Some brief notes on Fodor and Pylyshyn (2015) Some assumptions and where I agree or disagree with them:
(1)They assume the truth of belief/desire psychology and think that all branches of behaviourism are false. (I do not agree with this assumption and think that eliminative materialism may turn out to be true about the P/A. I think the proper course of action is to wait and see the truth is we don’t know the answer on this one yet.)
(2) They assume the truth of Naturalism. (I agree with this assumption)
(3) They assume the truth of the type/token distinction. (I agree with this assumption)
(4) They assume that grammatical discoveries have psychological reality. (I think that the grammatical discoveries in linguistics have psychological reality and are implemented by the brain so I accept their assumption)
(5) They assume that propositions have compositional structure. (I accept this assumption)
(6) They assume that mental representations have compositional structure. (I accept this assumption in so far as I agree with the RTM. But I have my doubts about the RTM)
(7) They assume that the Representational Theory of the Mind is true. (I am not sold on the RTM. I haven’t made up my mind yet but I am working towards a theory of embodied pragmatism which is anti RTM but I don’t think we can say for sure whether the RTM will remain true in a completed science of the mind/brain. My guess at this early stage is that the RTM is false.)
(8) They assume that the Computational Theory of the Mind is true. (I agree with this if the CTM includes connectionist models, and Bayesian models. I agree with Fodor that the CTM as traditionally conceived (Fodor’s isolated modules) cannot (and probably never will) handle things like global abduction. That is why I am don’t understand why Fodor and Pylyshyn don’t use connectionist models (they are wrong that connectionist models cannot handle compositionality see Chalmers [insert ref].
On another note while I accept the CTM (suitably modified) I think these models have to stop idealizing away from an organisms need to supply its own energy to fight off entropy (See Deacon ‘Incomplete Nature’). I think the idealisation inherent in some CTM models results in wildly inaccurate models. So while I accept the CTM I probably don’t do so in any sense that Fodor and Pylyshyn would agree with.
(9) They assume that Thought has priority to Language. (This is a complex one. I don’t accept the view that people always think in an internal Language of Thought. But strictly speaking this isn’t what Fodor and Pylyshyn asked. They are asking whether children develop thought prior to acquiring language? This is a massive area. I agree that thought is prior to language and not necessary for language. But I am not nearly as confident about the issue as Fodor and Pylyshyn are. )
So 1 the assumption that Propositional Attitude psychology must be maintained is the only one of their assumptions I disagree with. I don’t even say for certain that this assumption is false. I merely think that we don’t know whether it is true; and more research is needed until we can decide on the issue. Until such research in concluded I prefer to adopt Dennett’s intentional stance approach as opposed to making a-priori assumptions as Fodor and Pylyshyn do. I agree with them on assumptions 2, 3, 4, and 5. I agree with them on 8 and 9 though I am not as confident of my belief on the topic as they are. And in the way I interpret 8 they would probably have serious difficulties with me. I agree with 6 to the extent that I accept the RTM and I am not sure to what extent I accept the RTM. I think 7 is probably false but I am withholding judgement till I learn more.
I will examine how my differing assumptions relates to my difficulties with their central argument later on for now I want to note some quibbles I have with some claims they make in the preliminary part of their book.
Fodor and Pylyshyn engage in the grand old Chomskian tradition of dismissing scientific theories while presenting little justification. Thus we get the following:
“To begin with, we take it that behaviourism is false root and branch; in the paradigm cases, behaviour is the effect of mental causes… Wittgenstein and Skinner both held that the first-language acquisition is somehow the effect of “training” (“socially mediated reinforcement”, or the like) But that cannot be right either, since it turns out that children generally neither need nor get much language training in the course of learning a first language. Still more to the point, there is (as far as we know) no serious suggestion anywhere in either psychology or philosophy as to how training might work its putative effects…The thesis that (accepting, perhaps, occasional reflexes) new-borns are cognitively inert has become increasingly unattractive since the collapse of the Piagetian program in developmental psychology…” (Fodor and Pylyshyn: ‘Minds Without Meanings’ pp. 2-15)
This sort of hubris has become standard in cognitive science. Chomsky is the master of dismissing all psychologists, philosophers or linguists who disagree with him. Fodor and Pylyshyn show that they are more than a match for Chomsky when it comes to the rhetoric of dismissing the work of rivals. On the issue of behaviourism; it is now standard for cognitive scientists to act as though behaviourism is dead. This is simply not true. Behaviourism (in the guise of Applied Behavioural Analysis) is a flourishing field which is used in psychiatric and intellectual disabilities institutes regularly. The evidence for the success of using Applied Behavioural Analysis is overwhelming. Furthermore, the standard mode of attack against behaviourists have been refuted. I won’t go over the material here but refer my readers to my blog-posts (1) Poverty of Stimulus Arguments and Behaviourism, (2) PECS and Verbal Behaviour, (3) Some Behavioural Techniques and The Idea of a Blank Slate. In these blog-posts I demonstrate that standard arguments that behavioural techniques cannot work; poverty of stimulus arguments; the idea that children do not get corrected by their peers for incorrect grammatical assertions and even if they did they would not make use of these corrections are all false (See Pullum and Schulz 2002, Hart and Risely 1995, and Choinard and Clark (2002). I think my data clearly shows that Fodor and Pylyshyn’s confident assertions aside behaviourism is alive and well.
