This study investigates willingness to communicate (WTC) and its determinants through structural equation modelling (SEM). Building on models by MacIntyre and Charos (1996) and Yashima (2002), it addresses 3 apparent gaps in the current knowledge base: It is the first SEM-based WTC study in a Western European context, investigating French as a second language in a representative sample of over 1,000 grade 12 students in Flanders, the Dutch-speaking part of Belgium; it is among the few studies that compare classroom WTC with WTC outside the classroom; and it investigates the underexplored relationship between WTC and language proficiency through extensive standardized listening tests. Among the study's findings are that classroom WTC is a strong predictor of WTC outside the classroom; however, in naturalistic settings the role of integrativeness decreases while anxiety levels play a larger role. The study provides new support for several existing suggestions for L2 pedagogy.Keywords: willingness to communicate; structural equation model; French L2; secondary school; Flanders IN A GLOBALISING WORLD, IT IS A CLEAR asset for citizens to be taught a language other than their mother tongue. The key aims of such second language (L2) instruction should be not only to improve the learners' L2 competences, but also to stimulate their willingness to communicate (WTC) in the language they are learning. This need to stimulate WTC follows Swain'sThe Modern Language Journal, 99, 4, (2015 (2000) output hypothesis, which postulates that practice in using the L2 is a key factor in successful language acquisition. Since WTC is the final psychological step before starting L2 communication (MacIntyre, 2007), higher degrees of WTC are associated with larger amounts of this much-needed L2 communication practice. WTC has therefore even been proposed as the true purpose of L2 instruction: "The ultimate goal of the learning process should be to engender in language students the willingness to seek out communication opportunities and the willingness actually to communicate in them" (MacIntyre et al., 1998, p. 545).Katrijn Denies, Tomoko Yashima, and Rianne Janssen 719 In research on the topic, WTC is defined as "the predisposition toward or away from communicating, given the choice" (MacIntyre et al., 2003, p. 538). It seems self-evident that introverted learners would tend to be less willing to communicate; and, indeed, that assumption was confirmed in early studies on WTC in the mother tongue (McCroskey & Richmond, 1991). However, subsequent research that focused on L2 WTC found that other variables, such as anxiety, perceived competence, and attitudes, also relate to this construct. A now generally accepted pyramid model (MacIntyre et al., 1998) schematizes how different factors converge to determine the WTC of L2 learners. It draws together several trait and state variables that have been well established as key factors in the language learning process.Since the pyramid model was established, a substantial amount of research ha...