A pilot study was conducted on 22 children (19 boys and three girls) aged between four and eight years, who were selected as hyperactive on the basis of developmental history and clinical judgement. Conners' parent-teacher ratings, objective tests of attention, standard perceptualmotor tests and subtests from the Wechsler Intelligence Scale for Children (WISC), were used as response variables. The children were tested before and after four weeks on the elimination diet, after a tartrazine and placebo challenge, and, finally, after a four-week washout period on the diet. Results showed a statistically significant improvement in the mothers' ratings of the children's behaviour after the first four weeks of the diet. The improvement was maintained in a combined analysis of the initial four-week diet period and four-week washout period. This result was not substantiated by the statistical analysis of the results from objective tests. The rating scales and objective tests for the full sample did not show a statistically significant deterioration in the children's behaviour when they were challenged under double-blind test conditions with the Yellow Dye No. 5, tartrazine, and the tests were conducted the day after a two-week challenge period. A comparison of mother ratings of behaviour during challenge and placebo double-blind trial and in the 24 hours preceding tests, in a subgroup of the children who, while on the diet, showed a 25% reduction of symptoms on the Conner's rating scale, indicated a significant challenge effect (P less than 0.025), with mothers reporting more symptoms during the challenge period. Dietary infringements with suspected trigger substances occurred throughout the trial.