next up previous contents
Next: Entailment Tests Up: Semantic Tests for Argument Previous: Semantic Tests for Argument

Constancy of Semantic Contribution

  Pollard and Sag (1987) discuss the semantic basis for the distinction between arguments and adjuncts as follows:
In general, a given adjunct can co-occur with a relatively broad range of heads while seeming to make a more-or-less uniform contribution to semantic content across that range. A given optional complement, by contrast, is typically limited in its distribution to co-occurrence with a small (and often semantically restricted) class of heads (possibly even a single item); in addition, the semantic contribution of the complement is idiosyncratically dependent on the head. (p. 136)

The - and -datives seem to have a constant semantic contribution across the verbal heads with which they appear. In general terms, the -datives seem to indicate the intended recipient of some object, and the -datives seem to indicate the intended beneficiary of something. This semantic uniformity constitutes evidence in support of their treatment as adjuncts. This unchanging contribution was observed by jack:90 and formalised in terms of ``adjunct rules'' which identify the semantic contribution of particular elements in certain syntactic constructions and indicate how this contribution is to be integrated into the semantic representation for the overall construction. The contribution can also be observed in the similarity of the core structures of dative verbs as presented in verspoor:94.

The dative PPs, however, appear with only a restricted set of verbs which undergo dative alternation (see Levin (1993) for an outline of the semantic classes of verbs relevant to this alternation). These can be divided into subsets of semantically related verbs. As such, the conclusion from their limited distribution might be that these PPs are complements.

If one adds the criterion that the contribution of an adjunct to the semantic content of a phrase should not simply be the filling of some role in the head's relation, however, one finds that the dative PPs cannot be uniformly classified. There are some ``true'' -dative verbs, such as give and tell, for which the semantic information contributed by the dative PP is directly relevant to the meaning of the verb, fills a particular role and would therefore be construed as a ``true'' argument of the verb (Gawron gawron:86's argument PPs). On the other hand, there are ``adjunct'' -dative verbs, such as kick, for which the semantics of the dative PP provides additional information about the situation being described by the sentence, adding a semantic role instead of filling a pre-existing role in the verb's relation (Gawron's co-predicating PPs). We will see that this is so on the basis of the verbal argument structures below.


next up previous contents
Next: Entailment Tests Up: Semantic Tests for Argument Previous: Semantic Tests for Argument