The term ‘functionalism’ means different things in many different

disciplines from architectural theory to zoology. In contemporary

philosophy of mind, however, it is uniformly understood to stand for

the view that mental states should be explained in terms of causal

roles. So, to take a simple example, a functionalist in the philosophy

of mind would argue that pains are states which are normally

caused by bodily damage, and tend in turn to cause avoidance


Functionalism is often introduced by an analogy between mental

states and mechanical devices. Consider the notion of a

carburettor, say. For something to be a carburettor it need not have

any particular physical make-up. Carburettors can come in many

different materials and shapes. What makes it a carburettor is

simply that it plays the right causal role, namely that it mixes air

with petrol in response to movements of the accelerator and choke.

Similarly, argue functionalists, with the mind. The possession of

mental states does not depend on the physical make-up of the brain;

it depends only on its displaying the right causal structure. Since

organisms with very different sorts of biological make-up, like

octopuses and humans, can have states with the causal role of pain,

say, it follows from functionalism that octopuses and humans can

both be in pain.

There exists a number of different subspecies of functionalism. One

important division depends on how the relevant causal roles are

determined. ‘Common-sense’ functionalists take them to be fixed by

common-sense psychology; ‘scientific’ functionalists take them to be

fixed by the discoveries of scientific psychology. So, for example,

common-sense functionalists will hold that emotions play the causal

role that common-sense psychology ascribes to emotions, while

scientific psychologists will argue that scientific psychology

identifies this causal role.

Functionalism, of whatever subspecies, is open to a number of

well-known criticisms. One central objection is that it cannot

accommodate the conscious, qualitative aspect of mental life. Could

not a machine share the causal structure of someone who was in

pain, and thereby satisfy the functionalist qualification for pain,

and yet have no conscious feelings?

It might seem that functionalists can respond to this difficulty by

being more stringent about the requirements involved in the causal

role of a given human sensation. But there is a danger that

functionalism will then lose much of its appeal. The original

attraction of functionalism was that its ‘liberal’ specification of

causal roles allowed that humans could share mental states with

non-humans. This feature is likely to be lost if we switch to more

chauvinist’ specifications designed to explain why non-humans do

not share our conscious life.

Another objection to functionalism is that it cannot account for

mental representation. Functionalism focuses on the way mental

states enter into causal structure. But it is doubtful that mental

representation can be explained in purely causal terms.

Some philosophers argue that the issue of mental representation

can be dealt with by adding some teleology to functionalism, that is

by considering the biological purposes for which mental states have

been designed, as well as their actual structure of causes and

effects. However, once we do appeal to teleology in this way, it is

not clear that we still need a functionalist account of

representational states, for we can now simply identify such states

in terms of their biological purposes, rather than their causal roles.

1 Origins of functionalism

Originally functionalism was a response to philosophical behaviourism.

The behaviourists rejected the traditional Cartesian picture of the mind as

an essentially private realm accessible only to the conscious subject

(Descartes 1641). Instead they argued that mental states are dispositions

to behaviour, and so are publicly accessible. To desire an ice cream, said

the behaviourists, is to be disposed to eat one when the opportunity

presents itself (see Behaviourism, analytic; Private states and language).

The functionalists argued that behaviourism fails to distinguish

sufficiently between mental cause and behavioural effect. They argued

that there is no simple pairing of mental states with pieces of behaviour,

since which behaviour issues from any given mental state will depend on

the agent’s other mental states. My desire for an ice cream will make

me walk to the fridge if I believe it contains an ice cream, but it will

make me walk down the street to the shop if I believe I can buy an ice

cream there.

Because of this, functionalists argued that mental states are inner causes

distinct from their behavioural effects. In saying this, however, they did

not want to return to the Cartesian conception of mental events as

essentially private states. In their view, mental states are part of the

public world of causes and effects studied by science. They are

inner’ only in the sense that they are unobservable causes of overt

behaviour, in the same way as atomic structures are unobservable

causes of chemical reactions.

