Did language arise abruptly?
A question that has divided opinion perhaps more than many others is whether the emergence of human language was continuous, taking place in a gradual process over an extended period of time, or whether it happened in one abrupt leap from non-language to language. We have discussed this distinction in “Questions and approaches Previous work”, where we referred to the former as the gradualist hypothesis and to the latter as the leap hypothesis. The gradualist hypothesis tends to be based on the Darwinian notions of natural selection and adaptation and assumes a continuous evolution, where there may have been a number of intermediate stages (Pinker and Bloom 1990; Newmeyer 1998b; Bichakjian 1999; Corballis 2002b: 173; Jackendoff 1999, 2002; Givo̒n 2002a, 2002b; Heine and Kuteva 2002b). Supporters of the leap hypothesis tend to draw on alternative concepts of evolutionary biology such as exaptation, co-optation, or mutation, and/or invoke a modification of the brain structure to account for language evolution (Chomsky 1988; Piattelli-Palmarini 1989; Davidson and Noble 1993; Bickerton 1990, 1998, 2005; Gould 1997).
That these contrasting positions are to some extent due to differences in theoretical orientation, rather than to the facts examined, is suggested by the following example. Both Newmeyer (1998b) and Berwick (1998) invoke findings on syntax in support of their position; but whereas the former proposes a gradualist hypothesis, the latter insists on the leap hypothesis. Newmeyer argues that syntax is made up of various subsystems each of which is governed by a distinct set of principles which cannot be derived from a more fundamental principle or property. He concludes that these principles must have evolved in an incremental way, thus suggesting a gradualist scenario. In contradistinction, Berwick assumes that there cannot be a ‘‘partial syntax’’:1 Many syntactic relations and constraints can be derived from a combinatorial operation used in the derivation of sentences, called ‘‘Merge’’, and this operation cannot have evolved incrementally from non-combinatorial syntax but rather must be the result of a discontinuous process from a situation without syntax to one with syntax in all its glory (but see Burling 2002 for an alternative view).
Findings on grammaticalization suggest that the leap hypothesis is not really a plausible one and that it is essentially only the gradualist hypothesis that can account for grammatical evolution. We observed that grammaticalization is a cognitively motivated mechanism. But this mechanism is only one component of the process, the second component being regular human interaction and communication, in the course of which novel, and frequently more abstract concepts are proposed in specific contexts and are expressed in terms of old and less abstract concepts—with the effect that existing linguistic forms and constructions are extended to new contexts and new meanings. This process starts out with fluid discourse patterns acquiring a higher frequency of use, gradually crystallizing into new functional categories. Not all of the newly proposed concepts really end up in this way; in fact, it is only a tiny portion that, over time, are propagated and accepted as new grammatical structures. Accordingly, it takes, as a rule, centuries before a new functional category evolves (but see “A pidgin window on early language?”), and the time until such a category runs its full course, turning into an inflectional affix may be in the range of one millennium or more.
As the scenario of grammatical evolution suggests, the evolution of grammar did not involve just one kind of grammatical categories but rather a sequence of layers of different kinds of categories, with each layer expanding and refining the categories of the preceding layer. It is hard to estimate the timespan required for the evolution from layer I to VI to take place; judging on the basis of chronological observations from modern languages, it is reasonable to assume that several thousand years are required for a language to acquire a new grammatical makeup, but certainly much longer to go through all of the layers from I to VI.2
Our hypothesis of a gradual evolution is in agreement with that of other students of language evolution (e.g. Pinker and Bloom 1990: 721; Jackendoff 2002; Givo̒n 2002a, 2005). Agreement concerns the following points: First, language evolution was gradual and continuous rather than abrupt or catastrophic. Second, there was a series of intermediate steps or stages on the way from early language to the modern languages. Third, evolution led incrementally from less complex to increasingly more complex linguistic structures. And fourth, increase in complexity led from smaller to larger utterances and from non-hierarchical to hierarchical discourse organization.
In defense of the leap hypothesis, Crow (2002: 94, 107) observes that ‘‘[l]anguage is an embarrassment for gradualist evolutionary theory because, according to some authors, it requires a sa[e]ltation, that is, a discontinuous ‘speciation event’,’’ and he asserts that there are two problems with a Darwinian gradualist hypothesis. First, there is no evidence of a gradual accumulation of linguistic capabilities over a long period. And second, he asserts that the dispersal and geographical isolation of modern humans must have occurred after the appearance of language, with the propensity for language being present in all humans despite that dispersal and subsequent isolation.
