Source-based writing is becoming more common in tests of academic English, in part to make tests more reflective of authentic academic writing. However, the concern has been raised that over-reliance on language from the source texts in essays may mask gaps in proficiency and thus lead to an inaccurate assessment of writing skill. In this paper, we present results of a study designed to determine the extent to which students borrow source text language in an integrated reading/writing test. Sixty-three essays were transcribed verbatim and coded for source text borrowing using a scheme adapted from Shi (2004). The sample was selected to represent two different writing topics, two student groups (undergraduate and graduate), and four levels of proficiency. Each incidence of textual borrowing was coded for length (number of words), whether it was quoted, referenced, or not referenced, and whether it was copied, modified, or reformulated. Results suggest that only a small percentage of students borrowed extensively from the source texts and that there were only minor differences in patterns of borrowing across topics, student groups and proficiency levels. (Contains 11 tables.)