I have been thinking about my last post and Hubert Haiber’s argument that “Natural languages have the properties they have because they reflect the properties which our language-learning and language-using human brain capacities can cope with.” ( See An anthropic principle in lieu of a “Universal Grammar) This principle might seem self-evident, even circular, but it is a direct challenge to the concept of an innate, universal grammar. Indeed, Haiber is attacking Chomsky’s innate (a.k.a. nativist) principle directly: “Nobody has ever been able to produce immediate and compelling evidence in favour of the strong nativist hypothesis.” Disagreeing with Chomsky is hardly news, and I would probably let the paper pass by if it weren’t for a second feature, the metaphor of grammar as a virus. ("On the level of cognitive structures, grammars are self-reproductive in the same way as a virus…”) Viruses need a host to multiply and so, goes the metaphor, do grammars.
Only, Haiber insists he is not speaking metaphorically. “A Grammar is--even literally--a cognitive virus programme. It reproduces itself, but it needs a host that provides a replication environment, just as any virus does. Grammars ‘infect’ human brains as a result of language acquisition. The cognitive virus corresponding to the grammar of our mother tongue governs our language production behaviour.” But Haiber is confused. Is he talking about Grammar (as he specifies at the start of the quotation) or about language production behaviour (as he says at the end)? He thinks grammar can be a literal virus, so he starts talking about it, but at the end of the paragraph he describes language productions (utterances) going in and coming out. (“Children acquire their grammar on the basis of being exposed to language productions and they put it to use, Afterwards, their productions become part of the input for the next generation’s acquisition of grammar, and so on.”)
Haiber’s confusion over whether he is talking about language or grammar comes from his desire to be a reformer rather than a revolutionary. He hopes to replace the notion of an innate universal grammar with a learned mother tongue while leaving the rest of Chomsky in place. Sorry, professor. If utterances go in and utterances come out, it is utterances that are doing the evolving and, if that is so, Chomsky. who focuses exclusively on the non-utterances of internal language, collapses.
Is it so? Can we really talk about language as a virus? Yes, but only as a metaphor. Nonetheless, metaphors are extremely useful. You just have to be careful about where and when they apply and you can never argue ‘since X is part of the metaphor, X is part of language. Since Y is not part of the metaphor, Y cannot be part of language.’
One clear break in the metaphor is the process of replication. Viruses use a cell but replicate themselves. Utterances do not literally replicate themselves, yet they do change over time.
Viruses evolve via natural selection. Do utterances? Happily, the metaphor holds in this instance. We can say the brain that perceives the utterance is the equivalent of the cell that supports the virus. The brain’s cognitive functions include memory and perception. To be a candidate for selection, the brain must be able to perceive at least part of an utterance and remember it. Anybody who has tried to learn a new language has noticed how hard it is to perceive the details of the sounds rushing by the ear. “They speak so fast,” is a common complaint about any population of native speakers whose words overwhelm the newcomer. Perceiving an utterance is not a simple task. So we have at least two features that can play a role in selection. Can a speaker perceive an utterance and remember it?
There is also a feature of language that has no counterpart in a virus (or any other biological product for that matter): the topic. A conversation requires two or more people to pay attention to a shared topic. Without the topic and joint attention, language is impossible. This dramatic difference between language and evolving DNA strands explains many differences between a language and a virus. Viruses vary but only slightly, while language output varies greatly but not completely. We can produce never-before-uttered sentences, but the words are familiar. We can produce never-before-uttered words, but their syllables are familiar. We can produce never-before-uttered syllables, but their context had better be danged familiar. Utterances bear a family resemblance, suggesting they are part of a system, and utterances can be quite unusual, suggesting a creative or divine source for their content.
Haiber offers no hint as to how novel syntax structures (e,g,, phrases, clauses, sentences) emerge from utterances. I have been working for sometime on what I call attention-based syntax (see paper 1 and paper 2) and can therefore propose a way to use attention that has nothing to do with Haiber’s reliance on Chomskyan computations. Basically, a speaker uses language to direct attention from point to point, forming a gestalt (or whole). The task of the listener is to follow the utterance from point to point, and the task of the speaker is to shape the utterance so that it leads from point to point.
To sum up: using a language as a listener requires an ability to perceive an utterance in detail, remembering what is being said, and following the topic from point to point. Meanwhile, speaking requires an ability to form an utterance in detail while directing the topic from point to point. Selection of utterances thus depends on how much of an utterance a listener can perceive and how much can a speaker reproduce; how much of an utterance can a listener recognize and a speaker recall; how much of an utterance can a listener follow and how skilled is a speaker at directing a listener’s attention from point to point in a topic; and does a listener care enough to pay attention?
Is that all there is to language? It sounds too simple to credit. After all, the brain grew enormously over the past 2 million years. Was none of that an adaptation to language? I have thought so, but I may have been wrong.
Right now there are three competing theories: (1) language is entirely the product of cognitive operations (most famously supported by Chomsky, but there are many other versions of the idea); (2) language is the product of co-evolutionary adaptations by both the brain and language (most strikingly proposed by Terrence Deacon who has inspired many variations on his theme); and (3) language is entirely learned (most notoriously argued by B.F. Skinner who was beaten so badly that this very old idea has been recalled to life only by adopting radically different premises about how learning can proceed).
This blog has been pretty solidly on the side of theory #2, but now I want to consider #3 a bit. The idea of language as a kind of artificial virus that has been selected and replicated by the brain appeals to me, largely because I can see that, if true, the fantastic freedom of language is readily explained. People can speak with strange accents and disobey many of the formal rules insisted upon in school and still be understood. How is that possible? Theory 1 sees the brain as a rule-based set of behaviors, and is hard put to explain why we can follow a great deal of rule-violating speech. Computers certainly cannot stray from the rules. Theory number 2 is a bit less fixed, but still has trouble with speech’s rhetorical freedom. Theory 3, however, tosses aside rules (although there are habits), so the freedom is far less of a problem.
Thus, even with #3’s dubious simplicity, I think I will give it a closer look. I may find some undeniable road block, but it seems worth investigating. Next post, let’s take a look at language acquisition and the poverty of the stimulus. That was the issue that ruined Skinner and gave Chomsky staying power. So I might as well face that one head on.