Feature

In The Beginning...

...was the word, which just may have helped make us intelligent.

Dr Andrew Carstairs-McCarthy

What can we discover about the stages in the evolution of human language, and how these stages relate to other aspects of mental development?

Anthropologists, archaeologists, psychologists and philosophers debate this issue vigorously, but 20th-century linguists have tended to avoid it.

Many linguists would say that it is pointless to guess about a historical process which is inaccessible to direct investigation. We can dig up bones and tools, but we cannot dig up consonants and vowels. Even so, the origin of language is beginning to be a respectable topic among linguists.

I am currently exploring the relevance to it of an aspect of language which at first sight seems to have little connection: the internal structure of words, and especially the bits of words (prefixes and suffixes) which express grammatical information, such as the plural "-s" on "dogs", and the "-en" on "eaten" in a context like "I have eaten my tea".

Just as nature abhors a vacuum, so language abhors perfect synonyms --  expressions which mean exactly the same thing. Of course, synonyms exist (courgettes are the same vegetable as zucchini, for example), but nearly all terms which seem to mean exactly the same thing turn out not to be totally interchangeable after all (for example, each of us individually is likely to prefer either "zucchini" or "courgettes" in our own usage, rather than use both words equally). There is good evidence too that small children when learning language operate as if taking it for granted that there are no perfect synonyms, so that any new word must mean something different from all the words which they already know.

What has that got to do with language origins? I make the link in three steps. The first step is to show that the avoidance of synonymy is deeply embedded not only in the vocabulary but in the grammar -- specifically, in the grammatical prefixes and suffixes (collectively called "affixes") which many words carry.

At first sight this is an odd claim to make because it seems clearly untrue. The "-s" on "dogs" surely means just the same thing as the "-en" on "oxen" and the "-ae" on "formulae" -- they all indicate that the noun is plural. Likewise, the "-en" on "eaten" has just the same grammatical function as the "-ed" on "finished", in a context such as "I have finished my tea". Aren't these examples of perfect synonymy between affixes?

Anyone who has studied a language such as French, Russian or Swahili will know that this sort of affix synonymy can be much more pervasive than in English. In fact, the most frustrating aspect of learning many languages is sorting out the apparent superabundance of different ways of expressing the same grammatical information. Why does noun A have plural ending X ("dogs") while noun B chooses plural ending Y ("oxen"), in seemingly haphazard fashion? Why can't each meaning be expressed in just one way, and each affix have just one meaning -- a maximally "efficient" one-to-one pattern of linguistic coding?

Examining closely what happens in particular areas of grammar within a variety of languages, I have found evidence that things are not so haphazard after all. Linguists have observed that "synonymous" affixes are not freely interchangeable. We cannot choose to say "eated" or "finishen" instead of "eaten" and "finished". But more striking restrictions emerge in languages where word structure is more elaborate than in English and where much more grammatical information is carried by affixes.

Let us suppose that in some such language there are four different plural endings, all cropping up frequently in ordinary speech and writing -- by no means a rare situation. Not only will they not be freely interchangeable, it turns out, but at least three of the four will have a "meaning" which goes beyond just expressing plurality. This extra "meaning" will involve certain grammatical characteristics of the nouns to which they attach; in fact, it will involve information about all the other affixes which that noun will carry for all the other grammatical concepts which are expressed by means of affixes in that language.

What these plural endings encode, it seems, is not just information about the objects denoted by the words to which they attach ("this word denotes more than one object") but also precise information about how these words change their shape grammatically ("the affixes which this word carries in other grammatical contexts are precisely such-and-such"). And, because at least three of the available plural endings encode this extra grammatical information alongside simple plurality, the four plural endings will, after all, not be perfectly synonymous. The same goes for affixes which signal not just plural number but all other kinds of grammatical information -- past tense or imperative mood in verbs, possessive forms in nouns, and so on.

