Instructional support has been widely discussed as a strategy to optimize student-learning experiences. This study examines instructional support within the context of a multimedia language-learning environment, with the predominant focus on learners' perceptions of captioning support for listening comprehension. The study seeks to answer two questions: (1) do learners' perceptions regarding dependence on captions match their actual reliance on captioning for listening comprehension? and (2) which learners' perceptions are most influenced by proficiency: low-intermediate, intermediate, or high-intermediate? A total of 139 students from a high school English course in northern Taiwan, all accustomed to multimedia instruction that includes full captions, completed an English language proficiency test as well as a caption reliance test (CRT), and also provided their perceived degree of reliance on captions for English listening comprehension. The results show that overall perceived reliance was significantly related to actual reliance as assessed by the CRT. However, proficiency was also found to be a mitigating factor in the relationship: while low-intermediate level learners accurately perceived their reliance, no relation was found for either intermediate or high-intermediate learners, indicating that, at these levels, some learners may not accurately judge their reliance on captioning. Accordingly, the study offers pedagogical implications that captioning support, added or removed, based on learner self-reports, may not be inherently beneficial, as perceptions on the reliance of captioning may be inaccurate.