The primary purpose of study is to identify predictors of willingness to communicate (WTC) and of actual frequency of English communication at work inside and outside the foreign language classroom among 439 university students (male = 226, female = 213) learning English in Japan. Based on Wen and Clément’s (2003) theory of L2 WTC, I replicated Peng and Woodrow’s (2010) structural path model using the variables of state L2 communicative confidence, L2 learning motivation, positive classroom climate, L2 WTC, with the newly added variable of actual speaking frequency. A hypothesized structural model was examined in two contexts, WTC inside the classroom and WTC outside the classroom. Inside the classroom, communicative confidence was the predictor of L2 WTC. L2 WTC and L2 learning motivation were predictors of actual frequency of L2 communication. Positive classroom climate was a mediating variable that indirectly predicted L2 WTC through state L2 communicative confidence and task motivation. In contrast, outside the classroom, state L2 communicative confidence, L2 learning motivation, and positive classroom climate were the predictors of L2 WTC. State L2 communicative confidence, task motivation, and positive classroom climate were the predictors of actual frequency of L2 communication. The results supported Wen and Clément’s (2003) model and Peng and Woodrow’s (2010) study. Second, Dӧrnyei and Kormos’ (2000) study was replicated to investigate a significant difference for the four types of the students’ speaking behavior between pretest and posttest. A repeated-measures ANOVA was performed for English turns, Japanese turns, English words, and interjections with 13 students (male = 8 and female = 5) aged 18-19. The 13 participants were part of those who completed the first questionnaire. There were no significant differences for the four dependent variables. Finally, a qualitative content analysis was performed using transcribed interview data with nine university students (6 male and 3 female students), who completed the first questionnaire. Ten variables emerged from the interviews. Four variables—teacher support, group cohesiveness, L2 learning motivation, and perceived communicative competence—supported both quantitative (Peng & Woodrow, 2010) and qualitative studies (Cao, 2011; Peng, 2007, 2012). Four additional variables—security of speaking, interlocutors, small group, and topic familiarity—supported qualitative studies by Cao (2011) and Kang (2005). The other two variables—point system and tests—were new variables identified in this study. Positive classroom climate and task motivation (Dӧrnyei & Kormos, 2000) were key variables influencing state L2 communicative confidence, L2 WTC, and L2 Use. As a result, I propose that task motivation and positive classroom climate should be added into MacIntyre et al.’s (1998) L2 WTC model. / Teaching & Learning