This study conducted a mixed-method study of the influence of audio and video input mode on free-recall listening performance. It first explored quantitatively whether input mode significantly influenced 34 sophomores' performance in general and across two genres (passage and long dialogue) and three ranks of idea units (the discourse topic, main point, and supporting detail). Then it investigated qualitatively how four of the participants interacted with the audio and video input. T-test results showed the video mode significantly facilitated listening performance in general and for long dialogue in particular, as well as the recall of supporting details for long dialogue. Qualitative findings revealed that the participants' interaction with the video varied with their language proficiency and the visual-verbal relationship in the input, suggesting the threshold of language proficiency for the visuals to take effect and the intervening effect of visual input and task features.