On uniformitarianism
It has been argued, implicitly or explicitly, that early language evolution was different in kind from modern language change, and that the study of the former therefore requires assumptions of a type that are not necessarily relevant to the study of the latter. Such discussions relate in particular to a notion that we consider relevant to any methodology employed for reconstructing early language, namely that of uniformitarianism. It is thanks to Newmeyer (2002, 2003) that this notion has received some attention in works on early language. The term has been used in a number of different senses, and these senses are not necessarily compatible with one another. To our knowledge, the following are senses of uniformitarianism that have been proposed in linguistics, which may be referred to as the U1, U2, and U3 assumptions.
(12) Assumptions of uniformitarianism
U1 All modern languages are in some important sense equal.
U2 Since the general structure of human languages of 5000 years back was about the same as it is today, it must also have been the same in early language.1
U3 Linguistic change in early language was of the same kind as we observe in modern languages.
U1 is fairly widely accepted in contemporary linguistics. It is also the assumption adopted here, even if there are some languages that are claimed to lack salient properties of human languages, such as the Brazilian language Pirahã, which is said to show a lack of color words, numbers and counting, and recursion (Everett 2005; but see “Are there languages without recursion?”), or Riau Indonesian, which is said to lack syntactic categories (Gil 2001).
As we observed in (10) above, U3 is central to our framework (see also Christy 1983: 2). While assuming that early language differed in structure from modern languages, we argue that the processes of grammatical change were the same as those characterizing modern languages, both being shaped by parameters of grammaticalization.
U2 requires more detailed attention. Newmeyer (2002: 358) observes that most research into language origin and evolution has taken a position in accordance with U2, but he concludes that this position should be rejected:
In other words, we have no reason to believe, and every reason to doubt, that the functionally motivated aspects of grammar have remained constant over time. (Newmeyer 2003: 75)
We side with Newmeyer’s (2002, 2003) position on U2, in accordance with the approach described in “The present approach”. As we saw there, the perspective on language change that we adopt is in accordance with U3, that is, it is based on the assumption that the processes of language change were the same in the past as they are in the present—hence, that it is possible to use generalizations on documented cases of language change to reconstruct earlier language states. And with reference to language structure, that is, with reference to U2, we also agree with Newmeyer in adopting a distinctly non-uniformitarian position: We hypothesize that early language had a structure that was different from the one characterizing modern languages—thereby following Comrie (2002: 257), who argues ‘‘that it is a valid exercise to reconstruct stages in the development of human language typologically different from, in particular less complex than, attested human languages’’.
That early language was typologically less complex than modern languages has in fact been maintained by many students of language evolution, and a number of scenarios have been proposed to account for this fact (see especially Givo̒n 2002a, 2005; Jackendoff 2002; Johansson 2005, 2006). But this raises the question of how to reconcile this hypothesis with U3, according to which linguistic change was of the same kind throughout language evolution.
A possible piece of an answer is volunteered by Newmeyer (2002: 368-9), who argues that overall pre-literate societies make less use of clause subordination than literate societies. There is in fact some evidence to show that there are significant differences between spoken and written language use and reasons have been provided to account for this difference (e.g. Tannen 1982b: 3, 8; Chafe 1982: 39–45; Romaine 1992a: 147, 159; Harris and Campbell 1995: 310; Croft 2000: 83–4). Such observations might be taken to indicate that language use in pre-literate societies and/ or in spoken communication reflects an earlier stage of language evolution. However, there are problems with such a view. First, the quantitative evidence in support of this view is still not entirely satisfactory. Second, as also acknowledged by Newmeyer (2002: 369), there are many pre-literate societies having languages with a remarkable amount of subordination and other forms of recursive structures (see e.g. König in prep.). In fact, a number of languages of traditional hunter-gatherer societies do not differ markedly with regard to subordination from languages of societies having had a long history of literacy—speakers of the Central Khoisan languages of southwestern Africa, for example, probably make more use of subordination than speakers of Italian or Chinese. And third, while it seems plausible that written discourse is more likely to invite clause subordination than spoken discourse,2 this does not mean that written languages necessarily dispose of richer inventories of subordinating categories.
Our reason for rejecting assumption U2 is of a different nature. When introducing our methodology in “The present approach” we observed that there are auxiliaries in English or other languages that can be historically derived in a principled way from lexical verbs. We therefore proposed a reconstruction procedure in (9), leading to the hypothesis that at some stage in early language there were lexical verbs but no auxiliaries. On the basis of this hypothesis we are led to conclude that early language at that stage was structurally less complex than modern languages since the former is hypothesized to have lacked a grammatical distinction that is present in modern languages.
But the situation is more complex. It is not for the first time that the process from verb to auxiliary that we discussed in “The present approach” happened in English, or in other languages for that matter. New auxiliaries and other grammatical categories arise at all times in the history of a given language while old ones may co-exist with the new ones, or develop further into clitics and affixes, or disappear altogether, frequently—though not necessarily—being replaced by other new categories (see “An outline of grammatical evolution conclusion”). And much the same applies to the many other grammatical categories; accordingly, grammatical development in the modern languages is, at least to some extent, cyclical. The result is that modern languages are replete with grammatical categories, some of them less grammaticalized, others more strongly grammaticalized, and modern languages therefore exhibit a fairly high degree of grammatical complexity. Our concern, however, is exclusively with the situation that we hypothesize to have characterized early language, when these processes took place for the first time, that is, when there were, for example, verbs but no auxiliaries—hence, when human language was less complex than it is today. On the basis of this hypothesis, we see no reason to adopt assumption U2.
1 Comrie (2002: 252) provides the following analogy to question the logic underlying U2: ‘‘[F]rom the fact that my parents were humans, and that their parents were humans, and that their parents were humans, etc., one cannot logically conclude that there have been humans around from the beginning of time’’.
2 For example, Harris and Campbell (1995: 310) assert: ‘‘It is clear that hypotaxis is well suited to written language, for it provides a means for packing a great deal of information into a few words and has the potential of making the relationships among ideas specific. Complex hypotaxis is less appropriate in spoken language, since it places a greater burden on memory and processing ability’’.
