The effects of isolating the hippocampus from its neocortical inputs and outputs by damaging the deep layers of entorhinal cortex and subiculum were compared with direct removal of the hippocampus using acquisition of a complex radial maze task. A series of eight problems (four out of eight arms being correct) were learned under either massed (45 s) or distributed (10 min) practice conditions, thus varying contextual information. Performance of rats with subiculum/entorhinal cortex lesions was similar to that of controls in all aspects of the radial maze task; whereas animals with hippocampal lesions were impaired on nearly all dependent measures. Although the effects of varying the intertrial interval were generally small, distributed practice did serve to facilitate the performance of hippocampal rats in terms of working memory. These findings are discussed as they related to recent theorizing in the area.