This school-based study explored the role of collective and proxy efficacy beliefs in the performances of project-based learning teams comprising friends and acquaintances. Participants were 162 male students in Grade 8 who attended a Catholic high school, located in Sydney, Australia. Students were organized into 20 acquaintance groups and 21 friendship groups. Each group comprised 4 students who were completing project-based learning assignments in Geography, Religious Studies, and English. Data were self-reports and teacher-assessed group performance scores. Data collection occurred three times over a five-week period. Multilevel modeling was used to examine relationships between variables in the study. Statistically significant interactions involving group type, collective efficacy, and proxy efficacy were identified in Geography and Religious Studies. Implications are that it may be advantageous for teachers to assign students to friendship groups, provided they nurture collective efficacy, and that proxy efficacy may negatively affect group performance, depending on the context.