Sight translation requires silent reading of the text in the source language and the concomitant oral production of equivalent content in the target language. This task is often performed by professional interpreters at meetings when written texts need to be translated orally on the spot. Sight translation becomes particularly demanding when the source text contains syntactic structures that are not equivalent across languages, so they cannot be simply copied from the source language into the target language but require reformulation. As shown in earlier sight translation research, increased task demands (e.g. resulting from cross-linguistic differences) may influence not only the quality of the target text but they may also affect the source text viewing patterns (Chmiel & Lijewska, 2019; Jakobsen & Jensen, 2008; Korpal, 2015; Shreve et al., 2010). The purpose of the current study was to test whether working memory (WM) span (Engle, 2002) and the length of interpreting experience can modulate sight translation performance of a demanding text that contains cross-linguistically incongruent syntactic structures (requiring reformulation). To this end, a group of professional conference interpreters (N=24) was asked to sight translate a text from Polish (their A language/L1) into English (their B language/L2). The Polish text included compound sentences with object-verb-subject structures where reformulation was necessary or the sentences would become unnatural in English. During task completion, the participants’ output was recorded and their eye movements were tracked. A preliminary analysis of the output data showed that larger WM span but not experience lead to better task performance (more reformulation). In contrast, no effects of WM capacity or experience were found in the eye movement data. These findings seem to suggest that when sight translating a syntactically demanding text, professional interpreters may capitalize on their increased WM capacity rather than on a more extensive professional experience.