A corpus-based study on dative alternation and constructional semantics

Many English verbs that express transfer can be used in two different constructions, one with no preposition ("Rick gave Kate a coffee") and one with the preposition "to" ("Rick gave a coffee to Kate"). One approach to the study of this dative alternation (exemplified in Bresnan et al. 2007) assumes that speakers choose between these two constructions when they use such a verb, and that several features of the verb, the two objects, and the sentence or context may interact to determine that choice. One commonly assumed feature, following Gries & Stefanowitsch (2004)'s study of ICE-GB, is the preference for individual verbs to be used with one or the other construction: "give", for example, is used much more frequently with the preposition-less construction than with the prepositional construction; conversely, "bring" is used much more frequently with the prepositional construction.

This work-in-progress report will discuss results from a similar analysis of the British National Corpus. Focusing on bring, lend, offer, send, and show, we find that the preferences of the verbs can differ depending on whether all items, or only those where the alternating construction is judged to be possible/felicitous, are included in the analysis. Even following G&S's approach of including all items, some verbs differ: lend prefers the prepositional construction in G&S, but the double object construction in our data; send prefers the double object construction in G&S, but the prepositional construction in our data. This casts doubt on the verb meanings that G&S argue for.

We argue that a corpus study of verb meanings that ignores the context of occurrence ignores crucial information, and that careful consideration of alternating constructions must precede semantic and multifactorial analysis of alternation data.