Two types of translation equivalence

This paper looks at two examples of unexpected correspondences that were found in a translation corpus: the German word kaum and its equivalents in English, and the English word contain with its counterparts in French. The patterns that came to light were different: for kaum, the expected translation equivalent hardly is relatively rare, but the range of actual equivalents is small and tractable, whereas for contain, the expected equivalent contenir is relatively common but there are many unique equivalents. We claim that kaum is “lexicographically complex” with respect to English, whereas contain is “translationally under-specified” with respect to French. We examine why this might be, drawing on the notion of “modulation” in translation theory, and we consider the implications for lexicographers, translators and contrastive linguists. The data are taken from the INTERSECT corpus, consisting of about 1.5 million words in French and English and about 800,000 words in German and English.