What do you think?
Rate this book
[(The Elm and the Expert: Mentalese and Its Semantics)] [Author: Jerry A. Fodor] published on
The Elm and the Expert provides a lively discussion of semantic issues about mental representation, with special attention to issues raised by Frege's problem, twin cases, and the putative indeterminacy of reference. The book extends and revises a view of the relation between mind and meaning that the author has been developing since his 1975 book, The Language of Thought. Among philosophers, a general consensus exists that a referential semantics for mental representation cannot support a robust account of intentional explanation. This book is largely a reconsideration of the arguments that are supposed to ground this consensus. Fodor offers a theory sketch in which psychological explanation is intentional, psychological processes are computational, and the semantic properties of mental representations are referential.
Unknown Binding
First published August 1, 1994
5 people are currently reading
111 people want to read
About the author
Jerry A. Fodor
26 books87 followersJerry Alan Fodor is an American philosopher and cognitive scientist. He is the State of New Jersey Professor of Philosophy at Rutgers University and is also the author of many works in the fields of philosophy of mind and cognitive science, in which he has laid the groundwork for the modularity of mind and the language of thought hypotheses, among other ideas. Fodor is of Jewish descent.
Fodor argues that mental states, such as beliefs and desires, are relations between individuals and mental representations. He maintains that these representations can only be correctly explained in terms of a language of thought (LOT) in the mind. Further, this language of thought itself is an actually existing thing that is codified in the brain and not just a useful explanatory tool. Fodor adheres to a species of functionalism, maintaining that thinking and other mental processes consist primarily of computations operating on the syntax of the representations that make up the language of thought.
For Fodor, significant parts of the mind, such as perceptual and linguistic processes, are structured in terms of modules, or "organs", which are defined by their causal and functional roles. These modules are relatively independent of each other and of the "central processing" part of the mind, which has a more global and less "domain specific" character. Fodor suggests that the character of these modules permits the possibility of causal relations with external objects. This, in turn, makes it possible for mental states to have contents that are about things in the world. The central processing part, on the other hand, takes care of the logical relations between the various contents and inputs and outputs.
Although Fodor originally rejected the idea that mental states must have a causal, externally determined aspect, he has in recent years devoted much of his writing and study to the philosophy of language because of this problem of the meaning and reference of mental contents. His contributions in this area include the so-called asymmetric causal theory of reference and his many arguments against semantic holism. Fodor strongly opposes reductive accounts of the mind. He argues that mental states are multiply realizable and that there is a hierarchy of explanatory levels in science such that the generalizations and laws of a higher-level theory of psychology or linguistics, for example, cannot be captured by the low-level explanations of the behavior of neurons and synapses.
Fodor argues that mental states, such as beliefs and desires, are relations between individuals and mental representations. He maintains that these representations can only be correctly explained in terms of a language of thought (LOT) in the mind. Further, this language of thought itself is an actually existing thing that is codified in the brain and not just a useful explanatory tool. Fodor adheres to a species of functionalism, maintaining that thinking and other mental processes consist primarily of computations operating on the syntax of the representations that make up the language of thought.
For Fodor, significant parts of the mind, such as perceptual and linguistic processes, are structured in terms of modules, or "organs", which are defined by their causal and functional roles. These modules are relatively independent of each other and of the "central processing" part of the mind, which has a more global and less "domain specific" character. Fodor suggests that the character of these modules permits the possibility of causal relations with external objects. This, in turn, makes it possible for mental states to have contents that are about things in the world. The central processing part, on the other hand, takes care of the logical relations between the various contents and inputs and outputs.
Although Fodor originally rejected the idea that mental states must have a causal, externally determined aspect, he has in recent years devoted much of his writing and study to the philosophy of language because of this problem of the meaning and reference of mental contents. His contributions in this area include the so-called asymmetric causal theory of reference and his many arguments against semantic holism. Fodor strongly opposes reductive accounts of the mind. He argues that mental states are multiply realizable and that there is a hierarchy of explanatory levels in science such that the generalizations and laws of a higher-level theory of psychology or linguistics, for example, cannot be captured by the low-level explanations of the behavior of neurons and synapses.
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 3 of 3 reviews
April 28, 2025
It's a very important and surprising book where Fodor changes his mind about a few things (mostly narrow content, informational semantics, and inferential role semantics too a bit). What is particularly surprising is the boldness of the instrumentalist line of thought that runs through it (mostly lectures 2 and 4).
