Over the past few years, many studies have examined the cognitive processes of students when translating texts from one language to another. However, little is known about the cognitive processes of Arab translators when translating texts from their heritage language into English. This study aims to track the cognitive processes of students as indicated by their pauses when translating two texts from Arabic into English, and vice versa. To this end, 26 Arab native students were asked to translate medical text from English into Arabic and a literary text from Arabic into English. Data on these students' behaviour were collected using a keystroke logging programme Inputlog.07, a screen recording software and a stimulated recall interview. Results from quantitative data suggest that these students paused longer at word boundaries than between sentences, and their pauses within-words were significantly shorter when translating English into Arabic, indicating that their main concerns are on the low-cognitive level such as lexis, grammar, and spelling errors. Meanwhile, results from qualitative data suggest that these students pause when facing difficulties in understanding English text, were striving to accommodate the equivalent terms in the foreign language, and were struggling with their spelling when translating Arabic into English.