Difference between revisions of "Multiverbs and Complex Predicates"

Revision as of 15:31, 19 December 2009

Edited by Gautam Sengupta and Lars Hellan

This page is an attempt at classifying cross-linguistically a variety of types falling under the categories 'multiverb constructions' (MVC) and 'complex predicates' (CP).

For MVCs, a first set of salient examples includes,

(1) from Bangla (Bengali):

(2) from Akan:

(3) from Kistaninya (Kistane):

They have in common the expression of temporally successive events, patterns of argument and tense/aspect sharing, and lack of coordinating items, which is held as typical of 'serial verb' constructions. But they also differ in some respects:

In Bangla and Kistaninya, the last verb in the series has a different form than the preceding verbs, whereas in Akan they all have the same form. Although Bangla and Kistaninya have in common being 'verb last' languages, so that the last verb may be counted as being head of the constructions in question, and in Akan it is perhaps V1 which is head, that factor does not explain the difference in form between head and non-heads. Thus, we have to recognize the distinguishing factor I.1 below.

On the other hand, in Bangla only the last verb has a finite form, while the others are in an aspectually completive form. In contrast, in Kistaninya, all the verbs are finite; in this respect they are like in Akan, except for the special added suffix in all the non-final verbs. This has to be counted as a distinguishing factor I.2:

I.1. In the Akan example all verbs are in the same form, whereas in the Bangla and Kistaninya examples, the non-final verbs share a formative absent in the last verb.

I.2. In the Akan and Kistaninya examples, all verbs are finite, whereas in the Bangla example, only the last verb is finite.

A different group of serial constructions is the one where the verbs - typically just two - describe different aspects of one and the same situation: typically a 'main' verb, and then a VP which, in the total setting, expresses an instrument, or a beneficiary, or a few other options. An example of this type from Bangla, with an instrumental, is given in (4), and from Akan, with a beneficiary, in (5):

(4) Bangla:

(5) Akan:

As is clear, the formal patterns in these examples are like those observed in (1) and (2).

Thus, irrespective of whether temporally consecutive vs. not, a significant portion of the MVCs in the languages considered pattern according to the distinctions entered as I.1 and I.2.

Shifting to a possibly different type, consider (6):

(6) Bangla:

One traditionally calls the first verb here the POLE verb and the other verb the VECTOR verb, where the pole verb can be any transitive or intransitive verb, and the vector verb is one from a set of 15-20 verbs, in this connection determining aspect or orientation. Thus, also these constructions determine a single event, like the type instantiated in (4), and the 'role'-indicating function of the first verbs in (4) and (5) is not too unlike the 'vector' verb in function. The formal pattern in (6) conforms to the pattern already seen above for Bangla MVCs, the first verb being in a completive form, the second being finite. What still sets this type apart from the type in (4) is the restricted set of last verbs, and perhaps the circumstance that the 'vector'-like verbs in (4) and (5) come before the 'main' verb, whereas in (6), the vector verb is last. For short, we will refer to constructions like the one in (6) as VV-constructions, those in (4) and (5) as single-event serial constructions - SE-SV - and those in (1)-(3) as multi-event serial constructions - ME-SV. Cross-cutting this classification are the factors mentioned in I.1 and I.2.

With this as a point of departure, we now look into further types of MVCs. Two types are found in Ga: verbid constructions and extended verb complexes, exemplified in (7) and (8):