Mind the source data! Translation equivalents and translation stimuli from parallel corpora

Statements like ‘Word X of language A is translated with word Y of language B’ are incorrect, although they are quite common: words cannot be translated, as translation takes place on the level of sentences or higher. A better term for the correspondence between lexical items of source texts and their matches in target texts would be translation equivalence (Teq). In addition to Teq, there exists a reverse relation—translation stimulation (Tst), which is a correspondence between the lexical items of target texts and their matches (=stimuli) in source texts. Translation equivalents and translation stimuli must be studied separately and based on natural direct translations. It is not advisable to use pseudo-parallel texts, i.e. aligned pairs of translations from a ‘hub’ language, because such data do not reflect real translation processes. Both Teq and Tst are lexical functions, and they are not applicable to function words like prepositions, conjunctions, or particles, although it is technically possible to find Teq and Tst candidates for such words as well. The process of choosing function words when translating does not proceed in the same way as choosing lexical units: first, a relevant construction is chosen, and next, it is filled with relevant function words. In this chapter, the difference between Teq and Tst will be shown in examples from Russian–Finnish and Finnish–Russian parallel corpora. The use of Teq and Tst for translation studies and contrastive semantic research will be discussed, along with the importance of paying attention to the nature of the texts when analysing corpus findings.