Converging evidence from corpus and experimental data to capture idiomaticity

Abstract It is a by now established fact that idiomaticity cannot be equated with non-compositionality alone, but is a complex concept that is also associated with various aspects of formal flexibility. This raises the question to what extent speakers call up these different factors when judging the overall idiomaticity of a phrase. In the present paper, experimental and corpus-linguistic methodology are combined to address this question. For a total of 39 V NP-idioms of the kind make a point or take the plunge, comprising more than 13,000 tokens obtained from the British National Corpus, their compositionality, syntactic, lexico-syntactic, and morphological flexibility were assessed corpus-linguistically. The corpus-based results thereby obtained were then correlated with native speakers' overall idiomaticity judgments in a multiple regression analysis to determine each factor's impact on the overall judgments. The results indicate that speakers indeed rely on multiple factors simultaneously, with lexico-syntactic and morphological factors being even more important than compositionality, and verb-related being more important than NP-related information. Overall, the results back up the theoretical concept of a collocation-idiom continuum, and demonstrate how various, and sometimes competing, motivations determine a phrase's position on this continuum.