What’s your favourite metaphor for minds? If you’re an empiricist, or an associationist, or a connectionist, you probably favour webs, networks, switchboards, or the sort of urban grid where the streets are equidistant and meet at right angles: New York’s Midtown, rather than its Greenwich Village. Such images suggest a kind of mind every part of which is a lot like every other. Take a sample here and you find the concept dog associated to the concepts cat and animal; take a sample there and you find the concepts Antony and Cleopatra in close connection. Though mental content changes as you go from place to place in the web, the structure is everywhere the same. It’s concepts and connections wherever you look.
But there is also an older, rationalist, tradition of theorising about the mind; one that runs from Plato through Gall, Kant and the faculty psychologists, to Freud and Chomsky. It, too, has its proprietary metaphors, which are frequently architectural. The mind is like a building (Steven Mithen thinks it’s like a cathedral). Entrance and egress are variously constrained, and so are the paths through the interior. There are public places and private places, and places where the children aren’t allowed to go; and there are places in the attic that hardly anyone but Granny ever visits. With that sort of arrangement, you might expect that information would be quite unevenly diffused among the occupants; downstairs knows things that upstairs doesn’t, and upstairs doesn’t know that downstairs knows them. Also, buildings are often divided into lots of little spaces, with more or less specialised functions. These differences of function generally imply differences of structure, so you wouldn’t expect to learn much about how the kitchen is organised by examining the potting shed. The long and short of the building metaphor is that when the whole mind functions as a system, it does so through the interaction of heterogeneous parts.
The empiricist idea of a homogeneous mind has dominated psychological and philosophical thinking in the English-speaking countries since the 18th century. But now the fashions seem to be changing and the ‘modular’ organisation of cognition is widely promoted. Just what this amounts to is very far from clear, and some of the consensus is doubtless just in terminology. Still, people in a lot of fields of research have recently come to think that some of our intellectual competences are isolated and specialised in ways that the web/grid/network story doesn’t easily accommodate. There is even considerable agreement on which cognitive capacities are the best candidates for the status of modules. The learning and use of language is on everybody’s list; Chomsky put it there, thus initiating the new fashion in cognitive architecture. Commonsense-biology, commonsense-physics, commonsense-psychology and aspects of visual form perception are other favourite candidates. The bidding is still open, but these five are the paradigms. If none of them is a mental module, then probably nothing is.
Isolation and specialisation are the defining properties of a module, so a lot of the current discussion is about what precisely these consist in. A word about this to set the stage for Mithen’s book.
Specialisation: modules are supposed to be ‘domain-specific’ cognitive mechanisms. To say, for example, that there is a commonsense-psychology module is to claim that each of us has concepts and processes available for inferring from the behaviour of conspecifics to their states of mind that are not likewise available for balancing the family chequebook or deciding which omnibus to take to Clapham. Other people are a very special part of the world; maybe we use very special kinds of cognition when we try to understand them.
For instance: Brentano remarked on how odd it is that you can want (need/seek/believe in) a gold mountain even though there aren’t any (whereas, try climbing a gold mountain if there aren’t any). Apparently, the commonsense-psychology module countenances relations between minds and things-that-don’t-exist. It may be that only phenomena in the domain of the commonsense-psychology module exhibit this logical peculiarity; philosophers think so who hold that ‘intentionality’ is the mark of the mental. In similar spirit, Chomsky thinks that the grammatical structures of all human languages are of the same formal type, and that no communication system which failed to exemplify the universal format could engage the operation of the language module. Are there likewise eccentric constraints on the kinds of geometry that the commonsense-physics module can contemplate? Does physical space have to strike us as Euclidean even if it’s not? Kant thought that it does.
