Whirring away

P.N. Johnson-Laird

  • The Modularity of Mind: An Essay on Faculty Psychology by Jerry Fodor
    MIT, 145 pp, £15.75, January 1984, ISBN 0 262 06084 1

Who now remembers phrenology as anything other than a Victorian pastime? Yet it began as a serious scientific hypothesis. Its founder, the German anatomist Franz Joseph Gall (1758-1828), argued that the brain is the organ of the mind, and that particular parts of the brain function as the organs of particular mental faculties. He assumed that there are distinct intellectual powers for language, music, mathematics and other domains, and that there are also distinct propensities for humour, destructiveness and even murder. Each of these faculties has its own organ in the brain, and the size of this organ depends on the development of the faculty. If you are a good musician, your musical organ of thought will be large; if you have little pre-disposition to murder, your homicidal organ will be small, and so on. Gall made a crucial mistake, however, in assuming that the relative sizes of the various organs will be reflected in the contours of the cranium. In fact, the skull does not fit the cortex like a glove, and there is no evidence that the relative sizes of the bumps or bulges in either of them have any relation to mental abilities. The putative science degenerated into a fictitious geography of the brain, which was charted by fabulous maps with such evocative regions as ‘amativeness’ and ‘philoprogenitiveness’ – maps that are as outdated as the pages of a 19th-century atlas. The subject finally petered out in party games where young persons searched for appropriate irregularities on the heads of other young persons, preferably those of the opposite sex. Phrenology had begun as science but ended as bumps-a-daisy.

Yet there was some truth in the doctrine. The brain is now universally accepted as the organ of the mind, and certain parts of it are known to specialise in particular mental functions. As Jerry Fodor, the author of this splendidly provocative book, puts it, ‘the good that men do is oft interred with their doctoral dissertations.’ His aim is to disinter the idea that there are separate mental faculties and to develop it unencumbered by naive craniology. His book reflects a general revival in the fortunes of ‘faculty psychology’. The linguist Noam Chomsky has argued for some time that there are distinct mental faculties: innate abilities that develop as ‘mental organs’ in much the same way that heart, lungs and other physical organs grow to maturity. And the psycholinguist John Marshall has drawn attention to the resemblance between this ‘new organology’ and aspects of phrenological thinking. The educational psychologist Howard Gardner has refurbished the old idea that there are distinct forms of intellectual ability, not just a single ‘general intelligence’. Fodor, however, is a philosopher – perhaps the most charismatic figure in the philosophy of mind, and certainly the only one who regularly carries out psychological experiments – and his goal, as befits a philosopher, is to establish a proper theory of faculty psychology and to tease out all the implications of this particular architecture of the mind.

Consider the difference between tap-dancing and mental arithmetic. They are obviously distinct skills, because some people are prodigies in one and mere plodders in the other. But the two activities may depend on certain shared mechanisms. For example, both may rely on a common memory for immediate information, such as the beat of the bar or the numbers to be multiplied. Both may depend on a common mechanism for long-standing memories – for types of dancing-step or for multiplication tables. If so, the mind must have a two-dimensional lay-out: to follow Fodor’s terminology, it has ‘vertical’ faculties, such as tap-dancing and mental arithmetic, that concern an entire skill from top to bottom, and it has ‘horizontal’ faculties, such as memory and perception, that crop up as parts of many, if not all vertical skills.

There is nothing very novel in the idea of separate mechanisms for memory, perception and thought, which work in the same way for all the different skills into which they enter: remembering a number is the same process as remembering a dance-step, and all that differs is the content of the memory. What was revolutionary in Gall’s thinking, according to Fodor, was his insistence on the primacy of vertical faculties, such as tap-dancing and mental arithmetic. He took horizontal faculties like memory to be fictions – mere bundles of propensities and aptitudes. The cardinal distinction between his view and that of most, other psychologists is therefore that mental faculties are distinguished by their contents.

Fodor takes this idea of a vertical faculty and brings it up to date in his concept of a ‘cognitive module’. Roughly speaking – and Fodor deliberately speaks only in this manner – a cognitive module is a mental faculty that specialises in a particular content domain. A good example of a cognitive module is accordingly one that takes the output of a sensory organ like the eye and translates it into a representation of some aspect of the world, which thereby becomes accessible to conscious thought. Thus there are ‘input’ modules for the perception of colour, shape, music, language and so on. Fodor argues that each module depends on innately determined and distinct neural structures in the brain, which are specially designed and wired up for carrying out the relevant task. He also argues that each module is ‘informationally encapsulated’ and ‘computationally autonomous’. These are technical notions of some importance – if only because they drive a wedge between Gall and Fodor, on one side, and many contemporary psychologists, on the other. In order to grasp them, however, we need to consider first the relevance of computation to psychological theories.

Fodor has played a major role in formulating the philosophical doctrine that mental phenomena are definable, not in terms of the underlying physical constitution of the brain (neurons, transmitter substances and electrochemical events), but in terms of the causal or functional structure of the operations that the brain carries out. The first modern formulation of this idea is due to the British psychologist Kenneth Craik, who wrote of the mind as analogous to a computational device in which what was crucial was the computations it carried out, not its physical make-up. Some years later, the invention of the digital computer provided a beautiful metaphor for the doctrine. The mind is like the software of a computer – the program that controls the computations that are carried out – whereas the brain is like the hardware that is ultimately responsible for the execution of the program.