Because they take mental states to be unobservable, functionalists think

that we only have an indirect grasp of their nature, as playing a certain

causal, or ‘functional’, role in a cognitive system. One important

consequence of this is that different physical states might play the

relevant role in different beings, or even in the same being at different

times. (Indeed functionalism as such leaves it open that this role could be

played by some special non-physical state. In this sense functionalism is

compatible with dualism (see Dualism). However, most functionalists

also hold, for independent reasons, that mental roles are in fact filled by

physical states. I shall simplify the following discussion by adopting this


Some early functionalists, most notably Hilary Putnam (see Block 1980),

appealed to an analogy with computers to add precision to the idea of a

causal role. Putnam pointed out that any programmed computer can be

abstractly characterized as a Turing machine, independently of its

hardware’ or physical make-up; he then argued that any two systems

will share mental states as long as they have the same Turing machine

description (see Putnam, H. §6; Turing machines). Today, however, it is

more common to elaborate functionalism in terms of the analogy with

scientific unobservables.

Unobservable entities of any kind pose a prima facie problem which has

been much discussed in the philosophy of science. How should we

understand terms like ‘ionized’, ‘radioactive’, ‘diatomic’, and so on,

given that we have no direct access to their referents? Frank Ramsey

argued that our grasp of such terms derives from our theories about the

relevant unobservables. If we have a theory , about

how various unobservable properties, , relate to each other

and to observables O, then we can read claims involving these properties

as claims about whichever properties happen to play the relevant

theoretical roles. More precisely, a claim attributing property , to

individual , say, can be read as the claim

. (In words: ‘There exist

properties , which are related to each other and to

observables as T says, and has the ith one.’) Note that this way of

understanding claims about the Ps does not credit us with any prior

understanding of these terms, but only with an understanding of

existential quantification and of the observable terms O (see Ramsey,

F.P. §5).

Functionalists take an analogous attitude to terms for mental states, like

belief’, ‘desire’, ‘jealousy’, ‘pain’, and so on. Suppose that our

psychological theory contains such assumptions as that: anybody faced

with an ice cream will believe there is an ice cream in front of them;

anybody who is hungry and hot will desire an ice cream; anybody who

desires an ice cream and believes one is in front of them will reach out

for it; and so on. Then we can understand ascriptions of desires and

beliefs to particular people as claims that there are states which behave

as this theory claims, and that the people in question have them.

2 Common-sense and scientific functionalism

There are various brands of functionalism. One division depends on

whether the theory T used to introduce mental terminology derives from

common-sense psychology or from scientific research. This issue also

determines whether functionalism can claim to give an account of the

meaning of everyday mental terms like ‘desire’ and ‘pain’.

If the theory T is derived from common-sense psychology, then it is open

to functionalists to argue that the resulting Ramsey-style account of

mental discourse explicates what everyday people mean by terms like

desire’ and ‘pain’. That is, they can argue that these terms, in everyday

discourse, simply signify ‘the causal roles specified by T’.

On the other hand, if the theory T used to introduce mental terminology

is not part of everyday thinking, but rather some new scientific theory,

then it cannot plausibly be argued that everyday thinkers derive their

grasp of terms like ‘desire’ and ‘pain’ from this theory. So functionalists

who want to replace common-sense psychology by some new scientific

theory need to view the Ramsey-style account as fixing the meaning of

new technical terms, which signify the causal roles specified by the new

scientific theory. It is then open to them to argue that the states picked

out by their technical terms are in fact the same states as referred to by

the everyday terms ‘desire’ and ‘pain’. But, from the point of view of

this kind of scientific functionalism, this will be a synthetic matter, to be

supported by empirical evidence, and not a matter of definition.

The former kind of functionalism, which I am calling common-sense

functionalism, is sometimes also called ‘analytic’ functionalism, in

recognition of the fact that it makes it a matter of definition that

everyday mental terms stand for certain theoretical roles. It is perhaps

worth pointing out, however, that the common-sense theory T involved in

these definitions will not itself be analytic. What is analytic is this claim:

iif there are states which play the roles specified by T, then these states

are desires, pains, and so on. But the further claim, that there are in fact

such states linking sensory inputs with behavioural outputs, is obviously


One issue facing both common-sense functionalists and scientific

functionalists is how much of the relevant theory to count as contributing

to the meanings of mental terminology. It is obviously unsatisfactory to

include all the many assumptions of everyday or scientific psychology in

our Ramsey-style definitions of mental terms. For it would then follow

that, if any single one of these assumptions turns out to be false, then all

these terms will fail to refer to anything (since there will not then be any

sstates which play the precise causal roles specified by the original

incorrect theory).

