OBJECTIVE: To investigate the neuropsychological profile of individuals with an at risk mental state for psychosis (ARMS, N=60) compared to healthy controls (HC, N=51) and to identify those cognitive domains which discriminate best between groups. METHOD: Study subjects and controls were compared using a neuropsychological test battery covering the domains of intelligence (LPS3, MWT-A), executive functions (ToH, WCST, TAP - Go/NoGo), working memory (Tests for Attentional Performance (TAP) - Working Memory), and attention (CPT-OX). A multivariate analysis of variance (MANOVA) comparing ARMS subjects with HC was conducted. A stepwise logit regression procedure was performed in order to determine the subset of measures which best distinguish ARMS subjects from HC. RESULTS: ARMS subjects revealed deficiencies in intelligence, executive functions, working memory and attention. Verbal intelligence, executive functions, and, in particular, working memory discriminated best between the groups. CONCLUSION: Individuals with an at risk mental state for psychosis already show impairment of neuropsychological functions prior to the onset of the first psychotic episode and can best be distinguished from healthy controls on the basis of working memory.