A Theory of (C)overt Object Pronouns in Kwa

Abstract

In this talk, I argue that the distribution of null versus overt object pronouns in Gã and several other Kwa languages) can best be explained in structural terms; all overt pronouns are realised in a specifier position while null object pronouns are deleted in a complement position.

Kwa languages like Akan, Baule, Ga˜, Nzema, a.o. exhibit a null object pattern whose profile does not seem to fit any of the types traditionally acknowledged in theoretical linguistics lit- erature, i.e., the possibility of omitting an object pronoun does not depend on (a) agreement marking (contra Jaeggli 1982; Rizzi 1986), (b) topicality (contra Huang 1984), or (c) the mor- phology of their pronominal system (contra Neeleman & Szendroi 2007). Thus, in Ga˜, for instance, the realisation of third person object pronouns can be summarised as in (1).

CONTEXT +ANIM -ANIM

a. Clause-final overt null

(1) b. Argument of change of state predicates overt overt

c. Before verbs overt overt

d. Argument of depictive predicates overt overt

From (1), it is apparent that we are confronted with a new kind of null object, i.e., one that seems to be conditioned by clausal-finality and animacy.The question thus arises: What accounts for the natural classes; what explains the uniform realisation of the inanimate object pronoun in (1-b,c,d), and all animate object pronouns, to the exclusion of the inanimate object pronoun in (1-a). At first sight, a uniform explanation of (1) appears illusive, and existing proposals in the Kwa literature, e.g., Chinebuah (1976), Saah (1992), Osam (1996), and Larson (2005), for which (1) is revived problem appear not to provide a satisfactory solution either.

I claim, following Kayne’s (1994) Linear Correspondence Axiom (LCA), that we can trace the (non-)pronunciation of an object pronoun to where in the syntactic structure it is located (i.e., specifier or complement) at the point of linearisation. I will provide independently-motivated arguments to show that this analysis works for the Ga˜ data problem presented above, and that we can extend same to the analysis of objectpronouns in other ((un)related) languages, such as Ewe and Dagaare. The conclusion, thus, will be that the overt-covert object pronoun distinction that we find in Kwa languages is a reflection of a specifier-complement asymmetry in grammar.