Pillow talk for cavemen

The problem of the origin of language has come back into fashion after more than a century of neglect. In 1866 the Societe de Linguistique de Paris made it a standing rule that no paper on the subject was to be presented or discussed. For generations, linguists referred derisively to popular views by such terms as "the bow-wow theory", "the pooh-pooh theory" and "the ding-dong theory".

Now that has changed, and the conundrum of how it all began is back on the agenda of intellectually respectable topics. One wonders why, because on the face of it not much progress has been made towards providing a satisfactory answer.

The reason for the lack of progress, according to Terrence Deacon, is that linguists, psychologists and others have been asking the wrong questions. The right question, he claims, is deceptively innocent: why do animals not have simple languages? Deacon's eureka moment occurred when he was asked this by an eight-year-old schoolchild, and Deacon, who teaches biological anthropology at Boston University, realised that he had no answer. It led him to ponder the evolutionary paradox presented by the fact that "the unbroken continuity between human and non-human brains" is apparently accompanied by "a singular discontinuity between human and non-human minds". Language, allegedly, marks the divide.

That is already, some will say, a controversial way of formulating the issue. For while it is not open to doubt that my dog has a brain, whether he has a mind certainly is. But if he has, what that has to do with his inability to tell me about it is more controversial still.

The idea that language distinguishes human beings from non-humans has a venerable pedigree, going back to at least the fifth century BC. Isocrates states it quite plainly and even appears to take it for granted. But it did not occur to any Greek to ask what it is about the human brain that accounts for this. Or what it is about non-human brains that might make the development of language impossible.

Deacon insists from the start that the two questions cannot be separated. To answer one we have to answer the other. So the explanation of why humans have very complex languages whereas other species do not even have simple languages has to be sought in the evolution of their brains. Deacon identifies the key factor as the enlargement of the pre-frontal cortex in humans. But language was not simply a consequence of this. On the contrary, language itself played an essential role in the gradual development of that part of the brain. It was a joint bootstrapping operation, or, as Deacon prefers to call it, a case of "co-evolution".

What first triggered this co-evolution? Deacon is scornful of "just so" stories about language origins, including the Chomskyan myth of a genetically provided "language organ". But then he tells a "just so" story of his own which is scarcely more convincing. Our hominid ancestors needed to develop a primitive form of language in order to institutionalise marriage.

Why did they need to do this? The answer is the big surprise in the book. They needed marriage because they had just changed their diet to include eating meat. Hunting for meat, unlike gathering fruit and nuts, was a men-only business. It had repercussions on social organisation, division of labour and the daily timetable. But most important of all, it meant that the boys could not keep such a close eye on the girls, and vice versa. So some form of contract was necessary to ensure that Jack provided meat for Jill, and Jill did not cheat with other boys, thus diddling Jack out of his right to produce offspring carrying his genes. No males provide meat for females "unless there is significant assurance that the provisioning is likely to be of reproductive value to the provider". Jack-and-Jill contracts, however, because they are essentially about matters which have not yet happened, but might or might not happen, required a special form of communication hitherto lacking. They required "symbolic reference".

What exactly is symbolic reference? Here we come to the nub of Deacon's theory and, unfortunately, to its weakest part. Deacon knows a lot about brains but less about language, a subject on which he seems to be not very well informed at all. Diagnostic symptoms include his telling us that Saussure was French and called the study of language "semiology"; that the distinction between syntagmatic and paradigmatic relations was introduced into linguistics by Roman Jakobson, "the famous MIT structural linguist"; and an account of Frege's distinction between sense and reference that would hardly pass in a first-year undergraduate exam. More dubious still, however, is what Deacon has borrowed - or thinks he has borrowed - from Charles Sanders Peirce.

According to Deacon, Peirce seems to have established that there are only three kinds of reference: iconic, indexical and symbolic. Furthermore, these three are hierarchically related. The most basic is iconic. Why? Because noticing whether things are the same or different is more fundamental to living creatures than anything else. Deacon gives the bizarre example of the "camouflage" which makes a moth's wings look like tree bark. He tells us that this is an icon for any predatory bird who is thereby fooled into failing to notice the moth under its beak. It is not the resemblance but the bird's "interpretive process" (mistaking the moth for bark) which establishes the iconic relationship. The case is in principle no different, we are assured, from our looking at a portrait and noting its resemblance to the sitter. Swallowing this is hard enough, but there are even harder swallows to come.

Indexical reference Deacon holds to be "reducible" to iconic reference. How? Indexicality is based on physical contiguity (the smell of smoke indicates that something is burning), but this would not work unless the stimulus evoked the recollection of similar experiences in the past. The smell of smoke reminds us of these experiences "by iconically representing them": indeed, all those experiences were "icons of each other" (sic). The only new feature is the "repeated correlation" between the smell and the burning. This adds "a third higher-order level of iconicity" to the whole business and thus upgrades it to indexicality.

