Department of Psychiatry and Behavioral Sciences
University of Washington
Seattle, Washington 98195-1800 USA
wcalvin@u.washington.edu
http://faculty.washington.edu/wcalvin

In our book, Lingua ex Machina, Derek Bickerton and I offer three possible preadaptations for the big step up from protolanguage to syntax:

1. The cognitive categories needed for keeping track of "Who owes what to whom" (evolved for minimizing the freeloader problem in reciprocal altruism) are a nice setup for the agent-theme-goal role tags so useful for quickly communicating "Who did what to whom."
2. The prefrontal-premotor circuitry needed for planning novel throwing and hammering movements looks a lot like the binary trees of phrase structure.

The repeated payoffs of even more altruism, or even more accurate throwing, seem capable of adapting brain circuitry that could be occasionally borrowed for structured language.

The threshold for coherence provides a "capstone" candidate for fluent structured thought and talk – something that might have triggered the flowering of art and technology seen late in hominid evolution, after brain size itself had stopped growing.

Since corticocortical coherence is the one that builds atop the other two, and is the one that might have had a sudden emergence (the others look like gradual carryover so far, perhaps for millions of years), let me start by briefly explaining the first two pre-syntax "stones" of this arch so that the "capstone" role in enabling fluent syntax can be appreciated.

Bickerton’s preadaptation is from reciprocal altruism. What pays for the improvements in abstract cognitive categories for actor, recipient, and action are that they help keep track of who owes what to whom. This is not only handy for minimizing the cheater problem, telling the individuals possessing such cognitive traits when they ought to find a partner more likely to reciprocate occasionally, but it aids the furtherance of coalitions between nonrelatives.

Bickerton’s idea is that, once you have the cognitive categories for keeping track of "Who owes what to whom," you can start communicating "Who did what to whom" – and have a syntax-ready recipient with the preformed mental categories needed for disambiguating the longer utterances. You get the essentials of argument structure (categories for agent, goal, and theme) from the payoffs of ever more cooperation. You may further improve the underlying neuroanatomy with new payoffs from language itself, but the foundations came "free," paid for via the prior cooperation/altruism use.

Conversions of function are a familiar story in evolutionary theory (Darwin spoke of them before introducing his example of the fish swim bladder converting into an amphibian lung). And any conversion goes through an intermediate phase of multifunctionality.

Despite our mapmaking tendencies that lead us into erroneously assuming one region, one function (it’s another instance of the reification fallacy), cortical regions are notoriously multifunctional. While a region may have a specialization that is essential (you can’t do the function without it, as seen in the stroke and cortical surface stimulation results), the region may participate in other functions (as seen via its increase in blood flow when performing nonspecialist tasks). Even the common strokes illustrate the multifunctionality: most aphasic patients also suffer from hand-arm apraxia, suggesting a core of neural machinery for novel sequencing that is shared by novel hand-arm and oral-facial movements.

The other common way of disambiguating a long string of words is what, in pre-minimalist days, was called phrase structure. The nested embedding of phrases and clauses, one within another, is strongly reminiscent of the structured planning needed for planning multi-joint ballistic movements during "get set." The hand movement is embedded in the elbow rotation which is embedded in the shoulder rotation which is embedded in the body’s forward motion. Planning involves getting the whole thing right; there may be dozens of ways to hit the target, but they’re hidden in a sea of millions of wrong solutions, ones that would cause dinner to run away.

What pays for planning improvements is aimed throwing for hunting and precision hammering for toolmaking; when novel sequences of hand-arm movement are not being planned, such neural machinery may be available for structured planning of other things, such as long utterances or agendas. Again, once structured language pays for further improvements, it may in passing improve aimed throwing or precision hammering, to the extent that they continue to share neural machinery.

The important feature of ballistic planning is its use of cortex for novel sequences (i.e., not standardized target distances as in darts and basketball free-throws, where subcortical circuits probably take over) with demanding requirements for timing precision (small targets have very brief "launch windows" and so timing jitter must be minimized).

I recently refined Hebb’s 1949 notion of a cell-assembly whose spatiotemporal firing pattern (think of a short song) represents a concept, relationship, action – or a phrase or a clause. Even the complete sentence should have one, if it is to compete with alternatives. This firing pattern would be a "code" for the concept, but one would expect the code for, say, comb to be different in visual association cortex than near auditory or motor areas.

