Cognitive deficits are a core feature of established psychotic illnesses. However, the association between cognition and emerging psychosis is less understood. While there is some evidence that cognitive deficits are present prior to the onset of psychosis, findings are not consistent. In this article we provide an overview of the more general cognitive findings available from genetic high-risk studies, retrospective studies, and birth cohort studies. We then focus the review on neuropsychological performance in clinically "at-risk" groups. Overall, general cognitive ability as assessed by established batteries appears to remain relatively intact in these ultra-high risk cohorts and is a poor predictor close to illness onset relative to other vulnerability factors. Further decline may occur with illness progression, more consistent with state relative to trait factors. In addition, most established cognitive tasks involve several relatively discrete cognitive subprocesses, where findings from general batteries of subtests may mask specific deficits. In this context, our review suggests that relatively specific olfactory identification and spatial working memory deficits exist prior to illness onset and may be more potent trait markers for psychosis than cognitively dense tasks such as verbal memory. Suggestions for further research address the importance of standardization of inclusion criteria and the maintenance of basic neuropsychological assessment to allow better comparison of findings across centers. Further, in order to better understand the aetiopathology of cognitive dysfunction in psychosis, more experimental, hypothesis-driven measures of discrete cognitive processes are required. Delineation of the relationship between specific cognitive ability and symptoms from data-driven approaches may improve our understanding of the role of cognition during psychosis onset.