October 7, 16-17.30 h.,Erasmusbuilding 14.11Åshild Næss, University of Nijmegen
Abstract: It is well known that in English, Dutch, and many other languages, both European and otherwise, the verb 'eat' has the peculiarity of being ''ambitransitive'', that is, it is equally grammatical with a direct object (I'm eating a sandwich) as without one (I'm eating). What is perhaps less well known is that this is only one manifestation of a broader tendency for verbs meaning 'eat' to occur in formally intransitive constructions crosslinguistically. In some languages, 'eat' is a so-called extended intransitive verb, taking an oblique rather than a direct object; in others, it is simply intransitive, and must undergo morphological derivation if it is to be used with a direct object; in yet others, it occurs in causative constructions normally reserved for intransitive verbs, and so on.
Current theories of the semantic properties of transitive constructions have no obvious way of explaining these patterns. According to the generally accepted notion of a transitive construction as one referring to an event involving a volitonally causing agent and a highly affected patient participant, 'eat' should be a highly transitive verb, and in fact is treated as such in a number of accounts such as Andrews (1985). It has been argued that the relevant property is that of telicity, intransitive uses of 'eat' being necessarily atelic; but this is not in fact the case, as can be seen from English sentences such as I ate in five minutes before rushing off to work.
In this talk, I will argue that the semantic property responsible for the recurrent use of 'eat' in intransitive constructions is that of that of having an affected agentargument (Saksena 1980). A verb like ‘eat’ not only affects its patient, it also has a consistent and salient effect on its agent, and it is this latter effect which generally constitutes the motivation of the agent for performing the act. It is this fact which distinguishes 'eat' from fully transitive verbs like 'kill' or 'break', which typically affect their patient only.
Standard accounts of transitivity refer only to affectedness of the patient and to properties such as volitionality and animacy of the agent, and so have no way of explaining how affectedness of the agent would render a verb less semantically transitive. I will argue that a prototypical transitive clause must be defined in terms of maximal semantic opposition of its arguments: if it is a defining characteristic of an agent argument that it volitionally instigates the action, and of the patient that it is affected by the action, then it is also characteristic of the patient that it is not volitional and instigating, and of the agent that it is nonaffected. In other words, a prototypical transitive clause has one volitional instigator and one affected argument, and any deviation from this maximal semantic distinction between arguments leads to a less semantically transitive clause, which may be reflected formally in the encoding of such a clause differently from clauses which doshow such maximal semantic opposition of arguments.