A uniform-across-the-cortex code would be powerful but, for the same reasons as it took Europeans so long to invent the Euro, it isn’t a default solution. The corticocortical connection from concept-laden temporal lobe to movement-schema-laden frontal lobe via the arcuate fasciculus (second only the corpus callosum in size) is surely jumbled (neighboring axons may not remain neighbors at the other end) and the fanout of connections guarantees blur. So a different code arrives, equally good in most respects (just like changing money when border crossing) but when it gets sent back to the originating cortex, it is doubly distorted. For frequently used concepts, of course, an identity relation can soon be learned.

But when novel messages are being sent around ("‘A square green tomato’ – anyone recognize this phrase?"), it runs into the same problem as trying to get moneychangers to recognize a wooden nickel. You can’t get the virtues of a universal code for novel on-the-fly concepts without the equivalent of a coherent corticocortical connection.

What I demonstrated in Chapter 7 of The Cerebral Code was that enough clones of the code in the sending cortex (think of a plainchant choir recruiting additional members from neighboring cortex, all singing the same song) would suffice to create a small choir singing the same song at the destination. (Such choirs are needed for reducing jitter and are also an outcome of a Darwinian copying competition for achieving quality on-the-fly.)

This allows novel codes to be passed back and forth without slowly learned identities. While handy for association tasks in general, the role of corticocortical coherence in nested embedding is where the common code really shines. It seems capable of making syntax an everyday, subconscious task that operates in seconds.

The "meaning of the sentence" is, in this model, an abstract cerebral code which competes for territory with codes for alternative interpretations, often in the manner of a Darwinian cloning competition. Phrases and clauses require coherent corticocortical links to contributing territories, having their own competitions and tendencies to die out if not reinforced by backprojecting codes.

It starts to look like a choral work of many voices, each singing a different tune but with the requirement that it mesh well with all the others. Indeed, the symphonic metaphor might be appropriate for the more complex sentences that we can generate and understand. Certainly the reverse-order analogy to Benjamin Britten’s Young Person’s Guide to the Orchestra, the all-together version being succeeded by the various voices playing separately, is the best metaphor I know for the read-out process that converts the parallel-structured plan into serial-ordered speech.

Consider the implications of efficiently linking the concept-filled temporal lobe with the prepare-for-action frontal lobe, with a common code replacing the degenerate codes – and then dropping back to the old system, with now-incoherent paths forcing a reliance on slowly established identities. Without coherence, you’d still have a vocabulary (the temporal lobe still works). You’d still be able to plan some nonlanguage actions (you’d pass many of the neuropsychological tests for frontal lobe functioning), but your ability to quickly invent new trial run associations would suffer.

Not only couldn’t you form up a syntactic sentence to speak (except for stock phrases), but you couldn’t judge sentences that you heard someone else speak because you could no longer judge the quality of your trial interpretations, whether they were nonsense, good guesses, or sure things. Your quality associations would be too slow for the windows of opportunity, and the results would be of poor quality because not shaped up very far by Darwinian copying competitions in the brain. And so your performance on language tasks would drop back to something like protolanguage, a wide choice of words but with novel sentences limited to just a few words to avoid ambiguity.

Carryover from reciprocal altruism’s cognitive categories and ballistic movement’s planning circuits are both compatible with slow language improvement over a few million years. But corticocortical coherence should have a threshold (the size of the plainchant choir, achieved mostly by temporarily borrowing neighboring cortex, not via brain size increases).

Once borrowing abilities cross the coherence threshold, structured thought and talk would have become far more fluent – and thus a capstone candidate for what triggered the flowering of art and technology seen late in hominid evolution, after brain size itself had stopped growing.

A proper lingua ex machina would be a language machine capable of nesting phrases and clauses inside one another, complete with evolutionary pedigree. Such circuitry for structured thought might also facilitate creative shaping up of quality (figuring out what to do with the leftovers in the refrigerator), contingency planning, procedural games, logic, and even music. And enhancing structured thought might give intelligence a big boost. Solve the cerebral circuitry for syntax, and you might solve them all. None of us are there yet, but this "three-stone archway" provides an illustration of how our big questions – what, how, and why – might hang together.

_____________

I thank Derek Bickerton for persistently steering me in this direction, and the Rockefeller and Mathers Foundations for provocative venues.

 Conference site: http://www.infres.enst.fr/confs/evolang/