Knowing without learning

We know how to look at and listen to our nearest and dearest, how to sit up, crawl, walk, talk, not by instinct, but between the gut and the bones, in some way for which there is no clear expression. Plato recognised the problem here two and half thousand years ago. There is a word missing.

One of the most primitive abilities of living organisms is to distinguish the source of light – usually from on top. The most primitive organisms benefitted by being able to orient one side of themselves towards the sun and one side away from it, usually to hide from prey or predators or to get as much benefit as possible from the heat of the sun.

We may try to capture this sort of deeply embedded knowledge or perception as ‘knowing in one’s gut’ or ‘knowing in one’s bones’, perhaps trying to capture a sense that there are levels of depth in knowledge. Such a sense is not inappropriate. Most of human knowledge is defined within the cerebral cortex, the outermost aspect of the brain with its deep folds and nine billion or so neurones, each with five thousand or so synapses with other neurones. But there other significant functionalities which are more deeply buried within the brain, like the limbic system, allowing a fight or flight decision, and a system giving us the ability to navigate in some unfamiliar environment. I know about the new navigation ability having almost completely lost mine as a consequence of an almost-fatal heart attack some years ago.

There are instincts. And there are reflexes. Exactly what these are, what the differences are between them, what functionalities they control, even whether there is a difference between them, are all matters of dispute.

There is an especially hot dispute about whether human speech and language have any place in this schema. I personally believe that both speech and language would not be as they are without a set of one or more well-defined cognitive properties allowing them to be learnt and used they way they plainly are. This cognitive property is commonly known a ‘Universal Grammar’ or UG. But this idea is regarded as quite absurdly wrong, often with some derision, by those like Ben Ambrose trained in some version of what I call ‘radical sociology’, radical because it sees little or no interest in any ideas from the time of the so-called enlightenment.

Ambrose (2004) makes a number of predictions from UG as he understands it, shows them to be inexplicable by UG, but better explained by a model due mainly to Michael Tomasello (2003 and 2010) by which human language is defined exclusively by culture. Ambrose is particularly hostile to what he sees as the algebraic quality of UG.

The error in Ambrose’s reasoning, it seems to me, is that from any theory or model of the natural world, there may be any number of predictions. The predictions of Einstein’s theories of Special and General Relativity were not obvious at the time he first published them. It was Eddington who was one of the first to see the significance of Einstein’s work who saw the prediction that light would be bent around a massive entity such as the sun. When he showed by a clever and well-designed experiment that light was indeed bent around the sun, astronomers, physicists and mathematicians listened afresh. But this did not make Einstein’s theories true. As Einstein was the first to recognise, his theories could, in principle, be proved wrong, for instance by showing that there was no such thing as space-time. Other predictions followed. So far, the predictions have been shown to be correct.

But for any theory or model, derivable predictions may point in more than one direction. The notion of UG makes other predictions which Ambrose passes over. For instance, UG predicts that no language will ever be discovered that does not distinguish between functors like a and the and lexical items like dark and matter. Many artificial languages, designed for various purposes, make no such distinction. It would be easy to design an artificial, supposedly human-type language, to be read strictly from left to right, like a language for data-mining, modeled on the language of human two-year olds, without a single functor. Such a language would seem simpler than a language with different sorts of words. But no such language has ever been discovered. And that seems to me significant. It raises the obvious question: Why should this be?

As a first step towards answering that question, there are, I think, two strong, empirical sorts of reason for thinking that modern human speech and language must be organised in ways that go beyond culture and the powerful demands of communication. One sort of reason is from the evidence, elegantly set out by Stephen Anderson (2004), of how human language is uniquely different from any sort of non-human language or communication, call it what you will. The other sort of evidence is from what are known as ‘poverty of stimulus’ or POS phenomena where languages display what seem be systematic complexities, common to diverse, unrelated languages, many unnoticed until linguists working in the framework inspired by Noam Chomsky started looking at grammars in a new way, trying to explain both what the grammars allowed, and, just as importantly, what they disallowed. Traditional grammars did not bother with the disallowing.

POS phenomena are unaccountable in cultural terms. Attempts have been made to account for them by trying to teach normally competent adult speakers fragments of some invented language. But the simplest account of POS phenomena, first put by Chomsky in 1965 in a kind of programmatic way, is that they reflect some aspect of the human genome. That program has continued ever since.

The issus at stake here are not new. They occurred to Plato two and a half thousand years ago. But as far as I am concerned, there are two quite different sorts of reason for believing that the notion of UG is essentially true, that the mind of a human new-born is far from being a blank slate, and that Ambrose and Tomasello are mistaken.