physical event can cause another. Yet there is an important
difference: there are many clear cases of physical causation but not one clear case of nonphysical causation.
Physical interaction is something philosophers, like all other people, have to live with. Nonphysical interaction,
however, may be no more than an artifact of the immaterialist construal of the mental. Most philosophers now
agree that no argument has successfully demonstrated why mind-body causation should not be regarded as a
species of physical causation.
Dualism is also incompatible with the practices of working psychologists. The psychologist frequently applies
the experimental methods of the physical sciences to the study of the mind. If mental processes were different
in kind from physical processes, there would be no reason to expect these methods to work in the realm of the
mental. In order to justify their experimental methods many psychologists urgently sought an alternative to
In the 1920s John B. Watson of Johns Hopkins University made the radical suggestion that behavior does not
have mental causes. He regarded the behavior of an organism as its observable responses to stimuli, which
he took to be the causes of its behavior. Over the next thirty years psychologists such as B. F. Skinner of
Harvard University developed Watson's ideas into an elaborate world view in which the role of psychology was
to catalogue the laws that determine causal relations between stimuli and responses. In this "radical
behaviorist" view the problem of explaining the nature of the mind-body interaction vanishes; there is no such
Radical behaviorism has always worn an air of paradox. For better or worse, the idea of mental causation is
deeply ingrained in our everyday language and in our ways of understanding our fellowmen and ourselves. For
example, people commonly attribute behavior to beliefs, to knowledge and to expectations. Brown puts gas in
his tank because he believes the car will not run without it. Jones
The radical behaviorist is unmoved by appeals to such cases. He is prepared to dismiss references to mental
causes, however plausible they may seem, as the residue of outworn creeds. The radical behaviorist predicts
that as psychologists come to understand more about the relations between stimuli and responses they will
find it increasingly possible to explain behavior without postulating mental causes.
The strongest argument against behaviorism is that psychology has not turned out this way; the opposite has
happened. As psychology has matured, the framework of mental states and processes that is apparently
needed to account for experimental observations has grown all the more elaborate. Particularly in the case of
human behavior psychological theories satisfying the methodological tenets of radical behaviorism have
proved largely sterile, as would be expected if the postulated mental processes are real and causally effective.
Nevertheless, many philosophers were initially drawn to radical behaviorism because, paradoxes and all, it
seemed better than dualism. Since a psychology committed to immaterial substances was unacceptable,
philosophers turned to radical behaviorism because it seemed to be the only alternative materialist philosophy
of mind. The choice, as they saw it, was between radical behaviorism and ghosts.
By the early 1960s philosophers began to have doubts that dualism and radical behaviorism exhausted the
possible approaches to the philosophy of mind. Since the two theories seemed unattractive, the right strategy
might be to develop a materialist philosophy of mind that nonetheless allowed for mental causes. Two such
philosophies emerged, one called logical behaviorism and the other called the central-state identity theory.
Logical behaviorism is a semantic theory about what mental terms mean. The basic idea is that attributing a
mental state (say thirst) to an organism is the same as saying that the organism is disposed to behave in a
particular way (for example to drink if there is water available). On this view every mental ascription is
equivalent in meaning to an if-then statement (cal1ed a behavioral hypothetical) that expresses a behavioral
disposition. For example, "Smith is thirsty" might be taken to be equivalent to the dispositional statement "If
there were water available, then Smith would drink some." By definition a behavioral hypothetical includes no
mental terms. The if-clause of the hypothetical speaks only of stimuli and the then-clause speaks only of
behavioral responses. Since stimuli and responses are physical events, logical behaviorism is a species of
The strength of logical behaviorism is that by translating mental language into the language of stimuli and
responses it provides an interpretation of psychological explanations in which behavioral effects are attributed
to mental causes. Mental causation is simply the manifestation of a behavioral disposition. More precisely,
mental causation is what happens when an organism has a behavioral disposition and the if-clause of the
behavioral hypothetical expressing the disposition happens to be true. For example, the causal statement
"Smith drank some water because he was thirsty" might be taken to mean "If there were water available, then
Smith would drink some, and there was water available."
I have somewhat oversimplified logical behaviorism by assuming that each mental ascription can be
translated by a unique behavioral hypothetical. Actually the logical behaviorist often maintains that it takes an
open-ended set (perhaps an infinite set) of behavioral hypotheticals to spell out the behavioral disposition
expressed by a mental term. The mental ascription "Smith is thirsty" might also be satisfied by the hypothetical
"If there were orange juice available, then Smith would drink some" and by a host of other hypotheticals. In any
event the logical behaviorist does not usually maintain he can actually enumerate all the hypotheticals that
correspond to a behavioral disposition expressing a given mental term. He only insists that in principle the
meaning of any mental term can be conveyed by behavioral hypotheticals.
