There are few reports in the literature on the agreement between structured interviews and behaviour checklists in terms of specific diagnostic categories. The objective of this paper is to see how this agreement varies by source of information and diagnosis. One hundred children and their parents referred to a mental health clinic were given the Diagnostic Interview for Children and Adolescents (DICA). Parents and teachers also completed a checklist, the Survey Diagnostic Instrument, a modification of the Child Behaviour Checklist. Using likelihood ratios, the parent checklist agreed best with the parent interview, regardless of diagnosis. Among DSM-III diagnoses, the combined disorder anxiety-depression obtained the highest agreement with the DICA, attention deficit disorder with hyperactivity the next, and conduct disorder the lowest. For screening purposes, combining the parent and teacher checklists appears best, while, for diagnostic purposes, using the parent checklist alone may suffice. Nevertheless, positive predictive values remain low and may remain an inherent limitation of any checklist.