Impaired insight is common in schizophrenia. Etiological models focusing on single determinants have not succeeded in explaining insight deficits. More complex models seem promising. This study tests Startup's (1996) model of insight and cognition, predicting a curvilinear relationship and specific insight-cognition configurations. Patients with schizophrenia diagnoses (N 248) were assessed with the Dynamic Wisconsin Card Sorting Test (WCSTdyn) and measures of psychopathology and premorbid intelligence. In a regression model connecting insight and WCSTdyn, the linear and quadratic term accounted for a small but significant proportion of variance. Cluster analysis yielded two cognitively high-functioning groups differing in insight and a group with impaired cognition and reduced insight. Results support Startup's framework of multiple barriers to insight. Cognitive deficits seem to be one insight-limiting factor, but motivational influences on insight cannot be excluded. Research on therapeutic interventions should take these different pathways into account.