One advantage of using student ratings over observer ratings in measuring teaching quality is seen in the extended observation period they are based on (usually the school year so far). However, it has rarely been investigated to what degree student ratings are nevertheless influenced by occasion-specific characteristics; this is thus the aim of the present study. Teaching quality was rated three times by 606 students from 30 classrooms regarding three dimensions (classroom management, personal learning support, cognitive activation). Data was analyzed using generalizability theory. The analyses indicated that the dimensions differ in the amount of occasion-specificity. For measuring personal learning support, one measurement point was sufficient, whereas for classroom management and cognitive activation at least two measurement points were necessary.