The obvious remedy is to argue that only some core set of assumptions

from the relevant theory plays a part in fixing the meanings of its terms.

This would in effect make it an analytic requirement that desires, or

pains, or whatever other mental states are at issue, satisfy the relevant

core assumptions, while leaving it synthetic that they satisfy any further

assumptions. However, there are general doubts about any sharp

analytic-synthetic distinction of this kind. What principled reason could

there be for picking out certain theoretical assumptions as constitutive of

the meaning of ‘desire’, say, while excluding others? Perhaps the best

solution for the functionalist would be to postulate some more-or-less

vague distinction between the central and non-central assumptions of the

relevant theory, and then take it to be a matter of meaning that the

various mental terms will satisfy most of these assumptions. This will

admittedly make it a vague matter whether the relevant mental terms

apply to states which do not satisfy all the central assumptions; but this

vagueness is arguably a feature of all theoretical terminology in science,

and not necessarily vicious.

3 Roles and realizers

A further division among functionalists depends on whether they identify

mental states with ‘roles’ or ‘realizers’. Should we equate pain, say, with

the property of having-some-state-which-plays-the-requisite-causal-role;

or should we equate it with whichever physical state in fact realizes that

role? Suppose, to adopt the usual philosophical oversimplification, that the

pain role is realized in human beings by C-fibres firing. The question is

then whether our term ‘pain’ picks out the ‘realizer’ property C-fibres

firing, or whether it picks out the ‘role’ property

having-some-physical-state-which-plays-the-pain-role. Note that the

role answer implies that humans and octopuses, say, share the same

mental state when they are in pain, for it is true that they are both in a

physical-state-which-plays-the-pain-role, even if octopus pains are

realized by a quite different physical state from C-fibres firing. The

realizer answer, by contrast, implies that in this case human pains are

different states from octopus pains, since the physical states which

realize these roles are different.

The issue here is whether mental properties are first-order properties or

second-order properties. By way of comparison, consider the property of

being coloured. This is a second-order property, since you have it if you

have some first-order property (red, blue,…), which in turn has the

property of being a colour. Role functionalists maintain that mental states

are second-order properties in this sense, in that you have them if you

have some first-order physical property, which in turn has the property of

playing a given causal role.

The obvious argument in favour of the role answer is precisely that it

does allow humans and other beings to share pains. It seems odd to hold

that the state of pain can only be present in those beings that possess

C-fibres. Indeed one of the original attractions of functionalism was that,

by detaching mentality from physical make-up, it seemed to allow beings

of quite different physiologies to share the same mental state.

Those functionalist philosophers who favour the realizer view will allow

that there is a role property common to humans and other beings with

different physiologies, and that we understand the concept ‘pain’ by

associating it with this role property. But they nevertheless argue that

this term still refers to different realizer states in application to the two

species. By analogy, consider the word ‘eye’. All eyes have a common

role feature (namely, that they are sense organs which respond to visible

light), in virtue of which they are all eyes, despite their different physical

realization in different species. Nevertheless, when we use the term

eye’ in application to some individual organism, we seem clearly to use

it to refer to a physical part of that individual, not to the individual’s

instantiation of some abstract property.

In the end the debate between role and realizer versions of functionalism

hinges on whether we take physical instantiation or causal role to be the

essential feature of states like pain. Does difference in physical

instantiation, or only difference in role, imply that we have a different

state? This may seem an overly nice metaphysical issue. But it matters

to the question of whether functionalism supports the identity theory of

mind. The identity theory argues that mental states are identical with

physical states (see Mind, identity theory of). The realizer version of

functionalism agrees with this identity theory, since it identifies human

pains, say, with the physical state of C-fibres firing. But role

functionalism denies the identity theory, since it equates human and other

pains with the second-order role states.

