Seventy-eight children, referred to a diet clinic because of hyperactive behaviour, were placed on a 'few foods' elimination diet. Fifty nine improved in behaviour during this open trial. For 19 of these children it was possible to disguise foods or additives, or both, that reliably provoked behavioural problems by mixing them with other tolerated foods and to test their effect in a placebo controlled double blind challenge protocol. The results of a crossover trial on these 19 children showed a significant effect for the provoking foods to worsen ratings of behaviour and to impair psychological test performance. This study shows that observations of change in behaviour associated with diet made by parents and other people with a role in the child's care can be reproduced using double blind methodology and objective assessments. Clinicians should give weight to the accounts of parents and consider this treatment in selected children with a suggestive medical history.