An important aim in schizophrenia research is to optimize the prediction of psychosis and to improve strategies for early intervention. The objectives of this study were to explore neurocognitive performance in individuals at risk for psychosis and to optimize predictions through a combination of neurocognitive and psychopathological variables. Information on clinical outcomes after 24 months was available from 118 subjects who had completed an extensive assessment at baseline. Subjects who had converted to psychosis were compared with subjects who had not. Multivariate Cox regression analyses were used to determine which baseline measure best predicted a conversion to psychosis. The premorbid IQ and the neurocognitive domains of processing speed, learning/memory, working memory and verbal fluency significantly discriminated between converters and non-converters. When entered into multivariate regression analyses, the combination of PANSS positive/negative symptom severity and IQ best predicted the clinical outcomes. Our results confirm previous evidence suggesting moderate premorbid cognitive deficits in individuals developing full-blown psychosis. Overall, clinical symptoms appeared to be a more sensitive predictor than cognitive performance. Nevertheless, the two might serve as complementary predictors when assessing the risk for psychosis.