Abstract: 54 words
Text: 1018 words
Reference: 18 words
Total: 1090 words.
AFFERENT ISN'T EFFERENT, AND LANGUAGE ISN'T LOGIC,
EITHER.
Derek Bickerton
Department of Linguistics, University of Hawaii, Honolulu, Hawaii 96822.
bickertond@prodigy.net
Abstract: Hurford's argument suffers from two major weaknesses. First,
his account of neural mechanisms suggests no place in the brain where the two halves of a predicate-argument structure could come together. Second, his assumption that language and cognition must be based on logic is neither necessary nor particularly plausible, and leads him to some unlikely conclusions.
Text:
Hurford is to be commended for attempting to root his analysis of linguistic primitives in the workings of the brain. But there are some serious problems with his approach.
He demonstrates interesting parallels between, on the one hand, the dorsal and ventral pathways that carry information about the 'where' and the 'what' of objects, and on the other, the two halves, predicate and variable, of simple propositions. But a parallelism, intriguing though it be, falls short of an explanation, or even a description.
The dorsal and ventral streams carry information about real-world objects along afferent fibers that lead to very different areas of the brain. A proposition of the form PREDICATE(x) consists not of raw information about external objects but of internal representations of objects and their properties. This already suggests that a lot is missing from Hurford's account. His abstract tells us that "PREDICATE(x) is a schematic representation of the brain's integration of the two processes of delivery by the senses" (emphasis added) but of course what the brain integrates is not the raw sensory feels delivered to it but the already stored abstract representations that it decides (correctly or otherwise) are a match for those feels.
Moreover, for these representations to be translated into anything like a predicate-argument structure, there must surely be some place in the brain for predicate and argument to come together. But on Hurford's account, there is nowhere for this to happen. One half of the predicate-argument equivalent occurs in the parietal cortex, the other half in the infero-temporal cortex. There would have to be efferent fibers from parietal to infero-temporal, or vice versa (or from both of these to some third place) if the two halves were to be integrated into either a thought or a sentence.
Hurford might want to argue that such considerations lie beyond his scope, that he merely wished to demonstrate the existence of a distinction necessary though not in itself sufficient for even the most basic processes of language or thought. However, one can go still further back into his argument and question whether PREDICATE(x) is as fundamental to either thought or language as he supposes.
His insistence on approaching language from a logical point of view begins to look questionable when he doubts whether proper names existed in earlier forms of language. This doubt is based on the belief that "control of a proper name in the logical sense requires Godlike omniscience". So it may be 'in the logical sense', but what were our remote ancestors most concerned about, getting their FOPL straight or telling one another interesting things? I doubt (contra Dunbar 1993) that language was born for gossip, but gossip was surely a major function of language, or even protolanguage, from quite early on. How can you gossip without names for the people you're talking about?
Hurford's supporting arguments are quite weak. It's immaterial whether animals with discriminatory skills less subtle than ours can be fooled into thinking a cardboard cutout is their mother: they have a clear concept of a specific individual, 'Mother'-it's concepts, not things in the world, that matter here--and if they wrongly identify that individual once in a while, so what? Situations where A has sex with B under the impression that he or she is having sex with C have been a staple of farce for millennia, so logically speaking Hurford should deny us too the right to have proper names. Tribes claimed still not to use proper names do of course use them. What determines whether something is a proper name is not its internal structure but how it is used. An expression like 'knocked the hut over' is (part of) a sentence in "Last night the wind [knocked the hut over]" but in the context of "[Knocked the hut over] seduced your wife last night" it's every bit as much a proper name as is 'Jim Hurford'.
Further doubts about the logic-language connection spring from the division, in Figure 1, of nouns into the classes 'Proper' ( said to be arguments) and 'Common' (said to be predicates), and are reinforced by the statement in Section 2.3.1 that "the vast majority of words in a language correspond to predicates". The vast majority of words are common nouns, which in the vast majority of sentences are arguments, not predicates; they occur as predicates only in sentences of a type seldom uttered by non-logicians ("Socrates is a man") that were probably rarer still in the dawn of language.
To assume, as Hurford seems to, that such predications form invisible but ineradicable sub-parts of normal sentences takes us back to the heyday of generative semantics, when to derive a simple sentence like "Floyd broke the glass" required the integration of a dozen or so clauses including things like "There is someone", "Someone is Floyd", "There is something" and "Something is a glass". But the fact that you can transform the simplest sentence into such components if you try hard enough has no necessary connection with how those sentences actually get constructed (as most syntacticians quickly realized).
A more plausible (and more parsimonious) scenario might go something like this. Information from the dorsal stream alerts the organism to the fact that something of potential interest or importance is out there. Thereafter, it plays no direct role in cognition or language. The ventral stream carries richer information to (more or less) where concepts are stored. A match is made, or not, as the case may be. If it is, the existence of something out there matching something in here is simply presupposed. (In logic you may have to assert your presuppositions, but that's no reason to assume that the brain has to do it that way). Thoughts, or sentences, can then be assembled using a handful of predicates-verbs, prepositions and the like-that take common or proper nouns indiscriminately as their arguments.
Whether some such scenario or Hurford's is nearer the truth is an empirical question. Hopefully, someone will be able to come up with ways to test them empirically. Till then, none is more than a hopeful hypothesis.
Reference.
Dunbar, R.I.M. (1993) Coevolution of neocortical size, group size and language in humans. Behavioral and Brain Sciences 16.681-735.