Another Look at the Topology of Serial Verb Constructions: The Grammaticalization of Temporal Relations in Bislama (Vanuatu)

Abstract

This paper analyzes serial verb constructions (SVCs) in Bislama, the creole spoken in Vanuatu. The form of SVCs in Bislama does not appear to fit neatly into existing formal typologies of SVCs, and evidence is introduced in favor of this pattern being a direct transfer from some of the substrate languages in Vanuatu. It is proposed that Seuren's (1990) typology of SVCs be extended to include the Vanuatu data.

The paper notes that another property of Bislama and other languages of Vanuatu is that there is no grammaticalization of a past/nonpast temporal relation as in, for example, English. One way of analyzing this difference is to say that Bislama and English differ with respect to which relationship between temporal operators is grammaticalized in a finite clause: in English, the ordering of S(peech time) and E(vent time) is realized overtly; in Bislama, the ordering of S and R(eference time) is. If R is associated with a functional projection (FP) below tense (TP), a number of constraints on the second verb in Bislama SVCs follow from the claim that this FP is anaphorically dependent on the first clause and the second verb phrase is not finite (but also not nonfinite).

1. Introduction.

This paper analyzes series of verbs in Bislama (the pidgin/creole spoken in Vanuatu).¹ It shows that serial verb constructions (SVCs) in Bislama (and several substrate languages) do not fit neatly into existing typologies of SVCs, and the paper demonstrates how such typologies need to be extended to accommodate the Vanuatu data. It suggests that the form of these constructions in Bislama and several constraints barring modification of the second verb in series may be accounted for by positing an impoverished structure associated with the second verb. It is suggested that the second verb in series lacks an independent specification for what Reichenbach (1947) calls reference time (R). In Bislama, it is argued, R is required for a clause to be finite. A language-specific definition of finiteness is given (the verb may occur with all person and number agreement options); when it is not finite, an unmarked inflection occurs on the verb.

A terminological point is needed from the outset. Calling these series of lexical verbs SVCs perhaps suggests that (a) SVCs are a more unified class than is actually the case (Durie 1997, Senft 2001), and (b) that they differ from other syntactic constructions in a more systematic way than they in fact do. The class of constructions under consideration in this paper is a good deal narrower than those considered by Durie (1997) in his cross-linguistic review of SVCs. Narrowing the focus of investigation like this is not a covert claim about what is or is not a "proper" SVC; it is simply a pragmatic decision to try and make sure the constructions being compared are indeed comparable in the face of this heterogeneity. There is undoubtedly something to be said about the relationship between the Bislama SVCs discussed here and Bislama verbs that select bare complements, such as want and try, but this matter will only be touched on briefly. In addition, it seems likely that there are similarities between the SVCs to be discussed here and what have been called complex predicates and have been discussed in some detail for Indo-European languages, for example, in Romance (Kayne 1975, Zubizaretta 1985, Burzio 1986), Urdu (Butt 1997), Bengali (Fitzpatrick-Cole and Lahiri 1998), and Hindi (Bhatt 1998). The goal of this paper is not to provide a typology of all such complex verb constructions, but it is hoped that the data will add to the growing cross-linguistic corpus in a manner that contributes something to the development of more general theory.

What the paper focuses on is an unusual property of the second verb in SVCs in Bislama (and some other Vanuatu languages). Although the lower verb is inflected, the usual subject-verb agreement paradigm does not appear. Since this pattern for SVCs has not been discussed in the literature, these data are of some...