Tag Archives: Functionalism

Article Summary: “What is Functionalism?” by Ned Block

“What is Functionalism?” originally printed in Readings in Philosophy of Psychology vol.1, 171-184 (MA: Harvard University Press, 1980).


“What is Functionalism” is first and foremost an encyclopedic article; as such it is mostly descriptive rather than argumentative. Nonetheless, in its exploration of the theory of functionalism in the philosophy of mind, it does present an argument against the idea that functionalism supports physicalism – the view that all things and all properties are physical.

Block begins by delimiting several senses of ‘functionalism’ used by philosophers. The first is functional analysis, whereby a system is decomposed into functionally isolated parts and the function of the system explained in terms of these parts. Then there is computation-representation functionalism, which is an explanatory strategy in cognitive science whereby mental states are analyzed as computations taking place over representations. Finally, there is metaphysical functionalism, which examines what mental states are, and in virtue of what are states such as pain similar to one another. The article is concerned solely with this latter sense of functionalism.

Block states that metaphysical functionalism identifies mental states with functional states. A functional state is just a specific causal role in a larger system (in this case, a mind) and is intricately linked with stimuli, other mental states, and behavior. As such the functionalist answers the question “In virtue of what are X’s, X’s rather than Y’s?” (where X and Y are types of mental state, e.g., pain) by saying that X’s are all X’s in virtue of their functional role in the systems of which they are a part, regardless of their physical realization.

Then Block proceeds to describe the first version of functionalism, machine functionalism. Presented by Hilary Putnam in the early 1960’s, machine functionalism argues that mental states are not just functional states, but more specifically are machine table states of a hypothetical machine called a Turing Machine. Turing Machines are automatons which can, in principle, compute any problem and which do so in virtue of what are called ‘system states,’ which are tied to instructions for computational steps (e.g., “If in system state S, perform computation C and then transition into system state S2, and so on). (Readers are encouraged to look at my summary of Putnam’s article if they want more detail on his presentation of machine functionalism.)

Block points out that machine functionalism is not the most general formulation of functionalism available; as such, he presents a more general version, using the tools of formal logic (persons not familiar with these tools might want to skip this part of the summary unless they are overly curious):

1. The first step is to reformulate a theory T (either ‘folk’ or ‘scientific’) of psychology such that is is a long conjunctive sentence with all mental state terms as singular (i.e., in the form of ‘has pain’ rather than ‘is in pain’):

T (s1 . . .sn)

‘s1 . . . sn’ are sentences including mental state terms.

2. Then, variables should be placed in the spot of mental state terms, and existential quantifiers for these terms prefixed. This forms what is called the ‘Ramsey sentence’ of the theory (named after Frank Ramsey, the genius who formulated this methodology):

∃x1 . . . xn T(x1 . . . xn)

‘∃x1 . . . xn‘ is the set of existential quantifiers prefixed to the theory, and there should be as many of them as there are mental state terms in the theory. ‘(x1 . . . xn)’ is the set of variables which now sit in the theory where the mental state terms used to.

3. Thus, if ‘x1‘ is the variable that replaced ‘pain’ in the psychological theory, we can now define pain as follows:

Y is in pain iff: ∃x1 . . . xn [T(x1 . . . xn) & Y has x1].

Since the sentence formerly expressing the mental state term ‘pain’ described it in terms of its relations to stimuli, other mental states, and behavior, pain is found to be whatever fulfills the relations which x1 exhibits.

Pain can be expressed as the property ascribed when one says ‘X has pain.’ Then, pain can be identified with the property expressed by the predicate ‘∃x1 . . . xn [T(x1 . . . xn) & Y has x1]’. Block gives us an example: if our psychological theory defines pain as being caused by pin pricks, causing worry & loud noise emission, and if worry subsequently caused brow wrinkling, then pain can be identified with the property expressed by the predicate ‘∃x1∃x2[(x1 is caused by pin pricks and causes loud noise emission and x2 & x2 causes brow wrinkling) & Y has x1]’.

This is the essence of the Ramsey method for defining mental state terms causally. Block supplements this method with extra notation for convenience’ sake. He establishes that ‘%xFx’ should be understood as ‘being an x such that x is F.’ Once this notation is introduced, we can define pain as follows: pain = %y∃x1∃x2[(x1 is caused by pin pricks and causes loud noise emission and x2 & x2 causes brow wrinkling) & Y has x1]. This just means that pain equals being a y (a thing) and being in x1, where x1 is caused by pin pricks . . . etc.

