The study aims to identify potential neurocognitive indicators of an enhanced risk for developing psychosis. N=44 patients meeting clinical inclusion criteria for initial prodromal states (IPS) who developed psychosis within a median interval of 10 months were compared to N=39 IPS patients not developing psychosis within a minimum interval of 1 year (median 36 months), and to N=44 healthy controls on a comprehensive neuropsychological test battery (pattern recognition, divided and sustained attention, spatial and verbal working memory, verbal/visual memory, speed of processing, executive and intellectual functions). IPS patients who converted to psychosis performed worse than healthy controls on all broad neurocognitive domains. They were more impaired than IPS patients not developing psychosis on the Subject Ordered Pointing Task (SOPT; working memory), verbal memory functions, verbal executive, verbal IQ and speed of processing tests. After a Bonferroni-Holms adjustment for multiple testing differences on SOPT, Digit-Symbol Test, and verbal IQ remained significant (effect sizes d=0.54-0.88). Neurocognitive predictors had a sensitivity of 0.75 and a specificity of 0.79. Results support several cognitive domains as indicators of vulnerability to psychosis, and additionally suggest that subtle deficits in verbal abilities (working and long-term memory, executive and intellectual functions) and decreased speed of processing may help to predict conversion to psychosis in a clinically defined IPS group.