Combining Mutual Correspondence and Textual Variation in a Study of English GO and Norwegian GÅ

This article expands and nuances the procedures of calculating translation bias (TB) and mutual correspondence (MC) in contrastive studies. Both measures are intended to show how similar, or mutually translatable, two linguistic items are across languages. Traditionally, these procedures have been carried out without attending to variation across individual texts in a corpus. In a contrastive case study of English GO and Norwegian GÅ, we calculate the TB and MC of the fully congruent instances, i.e., instances where the verbs correspond to each other in the same lexicogrammatical patterns, both in the traditional way and in a newly developed procedure where textual distribution is integrated into the measures. The study substantiates the initial hypothesis that the distributive variants of TB and MC—Distributive Translation Bias and Distributive Mutual Correspondences—are more informative and reliable, as the values more accurately show the range and distribution of correspondences. The study also suggests that the strict definition of congruence adopted (requiring lexicogrammatical correspondence) is better suited to capture details of cross-linguistic similarity than simply to require formal correspondence at the level of word class. This is made evident by the varying degrees of mutual correspondence in the different verb categories, e.g., the generally higher correspondence measures for intransitive GO and GÅ compared to the phrasal verb use.