The generation of generativity: a response to Bloom

The generation of generativity: a response to Bloom

191 Cognition, 51 (1994) 191-198 OOlO-0277/94/$07.000 1994- Elsevier Science B.V. All rights reserved. Discussion The generation to Bloom of genera...

583KB Sizes 0 Downloads 20 Views

191

Cognition, 51 (1994) 191-198 OOlO-0277/94/$07.000 1994- Elsevier Science B.V. All rights reserved.

Discussion The generation to Bloom

of generativity:

a response

Michael C. Corballis Department

of Psychology,

University of Auckland,

Private Bag 92019, Auckland,

New Zealand

Recapitulation

Among the human activities that exhibit recursive generativity, or what Chomsky (1988) termed “discrete infinity”, language seems to demand priority. Unlike mathematics, manufacture, or music, language is truly universal, and it is also much more complex than tool use, for example (Wynn, 1991). Speech is also effortlessly generative, whereas these other generative activities seem to require more deliberate planning. It therefore seemed to me reasonable to suppose that, in evolution, generativity was manifest in language before it was manifest in other activities, although I shall suggest later that it may already have characterized human thought. The idea that all contemporary humans derive from an expansion out of Africa between 100,000 and 60,000 years ago continues to receive support (CavalliSforza, Menozzi, & Piazza, 1993). The brain had already reached its present size before this expansion. Cavalli-Sforza et al. suggest also that human language had probably already evolved to its present level of sophistication, and speculate further that language was a major support of the expansion. A recent, large-scale survey of the diversity of existing languages provides no evidence that linguistic universals have changed in the last 60,000 years (Nichols, 1992). By contrast, manufacturing technology varies enormously between cultures, and has manifestly evolved, even in our own time. Klein (1990) claims to have identified, in the Middle to Upper Palaeolithic transition (between about 47,000 and 35,000 years ago), the emergence of a stone tool technology with the same recursive, phrasestructure grammar as language. The earliest evidence for music, in the form of primitive instruments such as bone finger flutes or jaw-bone rattles, comes from the Aurignacian period of some 34,000-30,000 years ago, also remarkable for the SSDI 0010-0277(93)00586-V

192

M.C.

Corballis I Cognition 51 (1994) 191-198

emergence of body 1992). My inference Given dramatic

ornamentation and carved figurines (Bradshaw is that language preceded these developments.

& Rogers,

these broad premises, it seemed to me to make sense to suppose that the expansion of manual technologies may have resulted from a switch from

manual to vocal language, so that the hands were freed (Corballis, 1992). This would not only have allowed for more flexible communication, but would have enabled speech and manual activities craft, for example, might demonstrate

to occur in parallel. A teacher of a manual and explain at the same time. If language

did evolve first as a system of manual gestures, then gradually, along with the increase in brain size, from

it might have the emergence

developed of Homo

habilis some 2 million years ago. The switch to a predominantly vocal language, and the ensuing advance in manual technology, may have been precisely what distinguished H. sapiens sapiens from other hominids, such as those who had earlier migrated from Africa into various parts of the Old Word.

Capitulation The main point I wanted to make, then, was that the late (and continuing) explosion of technology might have been, initially at least, a result of the freeing of the hands. However, I also raised the further question of “whether generativity could have been somehow extracted from language and applied to other activities” Corballis (1992, p, 218, italics added). I went on to express some reservations about this possibility, and these have now been sharpened by Bloom (1994). Generativity, or more particularly discrete infinity, is evidently not a commodity that is easily extracted, like juice from a lemon, and sprinkled on other activities to make them generative too. Bloom entertains some alternative possibilities, notably a developmental scenario in which children come to understand the discrete infinity of the number system only by mapping their non-linguistic concepts of numerosities onto words. Discrete infinity is then supplied by language itself. One line of evidence against this proposal comes from so-called idiot savants who sometimes display prodigious feats of arithmetic despite poor linguistic and general intellectual ability (Howe, 1989). As does Bloom, I also considered the possibility that discrete infinity evolved as a more general property of human thought, and was exploited by language: If discrete infinity was not tied specifically to language, then, it may nevertheless have been first manifest in language-or perhaps first manifest in thought itself but expressed in language. My thesis that generativity in manufacture was the eventual outcome of a switch from manual to vocal language need not imply that generativity was somehow extricated from language and exapted for other domains. Rather, the freeing of the hands from language functions may simply have allowed discrete infinity a wider range of expression, leading to a new-found generativity in

M.C.

manufacture, (Corballis,

and perhaps

to the development

Corballis

of other

I Cognition

“open-ended”

193

51 (lYY4) 191-198

forms such as art and music.

