In an epidemiological sample of 411 13-year-old twins of normal intelligence, both parents and teachers reported more behavioural problems among children with lower IQs. This was not attributable to the effects of parental IQ or social class and was not entirely mediated by lower scholastic attainments. Different causal models are discussed: "rater bias" and "IQ is a consequence" explanations seem less plausible than "IQ is a cause" and "IQ is a marker" explanations. Higher parental IQ was associated with more emotional symptoms in the child, both by parental and school report. There was no evidence that being brighter or less bright than a (twin) sibling influenced behavioural deviance, casting doubt on the importance of contrast effects.