Of course, this does not mean that role functionalism is not

physicalist’ in some broader sense. For the role properties it identifies

with mental properties are still realized by physical properties. We

should not think of the role properties as akin to substantial non-physical

properties, like the properties a Cartesian dualist would ascribe to mental

substance. Rather, they are simply second-order properties - in particular

second-order properties that are guaranteed by the possession of

physical properties with certain causal roles.

4 Inverted spectra

Now to criticisms of functionalism: one family of objections focuses on

functionalism’s ability to deal with conscious, qualitative states - states

that it is ‘like something to have’, in Thomas Nagel’s 1974 phrase. These

states include sensory experiences, pains, itches, and emotions, and so

on, but arguably do not include such propositional attitudes as belief and

desire (see Qualia; Propositional attitudes.)

Let us start with the ‘inverted spectrum’ argument against functionalism.

Suppose that baby Matthew has an operation performed on his retina at

birth which switches the ‘red’ and ‘green’ messages from his retina to

his visual cortex: the central physiological state produced in Matthew by

red things is thus the state normally produced by green things in other

people, and vice versa. From then on Matthew is brought up normally,

learning how to discriminate between red and green things, to call them

red’ and ‘green’ respectively, and so on. Consider now the state (let us

call it A) produced in Matthew by red things. It seems likely, given

Matthew’s normal upbringing, that A will play the same causal role as

the physiologically different state which arises when other people are

presented with red things. Now let us ask: what kind of conscious

experience will Matthew have when he is in state A? Despite A’s

sharing a causal role with the state normally produced by red things, it

seems intuitively plausible that A will be consciously like the state

normally produced by green things. After all, when Matthew has A, his

brain is physiologically just like a normal brain presented with a green


But this now presents a prima facie problem for functionalism. For if the

functional role of A classifies it with the state normally produced by red

things, but it feels like the state which is normally produced by green

things, then functionalism has failed to capture the conscious aspect of

this state.

Note how this thought experiment differs from the simpler, traditional

inverted spectrum’, which postulates an individual - Millie, say - who is

normal in all physical respects but still has her colour experiences

inverted’. The physical state which gives normal people the conscious

experience of red gives her the conscious experience of green. Millie is

incompatible with the general physicalist assumption that there cannot be

mental differences without physical differences of some sort. The

retinal operation’ version of the inverted spectrum thought experiment,

by contrast, does not require us to deny this general physicalist

assumption. There are physical abnormalities in Matthew to account for

his abnormal colour experiences. The point of the retinal operation

thought experiment is rather that Matthew’s abnormality is only at the

physical level, and not at the functional level. Because of this, the retinal

operation thought experiment poses a problem specifically for

functionalism, which is committed to explaining mental differences in

terms of functional differences, but not necessarily for other versions of


5 Responses to inverted spectra

There are various responses to the retinal operation thought experiment

open to both ‘role’ and ‘realizer’ functionalists. Let us consider them in

turn, starting with role functionalism.

(1) The most direct response open to role functionalism is simply to deny

the intuition that Matthew will have different conscious experiences from

normal people. That is, role functionalists can argue that if Matthew’s

state A plays the same functional role as the state produced in normal

people by red things, then it will feel the same, even if it is physiologically

different. After all, they can point out, by hypothesis this state will make

Matthew react in just the way that normal people react to red

experiences, both in his behaviour and in forming further beliefs and

desires. If this is so, they will ask, then what substance is there to the

hypothesis that the experience is nevertheless consciously different?

However, this answer is less than wholly convincing. Maybe some

qualitative states, like pain, can plausibly be argued to depend on nothing

but functional role (if a state creates a pressing and intense desire to

move some part of your body, does that not show it is a pain?). But other

qualitative states, of which colour experiences are the paradigm, do not

seem to be nearly so closely tied to functional role (there is no specific

desire or behaviour which is typically prompted by an experience of

green). So if we take some such state, like normal people’s experience

of green, and imagine its functional role switched while its physiology

stays the same, as in Matthew, then most philosophers have a strong

intuition that will still feel the same, despite the fact this runs counter to

role functionalism.

