J-L Dessalles
excerpts from the English translation draft of
Aux origines du langage” - Paris: Hermès 2000
(to be published by Oxford University Press)



Chapter 8 Protolanguage


Protolanguage is an idea coined by the American linguist Derek Bickerton. His study of pidgin and creole languages and of the differences between the two has led him to posit the hypothesis that in the past the progenitors of our species spoke a less elaborate form of language, which he calls protolanguage. A vestige of this form of communication survives in our modes of behaviour. What type of communication was possible through protolanguage? And what can we know about the intelligence of those who spoke it?


8.1 Communicating just with words



Bickerton (1990) came to the idea of protolanguage through the study of Hawaiian pidgins, which he compared with other pidgins and with the creole languages that are often follow them. A pidgin is a simplified mode of verbal expression, typically spoken by adults who speak also a native language which is different. People may find themselves constrained by a variety of circumstances to communicate without using an established language. The best known of these circumstances is slavery, which brought people from very disparate linguistic backgrounds into contact with each other. In the Caribbean, for example, thousands of Africans transplanted from different ethnic groups found themselves suddenly thrown together; and in conditions of forced labour, they were cut off from all contact with other speakers of their own language. More recently in Hawaii, economic and commercial reasons brought together people from various parts of the Pacific, such as Japan, Korea, the Philippines, etc. Finding that they have to communicate with one another, or just responding to the normal human instinct which makes the normal person speak to other people, such adults very quickly devise a code of communication by adapting the only linguistic resource they happen to have in common, which in Hawaii happened to be either English or Hawaiian. Under normal circumstances, foreign immigrants soon pick up a modicum of competence in the local language. But when the number of simultaneous immigrants exceeds a certain level, the standard form of the language becomes less accessible to them and ceases to function as a model, and a form of pidgin soon emerges. Pidgin is a language apparently without syntax, reminiscent of the speech of the character of Tarzan. Here are a few sentences, unconnected with each other, from Taï Boï, a French-Vietnamese pidgin (Bickerton 1995: 163):


Moi faim. Moi tasse. Lui aver permission repos. Demain moi retour campagne. Vous pas argent moi stop travail. Monsieur content aller danser. Lui la frapper. Bon pas aller. Pas travail. Assez, pas connaître. Moi compris tu parler.[1]


The next extract is from Hawaiian pidgin (Bickerton 1990: 120), followed by English equivalents, the first of them word-for-word:


Aena tu macha churen, samawl churen, haus mani pei

And too much children, small children, house money pay.

And I had too many children, very young ones, amd {and} I had to pay the rent.


This type of rudimentary speech is not confined to colonial situations. Bickerton quotes an exchange from Russonorsk, a commercial language used for almost exclusively mercantile purposes among Russian and Scandinavian sailors (Bickerton 1990: 121):


R : What say ? Me no understand.

N: Expensive, Russian – goodbye.
R: Nothing. Four half.

N: Give four, nothing good.

R: No brother. How me sell cheap? Big expensive flour on Russia this year.
N: You no true say.

R: Yes. Big true, me no lie, expensive flour.

N: If you buy – please four pud (measure of 36 lbs). If you no buy – then goodbye.

R: No, nothing brother, please throw on deck.


The subject under discussion is bartering flour for fish. The structuring of the sentences, in its almost total lack of grammar, is very close to that of the two extracts from pidgins. What is remarkable is that the people who are speaking like this are all perfectly capable of expressing themselves properly in their native languages.




Bickerton’s structuring idea is that protolanguage, related in his view to pidgins, the speech of very young children, or the language used by Genie, is not a degraded form of normal language, but is a functional system of communication in its own right:


Genie knows what past tense means, knows when it is appropriate to use it, and even knows at least one of the ways of marking it in English. But she cannot incorporate this knowledge into her normal ongoing speech. . . . This suggests not that she has merely failed to acquire a full version of human language, but that she has acquired something other than full human language—an alternative means of communication that incorporates some features of language but rigorously excludes others. (Bickerton 1990: 117)


Bickerton sees protolanguage as a fossil, a behavourial vestige with which each of us is endowed. We are able, effortlessly and instantaneously, to adopt a pidgin form of speech, using words from our native language. Without the slightest reflexion, words come to us naturally, in an approximate order; we just spontaneously omit grammatical words, articles, prepositions, relative pronouns, markers of tense or aspect. In Bickerton’s view, this reveals the presence of a fossilized competence, an innate expertise which was once the normal form of communication among members of Homo erectus, the species from which our own derived. It is a vestige of their speech that survives in us and which we can fall back on at times when expression through normal speech is impossible.


