Cognitive Effort and Perceptions in Translation Revision and Post-Editing

Abstract

This study investigates the correlation between cognitive effort and translator perceptions during human translation (HT) revision and machine translation (MT) post-editing. Seven participants revised and post-edited two texts without knowing whether the texts had been produced by a human or a machine. Eye-tracking and keylogging captured fixation and pause durations (FD and PD) to measure cognitive effort, while a one-question survey assessed participants’ perceptions of translation type. Linear Mixed Models (LMM) were used to analyze fixation and pause duration data at both individual and group levels. FD results showed that MT revisions generally required greater cognitive effort for source texts but slightly less for target texts, while PD showed no uniform trend toward either lower or higher cognitive effort. Survey responses indicated a preference for identifying HT with higher confidence, while MT was frequently misclassified as HT. This suggests a possible correlation between misclassification of MT as HT and higher fixation durations, indicating that perception might influence cognitive processing. The study offers insights into translation practice, particularly by highlighting how assumptions about whether a text was produced by HT or MT may affect revision and post-editing decisions.