BackgroundIndividuals at ultra-high risk (UHR) for psychosis show reduced neurocognitive performance across domains but it is unclear which reductions are associated with transition to frank psychosis. The aim of this study was to investigate differences in baseline neurocognitive performance between UHR participants with (UHR-P) and without transition to psychosis (UHR-NP) and a healthy control (HC) group and examine neurocognitive predictors of transition over the medium to long term.MethodA sample of 325 UHR participants recruited consecutively from the Personal Assessment and Crisis Evaluation (PACE) Clinic in Melbourne and 66 HCs completed a neurocognitive assessment at baseline. The UHR group was followed up between 2.39 and 14.86 (median = 6.45) years later. Cox regression was used to investigate candidate neurocognitive predictors of psychosis onset.ResultsThe UHR group performed more poorly than the HC group across a range of neurocognitive domains but only performance on digit symbol coding and picture completion differed between the groups. The risk of transition was only significantly associated with poorer performance on visual reproduction [hazard ratio (HR) 0.919, 95% confidence interval (CI) 0.876–0.965, p = 0.001] and matrix reasoning (HR 0.938, 95% CI 0.883–0.996, p = 0.037). These remained significant even after controlling for psychopathology at baseline.ConclusionsThis study is the longest follow-up of an UHR sample to date. UHR status was associated with poorer neurocognitive performance compared to HCs on some tasks. Cognition at identification as UHR was not a strong predictor of risk for transition to psychosis. The results suggests the need to include more experimental paradigms that isolate discrete cognitive processes to better understand neurocognition at this early stage of illness.