8.2 A language that is not learned


The example of the deaf children of Nicaragua (cf. Chapter 3) fits very neatly with Bickerton’s hypothesis on the existence of a protolinguistic competence. The level of communication achieved by the deaf adolescents who have been to school is a form of pidgin. The signs they put together are more or less separate from each other and the sequences are brief, rather like statements in pidgin which are usually restricted to four or five words. Two other features of their communication that are reminiscent of pidgins are the fact that the protolanguage of signs was developed rapidly and spontaneously by the youngsters themselves, and of course the fact that they seem to have been already too old to learn how to make their protolanguage evolve towards a language with syntax.




 The form of language found in pidgins has two essential properties: it is functional and it is spontaneous. These two properties are lacking in any debased forms of language that we might observe or imagine. Protolanguage does enable hearers to construct meanings which roughly fit the meanings that speakers have had in their thoughts, though this requires that the contexts be sufficiently restricted. For instance, the protosentence ‘And too much children, small children, house money pay’ says enough for us to be able to grasp at least the gist of the speaker’s lament. That is a very positive quality, which might well be absent from other conceivable simplifications of normal syntax, such as omitting all nouns or every second word. If all the semantically weak parts of a statement are omitted, like the grammatical words or the inflections of nouns and verbs, this can often prevent the meaning from being plain. Such omissions from the sentence ‘‘The girl who was stolen the money has gone” would result in ‘Girl steal money go’, which might make it appear that it was the girl who stole the money. In protolanguage, the words ‘money’ and ‘girl’ would be juxtaposed and the statement would probably be made in two sentences: ‘Steal money girl. Girl go.’ There is still an ambiguity; but the wrong meaning is not so unavoidable. Protolanguage is not the result of a rough simplification of language; it is a tool for communicating meanings that has its own organization.




If we accept that conclusion, and if protolanguage was one of the characteristic behaviours of a species of hominids, it must be possible to show it was locally optimal (cf. Chapter 6), that is to say that no minor variation in the competence could have made it any better at fulfilling its biological function. The difficulty that arises when we propose to evaluate its local optimality is that we have still not defined what the function of protolanguage was. A glance at its structure tells us that the main feature of it is that it includes all the semantic elements essential to comprehension and only them. The result is an economical system which, though not nearly as accurate as language, none the less does provide a measure of efficacy.




8.3 Protosemantics


This hypothesis will achieve greater coherence if we try to understand what sort of meanings protolanguage can express and how it is adapted to convey them. In what follows, we shall argue that protolanguage is adapted to the expression of protosemantics, that is to say a field of meanings accessible to Homo erectus. What might such protosemantics consist of? In accordance with the principles established in Chapter 6, protosemantics cannot amount merely to a weaker version of Homo sapiens’s abilities in semantic representation. It has to be a mode of cognitive organization that is functional and locally optimal. So any arbitrary division, such as restricting its scope to concrete entities or to immediately visible objects, would be inappropriate. If protolanguage ever existed as a means of communication proper to a species, then we must assume that its existence necessarily involved a form of protosemantics. Members of that species communicated about something, and it is that something that we must try to reconstruct. This is an endeavour fraught with potential dangers, as what we are about to embark on is an attempt to reinvent if not the mind of Homo erectus, at least some aspects of the cognitive functioning of that mind. The main danger is that, in the absence of subjects on whom to test any hypotheses, we might get carried away and end up piling conjecture upon gratuitous conjecture in a world where the only limit to such things is imposed by authors’ lack of imagination. I suggest a more prudent course. The problem facing us (how are we to define protosemantics in relation to protolanguage?) is relatively constrained in four parameters, as follows: 1) protosemantics must be a functional field of meanings; 2) it must be locally optimal for a given biological function; 3) it must subsist in modern humans, either as a fossilized competence or as a functional subset of our semantic competence; 4) protolanguage, as we understand it from the study of pidgin, must be locally optimal for the expression of this protosemantics. It is clear that if we accept these constraints, the danger of fanciful conjecture will be greatly reduced. Added to that there is the fact that our objective is the relatively modest one of positing, if possible, a few minimal hypotheses aimed at making more sense of the existence of protolanguage, in the full knowledge that they may well be criticized and require revision.

