BACKGROUND: Chronic schizophrenia patients have been sampled for factor analytic studies to identify the factor structure of the Positive and Negative Syndrome Scale (PANSS). Many of these studies have identified a cognitive factor, which may provide useful information about patients for whom formal neuropsychological testing is unavailable. However, the relationship between the clinically rated cognitive factor and performance-based cognitive test scores has not been thoroughly examined, particularly in patients who are early in the course of illness. OBJECTIVES: The validity of the PANSS cognitive factor was examined in a sample of antipsychotic naïve, first-episode psychosis patients and the PANSS items that best predicted cognitive functioning were identified. METHOD: PANSS scores and performance on a battery of cognitive tests from the baseline assessment of a clinical trial of 167 neuroleptic naïve patients with schizophrenia-like illnesses were analyzed. RESULTS: Factor analysis revealed a five-factor model that was consistent with previously described factor models from samples of chronically treated patients. There were modest correlations (less than r=0.44) between the derived cognitive factor and six of the neuropsychological test variables (Wisconsin Card Sorting Test total errors and number of categories correctly sorted, WMS-R immediate and delayed recall scores, Category Fluency total score, and WAIS-R digit symbol). Five PANSS items (P6, N5, N6, G2, and G10) were identified that predicted global cognitive functioning; however, 66% of the variance in cognitive functioning remained unexplained. CONCLUSIONS: A PANSS cognitive factor can be identified in first-episode, neuroleptic naïve patients. However, clinical ratings of cognitive deficits lack sensitivity and specificity and thus should not be relied upon fully.