Testing The Literal Translation Hypothesis:An Empirical Study

Recent decades have witnessed the rapid growth of process-oriented translation studies,whose two main objectives are:(1)to describe the phenomena of translating and translation(s),and(2)to establish general principles that can explain and predict these phenomena(Holmes,1988,p.71).To date,there are several interesting hypotheses related to translation process and one of them is literal translation hypothesis,i.e.,the translator begins his search for translation equivalence from formal correspondence and it is only when the formal correspondence is not available or not desirable that he turns to structural and semantic shifts(Ivir,1981).So far,there remains a lack of empirical evidence for this hypothesis,especially in the English-Chinese language pair,and conditions under which this hypothesis would work have rarely been explored(e.g.,Pavlovic&Antunovic,2013).This study aims to test the literal translation hypothesis.It used multiple research methods including key-logging,eye-tracking,think aloud protocols(TAP),and interviews.All participants' behaviors during the translation process were recorded by a Tobii TX300 remote eye-tracker and Translog-II,a key-logging program.Following a study by Pavlovic and Antunovic(2013),we distinguished between three types of self-revisions,i.e.,T-move(involving a change from a more literal towards a freer rendering),S-move(involving a change from a freer towards a more literal rendering),and neutral change(involving a change neither closer nor far away from the source);we calculated the number and type of self-revisions made by participants(n=17)(a)in the whole process,(b)as a first revision of an initial solution,and(c)self-revisions in the post-drafting phase of the process(i.e.,the phase after the draft version is finished and until the whole process comes to an end).We compared the three aspects in terms of the proportion of T-moves in the number of self-revisions between two groups of participants—translation-track(n=10)and interpreting-track(n=7)graduate students—irrespective of the quality of their final translations.We also measured the literality of translations and assessed the participants' cognitive effort when they were producing translations deviating from a literal rendering.The results show that(1)the deliteralization process(moving from more literal to less literal solutions)exists in the two groups;(2)for the translation-track group,deliteralization(in terms of the proportion of T-moves)exists mainly in the "first revision of an initial solution";for the interpreting-track group,there is no significant difference among the three aspects(i.e.,the whole process,first revision of an initial solution,self-revisions in the post-drafting phase);(3)between the translation-track and interpreting-track groups,significant differences(in terms of the proportion of T-moves)are found in the first revision and post-drafting aspects,but not in the whole process;(4)literal translation entails less cognitive effort than free translation;the more semantically and syntactically similar between the source text and the target text,the less time needed for the translation;(5)compared with the translation-track group,the interpreting-track group tends to proceed in a more linear and controlled manner.