Rendering natural collocations in a translation task: The effect of direction, congruency, semantic transparency, and proficiency

Abstract

Second language (L2) research has extensively examined various factors that might influence the processing and learning of collocations, including congruency, semantic transparency, L2 proficiency, and (to a limited extent) directionality. However, research examining the effect of these factors on the natural rendering of collocations by translation trainees is scarce. The present study aims to apply the factors identified in L2 research in a natural translation context. The aim is to examine whether such factors may influence rendering collocations naturally from the source language to the target language. Ninety‐two Arabic–English translator trainees took part in the study. They translated short equivalent passages containing 40 collocations in both directions (Arabic—English and English—Arabic). A mixed‐effects model was fit to examine factors that may predict the production of a natural collocation: translation direction, L1 congruency, semantic transparency, and proficiency. Results showed no differences between the production of natural collocations to and from the native language. However, there was a significant effect of congruency and semantic transparency; congruent collocations and collocations that had a transparent meaning were more likely to be rendered naturally. Moreover, estimated proficiency level in English modulated congruency effects in that the difference between congruent and incongruent items declined as estimated proficiency increased. Results of the present study suggest that second language models and translation models should incorporate semantic transparency of collocations. Moreover, the results imply that collocations and their properties should be explicitly taught in translation classrooms.