In light of the findings presented here, neither of the problems discussed by Crow (2002) are necessarily an embarrassment for gradualists. With reference to his first assertion, we saw earlier that findings on grammaticalization suggest that there must have been a gradual accumulation of grammatical distinctions in the evolution of human languages (see also Newmeyer2002). Crow’s second assertion—supposing that there is evidence to substantiate it—opens a number of possible interpretations but does not lead to any clear conclusion. Irrespective of whether the appearance of language occurred prior to or after the dispersal and geographical isolation of modern humans, that appearance could have been discontinuous or continuous. If it had occurred after the dispersal, this would have meant that the sequence of grammaticalization processes took place more than once; as we will argue in “Looking for answers”, observations on grammaticalization can be reconciled with both a mono- and a poly-genetic hypothesis (for a similar conclusion using a different basis of argumentation, see Bickerton 2005). To conclude, neither of Crow’s arguments can be taken to be a challenge to the gradualist hypothesis.
But there is one observation suggesting that the two hypotheses sketched above are not necessarily mutually incompatible. As we observed, grammaticalization is a process that needs time to take place. Still, as we also observed, it is reasonable to assume that within a few thousand years a language can acquire a new grammatical makeup. For example, two millennia were enough in the development from Latin to modern Romance languages to introduce new modes of marking case, tense, aspect, modality, and subordination, and even less time was required for a similar development from Old English to present-day English. There are in fact scholars who argue that, while the transition to modern language was not abrupt, it nevertheless proceeded rapidly. Li (2002) is one of them:
Once words are sequenced to form larger linguistic units, grammar emerges naturally and rapidly within a few generations. The emergence of grammar in the first generation of speakers of a creole attests to the speed of the process. The speed of the emergence of the first grammar at the inception of language is astronomical in comparison to the speed of Darwinian evolution. (Li 2002: 90)
Note that Li’s concern is not with the transition from early language to modern languages but rather with the transition to a grammar consisting of a few word order principles and grammatical markers. While we agree overall with his scenario, we are hesitant to accept that a few generations were really sufficient to produce grammatical structure, or that observations on creole development provide an appropriate basis for reconstruction. Ignoring the fact that the neurological, cognitive, and socio-cultural conditions characterizing the situation of early language creators cannot have been the same as that of early creole speakers, there remains the fact that the former had no linguistic models to rely on whereas creole speakers were able to draw on multiple models provided by other language varieties, be that a ‘‘lexifier’’ language or ‘‘substrate’’ languages, or any other languages that presumably were available in the multilingual setting characterizing early creole development; the possibility that these contrasting conditions affected the relative pace at which grammars developed cannot be entirely ruled out.
Nevertheless, the evolution of language from the lexical structure as hypothesized for layers I and II to a well-grammaticalized morphosyntax of layer VI may have happened within a relatively short period of, say, ten to twenty thousand years: Assuming that human evolution from Homo erectus to Homo sapiens sapiens extended over a timespan of around 1.5 million years, the impression of a kind of leap or catastrophic shift can in fact arise.
However, this is but one possibility. An alternative scenario would be that the evolution of grammatical complexity from layer I to VI extended over a large period of time, possibly as much as 100,000 years. Considering the fact that the speakers of early language did not have any models to draw on in order to develop, for example, new structures of combining clauses, of establishing reference relations, of expressing abstract grammatical concepts of reference and modality, etc., such a scenario appears to be more plausible. But so far there is no appropriate evidence to decide in favor of one scenario over the other.
This means on the one hand that the evidence from grammaticalization does not allow for a firm conclusion on the gradualist vs. leap issue. On the other hand, this evidence is clearly in favor of the former hypothesis: It suggests that the evolution from concrete lexical to abstract inflectional structures, or from basic sentences to complex structures of clause subordination required a number of intermediate stages and must have taken millennia to be accomplished, and since such processes probably took considerably more time when they happened for the first time in human history, the timedepth of this evolution was presumably much larger. We therefore maintain that a hypothesis in terms of an abrupt, catastrophic transition from non-language to language, at least of the kind described above, does not provide an appropriate basis for reconstructing the evolution of grammar.
1 That there are good reasons to argue that there is something like ‘‘partial syntax’’ is suggested by the structure of restricted linguistic systems (see “Other restricted systems”). Also, Pinker and Bloom (1990) provide examples that can be interpreted as exemplifying ‘‘partial syntax’’ (Fritz Newmeyer, p.c.).
2 On the other hand, Fritz Newmeyer (p.c.) suggests that early language did not have fully-fledged modern languages spoken around it, which could possibly act as a conservativizing, braking force on the speed of language change.