The way the logical behaviorist has interpreted a mental term such as thirsty is modeled after the way many
philosophers have interpreted a physical disposition such as fragility. The physical disposition "The glass is
fragile" is often taken to mean something like "If the glass were struck, then it would break." By the same token
the logical behaviorist's analysis of mental causation is similar to the received analysis of one kind of physical
causation. The causal statement "The glass broke because it was fragile" is taken to mean something like "If
the glass were struck, then it would break, and the glass was struck."
By equating mental terms with behavioral dispositions the logical behaviorist has put mental terms on a par
with the nonbehavioral dispositions of the physical sciences. Thatis a promising move, because the analysis of
nonbehavioral dispositions is on relatively solid philosophical ground. An explanation attributing the breaking of
a glass to its fragility is surely something even the staunchest materialist can accept. By arguing that mental
terms are synonymous with dispositional terms, the logical behaviorist has provided something the radical
behaviorist could not: a materialist account of mental causation.
Nevertheless, the analogy between mental causation as construed by the logical behaviorist and physical
causation goes only so far. The logical behaviorist treats the manifestation of a disposition as the sole form of
mental causation whereas the physical sciences recognize additional kinds of causation. There is the kind of
causation where one physical event causes another, as when the breaking of a glass is attributed to its having
been struck. In fact, explanations that involve event-event causation are presumably more basic than
dispositional explanations, because the manifestation of a disposition (the breaking of a fragile glass) always
involves event-event causation and not vice versa. In the realm of the mental many examples of event-event
causation involve one mental state's causing another, and for this kind of causation logical behaviorism
provides no analysis. As a result the logical behaviorist is committed to the tacit and implausible assumption
that psychology requires a less robust notion of causation than the physical sciences require.
Event-event causation actually seems to be quite common in the realm of the mental. Mental causes typically
give rise to behavioral effects by virtue of their interaction with other mental causes. For example, having a
headache causes a disposition to take aspirin only if one also has the desire to get rid of the headache, the
belief that aspirin exists, the belief that taking aspirin reduces headaches and so on. Since mental states
interact in generating behavior, it will be necessary to find a construal of psychological explanations that posits
mental processes: causal sequences of mental events. It is this construal that logical behaviorism fails to
Such considerations bring out a fundamental way in which logical behaviorism is quite similar to radical
behaviorism. It is true that the logical behaviorist, unlike the radical behaviorist, acknowledges the existence of
mental states. Yet since the underlying tenet of logical behaviorism is that references to mental states can be
translated out of psychological explanations by employing behavioral hypotheticals, all talk of mental states
and processes is in a sense heuristic. The only facts to which the behaviorist is actually committed are facts
about relations between stimuli and responses. In this respect logical behaviorism is just radical behaviorism in
a semantic form. Although the former theory offers a construal of mental causation, the construal is
Pickwickian. What does not really exist cannot cause anything, and the logical behaviorist, like the radical
behaviorist, believes deep down that mental causes do not exist.
An alternative materialist theory of the mind to logical behaviorism is the central-state identity theory.
According to this theory, mental events, states and processes are identical with neurophysiological events in
the brain, and the property of being in a certain mental state (such as having a headache or believing it will
rain) is identical with the property of being in a certain neurophysiological state. On this basis it is easy to make
sense of the idea that a behavioral effect might sometimes have a chain of mental causes; that will be the
case whenever a behavioral effect is contingent on the appropriate sequence of neurophysiological events.
The central-state identity theory acknowledges that it is possible for mental causes to interact causally without
ever giving rise to any behavioral effect, as when a person thinks for a while about what he ought to do and
then decides to do nothing. If mental processes are neurophysiological, they must have the causal properties
of neurophysiological processes. Since neurophysiological processes are presumably physical processes, the
central-state identity theory ensures that the concept of mental causation is as rich as the concept of physical
The central-state identity theory provides a satisfactory account of what the mental terms in psychological
explanations refer to, and so it is favored by psychologists who are dissatisfied with behaviorism. The
behaviorist maintains that mental terms refer to nothing or that they refer to the parameters of stimulus-
response relations. Either way the existence of mental entities is only illusory. The identity theorist, on the
other hand, argues that mental terms refer to neurophysiological states. Thus he can take seriously the project
of explaining behavior by appealing to its mental causes.
