The Origin of Language

Evolution from Protolanguage to Language

by : Professor John Maynard Smith 

      Evolution from apes to humans has involved both an increase in general cog­ nitive ability, and in speci c competence for language. The steady increase in brain size over the past 2 million years suggests that the increase in cognitive ability also proceeded over this period. The main selective advantages conferred were improved ability in the manufacture and use of tools, and in social skills. Already, in other primates, there is evidence of an association between brain size and the complexity of the social system, and this association was almost cer­ tainly important in our own ancestors.

     The timing of the origin of language is more di cult to determine. However, the dramatic increase in technical inventiveness during the past 40000 years, discussed in the last chapter, is most readily explained if the nal stages in linguistic competence emerged at that time. Humans differ from apes not only in grammatical competence, which has been the main topic of this chapter, but also in the ability to produce and perceive sounds. One di culty in teaching apes to talk arises because they cannot make all the sounds used in human speech. Human evolution required both anatomical changes, and improvements in the brain mechanisms concerned with sound production and perception (Lieberman, 1989) . Anatomically, the descent of the larynx in humans has increased the range of sounds we can make, at the cost of increasing the chanceofchoking when we eat or drink. It would be helpful ifwe could date this change from the fossil record, but anatomists are divided as to how far this is possible: unfortunately, the hyoid cartilage, which would provide the information, rarely fossilizes .

      The changes in brain mechanisms are equally important. Although sound input is in principle a continuous variable, we unconsciously classify incoming speech sounds into discrete categories: in other words, we treat the input as digital. We are able to produce and to perceive sound 'segments' (roughly equivalent to the letters of the alphabet) at the astonishingly high rate of 25 per second. This rapid transmission is necessary: without it, we would forget the beginning of a sentence before reaching the end.

       The emergence of human language, then, required changes in anatomy, in motor control, in sound perception, and in grammatical competence. These changes could not have been instantaneous, but may well have been rapid. They must have predated the dispersal of Homo sapiens throughout the world, most probably from Africa, because all existing human populations are alike in these respects. Subject to this constraint, however, the events may have been fairly recent. 

Protolanguage

Language does not fossilize. There are, nevertheless, four sources that may in­ dicate what primitive language was like: ape languages, early child language, language of 'Caspar Hauser' children and pidgin languages. We discusses them in turn.

Consider the following utterances, from two different sources: 

     Although remarkably similar, (12) is from children at the two-word stage, and (13) is from the chimpanzee Washoe (Gardner & Gardner, 1974). They illustrate the following types of utterance: the attribution of qualities to objects; the possession of non-living objects by living beings; the location of actions; the relation of agents and patients to actions. The two sets are formally identical.

       In both apes and in children under two, one-word utterances dominate. There are no grammatical items and no trace of structure. The only real difference between (12) and (13) is that children seem to categorize for categorization's sake, whereas apes talk only about objects they want or actions they want to perform or have performed: children have a greater curiosity about the en­ vironment.

        There have been many examples of children growing up in isolation, but the best-documented case linguistically is that of 'Genie:, a 1 3-year-old girl who was found in California in 1970. Her father imprisoned her in her room alone when she was about 1 8 months old. She was incapable of speech. Although normal in her ability to form concepts, she remained stuck at the linguistic level illustrated below: 

      Genie is a little more advanced than (12) or (13): there is one grammatical item, at, and the last sentence is a fairly complex proposition. But the difference is slight: it seems that Genie has acquired something less than human language, which we can call protolanguage. In view of the similarity to ape language, protolanguage is phylogenetically ancient. There is apparently no critical period at which it must be acquired.

      Turning to our nal example, pidgins develop when people must communicate with one another without sharing a common language, and without one lan­ guage being the obvious medium. Slaves in island colonies are a good example. The following comes from Bickerton (1990): 

Luna, hu hapai? Hapai awl, hemo awl. 

Foreman, who carry? Carry all, cut all. 

(,Who will carry it, boss? Everyone will cut it, and everyone will carry it.') The speaker of the rst line was a Filipino, while the italicized words are Hawaiian. Although there are a few grammatical items such as hu, there are no articles, prepositions or tense markers.

Sometimes pidgin remains at a primitive level for several generations. Rus­ sonorsk, of which an example follows, was used by Russian and Scandinavian sailors for trade: 

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 say true.

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

N: If you buy-please four pud. If you no buy-then goodbye. R: No, nothing brother, please throw on deck.

        There is some advance here over (12) and (13), but there are few gramma­ tical items, the longest utterance has no verb, give lacks a goal, and throw a patient. The remarkable thing about pidgin is how little linguistic knowledge its speakers were able to employ when communicating across language barriers.

The crucial differences between the various forms of protolanguage and true language are as follows:

• Word order in protolanguage has nothing to do with syntax.
• In ordinary language, null elements indicate the points in sentences where we can infer that some constituent is notionally present. In protolanguage, any element may be missing from any position in an unpredictable way.
• Phrase structure is almost completely absent. The few apparent exceptions may have been rote-learned, such as lexical items and idioms.
• There are few, if any, grammatical items. 