Block clarifies functionalism’s relation to behaviorism and traces the historical origins of functionalism as well. Behaviorism is the doctrine that mental states are to be analyzed purely in terms of behavior and behavioral dispositions. So, desiring an ice cream cone is just the disposition to grab a cone when one is present; and nothing more. Functionalism is importantly related to behaviorism because it also analyzes mental terms using behavioral responses to stimuli. The differences is that functionalism also refers to other mental states; further, these other mental states are interlinked with each other, stimuli, and behavior in a web of causal relations.

Block traces the development of functionalism to two developments in the philosophy of mind. The first is the ‘multiple realizability’ argument, originating with Hilary Putnam in his presentation of machine functionalism and subsequently used by Jerry Fodor in developing a method of explanation in psychology. The multiple realizability argument states that a mental state is unlikely to be a physical structure (such as a brain state) because it is eminently plausible that different physical structures – non-brain structures, for example – can realize mental states. So, pain is unlikely to just be C-fiber stimulation (or some other appropriate brain state), because octopuses and other such creatures can probably feel pain, despite their not having C-fiber stimulatory capacity. This led to the development of functionalism, which promised to unify physically different phenomena under the banner of causal (functional) similarity.

The second historical development leading to functionalism’s rise was JJC Smart’s development of ‘topic-neutral’ analyses of mental states. Smart was arguing for the identity theory of mind (see my summary of his article), the view that mental states equal brain states in the strict sense of identity. He needed, however, to respond to the objection that even if mental states were brain states, certain properties of these states were irreducibly mental, such as ‘sharpness.’ Smart responded by analyzing such property concepts in topic-neutral terms, or terms which did not indicate whether the property was physical or not. Other philosophers expanded upon these analyses, turning them into functional definitions of mental state terms (see, e.g., my summary of David Lewis’s philosophy of mind articles).

The introduction to this summary stated that Block argues against the idea that functionalism at all implies physicalism. The argument is strewn about the text and takes the form of two mini-arguments: one against the idea that functionalism supports type physicalism, and the other against the idea that functionalism supports token physicalism. Type physicalism is the view that, e.g., pain is a type of physical thing. Token physicalism, meanwhile, is a weaker thesis, holding merely that each individual instance of pain is a physical thing. The former implies the latter, but not vice versa.

Block delineates those who think that functionalism implies that physicalism is false from those who think it is true with labels: functional state identity theorists think physicalism is false because mental states are equivalent to functional states, which are non-physical. Functional specification theorists, meanwhile, think that physicalism is true because functional definitions (where pain = occupant of Causal role R, in the manner specified by the Ramsey method) lead us to physical mechanisms which fulfill the causal role in an organism. This leads to the view that pain-in-humans is C-fiber stimulation, since it occupies causal role R, and so on for all species.

Blocks argues against the idea that functionalism supports type physicalism by pointing out that, even if functional reductions are correct in the sense that a physical mechanism will be found to occupy causal role R, there is some non-physical facet of the various physical mechanisms which makes them pains rather than, say, pleasures. So type physicalism is false because mental state types are irreducibly functional. Meanwhile, even the weaker thesis of token physicalism is not supported by functionalism, in Block’s eyes. This is because there are conceivable entities which have non-physical states (such as soul states) which function just as mental states do. With functionalism negating type physicalism and not at all implying token physicalism, it seems as if functionalism does not support physicalism at all.

Logical Outline

Functionalism does not support physicalism

  1. Functionalism negates type physicalism.

    S1. If type physicalism is true, then mental state types are physical, e.g. pain is a brain state.

    S2. If functionalism is true, then mental state types are functional, i.e., non-physical states.

    S3. A mental state type cannot be both physical and non-physical.*

    S4. Therefore, functionalism negates type physicalism. [S1-S3]

      2.  Functionalism does not imply token physicalism.

S1. In order for functionalism to imply token physicalism, all conceivable tokens of mental states types would need to be physical.

S2. Yet, non-physical mental state token are conceivable, e.g., soul states rather than brain states.

