Segmentation in Translation: A Look at Expert Behaviour

The present contribution examines an interesting but technically quite imperfect sample recording in the CRITT Centre’s TPR-database in an attempt to demonstrate how imperfect gaze data can be meaningfully reconstructed and to illustrate and explore details of translational keystroke and gaze behaviour in a single translator. The data clearly show that translation proceeds segment by segment. A source text (ST) segment is read, comprehended, and translated. As the translation is typed, we see it emerging segment by segment. Perfectly smooth production of target text across extended stretches of time is not frequently seen, but is often approximated. Highly expert performers are able to bind processing segments together into a flow of continuous production. From their recorded gaze behaviour, we can observe that experts do process text segment by segment, so how is it that they can manage to sometimes maintain fairly continuous production? Evidence of how reading, comprehension, translation, formulation and typing activities are coordinated is found in recorded gaze data, which provide detailed evidence of what ST text unit was being worked on at any given point in time, and evidence provided by keystrokes. These combined sources of evidence can be used to infer both what ST (sub)segment was being processed within what ST context, and in what manner, always with the big unknowns at play of the translator’s knowledge, memory, meaning construction intelligence and expressive power – and the suspicion that the human brain is doing a good deal more than eye movements and keystrokes reveal.