The existence of different varieties of English in target language use (TLU) domains calls into question the usefulness of listening comprehension tests whose input is limited only to a native speaker variety. This study investigated the impact of non-native varieties or accented English speech on test takers from three different English use contexts: Korea, Sri Lanka and Brazil. The findings showed that the variety of English or accented speech used had no impact on test taker's performance on a listening test for academic purposes. Also test takers from the three different countries performed similarly even when the speakers shared the same native languages as the test takers. Despite these findings students still perceived that a native variety of English should be used in listening comprehension tests. Though the study suggests the use of non-native varieties as test input, it also raises questions of fairness in the use of such varieties. (Contains 4 tables and 1 figure.)