Trouble reconciling computational implementation of psychological laws and informational semantics of mental states: computation has to do with the intrinsic properties of mental states, while informational semantics has to do with external properties. If we want to keep semantic externalism (broad contents) and computationalism together, plainly, we need a way of explaining how broad contents can be computed. Yet there seems to be no way of doing that, because the whole point about broad contents is that computational twins may have different ones. The only kind of computable content is narrow content, the one that doesn’t depend on external properties.
Fodor counterargues that the reason why we want psychological laws, and hence mental states, to be computationally implementable is that we want them to hold at the nomologically nearby worlds, and Twin-Earth scenarios involve no such worlds. What Twin-Earth-like cases arise in our world and the worlds nearby in nomological space can be brushed aside as accidents, and so psychological laws are safe: they are not something laws ought to capture. Hence, for all worldly purposes, sameness of computational states guarantees sameness of (broad) content, securing folk-psychological explanations. The idea is that “intentional psychology is a special (i.e., nonbasic) science, so its laws are ceteris paribus laws. And ceteris paribus laws tolerate exceptions, so long as the exceptions are unsystematic.” (39)
Then arises the trouble of having computationally indistinguishable states directed at different objects, yielding different broad contents, in elm/beech cases. This kind of cases can’t be brushed aside as merely accidental. It would seem broad content individuation does cut things more finely than computational state individuation after all. Fodor, however, will have none of it: he thinks so-called deferential concepts are not the business of semantics, but epistemology (34). To use his example, the fact that, when I can’t tell an elm from a beech, I defer to an expert, is on a par with the fact that, when I don’t know if it’s Monday or Tuesday, I defer to the calendar. Fodor’s point is that asking an expert is a way of knowing how to tell elms from beeches; it’s a form of extended, shared knowledge. Fodor’s own account of what happens in such cases is surprisingly interesting for an externalist—it is a striking bout of extended epistemology:
“As a matter of fact, I can tell acids from bases; I use the litmus test to do so. And I can tell elms from beeches too. The way I do it is, I consult a botanist. / What I do with the litmus, and with the botanist, is this: I construct environments in which their respective states are reliable indicators of the acidity of the sample and the elmicity of the tree; in the one case, I dip the litmus into the fluid, in the other case, I point the expert at the tree. I construct these environments with malice aforethought; with the intention that what color the litmus turns (mutatis mutandis, what the botanist says about the tree) will cause me to have true beliefs about whether the sample is an acid (mutatis mutandis, whether the tree is an elm). In effect, I contrive to replace the problem of determining whether the sample is an acid with the (de facto easier) problem of determining whether the litmus turns red. Likewise, mutatis mutandis, I contrive to replace the problem of determining whether the tree is an elm with the (de facto easier) problem of determining whether the expert calls it one.” (34-5)
It is quite clear in Fodor’s exposition that extended epistemology affords cognitive offloading:
“The correlation between elms and the botanist’s elm thoughts was hard earned; think of all the dreary years he must have spent in graduate school learning to be a reliable elm-detector. Whereas I can now correlate my thoughts with elms practically instantaneously: My mind-world correlation co-opts his, much as, in the other case, the correlation between my acid thoughts and acids co-opts the correlation between acidity and the color of litmus. What philosophers call ‘linguistic deference’ is actually the use of experts as instruments [here, Fodor quotes Smart 1962 in a footnote]; not Marxist division of labor in semantics but capitalist exploitation in epistemology.” (36)
The idea is that society makes a form of mediated perception possible, and language just bears the consequences of that. Later (in Lecture 4), Fodor subsumes this under the label of ‘cognitive management’, i.e. the tendency to put yourself in a position where you are going to be caused to believe that p if p (99-100). He writes: “You find it not just in our intentional behaviour, but also in our reflexes; and not just in us but also throughout quite a lot of the animal kingdom. If there are noises off, many organisms will orient reflexively to foveate the noise source…. In effect, the orienting reflex is designed to so position a creature that if (to borrow an example of Winnie The Pooh’s) it was a Heffalump that made the noise, then the creature will come (and promptly too) to believe that it was a Heffalump that made the noise.” (100)
Roughly speaking, Fodor avoids the threat posed by Putnam cases by saying that, for all worldly purposes, they don’t exist (38-9). The more threatening threat is that of Frege cases. The thing is that, for a = b, and Fa, if intentional psychology is going to have predictive power, we better hope that Fa predicts something regarding whether Fb (40). In his earlier years, Fodor used to think that Oedipus stories provided a good argument for the relevance of narrow contents for psychological explanations. What Oedipus stories are taken to be a counterexample to is the piece of folk-psychological wisdom that people try not to marry their mothers. A revision of this saying, to accommodate Oedipus cases, could run: people try not to marry their mothers ‘so described’. But that just won’t do: what is the case is that people try not to marry their mothers at all! (44) I’m not sure why the proposed revision won’t work: the mother people are known to try not marrying can only be the person people know to be their mother. That doesn’t seem to be built into the semantics of ‘mothers’, it is built into that of ‘try’. Fodor’s way of explaning away Oedipus cases is more straightforward though: they just don’t happen irl! What this does is avoid recourse to narrow contents in psychological explanation.