What about the isolation of modules? Two kinds of idea are current. On the one hand, modules are said to be ‘informationally encapsulated’; there are supposed to be architectural constraints on the ways that information can flow among the modules (and between them and whatever unmodularised cognitive capacities the mind may have; see below). The standard illustration is the persistence of perceptual illusions even when you know that they’re illusory; you know perfectly well that the moon is no bigger when it’s on the horizon, but try convincing your visual perception module that it’s not. On the other hand, modules are supposed to have proprietary developmental careers. Language learning, for example, runs off largely independently from the development of visual perception. Blind children are not, in general, linguistically impaired; not even in their talk about space. Indeed, it’s part and parcel of the autonomy of the modules that their normal development appears to make quite minimal demands on the character of a child’s environment. Children don’t have to be taught their native language; they just pick it up. Nor, apparently, do they have to be taught to attribute people’s behaviour to the mental states they’re in. Children tend, quite naturally, to anthropomorphise whatever moves. What they have to learn is which things don’t have minds, not which things do.
Such considerations lead to the most spectacularly controversial thesis that modularity theorists endorse: that the modules are innate cognitive structures. They are, in Chomsky’s phrase, ‘mental organs’, and the autonomy of their maturation is no more surprising than the maturational autonomy of, as it might be, legs from digestive systems. You wouldn’t expect fine-grained developmental measures to apply indifferently to knees and stomachs, nor would you expect their development to depend on special tuition. Rather, their maturation is internally driven and largely mutually independent because it is genetically preprogrammed. So, too, with mental modules; or so the story goes.
You might think that cognitive psychologists would reject this theory out of hand for fear of becoming unemployed. If ‘cognitive development’ is just a cover term for the independent ontogenetics of a heterogeneous bundle of mental organs, what general laws of cognitive development could there be for psychologists to discover? But on second thought, that reaction is precipitate. It is, after all, not plausible that the mind could be made only of modules; one does sometimes manage to balance one’s chequebook, and there can’t be an innate, specialised intelligence for doing that. Maybe cognitive development is the story of how the child somehow gets from isolated, genetically preprogrammed, domain-specific cognitive modules to the relatively unspecialised kind of intelligence that doing one’s chequebook requires.
So, finally, to Mithen, who has been reading a lot of psychologists who think that the child’s cognitive development proceeds from a modularised mind to one that’s ‘cognitively fluid’ (Mithen’s term). It’s Mithen’s daring and original suggestion that maybe this sort of ontogeny recapitulates human phylogeny. Maybe the evolution from pre-modern man to early man, to us, is the story of the emergence of cognitive fluidity from (or rather, alongside of) the modularised components of our minds. ‘The differences ... are analogous to those between Romanesque and the succeeding Gothic cathedrals ... In Gothic architecture sound and light can flow freely ... unimpeded by the thick walls and low vaults one finds in Romanesque architecture ... Similarly, in the phase 3 mental architecture, thoughts and knowledge generated by specialised intelligences can now flow freely around the mind.’
Well, it’s a fine idea, and a fine book. At a minimum, Mithen’s easy style is accessible to the casual reader who is looking for a overview, but his book is also chock-full of useful archaeological and anthropological detail. A word to professionals: do not skip the footnotes. They contain stuff that will fascinate and edify the working cognitive scientist. Mithen seduced me from the Anthony Powell that I was reading; I can’t imagine warmer commendation.
Yes, but is Mithen’s theory true? For that matter, how well does it fit the ontogenetic theory that he uses as a model? And, for another matter, is the ontogenetic theory that he uses as a model itself true? Mithen is three levels deep in speculation: he needs there to be mental modules; he needs the ontogeny of cognition to consist of the maturation of these modules and, eventually, of the emergence of cognitive fluidity from them; and he needs a phylogeny of cognition that recapitulates this ontogeny. It’s a lot to ask for, and I’m not convinced that it works.
By his own estimate, Mithen has two major epochs in the prehistory of our minds to account for. There’s the transition, ten million years or so ago, from ‘the ancestral ape’ (the presumed last common ancestor of us and chimpanzees) to ‘early man’; and there’s the transition, 100,000 years or so ago, from early man to ‘modern man’ (in effect, to us). This taxonomy is itself not untendentious. But Mithen finds evidence, from a variety of archaeological and comparative sources, that each of these major transitions implied a comprehensive increase in cognitive capacity; so comprehensive as to suggest a corresponding reorganisation of the underlying cognitive architectures.