The chief advantage of the identity theory is that it takes the explanatory constructs of psychology at face
value, which is surely something a philosophy of mind ought to do if it can. The identity theory shows how the
mentalistic explanations of psychology could be not mere heuristics but literal accounts of the causal history of
behavior. Moreover, since the identity theory is not a semantic thesis, it is immune to many arguments that
cast in doubt logical behaviorism. A drawback of logical behaviorism is that the observation "John has a
headache" does not seem to mean the same thing as a statement of the form "John is disposed to behave in
such and such a way." The identity theorist, however, can live with the fact that "John has a headache" and
"John is in such and such a brain state" are not synonymous. The assertion of the identity theorist is not that
these sentences mean the same thing but only that they are rendered true (or false) by the same
The identity theory can be held either as a doctrine about mental particulars (John's current pain or Bill's fear
of animals) or as a doctrine about mental universals, or properties (having a pain or being afraid of animals).
The two doctrines, called respectively token physicalism and type physicalism, differ in strength and
plausibility. Token physicalism maintains only that all the mental particulars that happen to exist are
neurophysiological, whereas type physicalism makes the more sweeping assertion that all the mental
particulars there could possibly be are neurophysiological. Token physicalism does not rule out the logical
possibility of machines and disembodied spirits having mental properties. Type physicalism dismisses this
possibility because neither machines nor disembodied spirits have neurons.
Type physicalism is not a plausible doctrine about mental properties even if token physicalism is right about
mental particulars. The problem with type physicalism is that the psychological constitution of a system seems
to depend not on its hardware, or physical composition, but on its software, or program. Why should the
philosopher dismiss the possibility that silicon-based Martians have pains, assuming that the silicon is properly
organized? And why should the philosopher rule out the possibility of machines having beliefs, assuming that
the machines are correctly programmed? If it is logically possible that Martians and machines could have
mental properties, then mental properties and neurophysiological processes cannot be identical, however
much they may prove to be coextensive.
What it all comes down to is that there seems to be a level of abstraction at which the generalizations of
psychology are most naturally pitched. This level of abstraction cuts across differences in the physical
composition of the systems to which psychological generalizations apply. In the cognitive sciences, at least,
the natural domain for psychological theorizing seems to be all systems that process information. The problem
with type physicalism is that there are possible information-processing systems with the same psychological
constitution as human beings but not the same physical organization. In principle all kinds of physically
different things could have human software.
This situation calls for a relational account of mental properties that abstracts them from the physical structure
of their bearers. In spite of the objections to logical behaviorism that I presented above, logical behaviorism
was at least on the right track in offering a relational interpretation of mental properties: to have a headache is
to be disposed to exhibit a certain pattern of relations between the stimuli one encounters and the responses
one exhibits. If that is what having a headache is, however, there is no reason in principle why only heads that
are physically similar to ours can ache. Indeed, according to logical behaviorism, it is a necessary truth that
any system that has our stimulus-response contingencies also has our headaches.
All of this emerged ten or fifteen years ago as a nasty dilemma for the materialist program in the philosophy of
mind. On the one hand the identity theorist (and not the logical behaviorist) had got right the causal character
of the interactions of mind and body. On the other the logical behaviorist (and not the identity theorist) had got
right the relational character of mental properties. Functionalism has apparently been able to resolve the
dilemma. By stressing the distinction computer science draws between hardware and software the functionalist
can make sense of both the causal and the relational character of the mental.
The intuition underlying functionalism is that what determines the psychological type to which a mental
particular belongs is the causal role of the particular in the mental life of the organism. Functional individuation
is differentiation with respect to causal role. A headache, for example, is identified with the type of mental state
that among other things causes a disposition for taking aspirin in people who believe aspirin relieves a
headache, causes a desire to rid oneself of the pain one is feeling, often causes someone who speaks English
to say such things as "I have a headache" and is brought on by overwork, eyestrain and tension. This list is
presumably not complete. More will be known about the nature of a headache as psychological and
physiological research discovers more about its causal role.
