This study reports the development of a new, modified delayed matching to sample (DMS) visual recognition memory task that controls the relative novelty of test stimuli and can be used in human and nonhuman primates. We report findings from normal humans and unoperated monkeys, as well as three groups of operated monkeys. In the study phase of this modified paradigm, subjects studied lists of two-dimensional visual object stimuli. In the test phase each studied object was presented again, now paired with a new stimulus (a foil), and the subject had to choose the studied item. In some lists one study item (the novel or isolate item) and its associated foil differed from the others (the homogenous items) along one stimulus dimension (color). The critical experimental measure was the comparison of the visual object recognition error rates for isolate and homogenous test items. This task was initially administered to human subjects and unoperated monkeys. Error rates for both groups were reliably lower for isolate than for homogenous stimuli in the same list position (the von Restorff effect). The task was then administered to three groups of monkeys who had selective brain lesions. Monkeys with bilateral lesions of the amygdata and fornix, two structures that have been proposed to play a role in novelty and memory encoding, were similar to normal monkeys in their performance on this task. Two further groups--with disconnection lesions of the perirhinal cortex and either the prefrontal cortex or the magnocellular mediodorsal thalamus--showed no evidence of a von Restorff effect. These findings are not consistent with previous proposals that the hippocampus and amygdala constitute a general novelty processing network. Instead, the results support an interaction between the perirhinal and frontal cortices in the processing of certain kinds of novel information that support visual object recognition memory.