Word order
What was the order of linear arrangement of meaningful elements in early language?1 The main hypotheses that have been proposed on this issue are listed in (1).
(1) Hypotheses on linear word arrangement in early language
a. Words were simply juxtaposed without any ordering principle.
b. Linear arrangement was determined by pragmatic principles, involving an arrangement topic–comment (or given–new), or non-focus–focus constituent.
c. Word order was based on semantic principles, where e.g. an agent precedes a patient/undergoer.
d. Word order was based on the syntactic notions subject, object, and verb.
Hypothesis (1a) has occasionally been mentioned but, to our knowledge, not been elaborated in any detail. It is implied, for instance, in Calvin and Bickerton’s (2000: 137) understanding of early language, according to which language began without any formal structure—being just handfuls of words or gestures strung together.
A number of students who have worked on this issue have come up with proposals in terms of (1b). What Bickerton (1990) proposed to call ‘‘protolanguage’’ is said to have had an arrangement of elements determined by pragmatic rather than syntactic mechanisms. This view can be reconciled with that of Jackendoff (1999, 2002), although his position is more difficult to locate, being a mixture of (1b) and (1c): On the one hand he invokes semantic principles, maintaining that in what he calls protolanguage there was no notion of the subject and object of a sentence, only semantically defined notions like agent and patient. On the other hand, he proposes principles that are suggestive of pragmatic notions, arguing that in the discourse coding of given and new information there were the principles ‘‘Agent First’’ and ‘‘Focus Last’’, where the agent was expressed in subject position and the informationally focal element resided in the last position. Claiming that these are ‘‘fossil principles from protolanguage, which mod ern languages often observe and frequently elaborate,’’ he finds evidence for this hypothesis not only in the Basic Variety of late second language acquisition, but also in pidgin languages and agrammatic aphasics.
A hypothesis in favor of (1c) is proposed by Newmeyer (2000): He argues that syntactic categories presuppose semantic categories, and that the latter preceded the former in time:
In particular, since there is a rough correlation between the semantic notions ‘predicate’, ‘argument’, and ‘proposition’ and the syntactic categories ‘V’, ‘NP’ and ‘S’, respectively, it seems reasonable to hypothesize that, as language evolved, the latter were grammaticalizations of the former. (Newmeyer 2000: 388)
There is in fact support from grammaticalization for (1c). For example, semantic head–dependent patterns may give rise to syntactic head–modifier constructions, and we discussed cases where a sequence of two clauses S1--- S2, having a semantic relation where S1 expresses the main proposition while S2 provides qualifying or modifying information on S1, may grammaticalize into syntactic main clause–subordinate clause constructions. Conversely, we are not aware of regular processes leading from syntactic constructions to semantic configurations. Accordingly, (1d) would seem to be a less plausible hypothesis.
But Newmeyer (2000, 2003) also proposes, at least for a stage following that of a ‘‘protolanguage’’, a hypothesis in terms of (1d), according to which word order was syntactically organized—in that the human language had verb-final (OV) word order. Unlike the other authors mentioned, he provides an impressive list of arguments in support of this hypothesis, in particular the following: (a) OV predominates among the languages spoken today; (b) the earliest languages were OV; (c) the historical change OV > VO (= verb before object) is both more common than the change VO>OV and more ‘‘natural’’; and(d)OV languages are more likely to have alternative orderings of S, V, and O than do VO languages. Nevertheless, this hypothesis has to be taken with care: With the exception of (a), for which there is sufficient evidence, none of the arguments listed is really uncontroversial, and we still lack appropriate typological and diachronic information to verify them.
Findings on grammaticalization suggest in fact that there is a fairly regular development leading from pragmatically motivated structures to syntactic structures—a process referred to as syntacticization (Givo̒n 1979b, 1979c). On the basis of our bridge hypothesis (“The present approach”), this would lead to the conclusion that (1b) preceded (1d) in time—in other words, that (1b) would be the most plausible hypothesis. While many languages provide some kind of evidence that pragmatically motivated structures tend to acquire syntactic properties, such as those characterizing clausal subjects or objects, we are not aware of any language undergoing a process in the opposite direction, where a subject or object developed into a topic or focus constituent.
But there is an alternative perspective to this issue. Pragmatic functions are most commonly encoded by means of variation in word order, supra-segmental forms (e.g. stress), morphological material, or periphrastic constructions that are non-pragmatic in nature. It would therefore seem that the unidirectionality sketched above describes only one aspect of the development of pragmatic structures such as topic and focus constructions, in that the latter themselves can be derived from non-pragmatic structures. These observations suggest an overall development of the kind sketched in (2). Note that while (2) depicts a common pathway of grammaticalization, this is not the only one leading to the rise of pragmatic and syntactic functions.
(2) Complex syntactic structure > pragmatic function > syntactic function
With reference to the reconstruction of early language, this means that grammaticalization theory does not allow for any general hypothesis that would be supported by substantial evidence. There are, however, two conclusions to be drawn from grammaticalization evidence. One is that, in accordance with Newmeyer’s (2000) hypothesis that semantic arrangements can give rise to syntactic ones while the reverse is unlikely to happen, hypothesis (1d) can be ruled out. The second conclusion is that the remaining hypotheses (1b) and(1c) are mutually compatible in the following: Since topics generally precede comments and agents precede predicates, the two hypotheses converge on the ordering topic and agent before comment and predicate, respectively. This convergence is also in accordance with the preferred argument structure in the management of information flow in discourse, where speakers tend to place given information in the agent slot of transitive clauses (Du Bois 1987).
Accordingly, irrespective of whether early language from layer II onward was shaped by pragmatic or semantic principles, the order must have been topic/agent before comment/predicate—as hypothesized by Givo̒n (2002a), Jackendoff (1999, 2002), and others.
1 Since it is only at layer II that a differentiation of meaningful elements arose, this question does not concern layer I.