It may seem odd to include within the "meaning" of a affix something so purely linguistic (and so useless from the point of view of communicating about the outside world!) as information about a word's grammatical behaviour. But it is odd only if we assume that outside-world meanings are the only meanings that there can be. As soon as we allow that grammatical and other purely linguistic information can count as part of meaning, then it turns out that the apparent synonymy of plural endings "-s", "-en" and "-ae", of verb endings "-en" and "-ed", and of similar sets of affixes in several other languages, is only apparent after all, because precise information about the distribution of other suffixes and prefixes can be deduced from them.

Languages relate words to things and events, and it is tempting to see this relationship as lopsided -- things and events have priority, and words are labels invented to apply to them. But what we have just observed about synonymy avoidance on the part of affixes suggests that the relationship may be more balanced.

Different linguistic units (words or affixes) need to have different meanings, it seems, and one obvious way to differentiate them is to associate them with different things or events in the world. But this way of differentiating them is not the only way. An alternative is to associate linguistic units with other linguistic units, so as to yield a purely linguistic pattern of predictability ("if word X has ending A in grammatical context C, then the set of endings which it has in all other relevant grammatical contexts is such-and-such").

Purely linguistic patterns of this kind may seem pointless from the point of view of communication about the world. But the fact that these patterns crop up in so many languages will be just what we expect if synonymy avoidance is a fundamental characteristic of human language -- a more fundamental characteristic, in fact, than any pressure for the sort of sleek communicative efficiency which hard-pressed students of "difficult" languages look for in vain.

From Gibbon to Jabbering

This leads us to the third step in the link with language origins. Let us conduct a thought-experiment, visualising an ape-like animal which shares our emphasis on synonymy avoidance but which has roughly the communicative capacity of a gibbon -- a repertoire of about 11 distinct vocalizations or cries. Such an animal does not have to be particularly intelligent in order to distinguish "meanings" for each of its cries.

In particular, synonymy avoidance does not require this animal to identify numerous objects or events, independent of itself, for these cries to denote. Instead, each cry can be associated holistically with a situation ("Beware predator!") or a physical or emotional state ("I am hungry", "I am afraid", "I want sex", "Go away!"). But let us suppose now that this animal's vocal apparatus changes radically so that, instead of being limited to 11 cries, it can produce an infinite range by combining in strings of indefinite length a repertoire of 40 or so relatively short and auditorily distinguishable sounds. In this new situation, how can the animal continue to avoid synonymy?

One way might be by assigning a distinct meaning to each of the forty-odd sounds. But let us suppose that many of these sounds are hard to produce or to recognise reliably in isolation, so that there is no escape from treating as distinct "cries" not each individual sound but rather each distinct string of sounds. Yet to learn an infinity of meanings (one for each string of sounds) would be impossible for a creature with a finite brain capacity. So, if the creature is to exploit the new capacity of its vocal apparatus, it will have to break up its "cries" into shorter chunks whose number will be finite but still larger than 11 or even 40, while combining these chunks into longer strings according to some principle which will allow these longer strings to have identifiable meanings.

The change in the vocal apparatus that we have postulated for this animal is in fact a change which took place in our own ancestors. In apes as well as in new-born human babies, the larynx (or voicebox) is relatively much closer to the root of the tongue than it is in human adults. This arrangement has certain physiological advantages -- apes and young babies can breathe while swallowing, and there is less risk of them choking by getting food stuck in the windpipe. But at some stage in the evolution of human beings there arose the modern adult human positioning of the larynx, relatively lower in the neck. This anatomical change, along with greater mobility of the tongue, allowed a greater variety of speech sounds to be produced.

This anatomical change, along with the assumption that our ancestors already observed the principle of avoiding synonymy in communication, would have been enough to force the development of a new communication system. In this new system, utterances ("cries") were divisible not only into individual speech sounds (analogous to the 40 or so phonemes of English) but also into meaningful chunks, analogous to words and phrases.

In other words, the anatomical change would have been enough to force the development of something like human language, one of whose basic features is what linguists call "duality of patterning" -- the fact that it consists simultaneously both of meaningful chunks (phrases, words and smaller units such as stems and affixes) and also of individually meaningless sounds.

