This paper explores gaze direction in responses to polar questions employing both quantitative and conversation analytic (CA) methods. The data stem from a new corpus of conversations in which participants wore eye-tracking glasses to acquire direct measures of their eye movements. The results display that while most preferred responses are generated with gaze toward the questioner, most dispreferred responses are generated with gaze aversion. It is further shown that gaze aversion by respondents can indice self-repair by questioners in the transition space between turns, pointing out that the relationship between gaze direction and preference is more than a sole statistical association. It is concluded that gaze direction in replies to polar questions acts as a signal of response preference. Data are in American, British, and Canadian English.