Adjective-noun collocations in elicited and corpus data: Similarities, differences, and the whys and wherefores
Iain McGee · Corpus Linguistics and Linguistic Theory · 2009
This paper aims to provide a theory to help explain the similarities and differences between corpus and elicited data in the area of frequent adjective-noun collocations. The study begins with an overview of existing data and theories from word frequency estimation studies and word association studies. This is followed by a critical analysis of three explanations for elicited-corpus data differences (Sinclair, Corpus, Concordance, Collocation, Oxford University Press, 1991; Bybee and Hopper, Frequency and the Emergence of Linguistic Structure, John Benjamins, 2001; and Wray, Formulaic Language and the Lexicon, Cambridge University Press, 2002). I then report on an experiment designed to compare British National Corpus (BNC) data and English language teacher intuitions about the most frequent collocates of some very common adjectives in the English language. It is argued that the data provide support for the theory that a key factor affecting the ‘quality’ of lexical intuitions may be the employment of an availability heuristic in judgments of frequency. It is argued that in an elicitation task some collocates of words (particularly those typically occurring together with the stimulus word in a larger language chain) may be more hidden from memory searches than other collocates which tend to occur with the stimulus word as a ‘bare’ dyad in typical usage.