BACKGROUND Patients with schizophrenia spectrum disorders have been increasingly recognised to form cognitive subgroups with differential levels of impairment. Using cluster analytical techniques, this study sought to identify cognitive clusters in a sample of first-episode psychosis (FEP) patients and examine clinical and developmental differences across the resultant groups. METHODS In total, 105 FEP patients in the University of California Los Angeles Aftercare Research Program were assessed for cognition, symptoms and premorbid developmental adjustment. Hierarchical cluster analysis with Ward's method and squared Euclidean distance was conducted, confirmed by discriminant function analysis and optimised with k-means clustering. The stability of the solution was evaluated through split-sample (random, 80 and 70% samples) and alternate method (average linkage method) replication via Cohen's κ analysis. Controlling for multiple comparisons, one-way analysis of variances examined group differences in symptom severity and premorbid adjustment. RESULTS Three groups were identified: severely impaired (n = 27), moderately impaired (n = 41) and relatively intact (n = 37). There were no significant differences in symptom severity across the groups. Significant differences were observed for scholastic performance at three different developmental stages: childhood, early adolescence and late adolescence, with the relatively intact group demonstrating significantly better scholastic performance at all three stages than both the moderately impaired and severely impaired groups (who did not significantly differ from each other). CONCLUSIONS The findings add to growing evidence that cognitive clusters in FEP mirror that of later-stage schizophrenia. They also suggest that premorbid scholastic performance may not just be a risk factor for developing schizophrenia, but is also related to cognitive impairment severity and potentially to prognosis.