(p.318) New Appendix B
(p.318) New Appendix B
Language of Thought, Connectionism, and Folk Psychology
There has been considerable recent philosophical controversy about whether there is a ‘language of thought’ (sometimes jokingly called ‘Mentalese’). What those who advocate this thesis1 seem to be urging is that each human is born with an interior language in which he thinks many private ‘thoughts’. In order to communicate with others he then learns to translate from Mentalese into some public language, such as English, or vice versa.
Those who espouse or criticize this theory often do not distinguish between ‘thought’ in the sense in which I am using the term, and ‘belief’; and, as not much for present purposes turns on this distinction, I will gloss over it in this appendix. The interpretation given by language‐of‐thought theory to organisms having beliefs in a language is that they have beliefs by ‘storing sentences’; to each belief there corresponds some inner state which represents it and the ‘system of internal representation’ is such that ‘the sentence‐analogs have significant grammatical structure’ that is, each inner state which represents a belief does so in virtue of having aspects or parts which are related to each other as are parts of a sentence—change the aspect and you change the belief in the way that you would change its meaning by replacing one word of a sentence by another.
There is good reason for supposing that when people acquire a public language, they internalize its concepts and express their thoughts not necessarily in the words of the public language but with its concepts. As we noted on pp. 82–3, people reason in (p.319) thoughts or change their beliefs in ways which can be captured by patterns of argument sensitive to the structure of the public sentences which express those beliefs. If someone comes to believe that p, and then a bit later comes to believe that q, they then come to believe that p‐and‐q—so a belief has been added to another belief to produce belief in a conjunction, and so there ought to be separate internal states corresponding to each conjunct, the second being formed after the first but not altering it. If someone believes that all men are mortal and comes to believe that Socrates is a man, they will (often) conclude that Socrates is mortal; and that can only be if their internal states exhibit differences which correspond to the parts of the sentences which capture their beliefs and are programmed to change in ways corresponding to the principles which govern the validity of syllogisms.
But the language‐of‐thought hypothesis holds that we have an internal language prior to our acquisition of a public language. According to the cognitive psychology favoured by this hypothesis, our making perceptual judgements (e.g. ‘there is a ball in front of me’) in response to sensory input is best explained as the formation of linguistically articulable hypotheses on the basis of more primitive linguistically articulable pieces of information. Since such judgements occur long before organisms acquire public languages, what is going on seems to be the processing of information in an internal language. The learning of a public language is best explained, according to language‐of‐thought theorists, as the testing of different hypotheses formulated in the internal language about the meaning of public sentences.
Against the suggestion of an inborn (as opposed to subsequently formed) language of thought, there stands the fact on which I commented in the text—that the sort of thoughts and beliefs people have are very much conditioned by the language and practice of their society. Relatively seldom do they seem to have thoughts and beliefs not so expressible; and the beliefs and thoughts of which they are capable grow with their understanding of the language and participation in the practice of their societies. This would be surprising if they were born having a whole language of thought independent of the public language.2 All this (p.320) is not to deny Chomsky's claim that humans have an in‐built capacity for learning public language—it is merely to doubt that they have much of a language apart from the public language (which in time they come to use privately).
Since processes of logical thought can be well represented by successions of states of digital computers, the language of thought hypothesis naturally suggests a hypothesis about the brain—that it operates like a digital computer and that to every thought their corresponds a different brain state whose different parts correspond to parts of a sentence which expresses the thought. We may call this view brain sententialism. But all the evidence is that many brain processes are not computational in character. They have a more complicated ‘connectionist’ structure, in which neurones with many inputs from and many outputs to other neurones inhibit and excite each other—there are no linear processes corresponding to the processes of deductive inference. It is however possible that while the systems so far studied are connectionist, and those connectionist systems are involved in non‐conscious processes such as recognition, other systems in the brain are computational, and these are the ones involved in conscious processes of logical thought.3
I wrote that the language of thought hypothesis ‘naturally suggests’ the sententialist hypothesis. It would actually entail it, if thoughts (or beliefs) were, i.e. were the same events as, brain‐states. For then brain‐states which were thoughts or beliefs would have to have the sentential structure of the language of thought. From this it follows that if there are no brain states with such structure whose interactions correspond to the interactions of thoughts and beliefs, that there are no thoughts and beliefs (and a similar argument would lead to a similar conclusion for mental events of other kinds). The public behaviour, which we ordinarily suppose to be determined by and reflect our beliefs, would in fact be controlled by and only by neural events, none of which could be identified with separate beliefs. ‘Folk psychology’—the view that (p.321) we have beliefs, thoughts, etc. and do what we do because of them—would have proved a theory as outdated as the scientific theory that burning consists in the emission of phlogiston, or the theory that all material objects are made of the four elements—earth, air, fire, and water. The view that ‘folk psychology’ is an outdated (scientific) theory which should be abandoned by scientific psychology, is eliminativism, championed in recent years by (among others) Paul and Patricia Churchland and Stephen Stich.4
But this is of course an absurd view. Obviously we sometimes have occurrent thoughts; and we have beliefs about how things are in the world (even if sometimes people's beliefs are inconsistent, or it is not obvious what their beliefs are). We are aware of our thoughts as we experience them, and beliefs—as I shall argue in Chapter 7—are not just events postulated to explain public behaviour, but ones to which the subject has privileged access. The absurd conclusion follows from two premisses—one a plausible but contestable neurophysiological claim that brain sententialism is false, and the other one identity theory (that any mental events are identical with physical events). If we reject identity theory, as I have given plenty of other reason to do in the text, we are not saddled with the absurd conclusion—whatever neurophysiology might discover. And if the brain processes underlying logical thought do not exhibit sentential structure while thought processes do often exhibit sentential structure, that seems to provide a yet further reason for rejecting identity theory. If thoughts are separate from brain events, even if partly dependent on them, the succession of brain events need not have the same structure as the succession of thoughts. And that is not even to be expected if, as I argue on p. 84, brain state successions are not the only determinants of the processes of thought (and so of the formation of belief).
(1) See J. A. Fodor, The Language of Thought, first published 1975, Harvard University Press: Cambridge, Mass., 1979.
(2) See G. Harman ‘Language Learning’ republished in N. Block (ed.), Readings in the Philosophy of Psychology, vol. 2, Methuen, 1981; and ‘Language, Thought, and Communication’ in (ed.) K. Gunderson, Language, Mind, and Knowledge: Minnesota Studies in the Philosophy of Science, vol. 7, University of Minnesota Press: Minneapolis, 1975.
(3) On all this, see William Bechtel, ‘Connectionism and the Philosophy of Mind: An Overview’, republished in W. G. Lycan (ed.), Mind and Cognition: A Reader, Basil Blackwell: Oxford, 1990.
(4) See P. M. Churchland, Scientific Realism and the Plasticity of Mind, Cambridge University Press: Cambridge, 1979, esp. ch. 4; P. S. Churchland, Neurophilosophy, MIT Press: Cambridge, Mass., 1986; and Stephen P. Stich, From Folk Psychology to Cognitive Science, MIT Press: Cambridge, Mass., 1983. Somewhat similar is the view of Daniel Dennett that ‘folk psychology’ provides a useful everyday way of talking but does not, to speak strictly, describe fundamental constituents of the world, or their interactions. To speak strictly, there are just brain events and their interactions. See his Consciousness Explained, Little, Brown: Boston, Mass, 1991.
(9) See New Appendix B.