Self-efficacy is a well-known individual difference variable within second language acquisition research (SLA), shown to have predictive validity in language learning contexts. Collective-efficacy (CE), a group level construct, has remained relatively untouched by researchers despite its potential importance in formal language learning contexts, where students spend a considerable amount of time interacting and completing tasks in small groups. This paper reports on a mixed-method longitudinal study investigating CE in a tertiary educational context in Japan. A measure of CE was developed and piloted, and a subsequent cohort of students worked in fixed small groups for an academic year, completing the CE measure seven times. Hierarchical linear modelling was used to determine whether CE showed the group-level variance necessary to be considered a group-level construct and to examine growth over time. Results showed that students made gains in CE in the first semester, but that patterns of growth were erratic in semester two. Interviews with students revealed students' views relating to CE, and reasons for changes that occurred. The implications of the findings are related directly to teachers using groups in the language classroom.