Willingness to Communicate (WTC), the most immediate antecedent of actual communication behavior in an L2, is widely believed to facilitate the acquisition of the target language. Higher WTC is generally associated with higher L2 use (MacIntyre, Clément, Dörnyei & Noels 1998), though it has not been empirically tested. This article reports on an attempt to fill this gap by investigating to what degree learners’ self-reported L2 WTC predicts actual L2 use in the classroom. A group of 372 Japanese university EFL learners majoring in science and human arts subjects participated in the study. Students’ oral performance data were collected three times during the group discussions in one academic semester. The number of words each student produced during each discussion was counted. The results of structural equation modelling analyses suggested that self-reported L2 WTC predicted actual L2 use in the classroom. Second, gender differences in L2 use in the language classroom has not been empirically explored in the quantitative studies, therefore, whether L2 WTC would predict observed L2 use equally among males and females, and whether males and females produce equal amount of L2 in the classroom will be explored. Implications of the research findings will be discussed.