Rats were trained on delayed matching-to-sample (DMS) with goalboxes containing complex objects as stimuli. On reaching the preoperative learning criterion, the rats were allocated to conventional fornix-lesioned or control groups. In a series of postoperative DMS experiments, different kinds of stimuli were used, ranging from complex object boxes to large, simple black or white goalboxes, with 3 transitional types in between. Lesions impaired choice accuracy whenever the rats were tested with large, simple goalboxes, but not with smaller boxes of otherwise identical construction. A brief, final experiment showed no amelioration of the lesion-induced impairment when complex objects were added to large, simple goalboxes. The results are discussed in terms of spatial and nonspatial accounts of hippocampal function.