The compositionality of English phrasal verbs in terms of imageability

Abstract

There has been considerable discussion about the degree to which English phrasal verbs (PVs) tend to be semantically “compositional” or “non-idiomatic” versus “non-compositional” or “idiomatic”. Much of this discussion relates to evidence that many learners of English as an additional language (L2) find PVs particularly difficult to acquire. This article reports a novel empirical study of the compositionality of PVs which makes use of published and newly collected subjective ratings of lexeme “imageability” (i.e., the extent to which a lexeme evokes mental imagery). The premise is that a PV cannot be wholly compositional if its imageability differs greatly from the imageability of its constituent words (CWs). The study focuses on 150 frequent PVs. Key data are meaning-specific imageability ratings of these PVs and their CWs. Results of multiple linear regression analysis indicate that for native-speakers about 13% of PV imageability is attributable to CW imageability. Literality, measured by two kinds of ratings, emerged as a much stronger predictor. These results suggest that any psychologically real PV compositionality in terms of imageability tends to be weakly accessible to native-speakers engaged in off-line reflection. The results suggest lines of enquiry with respect to treatment of PVs in contexts of instructed L2 acquisition.