An assessment schedule was used to determine the nature of insight in 91 mixed psychotic patients, and to examine its distribution and associations. While all the components of the schedule intercorrelated significantly, scores for compliance were only weakly related to those for ability to label psychotic phenomena as abnormal. Compliance and illness recognition were related to IQ. Total insight score was inversely correlated, moderately, with a global measure of psychopathology derived from the PSE, and was less in patients involuntarily committed. Age, sex, diagnosis, and the number of previous hospital admissions had little effect. The results support the notion that insight is not a unitary concept.