Student reports are a common approach to characterizing how students experience their classrooms. We used a recently developed method--multilevel confirmatory factor analysis--to determine whether commonly employed measures of achievement goal structure constructs (mastery and performance) typically verified at the student level can be verified at the classroom level as well, using a sample of 1,406 7th- and 1,822 8th-grade students in 188 mathematics classrooms. Nine multilevel confirmatory factor analyses were examined: (a) modeling each classroom goal structure dimension separately (mastery, performance--approach, performance--avoidance, combined performance), (b) 4 models allowing 2 latent factors to covary (mastery + performance-approach, mastery + performance-avoidance, performance-approach + performance-avoidance, mastery + combined performance), and (c) a model reflecting the trichotomous goals framework (mastery + performance-approach + performance-avoidance). Consistent with prior research, the theorized models of classroom goal structure adequately represented students within the same classroom at Level 1. Although all models were acceptable at the individual level, only the performance-avoidance model represented the expected goal structure at the classroom level. Future research and replication are necessary to examine whether self-reports of classroom goal structure, and by implication other measures of classroom climate, reflect an individual-level phenomenon rather than a classroom-level construct, which has important theoretical and applied implications.