(2) Other role functionalists adopt a different tack. They simply admit

that their theory of mind cannot deal with ‘qualia’ - that is, with the

qualitative aspects of colour and similar conscious experiences. Instead

they accept that these aspects are fixed by physical realization rather

than functional role. Role functionalists of this stripe can continue to

maintain a role functionalist account of non-qualitative states like belief

and desire, and indeed of the non-qualitative aspects of qualitative states.

But their role functionalism will be incomplete, in that they admit that

qualia themselves are not fixed by their functional role. In particular, they

cconcede that an individual who is functionally normal but physically

abnormal, like Matthew, will not experience normal qualia.

(3) The ‘retinal operation’ thought experiment is much less of a problem

for realizer functionalism than for role functionalism. After all, realizer

functionalists take mental terms to refer to physical states from the start:

they distinguish between human pains (one physiological state) and

octopus pains (a different physiological state), even though both states

play the pain role. Similarly, they can distinguish between Matthew’s

experience of red, which is one physical state, and a normal person’s

experience, which is a different physical state. True, both states play the

same functional role. But nevertheless the realizer functionalist counts

them as distinct mental states, and to this extent need not be disturbed by

intuitions that they are also qualitatively distinct.

(4) Both the realizer functionalist of the last paragraph, and the

incomplete role functionalist of the previous one, take qualia to be fixed

by physical realization. Different physics, different qualia. However,

some philosophers argue that this principle is excessively ‘chauvinist’,

since it implies that beings who lack human physiology cannot share

human experiences. Surely, more ‘liberal’ philosophers argue, we want

to allow that dolphins, say, or the inhabitants of Proxima Centauri’s third

planet, might be functionally organized in such a way as to experience

pain and sadness, say, even if they have non-human physiologies.

David Lewis has devised an ingenious version of functionalism, which

allows this kind of liberalism while still respecting the intuition that

Matthew’s experience when he is faced with something green is like a

normal person’s experience of red. Lewis argues that terms for mental

states (‘belief’, ‘desire’, ‘pain’, ‘sadness’, ‘experience of green’ and so

on) should be understood, in their application to any given being, to refer

to the first-order state that realizes the relevant functional role in normal

members of the species (or other group to which the being belongs).

Matthew is a human being, so ‘experience of green’ as applied to him

refers to the physical state that is produced by green things and plays the

corresponding functional role in normal human beings. In Matthew, of

course, this physical state (state A) plays the functional role that relates

to red things. But classified mentally it is still an experience of green,

because in normal people it plays the role appropriate to green things.

If we are talking about a Proxima Centaurian, by contrast, then

experience of green’ does not refer to the physical state produced by

green things in humans, but rather to the physical state (if there is one)

produced by green things in Proxima Centaurians. So it is possible for a

Proxima Centaurian to have an experience of green, Lewis argues, as

long as it is in the physical state that plays the relevant role in its normal

conspecifics, even if that is different from the state A that plays this role

in humans. (Note that there could also be an abnormal Proxima

Centaurian in whom the experience of green does not play the green

functional role, as long as it is the state that plays that functional role in

normal Proxima Centaurians.)

So, according to Lewis, qualitative experiences vary with physical

realizer states within a given group, but with functional role states across

groups. His theory thus accommodates the intuition that Matthew is

qualitatively different from normal humans, while avoiding the chauvinist

implication that animals and extraterrestrials cannot share our


Lewis is normally classified as a realizer functionalist. But in one respect

the mixed theory just outlined goes beyond realizer functionalism. A

different physical state realizes experiences of green in humans and

Proxima Centaurians. Yet Lewis counts them alike in respect of their

qualitative nature. This makes him different from the realizer

functionalists described in (3) above, who take the difference in physical

realization to lead to a difference in qualitative feel. Still, this only shows

that Lewis is not a straightforward realizer functionalist, not that he is


A more serious difficulty for Lewis’ theory is that it makes the

qualitative classification of peoples’ mental states depend on which

species or group we assign them to. Suppose a minority subspecies of

humans evolves a different pain mechanism from other humans. And

suppose Jane belongs to this subspecies. Qua member of the subspecies,

Jane will be in pain when her mechanism is activated, for she will be in

the state that plays the pain role in normal members of the subspecies.