S3. Therefore, functionalism does not imply token physicalism. [S1,S2]

     3. Therefore, functionalism does not support physicalism. [1,2]


Leave a comment

Filed under Article Summaries

Article Summary: “The Nature of Mental States” by Hilary Putnam


Putnam presents a version of functionalism – a position in the philosophy of mind which says that mental states can be defined in terms of their causal roles – as an empirical hypothesis worthy of investigation, especially in light of what he sees as the problems facing the identity theory of mind and behaviorism. Central to Putnam’s conception are the notions of the Turing Machine and of
a Probabilistic Automaton. A Turing Machine is a hypothetical device proposed by Alan Turing which computes various tasks in a deterministic fashion, given certain instructions and given a machine table which specifies how various states of the machine relate to one another and to inputs. (Readers are encouraged to investigate the machine more fully if they want the technical details. The points to understand are that the machine computes deterministically, and that it has its behavior entirely produced by input and states which ’react’ to the input in a specific fashion, producing computations and transitioning to other states automatically). A Probabilistic Automaton, meanwhile, is very similar to a Turing Machine, except that the transitions between states produced by input are determined
probabilistically rather than with certainty.

Putnam first addresses whether the question “Is Pain a brain state?” is meaningful at all. Putnam in quick succession defuses the objections to the thesis that the question has merit. One such objection is that pain cannot be a brain state because we can know we are in pain and simultaneously not know that we are in brain state ’S’. But this objection also applies to us knowing that there is heat in the oven: we can know that the oven is hot without knowing
that its mean molecular kinetic energy is high, despite the fact that heat just is mean molecular kinetic energy. He further argues that the notion of Pain being a brain state is at least intelligible, if not necessarily true, just as other, successful empirical reductions (e.g., water being reduced to H20) have been intelligible.

Two objections remain. The first states that the proposed reduction of Pain (and other mental states) to a brain state cannot be successfully engaged unless both things in the reduction are associated with a spatio-temporal area, and the spatio-temporal area is one and the same for both things. Putnam thinks this is wrongheaded and gives a counterexample: the reflected image off of a
mirror is merely reflected light, despite the fact that the image can at times appear behind the mirror. That is to say, empirical reduction is successful (the image is merely reflected light) despite the two things not sharing a spatio-temporal area.

The last objection states that, at best, brain states can be correlated with mental states, but never said to be equivalent to them. The predictions yielded from saying that the brain state and the mental event are correlated are equivalent to those which are yielded when we say that the state is the mental event, and there is thus no principled way of preferring the one over the other. Putnam agrees with this in a sense, but says that the two different views enable/prevent certain empirical questions from being asked. So there is a meaningful difference between the two stances. For example, if we assume that identity obtains, the question “What makes the pain accompany the brain state?” is rendered meaningless.

With objections to the very idea of reduction of mental to physical dealt with, Putnam presents his model of functionalism. The model deploys the concepts detailed above: the Turing Machine and the Probabilistic Automaton. The idea is that mental states are functional roles within a system, where a system is (in the case of humans) a Turing Machine with a probabilistic transfer between
various states. Putnam denotes the complete functional profile of a system – a complete account of all states and their probabilistic relations with one another and with sensory input – as the description of the system. Further, the behavior of the system is understandable without knowing how the system is realized. In other words, the mind can be understood perfectly well without understanding the particulars of the physical-chemical brain states which realize the mind’s functions.

Putnam admits that his model is vague on the details: the model says nothing, in fact, about how the functional states are to be defined (e.g., what does it take for an organism to be ’in pain’?), nor does it say how exactly the functional states relate to stimuli. Nonetheless, Putnam thinks his model is advantageous over the identity theory of mind because it presents researchers with an em-
pirical research program which is less vague, and easier to carry out, than any research program under the auspices of the identity theory. Putnam then espouses the virtue of his functionalism over the identity theory in more depth, but also its virtues over behaviorism.

Putnam thinks his functionalism presents a more tractable research agenda because in the case of the identity theory, there must be a singular physical-chemical structure which is, say, pain. If identity theory is true, then researchers will have to find the exact same physical-chemical structure in all the creatures which feel pain, and not find it in creatures which do not feel pain. Putnam admits that this is possible, but thinks that identifying a functional profile which is a commonality to all the organisms is more reasonable than trying to identify the exact same physical structure in all pain-feeling organisms.

