Numerical taxonomy was used in an attempt to identify subgroups in a sample of children labelled as hyperactive. In each of two analyses of Conners' questionnaire data (supplied by parents and by teachers), two classes were obtained, apparently differentiated on the basis of rated severity of behavioural symptoms. No aetiologically important differentiating attributes emerged. There was poor agreement between the two data sources. particularly in the classification of the more disordered group; and, over all, teachers perceived the children as less deviant than did parents. Comparison with earlier studies indicated the similarity in behavioural symptoms between this sample and those from other countries. However, the obtained data argued against the existence of a pure hyperactive disorder, since there was virtually no differentiation between conduct problems and items relating more specifically to hyperactivity. Hyperactivity as a diagnostic category appears to have little meaning either aetiologically, or descriptively, or prescriptively.