But qua human simpliciter, she will not be in pain, for she is not in the

state that plays the pain role in normal humans. Lewis accepts that there

will sometimes be no unique answer to the question of which group an

individual belongs to, and consequently that it will be indeterminate what

qualitative mental states it has. This is, to say the least, a surprising

consequence of his theory.

(5) Some role functionalists argue that the way to deal with the inverted

spectrum problem is to distinguish extra levels of functional organization.

Standard discussions only consider two levels - ‘macroscopic’ functional

role and physical realization. Because of this, standard discussions face a

dilemma: they either tie qualia to ‘macroscopic’ functional role, and fail

to accommodate the intuition that Matthew is qualitatively abnormal; or

they tie qualia to physical realization, and end up

chauvinistically’ denying pains to Proxima Centaurians. Lewis’ mixed

theory offers one way out. But another way out would be to uphold role

functionalism, but identify some intermediate level of

micro-functional’ organization, which will distinguish Matthew from

other humans, yet be common to humans and members of different

species. By hypothesis, Matthew’s states coincide with normal human

states in respect of such ‘macroscopic’ role features as which actions

they give rise to, which beliefs and desires they prompt, and so on. But

they need not coincide in more ‘microscopic’ respects such as the light

reflectance profile being computed, the firing pattern of the relevant

neural units, and so on. The important point is that these microscopic

features are still functional, in that they can be realized in systems of

different physical composition. So they could be shared by humans and

physically different species, while at the same time serving to distinguish

Matthew from normal humans.

Note that this ‘microfunctional’ solution is open to scientific

functionalists, but not to common-sense functionalists. Assumptions

about reflectance profiles and neuronal firing are not part of

common-sense psychology. Common-sense functionalism is therefore

forced to count Matthew as functionally normal, and so unable to explain

his intuitive qualitative difference by reference to functional factors.

Scientific functionalism is not so constrained, as there are many possible

levels of functional organization in the brain which might distinguish

Matthew from normal humans. The only question is whether this is an

embarrassment of riches. For once we have opened up the possibility

that qualitative features might be fixed by any level of functional

organization, what could possibly decide which level does in fact fix it? It

is no good appealing to the subject’s introspective reports, since they are

part of macro-functional role, which is already agreed not to fix qualia

(thus Matthew will describe state A as ‘seeing red’, yet nearly

everybody agrees he is experiencing green). On the other hand, it seems

highly unlikely that intuition can resolve the issue, for intuition is surely

not fine-grained enough to decide such questions as whether somebody

who is computing the normal reflectance profile for green things, but

using an abnormal pattern of neuronal firing to do so, is experiencing

green or not.

6 Absent qualia

The inverted spectrum thought experiment is consistent with the idea that

appropriate functional organization guarantees conscious states of some

kind. The only issue raised by the inverted spectrum argument is whether

functional identity suffices to fix the qualitative identity of those

conscious states. A more radical form of argument asks whether

functional organization can guarantee consciousness at all. Are there not

always going to be systems that display the appropriate functional

organization yet lack qualia altogether?

Clearly, answers to this question will depend crucially on what qualifies

as ‘appropriate functional organization’. Recall that earlier the notion of

a functional role was introduced via Ramsey’s strategy for defining

unobservable terms in terms of a theory. On this conception, you have

mental states with a certain functional role if you have a set of states

which interact in the way specified by some psychological theory


A number of thought experiments aim to show that any such notion of

functional role must yield too ‘liberal’ an account of qualitative states, in

that it will imply the presence of qualia in systems that in fact lack them.

Imagine that the sensory messages into your brain and the motor

messages going out are disconnected from your brain and instead linked

by radio to a complicated structure made of old beer cans. This system

of beers cans is arranged in such a way that its properties relate to each

other and to your sensory and motor messages in just the way that

psychological theory T says your brain states did before they were

disconnected. Then the system consisting of your debrained body plus

beer cans will have exactly the same functional organization as you had

before your brain was disconnected. Yet it seems implausible that this

system would be conscious, that there is something that it would be like

to be a debrained body plus beer cans.