From Protolanguage to Language 

         Before discussing the evolutionary transition from protolanguage to language, we will describe a comparable transition that can occur in a single generation. This is the transition om pidgin to creole (Bickerton, 1983). The latter is a transformed pidgin, created by children when learning pidgin from adults. Children from Hawaiian pidgin-speaking parents utter sentences such as: 

                                  They wen go up there early in the morning-go plant                              (15)

('They went up there early in the morning to plant (crops).') The null element refers to the missing subject, they.)

       Creole languages arose when children of immigrants were not exposed to any normal natural language, but to pidgin only. Table 1 7 . 3 gives some examples of the jump in complexity between pidgin and Hawaiian Creole English. Whereas almost everyone speaks his or her own pidgin, the rules of Creole are uniform from speaker to speaker. One might expect that the grammar comes either overwhelmingly from one source language, or that it is a mixture derived from all the available source languages, but that is not so. Hawaiian Creole differs from Chinese, Hawaiian, Korean, Portugese, Spanish, or the Philippine lan­ guages: some differences between English and Hawaiian Creole are shown in Table 17.4.

      Creole languages have arisen repeatedly in various parts of the world, with widely different vocabularies, but with a surprisingly similar grammatical structure. One characteristic feature of Creoles (and, as it happens, Hungarian) is that they permit the double negative, 'I don't want no cabbage'. A second common feature (also shared by Hungarian) is to distinguish statements and questions by intonation alone. A third grammatical feature shared by many Creoles is the method ofconjugating verbs (Table 17.5). Bickerton (1983) has drawn attention to the fact that, between the ages of 2 and 4, children learning English sometimes make mistakes that would be grammatical in Creole: for example, children often try out the double negative.

        Creoles are proper languages: for example, the proportion of purely gram­ matical items is about 50 per cent. Apparently, Hawaiians jumped from proto­ language to language in a single generation. This rapid transition was possible because no genetic change was needed: the change could not have been so rapid in phylogeny. 

       Given the rather sharp distinction between protolanguage and language, we must return to the question whether any functional intermediates between them are possible. We are tempted to point to Gopnik's dysphasics, who lack a speci c grammatical skill but are obviously better off than speakers like Genie, and leave it at that. However, the di culty is so often referred to by linguists that we will say a little more about it.

        The di culty arises, we believe, because people are asking the wrong ques­ tion. They aslc is there any grammatical rule that could be absent, without reducing the power of linguistic expression? The answer is probably no, but the question is irrelevant. We should ask: is there any grammatical rule that, if it were absent, would still leave a linguistic competence that was better than mere protolanguage? To this question, the answer is obviously yes.

        To give one example, Premack (1985) suggested a language in which the­ matic roles would map directly onto surface ordering. It would then be possible to say the dog bit John, but not John was bitten by the dog. Such a grammar would be less expressive than our existing grammar, but it would be better than no grammar. Language would be still simpler if it permitted only verbs with two arguments, agent and patient. One could then say the dog bit John, but not John sleeps, or John gave the book to Mary: again, the limitations are largely irrelevant.

Consider the follOWing possible extensions to protolanguage:

• Items for the negation of information, such as no.
• 'Wh-' questions, such as what, who and where.
• Pronouns (instead of the repetition of proper names: John asked Bill for the book and Bill gave it to him, instead of John asked Bill for the book and Bill gave it to John) .
• Verbal auxiliaries such as can and must. 
•  Expressions for earlier and later (e.g. before and after). 
• Expressions for orientation in space (e.g. up, down and into). 
• Quantities such as many and few.

         There seems no good reason why these features should not have been ac­ quired one by one. The hard question, however,· is how we came to use a three­ tiered phrase structure (Figs 17.3), and to construct sentences by recursively inserting phrases into phrases. The crucial invention was the tem­ plate for phrase assembly. The three layers (Fig. 17.3) represent, respectively, a generic class (cow), properties peculiar to particular members of that class (with a crumpled horn) and speci cation of the complete individual. Thus, phrases are machines for going from the class to the individual, an ability important for cognition as well as for communication. The template may have originated with noun phrases, and later been extended to other units, so that one could say the cow with the crumpled horn kicked the dog, but not the dog ran round the mulberry bush, which contains a verb phrase.

Fig. 17.3 Universal phrase structure (after Bickerton, 1990). X represents any lexical category (noun, verb, adjective, preposition, etc.) that can be expanded to form a phrase. The n after X' indicates that this level, unlike X and X', can be repeated (e.g. big bad wol . The parentheses round the specifier and the complement indicate that these items are not obligatory. The horizontal two·headed arrows indicate that the relative positions of the items can be exchange.

      It is probably foolish to attempt a reconstruction of the precise path that was followed from protolanguage to language, at least until we have more extensive genetic data. Our only aim has been to show that it is not particularly difficult, in principle, to imagine a stepwise transition.