A substantial proportion of schizophrenia-spectrum patients exhibit a cognitive impairment at illness onset. However, the long-term course of neurocognition and a possible neurotoxic effect of time spent in active psychosis, is a topic of controversy. Furthermore, it is of importance to find out what predicts the long-term course of neurocognition. Duration of untreated psychosis (DUP), accumulated time in psychosis the first year after start of treatment, relapse rates and symptoms are potential predictors of the long-term course. In this study, 261 first-episode psychosis patients were assessed neuropsychologically on one or more occasions. Patients were tested after remission of psychotic symptoms and reassessed 1, 2, 5, and 10 years after inclusion. The neurocognitive battery consisted of California Verbal Learning Test, Wisconsin Card Sorting Test, Controlled Oral Word Association Task, Trail Making A and B, and Finger Tapping. We calculated a composite score by adding the z-scores of 4 tests that were only moderately inter-correlated, not including Finger Tapping. Data were analyzed by a linear mixed model. The composite score was stable over 10 years. No significant relationship between psychosis before (DUP) or after start of treatment and the composite score was found, providing no support for the neurotoxicity hypothesis, and indicating that psychosis before start of treatment has no significant impact on the course and outcome in psychosis. We found no association between symptoms and the neurocognitive trajectory. Stable remission during the first year predicted neurocognitive functioning, suggesting that the early clinical course is a good predictor for the long-term course.