Even so, a significant number of functionalist philosophers are prepared

to bite this bullet, and maintain that the beer can system would be

conscious. After all, if some diseased part of your brain were replaced

by some microscopic silicon-based circuitry, this would not necessarily

stop you being conscious. What is the difference in principle, it could be

asked, between this and the beer can case?

The beer can thought experiment starts with a normal human being, and

then assumes the alteration of this human being’s brain. But there are

further arguments which query whether even behavioural similarity to

human beings is necessary for the relevant kind of functional

organization. These arguments focus on the contribution of the

observational terms O to the specification of functional roles. Note that it

is essential to any plausible functionalism that these terms somehow be

independently understood. For, if the O terms are themselves defined in

terms of theory T, there will be far too many systems the properties of

which satisfy T. Indeed, given a completely uninterpreted theory

, it is arguable that any physical system - the air

molecules in your house, say - will display some set of properties that are

related to each other as T says the Ms and O are related. So some prior

understanding of the O terms is needed to give us any chance of

ensuring that only conscious systems have the functional organization

specified by T.

The natural way to do this is to equate O terms with terms for perceptual

inputs (for example, ‘receiving visual stimuli from an ice cream’) and

behavioural outputs (for example ‘moving your arm towards an ice

cream’). But this, then, has the disadvantage of implying that some

presumably conscious human beings, such as experimental subjects

whose optic nerves are being stimulated by brain scientists, or paralyzed

people who display no behavioural outputs, will lack the relevant

functional organization.

The natural response to this difficulty (for a scientific functionalist at

least, even if not for a common-sense functionalist) is to appeal to

physiology to specify less peripheral inputs (kinds of stimulation of the

optic nerve, say) and less peripheral outputs (kinds of activity in the

motor cortex). But this physiological move threatens chauvinism. For it

will exclude from the category of conscious beings any extraterrestrials

and animals whose non-human physiologies do not include optic nerves

and motor cortexes.

There remains some room for functionalists to manoeuvre here. One

possibility would be to return to Lewis’ mixed theory, which would imply

that the experimental subject and the paralyzed person are conscious

because they are in the states that play the requisite causal role in their

normal conspecifics, even if not in themselves. Another increasingly

popular response is to appeal to teleological considerations, and

understand ‘functional organization’ as not only specifying a causal

structure but requiring in addition that this structure be a product of

biological design. By appealing to this further sense of ‘function’, such

teleofunctionalists’ can specify that beings with different physiologies

can nevertheless share the same functional organization, in virtue of

having causally similar structures which have been designed for similar

biological purposes.

7 Representational content

This further teleological sense of ‘function’ is also relevant to the

difficulties functionalists face in accounting for content. Many mental

states, and in particular such non-qualitative propositional attitudes as

belief and desire, represent states of affairs other than themselves. It is

unclear whether the functionalist theory of mental states can account for

such representational contents.

One immediate difficulty is that many representational contents are

broad’, in the sense that the content of many propositional attitudes

seems to depend not just on the thinker’s physical make-up but also on

features of the context (see Content: wide and narrow). For example, it

is arguable that the possession of beliefs about natural kinds depends not

just on the organization of the believer’s brain, but also on features of the

believer’s social context and on which natural kinds are present in the

believer’s world. This raises a prima facie problem for functionalism.

For functionalism makes the possession of mental states depend on the

subject’s brain having a certain causal structure. Yet by hypothesis the

possession of broad propositional attitudes is not fixed by facts about


However, functionalism has an obvious remedy. There is no obvious

reason why functionalism should only consider causal structures inside

the head. Why not have a ‘broad’ functionalism, which recognizes

broad’ causal structures in which mental states interact, not only with

each other and with sensory inputs and motor outputs but also with such

external factors as the social environment and objectively existing natural

kinds? This would then open the way to a theory which makes the

possession of broad propositional attitudes depend on broad causal


So perhaps broad contents present no special problem for functionalism.

A more radical criticism of functionalism, however, queries its ability to

account for representational contents of any kind, broad or not.

Functionalism identifies mental states as items that have certain causes

and effects. Yet it is doubtful whether representation can be explained in

any simple causal terms.

