Background: Recent studies have investigated the causal attributions for difficult pupil behaviour made by teachers, pupils, and parents but none have investigated the temporal stability or predictive validity of these attributions. Aims: This study examines the causal attributions made for difficult classroom behaviour by students on two occasions 30 months apart. The longitudinal stability of these attributions is considered as is the predictive validity of the first set of attributions in relation to teachers' later judgments about individual students' behaviour. Sample: Two hundred and seventeen secondary school age pupils (114 males, 103 females) provided data on the two occasions. Teachers also rated each student's behaviour at the two times. Method: A questionnaire listing 63 possible causes of classroom misbehaviour was delivered to pupils firstly when they were in Year 7 (aged 11-12) and then again, 30 months later. Responses were analysed through exploratory factor analysis (EFA). Additionally, teachers were asked to rate the standard of behaviour of each of the students on the two occasions. Results: EFA of the Years 7 and 10 data indicated that pupils' attributions yielded broadly similar five-factor models with the perceived relative importance of these factors remaining the same. Analysis also revealed a predictive relationship between pupils' attributions regarding the factor named "culture of misbehaviour" in Year 7, and teachers' judgments of their standard of behaviour in Year 10. Conclusion: The present study suggests that young adolescents' causal attributions for difficult classroom behaviour remain stable over time and are predictive of teachers' later judgments about their behaviour.