Light Verb Constructions

In this paper, we present our corpus-based, constructional account of light verb constructions in the Indo-Aryan language Odia (earlier known as Oriya). These light verb (LV) constructions are asymmetric complex verb predicates of the form V-i-v, combining a main verb (V) with a light verb (v), joined by the linking morpheme  i. In Odia, there are 10 ‘true’ LVs (see Lemmens & Sahoo (forthc.) for discussion):
(1) MOTION verbs:  jā ‘ go’,  tʃāl ‘ walk’,  paɖ ‘ fall’,  pakā ‘ drop’,  uʈh ‘ rise’,  ās ‘ come’
(2) STATIVE verbs:  bas ‘ sit’,  rah ‘ stay’
(3) TRANSFER verbs:  d̪e ‘ give’,  ne ‘ take’
While the LVs are form-identical with a lexical verb, they have lost their original lexical content as well as their argument structure. Taking a constructional approach (see Goldberg 2006), we argue that LV constructions present a coherent system of event modulators (cf. Butt & Lahiri 2001), as shown in Figure 1 (appendix). First, all LVs modulate the interpretation of the event encoded by the main verb by adding a particular aspectual (phasal) profile on the event (ONSET, DURATION or COMPLETION, see example (1)). Secondly, some LVs further add a mirative interpretation, i.e., they express that the event is unexpected or not supposed to have happened. These ‘aspectuo-mirative’ LVs can be characterised as non-parasitic expressions of mirativity (T. Peterson 2015) whose main purpose is to express surprise, as opposed to parasitic expressions where pragmatic inferencing may lead to a mirative interpretation. Where precisely the mirative interpretation comes from for these LVs still requires further (diachronic) research. Based on corpus evidence (drawn from the Emille Corpus), we further claim that these aspectuo-mirative LVs differ in the degree with which they express mirativity:  paɖ ‘ fall’ and  pakā ‘ drop’ express a higher degree of mirativity than do  jā ‘ go’ and  d̪e ‘ give’ (which are also the most frequent ones).
Strikingly, the four durative LVs  ās ‘ come’  rah ‘ stay’,  bas ‘sit’,  tʃāl ‘ walk’ never express mirativity. This is due, we hypothesize, to absence of differentiation between the states within the event profile, as opposed to the LVs profiling ONSET or COMPLETION, where the difference between successive states can be seen as triggering the surprise reading. The distribution of the four different durative LVs cannot be explained easily, as there do not appear to be any specific and strict constraints based on features such as transitivity, telicity, semantic class, agent or event type, etc. However, our corpus data reveal some specific tendencies (partly confirmed by a collostructional analysis).
As far as we know, our constructional account of these LVs presents the first comprehensive account of Odia LVs; moreover, it presents an innovative contribution to the study of LVs in South-Asian languages that is both theoretically coherent and descriptively adequate