Functionalism construes the concept of causal role in such a way that a mental state can be defined by its
causal relations to other mental states. In this respect functionalism is completely different from logical
behaviorism. Another major difference is that functionalism is not a reductionist thesis. It does not foresee,
even in principle, the elimination of mentalistic concepts from the explanatory apparatus of psychological
The difference between functionalism and logical behaviorism is brought out by the fact that functionalism is
fully compatible with token physicalism. The functionalist would not be disturbed if brain events turn out to be
the only things with the functional properties that define mental states. Indeed, most functionalists fully expect
it will turn out that way.
Since functionalism recognizes that mental particulars may be physical, it is compatible with the idea that
mental causation is a species of physical causation. In other words, functionalism tolerates the materialist
solution to the mind-body problem provided by the central-state identity theory. It is possible for the
functionalist to assert both that mental properties are typically defined in terms of their relations and that
interactions of mind and body are typically causal in however robust a notion of causality is required by
psychological explanations. The logical behaviorist can endorse only the first assertion and the type physicalist
only the second. As a result functionalism seems to capture the best features of the materialist alternatives to
dualism. It is no wonder that functionalism has become increasingly popular.
Machines provide good examples of two concepts that are central to functionalism: the concept that mental
states are interdefined and the concept that they can be realized by many systems. The illustration . . .
contrasts a behavioristic Coke machine with a mentalistic one. Both machines dispense a Coke for 10 cents.
(The price has not been affected by inflation.) The states of the machines are defined by reference to their
causal roles, but only one machine would satisfy the behaviorist. Its single state (SO) is completely specified in
terms of stimuli and responses. SO is the state a machine is in if, and only if, given a dime as the input, it
dispenses a Coke as the output.
The machine in the illustration has interdefined states (S1 and S2), which are characteristic of functionalism.
S1 is the state a machine is in if, and only if, (1) given a nickel, it dispenses nothing and proceeds to S2, and
(2) given a dime, it dispenses a Coke and stays in S1. S2 is the state a machine is in if, and only if, (I) given a
nickel, it dispenses a Coke and proceeds to S1, and (2) given a dime, it dispenses a Coke and a nickel and
proceeds to S1. What S1 and S2 jointly amount to is the machine's dispensing a Coke if it is given a dime,
dispensing a Coke and a nickel if it is given a dime and a nickel and waiting to be given a second nickel if it
has been given a first one.
Since S1 and S2 are each defined by hypothetical statements, they can be viewed as dispositions.
Nevertheless, they are not behavioral dispositions because the consequences an input has for a machine in
S1 or S2 are not specified solely in terms of the output of the machine. Rather, the consequences also involve
the machine's internal states.
Nothing about the way I have described the behavioristic and mentalistic Coke machines puts constraints on
what they could be made of. Any system whose states bore the proper relations to inputs, outputs and other
states could be one of these machines. No doubt it is reasonable to expect such a system to be constructed
out of such things as wheels, levers and diodes (token physicalism for Coke machines). Similarly, it is
reasonable to expect that our minds may prove to be neurophysiological (token physicalism for human
Nevertheless, the software description of a Coke machine does not logically require wheels, levers and diodes
for its concrete realization. By the same token, the software description of the mind does not logically require
neurons. As far as functionalism is concerned a Coke machine with states S1 and S2 could be made of
ectoplasm, if there is such stuff and if its states have the right causal properties. Functionalism allows for the
possibility of disembodied Coke machines in exactly the same way and to the same extent that it allows for the
possibility of disembodied minds.
To say that S1 and S2 are interdefined and realizable by different kinds of hardware is not, of course, to say
that a Coke machine has a mind. Although interdefinition and functional specification are typical features of
mental states, they are clearly not sufficient for mentality. . . .
An obvious objection to functionalism as a theory of the mind is that the functionalist definition is not limited to
mental states and processes. Catalysts, Coke machines, valve openers, pencil sharpeners, mousetraps and
ministers of finance are all in one way or another concepts that are functionally defined, but none is a mental
concept such as pain, belief and desire. What, then, characterizes the mental? And can it be captured in a
The traditional view in the philosophy of mind has it that mental states are distinguished by their having what
are called either qualitative content or intentional content. I shall discuss qualitative content first.
It is not easy to say what qualitative content is; indeed, according to some theories, it is not even possible to
say what it is because it can be known not by description but only by direct experience. I shall nonetheless
attempt to describe it. Try to imagine looking at a blank wall through a red filter. Now change the filter to a
green one and leave everything else exactly the way it was. Something about the character of your experience
changes when the filter does, and it is this kind of thing that philosophers call qualitative content. I am not
entirely comfortable about introducing qualitative content in this way, but it is a subject with which many
philosophers are not comfortable.