Still, we want our psychological theory to distinguish between to want to marry Jocasta and to want to marry Mother. But, Fodor retorts, we can have this distinction at the level of the acts (the wants) without making it at the level of the contents (47), i.e. a distinction between coextensive attitudes (49). This suggestion is backed by a standardly Fregean idea: let’s individuate psychological states three ways, i.e. with a creature, a proposition (with a broad content), and a MOP. Appeal to the different MOPs buys us different acts with the same content, the same way as, mutatis mutandis, ‘John is a bachelor’ is not the same sentence as ‘John is an unmarried man’ (48).
One worry then: if broad contents alone don’t buy us the psychological explanations we want, but only MOPs do, since MOPs, like narrow contents (and the difference, after all, is not clear), supervene on purely syntax-driven computational states (and Fodor does identify the difference the MOPs make with the difference the syntax, as opposed to the semantics of the mental state, makes), and can, for this reason, be reduced to them, then, it seems like content per se drops out as irrelevant for psychological explanations. This line of argument is familiar from Stich (1983) (see p. 50). Fodor’s only way out of this eliminativist impasse seems to be: allow broad contents to do more explanatory work than has been assigned to them so far, and this is just what he proceeds to do.
In Lecture 3, Fodor solves the problem of the indeterminacy of reference by buying into some kind of inferential role semantics.
In Lecture 4, he comes close to saying that mental representations are, by definition, Pushmi-Pullyu Representations of sorts: “it’s of the essence of mental representations thay they face two ways at once: They connect with the world by representing it, by and large, veridically; and they connect with behaviour by being its typical proximal cause.” (83)
Trouble reconciling computational implementation of psychological laws and informational semantics of mental states: computation has to do with the intrinsic properties of mental states, while informational semantics has to do with external properties. If we want to keep semantic externalism (broad contents) and computationalism together, plainly, we need a way of explaining how broad contents can be computed. Yet there seems to be no way of doing that, because the whole point about broad contents is that computational twins may have different ones. The only kind of computable content is narrow content, the one that doesn’t depend on external properties.
Fodor counterargues that the reason why we want psychological laws, and hence mental states, to be computationally implementable is that we want them to hold at the nomologically nearby worlds, and Twin-Earth scenarios involve no such worlds. What Twin-Earth-like cases arise in our world and the worlds nearby in nomological space can be brushed aside as accidents, and so psychological laws are safe: they are not something laws ought to capture. Hence, for all worldly purposes, sameness of computational states guarantees sameness of (broad) content, securing folk-psychological explanations. The idea is that “intentional psychology is a special (i.e., nonbasic) science, so its laws are ceteris paribus laws. And ceteris paribus laws tolerate exceptions, so long as the exceptions are unsystematic.” (39)
Then arises the trouble of having computationally indistinguishable states directed at different objects, yielding different broad contents, in elm/beech cases. This kind of cases can’t be brushed aside as merely accidental. It would seem broad content individuation does cut things more finely than computational state individuation after all. Fodor, however, will have none of it: he thinks so-called deferential concepts are not the business of semantics, but epistemology (34). To use his example, the fact that, when I can’t tell an elm from a beech, I defer to an expert, is on a par with the fact that, when I don’t know if it’s Monday or Tuesday, I defer to the calendar. Fodor’s point is that asking an expert is a way of knowing how to tell elms from beeches; it’s a form of extended, shared knowledge. Fodor’s own account of what happens in such cases is surprisingly interesting for an externalist—it is a striking bout of extended epistemology:
“As a matter of fact, I can tell acids from bases; I use the litmus test to do so. And I can tell elms from beeches too. The way I do it is, I consult a botanist. / What I do with the litmus, and with the botanist, is this: I construct environments in which their respective states are reliable indicators of the acidity of the sample and the elmicity of the tree; in the one case, I dip the litmus into the fluid, in the other case, I point the expert at the tree. I construct these environments with malice aforethought; with the intention that what color the litmus turns (mutatis mutandis, what the botanist says about the tree) will cause me to have true beliefs about whether the sample is an acid (mutatis mutandis, whether the tree is an elm). In effect, I contrive to replace the problem of determining whether the sample is an acid with the (de facto easier) problem of determining whether the litmus turns red. Likewise, mutatis mutandis, I contrive to replace the problem of determining whether the tree is an elm with the (de facto easier) problem of determining whether the expert calls it one.” (34-5)