It might seem that, once we are allowed to appeal to ‘broad’ causal

structures, we can identify the contents of beliefs as those external

circumstances that typically cause them, and the contents of desires as

those external states of affairs to which they typically give rise. But this

simple strategy is afflicted by the problem of ‘disjunctivitis’. Take the

belief with the content that an ice cream is in front of you. This can be

caused, not only by a real ice cream, but also by a plastic ice cream, or a

hologram of an ice cream, and so on. Similarly the results which follow

any given desire will include not only the real content of the desire, but

also various unintended consequences.

So, even if we are allowed broad causal roles that include external

causes and effects, we still need somehow to identify, among the various

causes that give rise to beliefs, and the various results that eventuate

from desires, those which comprise the beliefs’ and desires’ real

contents. There are a number of possible ways of doing this. One of the

most promising is to appeal to teleological considerations once more. For

then we can pick out a desire’s content as that effect which it is the

desire’s biological purpose to produce. And, similarly, we can pick out a

belief’s content as that condition with which it is the biological purpose of

the belief to be co-present (see Semantics, teleological).

One question raised by this appeal to teleology is whether the original

functionalism is still doing any work. We started with the functionalist

idea that mental states can be identified by their causal roles. But it now

seems that, for contentful mental states at least, causal roles are not

enough, and need to be supplemented by biological purposes. The

obvious question is whether biological purposes would suffice by

themselves. Perhaps we can identify contentful mental states by their

purposes alone. The answer depends on whether a common biological

purpose, but different causal roles, implies different representational

states, and relates to a number of currently controversial questions in the

philosophy of representation.

See also: Materialism in the philosophy of mind; Reductionism in the

philosophy of mind


Routledge Encyclopedia of Philosophy, Version 1.0, London: Routledge

References and further reading

Block, N. (ed.) (1980) Readings in the Philosophy of Psychology,

London: Methuen, vol. 1.(The section on functionalism contains many

important articles, including early versions of functionalism by H.

Putnam, D. Lewis and D.M. Armstrong. See also Block’s

Introduction to this section, and the difficulties he raises in

Troubles with Functionalism’.)

Descartes, R. (1641) Meditations on First Philosophy in The

Philosophical Writings of René Descartes, trans. J. Cottingham, R.

Stoothoff, D. Murdoch, Cambridge: Cambridge University Press, vol.

2, 1984.(Classic statement and defence of dualism.)

Lewis, D. (1980) ‘Mad Pain and Martian Pain’, in N. Block (ed.),

Readings in the Philosophy of Psychology, London: Methuen,

1980, vol. 1; repr. with a postscript in D. Lewis Philosophical

Papers, Oxford: Blackwell 1983, vol. 1.(Lewis’ ‘mixed’ version of


Millikan, R. (1986) ‘Thoughts without Laws’, Philosophical Review

XCV: 47-80.(Argues that mental states should be analysed in terms

of biological purposes rather than causal roles.)

Nagel, T. (1974) ‘What is it like to be a Bat?’, Philosophical Review

83: 435-50, repr. in Block (1980).(Argues that the subjective

character of experience is not captured by physicalist theories.)

Papineau, D. (1991) ‘Teleology and Mental States’, Aristotelian

Society Supplementary Volume LXV: 33-54.(Argues that

functionalism without teleology cannot account for representation.)

Ramsey, F.P. (1931) ‘Theories’ in Foundations of Mathematics,

London: Routledge & Kegan Paul.(The original statement of

Ramsey’s account of theoretical terms.)

Rosenthal, D.M. (ed.) (1991) The Nature of Mind, London: Oxford

University Press. (Good general collection on contemporary

philosophy of mind. Papers by W.G. Lycan and E. Sober argue for

microfunctionalism’ (Lycan) and ‘teleofunctionalism’ (Lycan and


Shoemaker, S. (1984) Identity, Cause and Mind, Cambridge:

Cambridge University Press.(This collection of Shoemaker’s papers

includes discussions of the different species of functionalism and of

the inverted spectrum and absent qualia objections.)

Routledge Encyclopedia of Philosophy, Version 1.0, London: Routledge