Corpus evidence and the limits of intuition in the study of collocation

Drawing on data from the Corpus of Contemporary American English (COCA) and the Russian National Corpus, the article examines how reliably native-speaker intuition describes collocation and semantic prosody. It is shown that speakers can usually judge whether a word combination is acceptable, but they are much less successful in assessing its frequency, its typical lexical environment, and the positive, negative or neutral evaluative colouring that a word acquires from its habitual contexts. Such features become visible only when large amounts of text are analysed. A contrastive analysis of English and Russian material demonstrates that these patterns are language-specific and cannot be transferred from one language to another through translation or introspection. On this basis, conclusions are drawn for Uzbek: the need to build and use corpora, to verify phraseological and lexicographic descriptions against corpus evidence, and to take collocational and prosodic information into account in teaching Uzbek and in translation.