The present study compared the effects of fornix lesions and medial prefrontal lesions on a test of object recognition memory, delayed non-matching-to-sample. Neither lesion impaired the acquisition of this non-spatial test of working memory, indeed there was clear evidence that fornix damage resulted in improved non-matching performance during initial acquisition. This improvement in performance could be related to the loss of a spatial bias during the early stages of training. A series of experiments then systematically increased the familiarity of the stimuli (i.e. testing recency rather than recognition judgements). Neither the fornix nor the prefrontal lesions disrupted performance under these conditions, even though this manipulation affected nonmatching in a predictable manner. The same animals were also tested on a spatial forced-alternation task in a T-maze (spatial delayed non-matching-to-sample). The animals with fornix lesions performed at chance while the prefrontal animals were mildly, but significantly, impaired. The present findings are considered in the light of a number of seemingly contradictory findings regarding the effects of hippocampal system damage on nonspatial tests of recognition memory.