We have not assumed any increase in intelligence as a starting-point for this development. All that was necessary for a language-like communication system to arise, once the repertoire of speech sounds had been increased, was an insistence that every different cry should have a different meaning. Complying with that insistence meant that more "meanings" had to be found quickly for chunks into which utterances had to be divided. Quite simply, more things had to be found to talk about. So it was virtually inevitable that meanings should be sought not just among the speaker's emotional and physical states (such as hunger, fear, lust, and anger) but among things and events in the outside world.

Language Leading Intelligence?

This leads us to a picture of our pre-human ancestors which turns some popular notions on their head. We are inclined to think of language as an outgrowth of intelligence -- our ape-like ancestors became so clever that communicating by means of a small repertoire of grunts and cries came to be inadequate for their needs.

The alternative view which I support is that the increase in intelligence associated in the archaeological record with increasing brain volume may be, to a large extent, an outgrowth of language. Before the anatomical changes occurred, the number of ways in which unusually intelligent pre-humans could use their intelligence to help their family or tribe and give it an advantage over competing families or tribes was limited.

In particular, the person with bright new ideas could not communicate them by speech, because the repertoire of things which could be said was limited by the finite set of cries in use. But once there had arisen a communication system with the duality of patterning characteristic of human language, intelligent pre-humans had more scope to communicate their bright new ideas through combining the meaningful chunks of speech in new ways. The whole tribe could therefore benefit from an individual's brainwave, and do better in competition with other tribes. This in turn would mean that the intelligent individual would have a better chance of passing his or her genes on to future generations, rather than dying before having a chance to procreate. Thus the general intelligence level of the community would tend to rise.

The idea that language is one of the main factors which stimulated an increase in human brain capacity is not new. What I think I have discovered is a new kind of evidence in support of it.

For any species, insistence on synonymy avoidance in its communication system will confer an advantage, by making the system easier to learn. What happened when our ancestors acquired the modern vocal repertoire was that, in order to continue avoiding synonymy, they had to find many new things for distinct utterance-chunks to mean. Most of these new meanings resided in the outside world, and it was obviously the ability to manipulate outside-world meanings which would be most useful to the community and would provide leverage for intelligence to develop.

But there is no reason why meanings should reside only in the outside world. They could be got from anywhere, and at least some of them could reside in the communication system itself. It is meanings of this kind which are perpetuated in the precise information carried by seemingly synonymous grammatical affixes about the shape in other contexts of the words to which they attach. The apparent grammatical complexity which such "meanings" may give rise to is frustrating for adult learners of foreign languages; but is evidently something which children learning their mother tongues in infancy take in their stride, thanks to their strong reluctance to tolerate perfect synonymy.

Much work remains to be done in three areas. First, we must check how synonymy avoidance works in the grammatical systems of more languages. So far, I have found confirmatory evidence in English, German, Icelandic, Latin, Rumanian and certain African and Caucasian languages. This sample is sufficiently varied so that the patterns I have observed are unlikely to be due to coincidence or to the sharing of inherited characteristics by historically related languages.

Secondly, we need to investigate how synonymy avoidance does or does not figure in the communication systems of other animals. Many birds, for example, can produce a large repertoire of sounds which to human ears sound distinct, yet no bird "language" resembles a human language at all closely in its structure and use. Is this because birds are not subject to the same urge to avoid synonymy as humans and their ancestors have been? And what about marine mammals, which are among the most intelligent non-human animals but whose vocal communication systems are still not well understood?

Thirdly, we need to study how synonymy avoidance and vocal tract changes relate to other changes which have been claimed as relevant to human mental development: upright gait, the opposible thumb, tool-making, changes in diet and changes in social organisation. For the time being, however, it looks as if a rather frustrating and apparently accidental characteristics of many human languages -- the superabundance of "synonymous" grammatical affixes --  may provide unexpected reinforcement for the view that language was a trigger rather than a byproduct in the evolution of human mental capacities.

Dr Andrew Carstairs-McCarthy works and undertakes research in the Linguistics Department at Canterbury University.