Translation effort is influenced by various factors, with source text (ST) type being a significant one. While previous research has explored lexical and syntactic characteristics' impact, this study delves deeper into the mechanisms influencing the translator's transfer process. The translation process involves ST comprehension and target text (TT) reformulation. Each ST item may have multiple translation options, demanding comparison and selection, a process contributing significantly to translation effort. Translation uncertainty, reflecting the effort in choosing among alternatives, is central to this study. Translation entropy, a measure of uncertainty, is used as a metric. High entropy suggests numerous TT options for a single ST item, correlating with increased translation effort. This research investigates how text type impacts Chinese-English translation effort by considering translation entropy.

Existing research highlights the impact of text type on translation effort, often focusing on lexical or syntactic characteristics. Studies have examined various text pairings and their effect on cognitive processes. For example, legal texts have consistently proven more demanding than news texts. However, a gap exists in exploring the underlying reasons for these differences and the role of translation products in assessing effort. Reiss's text typology, while criticized, provides a framework for categorizing texts (informative, expressive, operative) and their translation criteria. Different text types have varying linguistic features. For instance, news texts are generally concise, while legal texts are characterized by jargon and complex structures. Previous studies have mainly focused on the process aspect of translation, examining data from eye-tracking and key-logging. This research aims to address the gap by incorporating translation products (translation entropy) to understand the "why" behind the observed differences in translation effort across text types.

This study employed a mixed-methods approach combining eye-tracking, key-logging, and questionnaires. **Tools:** A Gazepoint GP3 HD Desktop Eye Tracker (150 Hz sampling rate) and Translog II software were used for data collection. **Participants:** 31 postgraduate translation students (22-28 years old, mostly female) proficient in both Chinese and English participated. They had approximately one year of translation study experience and were proficient in touch typing. **Materials:** Four text types—news, legal, poetic, and advertising—were selected to represent different textual functions (informative, expressive, operative) and linguistic features. Each text contained around 50 Chinese characters, pre-tested for difficulty with undergraduates. **Procedure:** Participants completed a warm-up task followed by four translation tasks in randomized order. Eye-tracking and key-logging data were collected, along with subjective ratings (SR) of translation effort via a post-task questionnaire. **Data Processing:** Eye-tracking data were filtered to remove noise (fixation duration threshold of 100ms). Key-logging data were analyzed using a pause threshold of 300ms. Normalized indicators (fixation duration and count per word, translation time per word, edits per word, and pause duration/count per word) were employed. Translation entropy was calculated at the phrasal and syntactic levels through manual alignment of ST and TT to determine translation variants. **Statistical Analysis:** Non-parametric tests (Friedman tests, Dunn's tests, Kruskal-Wallis tests) were employed due to the small sample size and unknown population distributions. Linear mixed-effects models (LMMs) analyzed the relationship between translation entropy and effort indicators, accounting for individual participant variability.

The results showed significant differences in translation effort across text types in all indicators: subjective ratings, translation time, edits, fixation duration and count, and pause duration and count. The legal text consistently required the most effort, followed by poetic, advertising, and news texts. **Translation Entropy:** Significant differences in both phrasal and syntactic entropy were observed across text types, with the legal text exhibiting the highest entropy and the news text the lowest. **Correlation Analysis:** LMM analysis revealed significant positive correlations between translation entropy (both phrasal and syntactic) and most effort indicators (subjective ratings, translation time, fixation duration and count, pause duration and count). The effect sizes were large for subjective ratings and moderate for fixation duration/count. There was no significant relationship with the number of edits.

The findings demonstrate the significant impact of text type on translation effort, attributable to its inherent linguistic features and textual functions. The substantial differences between legal and news texts highlight the impact of structural complexity and specialized terminology. The difference between news and advertising texts underscores the additional cognitive demands of conveying operative functions. The relationship between translation entropy and effort indicators provides evidence for translation uncertainty as a key contributor to translation difficulty. Higher entropy (more translation options) correlates with higher effort. The study's results extend previous findings to Chinese-English translation, a pair from different language families, suggesting broader applicability of the translation entropy concept. The consistent pattern observed between entropy levels and effort across text types supports the view that translation uncertainty is a crucial factor influencing translation effort.

This study confirms the significant impact of text type on translation effort in Chinese-English translation. Translation entropy effectively measures the uncertainty involved and predicts effort levels. Future research should involve professional translators and longer texts to further validate these findings and explore nuanced differences.

The study's limitations include the use of student translators (potentially limiting generalizability to professional settings) and relatively short text passages (which might have underrepresented the variation in linguistic complexity across text types).