1992, p. 219)l

But if generativity evolved as a general property of human thought which can be duplicated in various parts of the brain, including those connected with language, evolve ? Bloom suggests that the capacity for in what context did it originally symbol

manipulation

may have arisen

or spatial cognition. But generativity must somehow applied to language.

Generativity

initially

in logical inference,

motor

control,

here we are back to the original difficulty be extracted or duplicated from these activities

that and

and brain size

A more general (if vague) solution is to appeal to some property of the human brain. Chomsky (e.g., 1972, 1980, 1988) has many times written in this vein. Once the brain reached a certain level of complexity, he suggests, the “conceptual apparatus would be freed for the construction of new thoughts and operations such as inferences involving them, and it would be possible to express and interchange these thoughts” (Chomsky, 1988, p. 170). The capacity for symbol manipulation may therefore have arisen as a consequence of the increase in the size and complexity of the brain that took place between the emergence of H. habilis some 2 million years ago and H. sapiens about 300,000 years ago. This is not to say that language itself is simply a by-product of increased brain size, as implied by Gould (1979), for example, who proposed that the enlarged human brain constitutes a general-purpose computer that might be programmed for language, as well as for other “computational” activities. Such arguments fly in the face of arguments for the “modular” structure of human thought, including specific language modules (Fodor, 1983); Pinker and Bloom (1990) similarly point out that the rules for language are highly specialized, and are often at crosspurposes with those involved in other kinds of knowledge. Is it possible, however, that generativity itself emerged with increased brain size, not as a by-product, but as the primary basis of selection? That is, larger brains may have been progressively selected because they permitted more flexible thought, and language may have evolved as a subsystem, with its own idiosyncratic features, within this more general development.

‘Although I was unaware of his identity at the time it was Paul Bloom, in his role as reviewer, who questioned the idea that generativity could be extracted from language, and who pushed me to this conclusion.

194

M.C.

Corballis

I Cognition

51 (1994) 191-198

The role of development Besides the sheer increase in brain size, another critical ingredient in the development of hierarchical structures necessary for recursion may be the prolongation of childhood, allowing the environment to influence brain structure in progressive fashion. Elman (1991) has begun to explore the power of this principle within a connectionist framework. Networks that failed to learn a complex grammar when both the network and the input remained constant succeeded in doing so when the input was presented incrementally, or when the memory capacity of the network was progressively increased. This idea need not be restricted to the learning of language. Elman concludes as follows: Higher primates are distinguished by their extended period of development, and by their relatively diminished capacities at early stages of development. Ultimately, of course, the higher primates achieve a high level of cognitive functioning and exert extraordinary control over their environment. But that environment is a complex domain, and not easily mastered. It may well be that the early developmental limitations play a key role in learning. The so-called critical period may be critical not by virtue of special capacities which are present at childhood and magically and lamentably lost at puberty. Rather, the critical period may be special because the later abilities which are found in adults are not fully developed. The simpler view of the world which this affords makes learning tractable. (p. 10)

there is a marked discontinuiDespite Elman’s reference to “higher primates”, ty between humans and the other apes in both brain size and the extent of postnatal cerebral growth (Passingham, 1982). It may only be in humans, then, that symbol manipulation has evolved to the point that permits discrete infinity. Notwithstanding Elman’s connectionist modelling, there may still be objections, based on the “poverty of the stimulus” (Chomsky, 1980; Crain, 1991), to the idea that language is wholly structured by the linguistic input, even allowing for the role of growth. There may be other aspects of neural structure, perhaps associated with the left hemisphere, that contribute to the ultimate shape of language. However, the progressive growth of brain structures, with environmental shaping of connections, may be what is required to build up the multilayered hierarchical structures for recursive, generative thought per se, whether in the context of language or of other skills, such as manufacture, music, or object representation. The parallel development of different circuits might permit each while preserving a generative and possibly function its own idiosyncrasies, recursive structure to all functions established in this way. Similar ideas were developed by Greenfield (1991), who suggests that the same structures in the vicinity of Broca’s area serve both language and object combination in children up to the age of about 2, but that with the further growth of connections involving more anterior regions in the frontal lobes, the structures involved in the language and object combination are progressively differentiated. It is perhaps worth reiterating that the left-hemispheric growth spurt in the vicinity of Broca’s area between the ages of 2 and 4, documented by Greenfield,

M.C.

