A multifactorial study of English-to-Polish translation of reporting verbs in literary novels: A negative binomial regression with mixed effects

Situated at the intersection of corpus stylistics, translation studies, and multifactorial statistics, this paper focuses on the identification of the predictors of preservation or avoidance of repetition in English-to-Polish translation of repeatedly used reporting verbs signalling direct speech. Using a sample of 20 literary texts, we fit multiple negative binomial regression with mixed effects models to assess the effect that seven predictor variables (i.e. frequency of a source-text verb, number of its translation equivalents in lexical databases, its number of senses, its semantic type, its length in characters, date of translation of a novel, and individual translators) have on the response variable: the number of Polish target-text reporting verbs (types) an English source-text reporting verb is translated into. The overall model fit per the lowest AIC (Akaike Information Criterion) and BIC (Bayesian Information Criterion) values obtained through backward elimination reveals that the frequency of a source-text reporting verb, its semantic type as well as the individual translators have the largest incremental contribution to the model’s fit. More precisely, the proportion of variance in the outcome variable explained by both fixed and random effects (76%) is higher than the proportion of variance explained by fixed effects alone (72%). Without the translators treated as a random intercept, some variation would have been unaccounted for. The findings attempt to explain the translator’s decisions, whether to avoid or preserve patterns of repetition in source texts, with respect to rendering reporting verbs, which play an important stylistic effect in literary prose.