On the issue of the Piagetian paradigm being dead and buried again Fodor and Pylyshyn are massively exaggerating. It is true that many contemporary developmental psychologists disagree with some central claims of Piaget. This though is standard in science. Darwin didn’t have a mechanism with which to explain heredity; though we do now. Nobody thinks that this fact refutes the theory of evolution. Piaget was wrong in his views of pre-linguistic child development; children have sophisticated concepts of objects which go beyond anything that Piaget imagined. Piaget didn’t really make use of the preferential looking tasks developed by Fantz. For this reason his research underestimated the children’s cognitive abilities. But it is worth noting that the researchers who have refuted aspects of Piaget’s developmental theory don’t consider themselves as smashing a paradigm. Writers like Markman, Carey, Spelke et al typically start their books by stating their admiration for the pioneering work of Piaget before stating where they disagree with him. I have written about some of this research in my blog post ‘Indeterminacy of Translation and Innate Concepts’, here I argued that Spelke et al have shown that aspects of Piaget’s theory was wrong (though I don’t think that any of this evidence demonstrates that children have innate concepts). But neither I nor any of the authors working in developmental psychology argued that Piaget’s developmental psychology was destroyed. I think Fodor and Pylyshyn really should think before using such wildly exaggerated assessments of various psychological schools. Somehow I doubt that this will happen though; because pretending that rival schools of thought have been smashed in to nothing helps them with their I am the only president you’ve got style arguments:
“But not liking a hypothesis is not, in and of itself, grounds for rejecting it; one has the obligations of refuting it and offering serious alternatives. To our knowledge, neither of these burdens has been met by the (many) cognitive scientists who disapprove of BCS; and we doubt that either is likely to be in any future that we are now able to foresee. And any old theory is better than none: out of nothing comes nothing” (ibid p.17)
Their claim that there are no serious opponents to their views is dubious at best. It is a political manoeuvre where you throw mud at your rivals and declare to the voters look I am not great but at least I am not as bad as that guy. If Fodor and Pylyshyn want to declare all opponents of their view dead they owe it to their readers to seriously engage with their rivals. This is science not politics. What is important is evidence not rhetoric.
From page 20-32 they present a series of arguments as to why concepts cannot be images. I think there are some serious arguments in this section. And I find myself largely in agreement with their arguments. They have four arguments against concepts being images (1) Concepts can’t be images: Some concepts apply to things that cannot be pictured. (2)Black Swan Argument: We cannot form images that show that all swans are white. (3) Concepts have Constituent Structure but Images only have parts. (4) Leibniz Law: concepts cannot be images because there are no images in the brain. I think that in this section they do a good job of arguing that concepts are not images. However I have one or two quibbles that I want to briefly discuss here.
They begin their critique of the idea that concepts are not images by discussing George Berkeley’s famous critique of John Locke who claimed he could form an image of the abstract idea of a triangle. Berkeley argued that it is impossible to form an abstract idea of a triangle:
“If any man has the faculty of framing in his mind such an idea of a triangle as is here described, it is in vain to pretend to dispute him out of it, nor would I go about it. All I would desire is that the reader would fully and certainly inform himself of whether he has such an idea or no. And this me thinks can be no hard task for anyone to perform. What is more easy than to look into his own thoughts, and there to try, whether he has, or can attain to have, an idea that would correspond with the description that is here given of the general idea of a triangle, which is neither, Oblique, nor Rectangle, EquilateraL, Equicrual, nor Scalene but all and none of these at once.” (Berkeley: Principles of Human Knowledge Sec 10)
Pylyshyn and Fodor use this argument of Berkeley’s as a demonstration that there are some concepts we have that we have no image of. However they extend Berkeley’s argument further and claim that it shows that concepts cannot be images. They argue for this by noting that our concept of a particular person John is too abstract to be captured by a particular image of John as we know him through all of his changes. I think this is largely right. Even though people with Synaesthesia do typically associate various ideas with particular images there is no evidence that I am aware of that the particular concepts they have are exhausted by the images they associate with the concepts.