What sort of reorganisation? Here’s where Mithen wants phylogeny to borrow from ontogeny. The ancestral ape had a mind that ran on ‘general intelligence’ together with, at most, a rudimentary modularised ‘social intelligence’; whereas, in the next stage, the mind of early man had modularised ‘technical intelligence’ and ‘natural history intelligence’ as well. Finally, the emergence of cognitive fluidity from a relatively compartmentalised, modular mind is what Mithen thinks distinguishes early man from us. It’s this latter difference that he likens to the change from Romanesque to Gothic in the passage I quoted.
I do admire the attempt to find places for so many pieces. But it takes some forcing to make them fit, and what Mithen ends up with shows the signs of strain. For example: if cognitive ontogeny is indeed the development of a fluid mind from a modular mind, then it offers a clear precedent for the hypothesised shift from ‘early’ to ‘modern’ intelligence, much as Mithen says. But what, exactly, is it in cognitive development that’s supposed to correspond to the emergence of early man’s modular mind from the rudimentary general intelligence of the ancestral ape? Most cognitive psychologists who are seriously keen on cognitive modules think that they are what children start with; certainly, that’s the natural view if you think that a modular intelligence is likely to be innate. And presumably it has to be innate if, as Mithen supposes, modularity is a product of the evolution of cognitive architecture.
Mithen relies heavily on some views of Annette Karmiloff-Smith, who does indeed envision three phases of ontogeny, including both the assembly of rudimentary, task-specific capacities into modules, and also transitions from modular intelligence to something that’s supposed to be more fluid. But, precisely because she doesn’t think modular architecture is genetically programmed, Karmiloff-Smith isn’t, by any means, a paradigm modularity theorist. In effect, Mithen is trying so hard to construct an ontogenetic theory to ground phylogeny, that he finds himself agreeing with people who don’t really agree with one another; or with him.
Still, Karmiloff-Smith might be right that ontogeny doesn’t actually start with modules; or Mithen might be right about the prehistory of the mind even if the psychologists are wrong about its development. More worrying than whether Mithen can get phylogeny to run with ontogeny are the internal tensions in his account of the latter.
For example, for his phylogenetic story to work, modules have to be smarter than generalised intelligence in order to account for the transition from the ancestral ape to early man; and cognitive fluidity has to be smarter than modules in order to account for the transition from early man to us. But, in fact, it seems that Mithen takes ‘generalised intelligence’ and ‘cognitive fluidity’ to be much the same sort of things; indeed, he sometimes describes the phylogeny of mind as an ‘oscillation’ from these to specialised intelligence and back again. But that can’t be right; the same cause can’t explain opposite effects. There must be something very different between, on the one hand, the kind of general intelligence that modular minds are supposed to have supplanted, and, on the other hand, the kind of fluid cognition that Mithen thinks allowed us to supplant Neanderthals. Mithen doesn’t say what this ingredient X might be. Nor does he say how it could be that, whereas modules emerge from general intelligence in the first transition, general intelligence emerges from modules in the second. Prima facie, one would think ‘emerging from’ ought to be a one-way street.
The trouble – and it’s not just Mithen’s trouble; it’s a scandal in the whole cognitive science literature – is that nobody knows what ‘general intelligence’ is or what ‘emerging from’ is, either. For example, there’s a strong inclination to say that minds that are specialised must somehow evolve from minds that are less so; that sounds like good Darwinism. And there’s also a strong inclination to say that it’s because our intelligence is generalised and theirs is not that we’re so much cleverer than other species. But, to put it mildly, it’s not clear that you can have it both ways. I suspect that cognitive scientists want ‘general intelligence’ and ‘emerging from’ to paper over this crack, but that no coherent concepts could. No wonder if nobody can figure out what ‘general intelligence’ and ‘emerging from’ are.