It is quite clear in Fodor’s exposition that extended epistemology affords cognitive offloading:
“The correlation between elms and the botanist’s elm thoughts was hard earned; think of all the dreary years he must have spent in graduate school learning to be a reliable elm-detector. Whereas I can now correlate my thoughts with elms practically instantaneously: My mind-world correlation co-opts his, much as, in the other case, the correlation between my acid thoughts and acids co-opts the correlation between acidity and the color of litmus. What philosophers call ‘linguistic deference’ is actually the use of experts as instruments [here, Fodor quotes Smart 1962 in a footnote]; not Marxist division of labor in semantics but capitalist exploitation in epistemology.” (36)
The idea is that society makes a form of mediated perception possible, and language just bears the consequences of that. Later (in Lecture 4), Fodor subsumes this under the label of ‘cognitive management’, i.e. the tendency to put yourself in a position where you are going to be caused to believe that p if p (99-100). He writes: “You find it not just in our intentional behaviour, but also in our reflexes; and not just in us but also throughout quite a lot of the animal kingdom. If there are noises off, many organisms will orient reflexively to foveate the noise source…. In effect, the orienting reflex is designed to so position a creature that if (to borrow an example of Winnie The Pooh’s) it was a Heffalump that made the noise, then the creature will come (and promptly too) to believe that it was a Heffalump that made the noise.” (100)
Roughly speaking, Fodor avoids the threat posed by Putnam cases by saying that, for all worldly purposes, they don’t exist (38-9). The more threatening threat is that of Frege cases. The thing is that, for a = b, and Fa, if intentional psychology is going to have predictive power, we better hope that Fa predicts something regarding whether Fb (40). In his earlier years, Fodor used to think that Oedipus stories provided a good argument for the relevance of narrow contents for psychological explanations. What Oedipus stories are taken to be a counterexample to is the piece of folk-psychological wisdom that people try not to marry their mothers. A revision of this saying, to accommodate Oedipus cases, could run: people try not to marry their mothers ‘so described’. But that just won’t do: what is the case is that people try not to marry their mothers at all! (44) I’m not sure why the proposed revision won’t work: the mother people are known to try not marrying can only be the person people know to be their mother. That doesn’t seem to be built into the semantics of ‘mothers’, it is built into that of ‘try’. Fodor’s way of explaning away Oedipus cases is more straightforward though: they just don’t happen irl! What this does is avoid recourse to narrow contents in psychological explanation.
Still, we want our psychological theory to distinguish between to want to marry Jocasta and to want to marry Mother. But, Fodor retorts, we can have this distinction at the level of the acts (the wants) without making it at the level of the contents (47), i.e. a distinction between coextensive attitudes (49). This suggestion is backed by a standardly Fregean idea: let’s individuate psychological states three ways, i.e. with a creature, a proposition (with a broad content), and a MOP. Appeal to the different MOPs buys us different acts with the same content, the same way as, mutatis mutandis, ‘John is a bachelor’ is not the same sentence as ‘John is an unmarried man’ (48).
One worry then: if broad contents alone don’t buy us the psychological explanations we want, but only MOPs do, since MOPs, like narrow contents (and the difference, after all, is not clear), supervene on purely syntax-driven computational states (and Fodor does identify the difference the MOPs make with the difference the syntax, as opposed to the semantics of the mental state, makes), and can, for this reason, be reduced to them, then, it seems like content per se drops out as irrelevant for psychological explanations. This line of argument is familiar from Stich (1983) (see p. 50). Fodor’s only way out of this eliminativist impasse seems to be: allow broad contents to do more explanatory work than has been assigned to them so far, and this is just what he proceeds to do.
In Lecture 3, Fodor solves the problem of the indeterminacy of reference by buying into some kind of inferential role semantics.
In Lecture 4, he comes close to saying that mental representations are, by definition, Pushmi-Pullyu Representations of sorts: “it’s of the essence of mental representations thay they face two ways at once: They connect with the world by representing it, by and large, veridically; and they connect with behaviour by being its typical proximal cause.” (83)
July 30, 2024
This book is very interesting, very niche, and I was awefully ill-equipped to start reading it when I did. This is a book for insiders, a lot is assumed to be known. Insofar as I can assess it, this book is worthwhile reading, but I'm not going to pretend that I understood most of it--maybe in a few years I'll be able to reread it with proper understanding.
June 25, 2024
Witty, characteristically.
Displaying 1 - 3 of 3 reviews