Within the context of increasing numbers of second language (L2) learners in Canadian schools and expanding standards-driven testing frameworks, a passing score on the Ontario Secondary School Literacy Test (OSSLT) is a recently imposed secondary school graduation requirement in Ontario. There is evidence, however, that tests designed on the basis of first language (L1) populations may have lower reliability and validity for L2 students. This study elicited accounts of the OSSLT in 33 focus groups of 22 L1 students and 136 L2 students, attending 7 Ontario secondary schools, prior to and immediately after the March 2006 test administration. The results suggest important differences in L1 and L2 accounts of test constructs and suggest a gap between what is valued as literacy on the test and what is valued in classroom literacy practice, raising some concern regarding the test's consequential validity. By examining how different groups of test-takers interpret test constructs and the interaction between these interpretations, test design, and accounts of classroom practice, we may better address issues of fidelity in test construct representation (i.e., understand what may constitute construct under-representation and construct-irrelevant variance). This study highlights what may make a test more "L2-friendly," i.e. what supports (or impedes) L2 test performance. Although in the washback literature test-taker accounts of tests have been the least researched, the results of this study suggest that such accounts have the potential to increase test fairness, enhance the validity of inferences drawn from test performance, improve the effectiveness of accommodation strategies, and promote positive washback. (Contains 3 notes.)