The aim of this study was to investigate the underlying structure of eight working memory tests used to assess prefrontal dysfunction in schizophrenia research [Letter-Number Span (LNS), Digit-Symbol Test (DST), Trail-Making Test B (TMT-B), Delayed Response Task (DRT) for spatial working memory, Subject Ordered Pointing Task (SOPT), Dual Tasking (DUAL), Continuous Performance Test (CPT)-Identical Pairs, Wisconsin Card Sorting Test (WCST)]. Sixty-six patients with schizophrenia showed significant working memory performance deficits in all tests when compared with 45 healthy controls. Performance was not systematically related to psychopathology. When differences in IQ were controlled, working memory deficits remained stable except in the WCST. Principal components analyses yielded three components for healthy controls: a comparator function of the central executive defined by a comparison of working memory content with information from the environment, an allocation of attentional resources function, and a maximum storage capacity function. The comparator and maximum storage functions could be replicated in the schizophrenia sample. However, the allocation function did not emerge as an independent component and was replaced by a component defined by the WCST. These findings suggest that working memory is not a unitary concept but rather should be conceptually differentiated as functions of transient storage/active rehearsal capacity and central executive manipulation supporting a previous suggestion proposed by Perry et al. [Schizophr. Bull. 27 (2001) 157].