Linear vs. non-linear translation in parallel text reading

Abstract

This study investigated the effect of differences between linear and non-linear translations on reading behavior in a situation where both the source and target texts were presented for revision. Linear translations, which are relatively close to the source text in word order (with a low cross value), and non-linear translations, which are relatively remote (with a high cross value), were created. An experiment was conducted in which participants compared these two types of translations with the source text to detect translation errors. Using the collected data, models were constructed to predict the reading time of the target text and the number of regressions to the source or target text from the difference between the linear and non-linear translations. When reading both the source and target texts, the difference between the linear and non-linear translations significantly affected the reading time of the target text. In other words, when the word order of the target text differed significantly from that of the source text, readers’ cognitive effort was greater. The difference between the linear and non-linear translations was also significantly associated with the number of regressions, such that if the difference in word order between the target text and source text is large, readers needed significant regression.