The cognitive reality of frequent verb-noun combinations: An empirical study

The cognitive reality of frequency is often taken for granted by linguists. If a particular linguistic phenomenon is frequent, the argument goes, it must have some relevance for cognition. Several studies have indeed suggested that frequency of occurrence has a place in the cognitive system and that information about frequency might actually be stored in memory by an automatic encoding process (cf. Hasher 1984). This would lead to the expectation that frequent items are somehow more salient or, as Schmid (2000: 39) puts it in the “From-Corpus-to-Cognition Principle”, that “[f]requency in text instantiates entrenchment in the cognitive system”. Some recent studies, however, have emphasised the discrepancy that may exist between frequently attested items and cognitively salient ones (e.g. Shortall, in preparation). In an earlier study (Gilquin 2005, 2006), I showed that the most frequent sense of a word does not necessarily correspond to the sense that comes first to mind when prompted for this word. More particularly, I demonstrated that, while highly polysemous verbs such as take or give are mostly used in their delexical sense in a corpus of spontaneous conversations (e.g. 'take a walk' or 'give a smile'), people seem to access more concrete senses first (e.g. 'take somebody home' or 'give somebody a book'). In this presentation, the focus will be on combinations of a high-frequency verb and a noun phrase, and the approach will be more lexical. Applying Gries & Stefanowitsch’s (2004) method of distinctive collexeme analysis to the comparison of one construction in two language varieties (rather than two constructions in one language variety), I will compare the noun phrase of such combinations in free language production (corpus data) and in controlled language production (elicitation data). The degree of overlap and the rate of lexical variation will be measured, and the grouping of distinctive collexemes into semantic fields will be examined. The observation of the differences and similarities between the collostructions in corpus and elicitation data will provide greater insight into the question of the cognitive reality of linguistic frequency