Abstract The development of translation and interpreting competence and expertise is the outcome of, among other things, sustained formal training and deliberate practice. Many studies have attempted to examine the training effects on different aspects of cognitive processing and translation products, hoping that a good understanding of this issue will inform translation curricular innovations. Unfortunately, the results thereof have so far been inconsistent, thus hindering related curricular efforts. Intending to solve the problem, the present study utilized eye-tracking technology to gather empirical evidence to ascertain how training and experience affect student interpreters’ cognitive load, interpreting speed and output quality. To achieve this, beginner and advanced interpreting students were invited to perform unrehearsed L1 and L2 sight interpreting tasks between English and Chinese. The results show that the advanced students demonstrated both more efficient processing and better output quality than the beginners in the processing of entire texts and the problem trigger of low-frequency words (LFWs), but not complex noun phrases (CNPs). No greater training effect was found for the advanced students than for the beginner students in L1 translation, but both more efficient and better output were found for the former than for the latter group in the more challenging L2 translation.