In educational research, characteristics of the learning environment are generally assessed by asking students to evaluate features of their lessons. The student ratings produced by this simple and efficient research strategy can be analysed from two different perspectives. At the "individual level", they represent the individual student's perception of the learning environment. Scores aggregated to the "classroom level" reflect perceptions of the shared learning environment, corrected for individual idiosyncrasies. This second approach is often pursued in studies on teaching quality and effectiveness, where student-level ratings are aggregated to the class level to obtain general information about the learning environment. Although this strategy is widely applied in educational research, neither the "reliability" of aggregated student ratings nor the within-group "agreement" between the students in a class has been subject to much investigation. The present study introduces and discusses different procedures that have been proposed in the field of organisational psychology to assess the reliability and agreement of students' ratings of their instruction. The application of the proposed indexes is demonstrated by a reanalysis of student ratings of mathematics instruction obtained in the Third International Mathematics and Science Study (N = 2,064 students in 100 classes).