Thus far in the story, my dog's brain seems to be keeping pace with mine. Deacon tells us nothing about the difference between brains that can handle iconic reference and brains that can handle indexical reference. And my dog's brain can apparently do the latter. It can, for instance, recognise smells as indexically referential of the presence of other animals. So can I, of course; although perhaps not quite as reliably or as subtly as my dog can.

So far, so good. But it raises a nagging doubt about what Deacon means by "reference". He seems to use the word as a catch-all term for any kind of recognised association. Thus, for anyone who knows what skunks smell like, the smell "refers to" the skunk. What worries me here is the naivety (or bravado?) with which Deacon appears to brush aside whole areas of rather well-documented discussion in the western tradition about problems of reference. It suggests that semantics (a term he rarely uses) is not his strong suit.

As far as one can identify it on the basis of this book, Deacon's underlying epistemology combines a behaviourist theory of learning with the medieval surrogational dictum aliquid stat pro aliquo. His surrogationalism is of the reocentric variety. But his assumptions about linguistic communication are telementational and Lockean. They are allied to a familiar Baconian semantics of the senses (which supposedly bring us "information" from the unknown world outside). These are not, perhaps I should add, descriptions that Deacon himself uses or, as far as I can see, would even recognise as applicable in his case. But they need to be mentioned in order to make sense of Deacon's view of "symbols".

Symbols, he tells us, are "constituted by relationships among indices", just as indices are constituted by relationships among icons. But things get tricky in these murky depths, and "more complex forms of indexical association" are easily "confused with" symbolic associations. So what exactly is the distinction? Well, symbolic reference, unlike indexical reference, requires the recognition that signs are not just individual one-off items, connecting to specific bits of the external world. They are first and foremost related to other signs. These relationships are systematic, both allowing and requiring signs to be combined in regular patterns in order to articulate complex messages for communication. Non-human species can communicate, but their messages have no units corresponding to our specialised "parts of speech"; no primitive nouns, verbs, etc.

It is already evident from this where the trouble lies. Deacon is making the common mistake of confusing languages with grammars. He does not seem to see that having a grammar does not automatically make a communication system into a language; or that the simplest grammars are very simple indeed - so simple that it is difficult to believe that the pre-frontal cortex in at least some other species would not have been capable of managing them, whether or not the species in question thought marriage and meat-eating were good deals.

Deacon himself concedes that ape brains can handle symbolic reference, although the difficulty trainers experience in teaching monkeys linguistic tricks shows that it does not come "naturally" to the beasts. Exactly how the human brain crossed the threshold from indexical reference to symbolic reference Deacon does not explain very clearly. Here his account tends to slide into tired metaphors that have a whiff of regress about them: we hear of shifts in "mnemonic strategy", "higher order" associations, the "recoding" of information (one of Deacon's favourite cliches), or even its "multiple coding", of knowledge being "re-presented" and "re-represented", and so on. All of which look suspiciously like arrows pointing back to a familiar place called Square One.

What Deacon fails to see in all this is that languages are not just sign systems with internal structure and combinatorial rules. Languages enable their users to state propositions which may be judged true or false, to express such judgements, and to give their reasons (as distinct from mere indications of approval or disapproval). If indexical reference cannot do this, then no mere "recoding" of indexical signs or improvements in "mnemonic strategy" will achieve it either.

As the book goes on, the word "symbol" begins to extend its range of meaning in ways that would have made Peirce's hair stand on end. The original equation between the symbolic and the grammatical is gradually forgotten, and the cadences of the discussion begin to sound less like those of neurobiology and more like those of New Age guru-speak. Prehistoric art is symbolic, rituals are symbolic, the universe is a symbol and we ourselves are symbols. Indeed, the symbolic self is independent of the particular brain and body that support it: it is "capable of true transmigration, though not necessarily as a unified whole".

In the end it is difficult to know what to make of it all. The biggest disapppointment in the book is that it fails to answer its own initial question. We end up with no very compelling reason why animals do not have simple languages, other than the rather lame hypothesis that they never found themselves in a situation where dietary change and partnership problems loomed large enough to force their brains into the labyrinths of symbolic reference. They managed to survive without systematising their indices.

As for Deacon's over-ambitious excursions into cosmic philosophy and the meaning of life, is it comforting to learn that ultimately my self is something constituted by relationships among indices? Or that it has the capacity to transmigrate? Neither comforting nor worrying, because I do not know what it means. Which leads one to ask whether one of the pay-offs in the co-evolution of brain and language might not be greater facility in the production of mumbo-jumbo. Has the pre-frontal cortex developed a dangerous tool we do not yet know how to handle?

Roy Harris is editor, Language and Communication.