An object-name selection test was administered to 16 newly admitted schizo phrenic, non-schizophrenic psychotic and non-psychotic patients on admission when they were drug-free, and again 4-6 weeks later, following treatment with neuroleptics. Sixteen healthy, drug-free age-matched and sex-matched controls were similarly tested on two occasions 4-6 weeks apart. Schizophrenic and non-schizophrenic psychotic subjects made significantly more association-type errors than non-psychotic subjects and controls. A signifi cant reduction in association-type errors which was observed in both psychotic patient groups on the second testing occasion may be attributed to medication. The data suggest that overinclusive thinking, as measured by the object-name selection test, is a feature of both schizophrenic and non-schizophrenic psychosis. It is concluded that tests of overinclusion are therefore of limited diagnostic utility although they may be of value in monitoring response to treatment.