GO (a)round and V vs. GO (a)round Ving: A multivariate distinctive collexeme analysis based on association rules

Distinctive collexeme analysis has proven very useful in identifying distinctive patterns of lexemic attraction among multiple constructions. However, if construction grammar is to take seriously the usage-based tenet that context is a crucial component of the language system, multivariate methods are required. In this article, one such analytical approach is proposed as we apply an extension of distinctive collexeme analysis – named multivariate collexeme analysis – in an analysis of the GO (a)round and V and GO (a)round Ving constructions. Based on the data-mining technique known as association rules, multivariate collexeme analysis can identify not just singular distinctive features but also establish entire ‘collo-profiles’ of multiple features going far beyond individual collexemes. Our analysis takes into account no less than eight features (including collexemic, colligational, pragmatic, and discursive ones), and it is found that this approach offers a much more informative account of the two constructions than a traditional distinctive collexeme analysis would.