The inferior performance of DBA/2 mice when compared to C57BL/6 mice in hippocampus-dependent behavioral tasks including contextual fear conditioning has been attributed to impaired hippocampal function. However, DBA/2J mice have been reported to perform similarly or even better than C57BL/6J mice in the passive avoidance (PA) task that most likely also depends on hippocampal function. The apparent discrepancy in PA versus fear conditioning performance in these two strains of mice was investigated using an automated PA system. The aim was to determine whether these two mouse strains utilize different strategies involving a different contribution of hippocampal mechanisms to encode PA. C57BL/6J mice exhibited significantly longer retention latencies than DBA/2J mice when tested 24 h after training irrespective of the circadian cycle. Dorsohippocampal NMDA receptor inhibition by local injection of the selective antagonist DL-2-amino-S-phosphonovaleric acid (AP5, 3.2 μg/mouse) before training resulted in impaired PA retention in C57BL/6J but not in DBA/2J mice. Furthermore, nonreinforced pre-exposure to the PA system before training caused a latent inhibition-like reduction of retention latencies in C57BL/6J, whereas it improved PA retention in DBA/2J mice. These pre-exposure experiments facilitated the discrimination of hippocampal involvement without local pharmacological intervention. The results indicate differences in PA learning between these two strains based on a different NMDA receptor involvement in the dorsal hippocampus in this emotional learning task. We hypothesize that mouse strains can differ in their PA learning performance based on their relative ability to form associations on the basis of unisensory versus multisensory contextual/spatial cues that involve hippocampal processing. © 2007 Wiley-Liss, Inc.