The reason qualitative content is a problem for functionalism is straightforward. Functionalism is committed to
defining mental states in terms of their causes and effects. It seems, however, as if two mental states could
have all the same causal relations and yet could differ in their qualitative content. Let me illustrate this with the
classic puzzle of the inverted spectrum.
It seems possible to imagine two observers who are alike in all relevant psychological respects except that
experiences having the qualitative content of red for one observer would have the qualitative content of green
for the other. Nothing about their behavior need reveal the difference because both of them see ripe tomatoes
and flaming sunsets as being similar in color and both of them call that color "red." Moreover, the causal
connection between their (qualitatively distinct) experiences and their other mental states could also be
identical. Perhaps they both think of Little Red Riding Hood when they see ripe tomatoes, feel depressed
when they see the color green and so on. It seems as if anything that could be packed into the notion of the
causal role of their experiences could be shared by them, and yet the qualitative content of the experiences
could be as different as you like. If this is possible, then the functionalist account does not work for mental
states that have qualitative content. If one person is having a green experience while another person is having
a red one, then surely they must be in different mental states.
The example of the inverted spectrum is more than a verbal puzzle. Having qualitative content is supposed to
be a chief factor in what makes a mental state conscious. Many psychologists who are inclined to accept the
functionalist framework are nonetheless worried about the failure of functionalism to reveal much about the
nature of consciousness. Functionalists have made a few ingenious attempts to talk themselves and their
colleagues out of this worry, but they have not, in my view, done so with much success. (For example, perhaps
one is wrong in thinking one can imagine what an inverted spectrum would be like.) As matters stand, the
problem of qualitative content poses a serious threat to the assertion that functionalism can provide a general
theory of the mental.
Functionalism has fared much better with the intentional content of mental states. Indeed, it is here that the
major achievements of recent cognitive science are found. To say that a mental state has intentional content is
to say that it has certain semantic properties. For example, for Enrico to believe Galileo was Italian apparently
involves a three-way relation between Enrico, a belief and a proposition that is the content of the belief
(namely the proposition that GaIileo was Italian). In particular it is an essential property of Enrico's belief that it
is about Galileo (and not about, say, Newton) and that it is true if, and only if, Galileo was indeed Italian.
Philosophers are divided on how these considerations fit together, but it is widely agreed that beliefs involve
semantic properties such as expressing a proposition, being true or false and being about one thing rather
It is important to understand the semantic properties of beliefs because theories in the cognitive sciences are
largely about the beliefs organisms have. Theories of learning and perception, for example, are chiefly
accounts of how the host of beliefs an organism has are determined by the character of its experiences and its
genetic endowment. The functionalist account of mental states does not by itself provide the required insights.
Mousetraps are functionally defined, yet mousetraps do not express propositions and they are not true or
There is at least one kind of thing other than a mental state that has intentional content: a symbol. Like
thoughts, symbols seem to be about things. If someone says "Galileo was Italian," his utterance, like Enrico's
belief, expresses a proposition about Galileo that is true or false depending on Galileo's homeland. This
parallel between the symbolic and the mental underlies the traditional quest for a unified treatment of language
and mind. Cognitive science is now trying to provide such a treatment.
The basic concept is simple but striking. Assume that there are such things as mental symbols (mental
representations) and that mental symbols have semantic properties. On this view having a belief involves
being related to a mental symbol, and the belief inherits its semantic properties from the mental symbol that
figures in the relation. Mental processes (thinking, perceiving, learning and so on) involve causal interactions
among relational states such as having a belief. The semantic properties of the words and sentences we utter
are in turn inherited from the semantic properties of the mental states that language expresses.
Associating the semantic properties of mental states with those of mental symbols is fully compatible with the
computer metaphor, because it is natural to think of the computer as a mechanism that manipulates symbols.
A computation is a causal chain of computer states and the links in the chain are operations on semantically
interpreted formulas in a machine code. To think of a system (such as the nervous system) as a computer is to
raise questions about the nature of the code in which it computes and the semantic properties of the symbols
in the code. In fact, the analogy between minds and computers actually implies the postulation of mental
symbols. There is no computation without representation.
