Are valency patterns item-based constructions?
This paper investigates the nature of valency patterns in the light of the theoretical issues that arose during the compilation of the Valency Dictionary of English (Thomas Herbst, David Heath, Ian Roe, Dieter Götz; Berlin/New York: Mouton de Gruyter 2004), which provides a detailed valency description of more than 500 English verbs.
For a number of reasons it seems appropriate to describe the valency of a verb etc. not just in terms of an inventory of valency complements but in terms of valency patterns. One such reason is that different patterns containing complements expressing the same (semantic) arguments do not necessarily permit the same lexical items. Valency patterns in this sense can then be regarded as item-based constructions. This has the further advantage that certain item-specific constructions of certain adjectives or of shell-nouns, which strictly speaking would not be part of a valency analysis, can be subsumed under such valency patterns.
What remains doubtful, however, is to what extent it can be claimed that each of the several hundred valency patterns of English can be attributed a particular meaning. While in many cases it can certainly be shown that the various uses of one pattern are semantically related (e.g. by metaphorical extension), it is much more difficult to provide a description that is sufficiently precise to cover all possible realizations of complements in a particular pattern and exclude those that do not occur. In many cases, VDE provides general semantic descriptions corresponding to inherent semantic features or semantic roles of the complements occurring in a pattern, in other cases lists of collocates are given since it did not seem possible to subsume all uses under a satisfactory label. It will be asked whether it is reasonable to assume that this gradience in terms of lexicographical describability is a reflection of what happens in language acquisition. If one applies the ideas central to usage-based approaches such as those by Tomasello (2003) or Bybee (1985, 2002) to valency, it might make sense to assume that storage plays a key role in the acquisition of valency patterns and that on the basis of stored information generalizations about the semantic features of complements are made where this is possible. In cases where the material does not allow such generalizations, the combinatory possibilities may remain stored as idiosyncratic or item-specific lexical combinations or collocations.