Whereas studies show that comprehensible output facilitates L2 learning, hypothesis testing has received little attention in Second Language Acquisition (SLA). Following Shehadeh (2003), we focus on hypothesis testing episodes (HTEs) in which learners initiate repair of their own speech in interaction. In the context of a one-way information gap task, we examine the linguistic categories targeted for HTEs, the proportion of grammatical and ungrammatical output, and the reaction of interlocutors. Results showed that high frequency of HTEs within a linguistic category did not necessarily result in grammaticality. Syntactic HTEs were quite frequent but rarely resulted in grammatically correct output. Moreover, interlocutors questioned phonological/lexical and syntactic HTEs more often than morphological HTEs, regardless of grammatical correctness. These findings indicate that learners target a variety of linguistic categories in HTEs, but interlocutors may ignore those that are not crucial for comprehension. We suggest pedagogical implications for English language teachers and the influence that the task may play on learners' hypothesis testing.