The representational account of the mind, however, predates considerably the invention of the computing
machine. It is a throwback to classical epistemology, which is a tradition that includes philosophers as diverse
as John Locke, David Hume, George Berkeley, Rene Descartes, Immanuel Kant, John Stuart Mill and William
Hume, for one, developed a representational theory of the mind that included five points. First, there exist
"Ideas," which are a species of mental symbol. Second, having a belief involves entertaining an Idea. Third,
mental processes are causal associations of Ideas. Fourth, Ideas are like pictures. And fifth, Ideas have their
semantic properties by virtue of what they resemble: the Idea of John is about John because it looks like him.
Contemporary cognitive psychologists do not accept the details of Hume's theory, although they endorse much
of its spirit. Theories of computation provide a far richer account of mental processes than the mere
association of Ideas. And only a few psychologists still think that imagery is the chief vehicle of mental
representation. Nevertheless, the most significant break with Hume's theory lies in the abandoning of
resemblance as an explanation of the semantic properties of mental representations.
Many philosophers, starting with Berkeley, have argued that there is something seriously wrong with the
suggestion that the semantic relation between a thought and what the thought is about could be one of
resemblance. Consider the thought that John is tall. Clearly the thought is true only of the state of affairs
consisting of John's being tall. A theory of the semantic properties of a thought should therefore explain how
this particular thought is related to this particular state of affairs. According to the resemblance theory,
entertaining the thought involves having a mental image that shows John to be tall. To put it another way, the
relation between the thought that John is tall and his being tall is like the relation between a tall man and his
The difficulty with the resemblance theory is that any portrait showing John to be tall must also show him to be
many other things: clothed or naked, lying, standing or sitting, having a head or not having one, and so on. A
portrait of a tall man who is sitting down resembles a man's being seated as much as it resembles a man's
being tall. On the resemblance theory it is not clear what distinguishes thoughts about John's height from
thoughts about his posture.
The resemblance theory turns out to encounter paradoxes at every turn. The possibility of construing beliefs
as involving relations to semantically interpreted mental representations clearly depends on having an
acceptable account of where the semantic properties of the mental representations come from. If resemblance
will not provide this account, what will?
The current idea is that the semantic properties of a mental representation are determined by aspects of its
functional role. In other words, a sufficient condition for having semantic properties can be specified in causal
terms. This is the connection between functionalism and the representational theory of the mind. Modem
cognitive psychology rests largely on the hope that these two doctrines can be made to support each other.
No philosopher is now prepared to say exactly how the functional role of a mental representation determines
its semantic properties. Nevertheless, the functionalist recognizes three types of causal relation among
psychological states involving mental representations, and they might serve to fix the semantic properties of
mental representations. The three types are causal relations among mental states and stimuli, mental states
and responses and some mental states and other ones.
Consider the belief that John is tall. Presumably the following facts, which correspond respectively to the three
types of causal relation, are relevant to determining the semantic properties of the mental representation
involved in the belief. First, the belief is a normal effect of certain stimulations, such as seeing John in
circumstances that reveal his height. Second, the belief is the normal cause of certain behavioral effects, such
as uttering "John is tall." Third, the belief is a normal cause of certain other beliefs and a normal effect of
certain other beliefs. For example, anyone who believes John is tall is very likely also to believe someone is
tall. Having the first belief is normally causally sufficient for having the second belief. And anyone who believes
everyone in the room is tall and also believes John is in the room will very likely believe John is tall. The third
belief is a normal effect of the first two. In short, the functionalist maintains that the proposition expressed by a
given mental representation depends on the causal properties of the mental states in which that mental
The concept that the semantic properties of mental representations are determined by aspects of their
functional role is at the center of current work in the cognitive sciences. Nevertheless, the concept may not be
true. Many philosophers who are unsympathetic to the cognitive turn in modem psychology doubt its truth, and
many psychologists would probably reject it in the bald and unelaborated way that I have sketched it. Yet even
in its skeletal form, there is this much to be said in its favor: It legitimizes the notion of mental representation,
which has become increasingly important to theorizing in every branch of the cognitive sciences. Recent
advances in formulating and testing hypotheses about the character of mental representations in fields ranging
from phonetics to computer vision suggest that the concept of mental representation is fundamental to
empirical theories of the mind.
The behaviorist has rejected the appeal to mental representation because it runs counter to his view of the
explanatory mechanisms that can figure in psychological theories. Nevertheless, the science of mental
representation is now flourishing. The history of science reveals that when a successful theory comes into
conflict with a methodological scruple, it is generally the scruple that gives way. Accordingly the functionalist
has relaxed the behaviorist constraints on psychological explanations. There is probably no better way to
decide what is methodologically permissible in science than by investigating what successful science requires.