The basic idea is that the words put together in protolanguage are a way of bringing to mind concrete scenes, things experienced or imagined. On hearing a word like ‘cat’ we may picture either a prototypical cat or else a particular one that is familiar to us. Similarly, ‘door-mat’ readily brings to mind the image of an object. In a particular context familiar to two speakers, both words would very likely convey the image of a particular animal, the cat of the house, and the image of a particular object, the mat at the front door. Combining the two words into the statement ‘cat mat’ requires us to join the two images together. We possess the ability to combine images in a way that is not arbitrary. Clearly, there was a strong chance that the cat in the example might be on the mat, if they both belong to the same house. But there are an infinite number of other possibilities: the cat might have been lying under the mat or to one side of it; the mat might well be under the cat but in the bedroom; both the cat and the mat might be floating about inside the kitchen; the cat might be either walking towards the  mat or away from it; or the cat might even have changed colour or shape to look like the mat; and so on. Most people, however, if the circumstances are right, will spontaneously picture the first image of the cat dozing on the mat by the front door. This human ability to combine images in a particular way that will be foreseeable by another person is largely a mystery. It must rely on the use of actual situations seen as more or less prototypical, such as a cat that is in the habit of dozing on the mat by the front door. However, we also readily create scenes that we have never experienced by combining images that are purely imaginary, such as the cat lying under the mat or balancing the mat on its nose. This astonishing competence no doubt uses our ability to associate in order to recall memorized visual elements and gauges the scene envisaged with reference to constraints inherent in each entity (the respective sizes and weights of the objects, their power of autonomous movement, etc) and to their expected behaviour (typically, a motionless cat is sitting and asleep). Nevertheless, it must be acknowledged that this process of synthesis of mental images remains largely obscure. No doubt future advances in psychology and the technologies of virtual reality systems will shed light on this.

Let us assume that this ability to construct scenes out of our combining of mental images is real and that Homo erectus also possessed it. We can now understand how the link between protolanguage and the making of images can happen.




It is clear that combination of images is an unreliable process; and for this there are two reasons. The hearer risks combining images in a completely unexpected way or failing to combine any at all. It is therefore the responsibility of the speaker to describe the scene in a way that facilitates the hearer’s task. Without the active cooperation of the speaker, this mode of communication cannot work. In the example, if the meaning was that [196] somebody stole money belonging to the girl, the speaker will prefer to say ‘Steal money girl. Girl go.’ By putting ‘money’ adjacent to ‘girl’, the speaker biasses the hearer towards making an image in which the money belongs to the girl; by separating ‘steal’ from ‘girl’, the speaker biasses the hearer away from imagining a scene in which the girl does the stealing; and by repeating ‘girl’ in front of ‘go’, the speaker helps the hearer to make a scene in which it is the girl who leaves.

Can we say that protosemantics, as discussed above, meets the four constraints we imposed? The answer is certainly positive for constraints numbers 1 and 3: if we accept that our ancestors possessed the mechanism of combination of images and scenes, then  it is obviously a functional system still extant in modern human beings. Constraint number 4 requires that protolanguage must be locally optimal for conveying representations of protosemantics. This is not an easy point to verify, though we do possess two pieces of evidence pointing in the right direction. One is, as Bickerton observes, that the words of protolanguage as suggested by pidgin are all words with semantic content. Even modern human beings, who are able to use words that are more or less empty of separate meaning, such as prepositions, conjunctions, or relative pronouns, omit them when expressing themselves in protolanguage. The second comes from word order. Although protolanguage is supposedly devoid of syntax, its word order is not entirely arbitrary. The grouping of words into semantic components is crucial for the drawing of any proper interpretation. It should be noted that this constraint is greatly relaxed in normal speech. In the sentence ‘I sent, on the day before she came, John’s book which was on the table to Mary’, the words ‘sent’ and ‘Mary’ are separated by five words with semantic content, a thing which would be quite impossible in protolanguage. The structure of protolanguage looks as though it is determined by the requirements of protosemantics, which strongly suggests that it is a linguistic system well adapted to its function, in accordance with constraint number 4.

Constraint number 2, relating to the locally optimal character of protosemantics, is much more difficult to verify. The fact is we have not yet broached the function of protosemantics. Can we define what usefulness individuals might see in using the words of protolanguage to communicate an image or a concrete scene to their fellows?




Does the limitation of mental imagery to concrete entities rule it out, as Bickerton seems to suggest? An answer to that question should not be based on the abilities of present-day humans. Concrete protosemantics might have been quite adequate to the needs of hominids who had no access to abstractions. Abstractions like the fidelity or trust cited in Bickerton’s example are qualitatively different from the images that concrete words can bring to mind in a reliably systematic way. If I say ‘cat’, I can make a pretty sure guess at the type of mental image you will form, especially if the context is clear, for instance if we have both just seen a cat walk past. If I say ‘fidelity’, I have no way of knowing what image you will form, if you do in fact form one. In the first case, communication is possible, because the speaker can foresee the signified constructed by the hearer; in the second, Bickerton is right to hold that communication cannot take place if it has to rely solely on the construction of an image. Hominids must have been able to function in the first of these two modes, communicating through the exclusive use of concrete words which were adequate to conveying scenes in a way that was more or less deterministic. If one accepts this description of communication among hominids, it is reasonable to conclude that they had no abstract representational abilities and that this constitutes a fundamental difference between protosemantics as used by them and semantics as used by us.


8.4 Prelanguage, a language without sentences




8.5 The lexicon of protolanguage




[1] Approximate equivalents in English-based pidgin: Me hungry. Me cup. Him have permission rest. Tomorrow me return country. You no money me stop work. Sir happy go dance. Him hit her. Good not go. Not work. Enough not know. Me understand you speak.