Australian complex predicates

Claire Bowern, Rice University

In numerous language families of the world we find structures in which information usually associated with a single verbal head is instead spread over several constituents in the predicate. These ‘complex predicates’ are of many different types, and range from the restructuring predicates of Romance languages (e.g. Samek-Lodovici 2003, Wurnbrand 2001) to the serial verb constructions of Asia and West Africa (cf. Crowley 2002, Lord 1993). One particular type of complex predicate involves a bipartite structure with a usually uninflecting ‘coverb’ (or ‘preverb’) and an inflecting ‘light verb’.

This last type of complex predicate is frequent among Australian languages. The languages under consideration are (or were) spoken mostly in the north of the country, from the Kimberley region of Western Australia to the Daly river area of the Northern Territory, through central Arnhem Land and south to the Central Desert (McGregor 2002). Detailed studies have been completed for several languages, including Warlpiri (Nash 1982, 1986), Jaminjung (Schultze-Berndt 2000), Wagiman (Wilson 1999), Ngarinyin (Saunders 1997) and Bardi (Bowern 2004), however there has been little work on comparing the languages and quantifying the extent to which the constructions in these languages differ. In this talk I aim to answer the following question: Do the complex predicates that we see in Australian languages have the same analysis? That is, are they the same construction? In order to answer this question I examine the following parameters:

• - The extent to which the coverb may appear without an accompanying light verb (possible in Wagiman and Jaminjung, impossible in Bardi, for example)
• - The possibility of inflection on the coverb (aspect marking occurs in Yawuru (Hosokawa 1991) and Wagiman; direction marking occurs in Warlpiri (Nash 1982), but no inflection is possible in Bardi other than reduplication)
• - The sources of coverbs (other verbs, nouns, adjectives, adverbs, or their own word class)
• - The productivity of complex predicate formation
• - The semantics of the light verbs used in these constructions

Finally, I make some comments on the theoretical implications of these data, particularly in relation to argument structure. Complex predicates are problematic for many theories of argument structure (see, for example, Ackerman and Webelhuth 1998), for example because of theta-role assignment from more than one place in the predicate. Central to these problems are whether the argument structure of the predicate as a whole results from the unification of information in the components of the predicate, or from the transfer of argument from the coverb to the light verb. I argue that the Australian data supports a unification-based approach.