The brain appears to consist of what are, in effect, a large number of different computers all whirring away together. And if Fodor is right, there is a separate computer for every vertical faculty, and each of these ‘cognitive modules’ is independent of the rest. A module operates like the simple eye-blink reflex to a looming object: you blink rapidly and involuntarily, and try as you will, you cannot prevent yourself blinking even when you know that there is no danger of a real poke in the eye. Hence, in Fodor’s terms, the operations of a module are ‘informationally encapsulated’ from beliefs, expectations or any other conscious attitude. Similarly, if Gall and Fodor are right – the latter with some reservations – then none of the modules makes use of common horizontal mechanisms, such as immediate or long-term memory. These claims are the most controversial in the whole book, since, as Fodor recognises, there are psychologists who have made their reputations defending the opposite point of view.

Not everything in Fodor’s architecture of the mind is modular. There is a non-modular system that takes the results of the various ‘input’ modules and cogitates about them. It is responsible for thought and judgment; and it relies on traditional horizontal faculties like memory. In a sudden access of pessimism, Fodor writes: ‘It may be that, from the point of view of practical research strategy, it is only the modular cognitive systems that we have any serious hope of understanding.’ This claim is strange because there are no obvious signs that the psychological understanding of language, allegedly a module, is any more advanced than that of thought and judgment.

Is Fodor’s architecture correct? And, if it is, why should there be specialised modules for ‘input’ (and perhaps ‘output’) but a non-modular system for thought? Why not specialised modules for thought too? It is easy to advance a case for some degree of specialisation in perceptual processes, and for the need for some general machinery which integrates all the resulting information, presumably what Aristotle intended by his idea of a ‘common sense’. It is rather harder, however, to establish decisive evidence in favour of Fodor’s detailed plans. On the one hand, damage to certain areas of the brain disrupts the memory for all types of immediate information. This evidence suggests that the ‘input’ systems may not be quite as modular as Fodor believes. He is happy to entertain the hypothesis that they may depend on common horizontal mechanisms of memory, but the trouble with this weakening of Gall’s position is that almost no possible evidence could refute it. On the other hand, studies of reasoning suggest that the content of an inference exerts a decisive effect on the form of the conclusions that are drawn. There may indeed be elements of modularity in the central processes of thought itself.

One of the great dangers of Nativism is the seductive power of supposing that an appeal to innate factors constitutes an explanation. Such invocations are really no more powerful than appeals to the great Architect in the sky, unless there is some testable way of explaining how the relevant processes evolved, and independent evidence for their existence. A perfect illustration of these dangers occurs in Fodor’s arguments about the computational basis of cognitive modules. Long before digital computers were first built, logicians had shown (in the theory of ‘recursive functions’) how any computable procedure could be constructed from a small set of elementary computational building-blocks combined according to a set of basic assembly rules. If, as Fodor and many psychologists suppose, the mind is essentially a computational device, then it follows that its computations could similarly be constructed from some elemental set of building-blocks and assembly rules. This set could be innate, and everything else could be acquired by a process of learning that consists in assembling new computations out of old.

Such a view, of course, is quite at odds with Fodor’s hypothesis that cognitive modules are separate ‘hard-wired’ computational devices: i.e. devices in which a special neuronal wiring diagram is laid down innately. Yet, as he admits, proponents of the two sorts of theory might agree to the letter on the nature of the computations carried out by a module and disagree solely about whether they are hard-wired into the module or assembled there from more primitive operations.

Why does Fodor object to the construction of computations out of elementary operations? The answer is, rightly, that the theory so far lacks constraints on the sorts of mental structure and computation that it can account for. The theory explains everything, and therefore nothing. Fodor contrasts the theory with the traditional idea of an association, which he claims at least had the constraint that mental events should mirror corresponding events in the world. Although I think that he has a point, and that computational psychologists need to worry about the constraints on their theories, the argument cuts both ways. A theorist who invokes the evolutionary hardwiring of a computational device is at present free of constraints. If a complex process is hard-wired, it almost certainly will require complicated wiring, not just a simple and unanalysable circuit. Yet Fodor asserts that ‘a natural interpretation of neural hard-wiring is that it packages into unanalysed operations what may be quite powerful primitive computational capacities.’ How this could be done seems more magical than natural. Just as computational psychology, as Fodor claims, needs a theory of learning, so his theory needs constraints on the possible wiring diagrams that evolution could produce.

The great virtues of Fodor’s book are its brevity, wit and intellectual bravura. Modern architects of the mind have recognised that there are aspects of its design that call for specialised processors. Fodor has taken this idea, hammered it into shape with a few phrenological bumps, and, like a Le Corbusier of cognitive psychology, pushed modularity almost as far as it will go. His numerous caveats reveal that he is quite aware that his modules have yet to be assembled into the final definitive plan.