coincides with the period emerges (see also Corballis,

Corballis

I Cognition

during which morphologically 1992), and this is also a period

51 (1994) 191-198

195

complex language in which children are

likely to play with construction toys. Bloom (1994) objects that there is no evidence for a correlation in development between language and object combination once factors such as age and memory load are factored out. However, this correlation is not critical to the more general argument that hierarchical and perhaps

ultimately

recursive

structures

are built up progressively

in the develop-

ment of the brain. Bloom (1991, 1994) also objects to Greenfield’s theory on the grounds that children understand language at a sophisticated level well before they can use it at that level. It may be that there are similar processes in posterior cortical regions that occur earlier in development. The same principle may apply to domains unconnected with language or the manipulation of objects. For example, there is evidence that the development of cells in the visual cortex selective to different visual features depends on specific visual input and on critical periods of growth (Wiesel, 1982). The staging of critical periods may depend on genetically programmed production of NMDA (N-methyl-D-aspartate) receptors, which act as a “gate” for controlling alterations to synaptic efficiency (Kalil, 1989). The progressive elaboration of visually modulated structures beyond the visual cortex may conceivably underlie higher-order units, such as the “geons” postulated by Biederman (1987) to account for human image understanding. Geons are combined to form representations of objects, which may in turn be combined to form representations of scenes. Again, this hierarchical, multilevel organization may be sufficiently complex to permit discrete infinity - and it may be uniquely human. It is unlikely that these developments took place in an evolutionary vacuum, so that generative activities were simply by-products. Rather, progressive increases in brain size were presumably selected precisely because they permitted more flexible thought. Despite my arguments for the priority of language, it must be considered unlikely that language evolved as a uniquely generative system, full of sound and fury but signifying nothing. That is, generative language presumably evolved for the expression of generative thought. Indeed, rather than supposing that one preceded the other, it might be supposed that they evolved together.

Generativity

and episodic memory

One ingredient of thought that has generative properties and that may be closely linked to language is episodic memory. Episodes comprise combinations and recombinations of elements, such as who did what to whom, and when, and how, and where, and why. Tulving (1983) has speculated that episodic memory is

196

M.C.

Corballis

I Cognition

51 (1994) 191-1%

unique to humans,* and although it has been objected that other species delayed matching-to-sample problems that require access to information

solve pre-

sented at an earlier point in time (Olton, 1984), the essence of episodic memory seems to require more specification than this. As Tulving (1984) points out, there is no evidence that memory for the past in non-human animals rather than merely causal, to borrow Dretske’s (1982) distinction.

is informational Indeed it seems

to me virtually impossible to test for episodic memory, in the informational sense, without using language, which contains precisely the sorts of structures required (Pinker episodic

& Bloom, 1990).3 While this could mean that other species do have memory but we could never find out about it, a more parsimonious

conclusion is that language and episodic memory evolved hand in hand, as it were, in humans, and that the usefulness of episodic memory may derive at least in part from the fact that it can be communicated. People can then learn from the experiences of others. But the generativity of episodic thought clearly extends beyond memory. We can invent episodes, and it is often difficult to disentangle real memories from those that derive from the words of others (Loftus, 1993; Loftus & Ketcham, 1991). We are compulsive tellers of tales, whether by mouth, through novels and movies, or in our own fantasies-even in our dreams. The generation of possible episodes also plays a role in the planning of future events, or in rehearsing different ways to deal with anticipated events. Donald (1991) argues that the final stage in human evolution was the development of external symbol storage. Speech provides one such system in that events and commentaries on them can be preserved in cultures by the telling of stories. But the more lasting and accurate of external storage devices (at least prior to magnetic storage) were visuographic, and these may again have depended on the switch from manual to vocal language, freeing the hands for the and writing. The importance of external development of drawing, sculpture, symbol storage in modern human life can scarcely be underestimated.

Conclusion The main point I wish to reaffirm is that a switch from manual to vocal language could help explain the relatively late flowering of manually based ‘Donald (1991, p, 149) seems to make a similar suggestion when he writes that the lives of apes “are lived entirely in the present, as a series of concrete episodes”. He goes on to argue, however, that they do have episodic memory, and indeed describes their culture as “episodic culture”. Only humans, he suggests, have gone beyond episodic memory to semantic memory, and so developed a culture that is not bound by situational specificity. This implies that semantic memory evolved more recently than episodic memory, whereas Tulving (1983) argued the opposite. 31t does not follow that episodic memory itself is linguistic. Pinker and Bloom (1990) make it clear that, although language may be optimal for the communication of internal knowledge, it is “hopeless” for the actual representation of that knowledge (see also Fodor, 1975).

M.C.