Nonetheless there are particular aspects of their argument which I have concerns with. Firstly the argument cites Berkeley who claims that he cannot form a particular type of mental image; a vague image of an abstract idea of a triangle. David Berman (2009) has argued convincingly that George Berkeley was an eidetic imager and part of his critique of Locke was derived from the fact that he was relying on his own introspection of how HE thought. As an eidetic imager Berkeley used vivid images in thinking about abstract ideas hence the idea of forming a vague image of an abstract triangle seemed impossible to him. Locke who wasn’t an eidetic imager would not need to a form a vivid image an abstract idea of a triangle to think about it. Hence the dispute between him and Berkeley was partly to do with their different abilities to form mental imagery. The interested reader should read Berman’s ‘Berkeley and Irish Philosophy’, and his ‘Philosophical Counselling for Philosophers’ for evidence that Berkeley was an eidetic imager and that this fact had serious implications for his type of philosophy. I won’t try to prove the point about Berkeley here I bring it up merely to point out that we shouldn’t just assume that all people have or use mental imagery in the same way. We know since Galton, and James that people have different abilities to form mental imagery, Kosslyn has added some interesting neuroscientific data to support the idea that people’s ability to form mental imagery varies. Ultimately I don’t think that people’s ability to form mental imagery substantially affects the point that Fodor and Pylyshyn are making. I am not aware of any eidetic imagers whose images exhaust the concepts that they have. As Wittgenstein in the ‘Philosophical Investigations’ showed convincingly any mental images have to be interpreted and these interpretations are conceptual in nature. So the images themselves are not the concepts but are interpreted in light of concepts.
Nonetheless I think that Pylyshyn who more than Fodor has worked on mental imagery over the last 30 years should at least deal with variations in people’s abilities to use and form mental imagery before claiming definitively what images are or are not. The connection between say a weak imagers concepts and images, and an eidetic imager’s concepts and images may be very different. A person who has no mental imagery may have different types of concepts to a person who is a mental imager. When Fodor and Pylyshyn argue that concepts are not images they need to deal with the empirical fact that people have different abilities to form mental images.
Fodor and Pylyshyn operate with a classical view of concepts and seem happy to ignore the mountains of empirical evidence that shows that the classical concept of ‘concept’ is false. In their discussion of why concepts cannot be mental images they make the following claim:
“Likewise, you can picture a chair, but not the property of being a chair, which is what all chairs have in common as such, and so on. The problem, in all such cases, is that the property shared by all and only things in the extension of the concept (i.e. Things the concept applies to) can’t be pictured” (‘Minds without Meanings p. 24)
Now whatever one may think of their views on the nature of imagery their view that concepts of x apply to all and only things that have the property of x seems like a classical view of concepts. Here is George Lakoff’s gloss on the classical theory of concepts:
“…The common idea of what it means to be in the same category: Things are categorised together on the basis of what they have in common. The idea that categories are defined by common properties is not only our everyday folk theory of what a category is, it is also a principal technical theory-one that has been with us for more than two thousand years” (Lakoff: Women, Fire and Dangerous Things p. 5)
Lakoff notes that this view of concepts is folk assumption of the nature of concepts one that we have had with us since at-least the time of Plato (more than likely this view predates Plato). However, while this folk view of concepts has a long history and tradition behind it; there is little empirical evidence to support it. There are key aspects to categorising things in the world; three important aspects are metaphor, metonymy and chaining.
In his ‘Women, Fire and Dangerous Things’ Lakoff noted that there eleven key features of the non-classical theory of concepts. (1) Concepts are Family Resemblances, (2) Concepts have Centrality, (3) There is polysemy as categorization (4) Generativity as a prototype phenomenon (5) Membership gradience (6) Centrality gradience (7) Conceptual Embodiment (8) Functional Embodiment (9) Basic-level categorisation, (10) Basic level Primacy (11) Reference-point or Metonymic Reasoning. (ibid p.13).
Now obviously I cannot go into the evidence for all of these features of concepts that Lakoff argues exists, that would take an entire book as opposed to a blog. But I do think that some salient points are raised by Lakoff that are directly relevant to Fodor and Pylyshyn’s position on the nature of categorisation. Firstly Lakoff appeals to the fact that Wittgenstein showed that concepts were family resemblances, members of a category are related to each other but they don’t all share common properties with each other. Wittgenstein supported his conclusion by an exhaustive examination of ordinary language concepts, and by the fact that philosophers since Plato have been searching for the essence of various concepts and the search has been in vain. Psychologist Eleanor Rosch empirically confirmed Wittgenstein’s view that concepts are family resemblances.