Mithen does suggest that the transition from modularity to fluidity is somehow mediated by the use of language. Here again, he is explicitly echoing a literature in cognitive development according to which the transition from modular to generalised intelligence in children is linguistically mediated. But neither Mithen nor anybody else has been able to make clear how language manages to do the trick. The problem is that, if specialised concepts are the only ones that you have, words that express specialised concepts are presumably the only ones that you can learn. In effect, the linguistic mediation story replaces the question ‘how do you get from specialised concepts to generalised concepts’ by the question ‘how do you get from specialised concepts to words for generalised concepts.’ There’s nothing to indicate that the second question is easier to answer than the first.
Well, any number can play this game. Here’s my guess, for what it’s worth, at how the pieces fit together. The phylogenetically primitive form of mind is not ‘generalised intelligence’ but the reflex; i.e. the more or less instantaneous, unmediated elicitation of an organic response by an impinging environmental stimulus. Two things have happened to reflexes in the course of the phylogeny of mind, both tending to decouple the stimulus from the response.
The first is the evolution of complicated, inferential processes of stimulus analysis, which parse an organism’s world into a wide variety of behaviourally relevant categories; roughly, the higher the organism the more such categories its perceptual repertoire acknowledges. Reflexes respond to mechanical forces and to chemicals, light, gravity and such, so unless an opportunity for behaviour is signalled by these sorts of stimuli, a mind that’s made of reflexes is ipso facto blind to it. By contrast, our kinds of minds envision a world of objects, persons and relations among them, on which the success of our behaviour depends; so it’s in these categories that we want perception to articulate the world that we act on. Phylogeny obliged, and apparently it started to do so early. Mechanisms for high-level perceptual analysis are much older than anything that could reasonably be called even rudimentary generalised intelligence. For example, insects use them when they navigate. Wherever they turn up, they exhibit the characteristic properties of modules: computational isolation, specialisation and genetic specificity. We probably have half a dozen or so of these modularised perceptual analysers, as I remarked above.
Modules function to present the world to thought under descriptions that are germane to the success of behaviour. But, of course, it’s really the thinking that makes our minds special. Somehow, given an appropriately parsed perceptual environment, we manage to figure out what’s going on in it and what we ought to do about it. It’s a hobbyhorse of mine that the thinking that’s involved in this ‘figuring out’ is really a quite different kind of mental process from the stimulus analysis that modules perform; and that it is deeply mysterious. Suffice it that Mithen’s model of phylogeny – that language somehow mediated a seepage of information from module to module – surely underestimates the difference between specialised intelligence and thought. Even if early man had modules for ‘natural intelligence’ and ‘technical intelligence’, he couldn’t have become modern just by adding what he knew about fires to what he knew about cows. The trick is in thinking out what happens when you put the two together; you get steak au poivre by integrating knowledge bases, not by merely summing them. But ‘integration’ is just a word, and we haven’t a smidgen of a theory of what it is or how it happens.
On my view, the phylogeny of cognition is the interpolation, first of modularised stimulus analysis and then of the mechanisms of thought, between the stimuli and responses of aboriginal reflexive behaviour. If that’s right, then perhaps cognitive phylogeny hasn’t much to learn from cognitive ontogeny after all. (Piaget taught that infants start out as sensory-motor reflex machines and develop into formal intelligence, but practically nobody outside Geneva believes that any more.) Quite possibly, the basic architecture of perception and thought are in place at birth, and what mostly happens to the child’s mind in development is what also mostly happens to its body: both get bigger.
I present this sketch as one kind of alternative to Mithen’s picture. But, in fact, I wouldn’t bet on either. The synthesis of cognitive ontogeny with cognitive phylogeny is certainly premature; if somebody were to tell us the right story right now, we probably wouldn’t understand it. But then, synthesis has to be premature if it’s to be of any use to guide research. We’re not very good at thinking about minds; we need all the help we can get. Mithen has made a convincing case that ‘if you wish to know about the mind, do not ask only psychologists and philosophers: make sure you also ask an archaeologist.’ And he’s drawn a first map of what might emerge from putting all their answers together. No doubt, we’ll have to do the geography again and again before we get it right; but Mithen has made a start, and we are all in his debt.