Putnam extends his previous point to apply to the identity theory’s application to all mental states (and not just pain). Indeed, identity theorists have argued that all mental states are merely brain states. Putnam thinks this is almost certainly false, because in order for all mental states to be brain states, the entire physical-chemical structure of all organisms who experience such mental states will have to be the same. And Putnam thinks that this is most unreasonable: if even a single mental state is found to be experienced in another creature, and that creature has a different physical correlate of the mental state than any other creature, then the identity theory is false.

Putnam then advances reasons for his theory. The first is that we identity mental states in creatures in light of their behavior, and the commonality in behavior across organisms suggests a similar functional organization, but not a similar physical-chemical organization. This fact is Prima facie evidence for functionalism. Further, we tend to identity mental states in organisms based upon the behaviors they produce and the transitions into other states which they presage. So, we identity ’thirstiness’ partly by the fact that it tends to produce a certain behavior, which then tends to transition the system to another state, namely, ’not needing liquid anymore.’ And this fact about how we delimit mental states from one another suggests the appropriateness of a functional approach to such states.

Putnam then compares the functional approach to behaviorism, another
theory of identifying mental states. In the case of behaviorism, the mental states are defined as a behavior, or disposition to behave, on the part of an organism. Putnam gives a reason for supposing behaviorism is not particularly advantageous: despite the fact that we identify mental states by using behavior, this fact does nothing at all to suggest that the mental state actually is the behavior. At the time the article was written, behaviorism was facing serious conceptual difficulties (from which it would never recover), and Putnam (as he recognizes) is thus able to recite these difficulties. One is that it seems impossible to identity a behavioral disposition without referencing the concept one is attempting to define. So when the disposition of pain is defined, one must say something to the effect of ’pain is a disposition of X to behave as if X were in pain.’ In contrast, a functionalist can simply identity a functional state which pain can then be equated to. Further, there can be animals (or persons, if you wish!) who show no difference in behavior and yet are experiencing different mental states. A person X might be in pain and yet be suppressing their pain behavior, whereas a person Y may be in excruciating pain and yet have their motor nerves cut so that they cannot engage in pain behavior. And so pain is not the behavior in either case; it is something else.

Putnam concludes by reiterating methodological reasons for accepting a functional approach to mental states: One, the functional approach precludes certain questions from being asked (e.g., “Why does pain always accompany functional state S?”); two, the functional approach actually explains the behavior in question, rather than identifying mere correlations, and third, laws of psychology can be developed once an organism’s description (in Putnam’s sense) is known and a mental state is identified with a functional state.

Logical Outline

Primary Argument: Functionalism is preferable to the identity theory and behaviorism as a theory of mental states

1. Either functionalism, the identity theory, or behaviorism is the preferable theory of mental states.*

2. The identity theory is not a preferable theory of mental states.

S1. For the identity theory to be viable, a mental state needs to have
the exact same physical correlate in all organisms which experience that mental state.
S2. Very probably, organisms experience the same mental state with
different physical correlates of the same state.
S3. Therefore, the identity theory is probably not viable. [S1,S2]

3. Behaviorism is not a preferable theory of mental states.

S1. For behaviorism to be viable, one must be able to define mental
state dispositions without referencing the mental state itself, and thus
avoid circularity.

S2. It is not possible to define a mental state disposition without
referencing the mental state in question, and thus avoid circularity.

S3. Therefore, behaviorism is not viable. [S1,S2]

4. Therefore, functionalism is the preferable theory of mental states. [1,2]
5. There exist reasons for adopting functionalism, independent of the problems of the identity theory and behaviorism.

S1. A theory of mental states which enables a robust research pro-
gram is to be preferred.*
S2. Functionalism provides a reasonable commonality to identify
among organisms – functional states.
S3. Functionalism successfully prevents a number of questions from
being asked (questions which can waste researchers time).
S4. Functionalism – through its description of organisms’ functional
profiles plus identifications of mental states with functional states – promises to provide laws of psychology.
S5. Therefore, functionalism enables the most viable research pro-
gram [S2-S4].
S6. Therefore, functionalism is preferrable. [S5,S1]

6. Therefore, functionalism is preferable over the identity theory and over
behaviorism as a theory of mental states, and there are reasons to adopt
it independent of the other theories’ difficulties. [5,6]

*Asterisks denote enthymemes, or ‘suppressed premises.’ All these are are assumptions which must be made in order to make the explicitly stated argument work.


Filed under Article Summaries