Now Fodor doesn’t deny Wittgenstein’s point that we cannot find necessary and sufficient conditions for the nature of particular concepts. But he argues that this merely shows that concepts are not definitions. He does however reject the second strand of Rosch’s argument that concepts are individuated by prototypes. Rosch noted that if the classic theory of concepts was actually correct; then since a concept is supposed to have the same essential property as all other members of the category, then any member would be as good an exemplar of the concept as any other. So, for example, with the category ‘Pet’ a spider would be as good an example of a pet as a dog. However when Rosch empirically tested subjects she discovered that people do in fact sort their categories interms of exemplars. People judge that some categories are more central members than others. Thus dogs are a better exemplar of Pet than are spiders. This finding showed that despite the fact that the classical theory of concepts is entrenched in our tradition the empirical evidence favours a non-classical theory of concepts.
Given this experimental evidence why do Fodor and Pylyshyn seem to accept the classical view of concepts? Strictly speaking they don’t. They think that the evidence shows that since concepts are neither definitions nor prototypes we have to stop thinking of them as intensions that determine extensions. Fodor uses the lack of definitions discovered by philosophers to argue against the view of concepts as definitions. When it comes to concepts conceived as prototypes Fodor argues that concepts are compositional and prototypes don’t compose so they cannot be concepts.
Fodor gives two reasons why concepts cannot be prototypes: (1) The Uncat Problem: Fodor argues that the concepts cannot be prototypes because there are countless concepts which do not have proto-types. So he asks us to think of the complex concepts ‘The Uncat’ this appears to have no prototype. Fodor notes that if we were to say that, for example, a stapler was a prototypical example of the ‘uncat’ concept we would be led to absurdities. So we would be forced to assume that the more something is like a stapler the less it is like a cat and the less something is like a stapler the more it is like a cat. Fodor argues that since this conclusion is absurd we are forced to conclude that for a substantial amount of concepts they have no prototypes. Hence the theory that concepts are prototypes is false (Fodor: Concepts p. 101). (2) The Pet Fish Problem: According to Fodor prototype theorists explicate notions like falling under a concept interms of being similar to its exemplar. He further argues that a prototype theory can only respect the compositional nature of a theory if they accept that a thing’s similarity to the exemplars of a complex concept is determined by its similarity to the exemplars of its constituents (ibid p.102). But Fodor notes that this does not work for many many concepts. Thus he notes that with ‘PetFish’, a guppy maybe a prototype of a pet fish, it is not a prototype of either ‘a pet’ or of ‘a fish’. So Fodor argues that ‘PetFish’ shows either that the complex concept is not constructed compositionally or that concepts are not prototypes.
Fodor’s arguments rely on a fundamental misunderstanding of prototype theory. When setting out both arguments made he made the false assumption that prototype theory requires that for something to fall under a concept it must be similar to the exemplar in the concept. This is not necessarily the case. In his ‘Women Fire and Dangerous Things’ Lakoff examined the work of R. M. W. Dixon (1982) who analysed the categorisation systems used by the Dyirbal an aboriginal tribe in Australia. Dixon noted that though at first glance the categories made little sense when one analyses them closely one finds that they conform to the pattern described by prototype theorists.
I will give one example of how the Dyirbal categorise:
“Balan: women, bandicoots, dogs, platypus, echidna, some snakes, some fish, most birds, fireflies, scorpions, crickets, the hairy mary grub, anything connected with water or fire, sun and stars, shields, some spears, some trees, etc” (Dixon 1982 taken from Lakoff ‘Women, Fire and Dangerous Things’ p. 93)
Dixon analysis of the concepts of the Dyirbal found that they followed the following rules (1) Centrality: Basic members of the category are central, so things like ‘gar fish’ and ‘hairy mary’s’ are less central than ‘women’ in the balan category. (2) Chaining: Complex categories are structured by chaining; central members are linked to other members, which are linked to other members and so on, for example women are linked to the sun, which is linked to sunburn, which is linked to the hairy mary grub. (ibid p. 94) Note number 2 contradicts what Fodor’s claim that prototype theorists argue that concepts must be similar to the exemplar of the concept. This claim is simply false. What prototype theory actually says is that the members of a category can be traced back to the main exemplar via a chaining member not that all members of a category have to be similar to the exemplar in a relevant way. Fodor’s objections to prototype theory are really just objections to his incorrect version of what proto-type theory is. (3) Experiential Domains: There are basic domains of experience, which may be culture-specific. These can characterize links in category chains. (4) No common properties: Members of a category do not need to share a common property with each other. I won’t list all of the rules Dixon uncovered but the ones I have listed show that Fodor fundamentally misunderstands the nature of proto-type theory.
I have briefly discussed Fodor and Pylyshyn’s major assumptions their ‘Minds Without Meanings’ and have shown which assumptions I agreed with and disagreed with. I have also outlined three minor quibbles I have with their negative critique of contemporary theories of concepts. (1) They caricature rival theories. (2) They ignore variation in people’s ability to form mental imagery. (3) They misunderstand the nature of the proto-type theory of concepts and hence are attacking a target they don’t understand. In my next blog I will evaluate their positive theory of the nature of reference.