In lexical studies collocation and semantic prosody/preference can only be quantified reliably on the basis of corpus data.