Corballis I Cognition 51 (1994) 191-198

197

technologies in human evolution. It might also explain why H. sapiens sapiens had an adaptive edge over other hominids, such as the Neanderthals, and indeed why H. sapiens sapiens has been so successful as an expansionist (even reaching the moon). The generativity of manual (and other activities) need not have derived from language, however; already there in episodic new

ways

in which

rather, generativity may have been a capacity that was thought, and the freeing of the hands simply provided

it could

be exploited

and expanded.

In my earlier

article

(Corballis, 1992) I discussed the role of increased brain size and the prolongation of childhood in the evolution of language; here, I suggest that these developments may have played a more general role in the evolution of hierarchical, generative thought. I realize that my additional speculations may arouse further controversy. In 1866, the Societe de Linguistique de Paris banned all presentations on the origins of language, and perhaps the time has come to reaffirm that ban. After this note is published, of course.

References Biederman, I. (1987). Recognition-by-components: a theory of human image understanding. Psychological Review, 94, 115-147. Bloom, P. (1991). What does language acquisition tell us about language evolution? Behavioral and Brain Sciences, 14, 553-554. Bloom, P. (1994). Generativity within language and other cognitive domains. Cognition, 51, 177-189. Bradshaw, J.L., & Rogers, L.J. (1992). The evolution of lateral asymmetries, language, tool use, and intellect. New York: Academic Press. Cavalli-Sforza, L.L., Menozzi, P., & Piazza, A. (1993). Demic expansions and human evolution. Science, 259, 639-646. Chomsky, N. (1972). Language and the mind. New York: Harcourt, Brace & World. Chomsky, N. (1980). On cognitive structures and their development: a reply to Piaget. In M. Piattelli-Palmarini (Ed.), Language and learning: The debate between Jean Piaget and Noam Chomsky (pp. 35-52). Cambridge, MA: Harvard University Press. Chomsky, N. (1988). Language and the problem of knowledge: The Managua lectures. Cambridge, MA: MIT Press. Corballis, M.C. (1992). On the evolution of language and generativity. Cognition, 44, 197-226. Crain, S. (1991). Language acquisition in the absence of experience. Behavioral and Brain Sciences, 14, 597-650. Donald, M. (1991). Origins of the modern mind: Three stages in the evolution of culture and cognition. Cambridge, MA: Harvard University Press. Dretske, F. (1982). The informational character of representations. Behavioral and Brain Sciences, 5, 376-377. Elman, J.L. (1991). Incremental learning, or the importance of starting small. Center for Research in Language Technical Report No. 9101. University of California, San Diego. Fodor, J.A. (1975). The language of thought. New York: Thomas Crowell. Fodor, J. (1983). Modularity of mind. Cambridge, MA: MIT Press. Gould, S.J. (1979). Panselectionist pitfalls in Parker and Gibson’s model of the evolution of intelligence. Behavioral and Brain Sciences, 2, 385-386. Greenfield, P.M. (1991). Language, tools, and the brain: the ontogeny and phylogeny of hierarchically organized sequential behavior. Behavioral and Brain Sciences, 14, 531-595.

198

M.C.

Corballis I Cognition 51 (1994) 191-198

Howe, M.J.A. (1989). Fragments of genius. London: Routledge. Kalil, R.E. (1989). Synapse formation in the developing brain. Scientijk American, 261(6), 76-87. Klein, S. (1990). Human cognitive changes at the Middle to Upper Palaeolithic transition: the evidence of Boker Tachtit. In P. Mellars (Ed.), The emergence of modern humans. Ithaca, NY: Cornell University Press. Loftus, E.F. (1993). The reality of repressed memories. American Psychologist, 48, 518-537. Loftus, E.F., & Ketcham, K. (1991). Witness for the defence. New York: St. Martin’s Press. Nichols, J. (1992). Linguistic diversity in space and time. Chicago: University of Chicago Press. Olton, D.S. (1984). Comparative analysis of episodic memory. Behavioral and Brain Sciences, 7, 250-251. Passingham, R.E. (1982). The human primate. San Francisco, CA: W.H. Freeman. Pinker, S., & Bloom, P. (1990). Natural language and natural selection. Behavioral and Brain Sciences, 13, 707-784. Tulving, E. (1983). Elements of episodic memory. Oxford: Clarendon Press/Oxford University Press. Tulving, E. (1984). Precis of Elements of episodic memory. Behavioral and Brain Sciences, 7, 223-268. Wiesel, T.N. (1982). Postnatal development of the visual cortex and the influence of the environment. Nature, 299, 583-591. Wynn, T. (1991). The comparative simplicity of tool-use and its implications for human evolution. Behavioral and Brain Sciences, 14, 576-577.