This study was aimed at testing the effects of a reversible inactivation of the hippocampal formation on long-term and short-term acquisition of spatial information. Rats chronically equipped with either bilateral cannulae into the ventral hippocampus or a single cannula into the medial septum had to locate, in a circular platform with 18 holes on the periphery, the unique hole leading to a hidden shelter in order to avoid bright light. In Expt. 1, following 16 days of training (1 trial/day, 24 h ITI) without physical intervention, the location of the correct hole was changed on both Days 17 and 23, and the rats were either sham-injected or injected with lidocaine. Both hippocampally and septally lidocaine-injected rats relearned the new location at a rate similar to corresponding sham-injected animals. In Expt. 2, a massed-trial version of the task was used, in which the rats had to learn a new hole location on each daily session (3 trials, ITI = 1 min). Animals were sham-injected or lidocaine-injected on alternate sessions. While sham-injected rats improved in orientational accuracy over successive trials, both hippocampally and septally lidocaine-injected rats failed to display any between-trial improvement. The impairment displayed by lidocaine-injected rats when their hippocampus was inactivated confirms the role of the hippocampus in short-term spatial memory (Expt. 2). In contrast, short-lasting inactivation of the hippocampus did not prevent long-term spatial learning (Expt. 1). These results suggest that the hippocampus could process information 'off-line' in the delay between temporally discontiguous learning trials, and show that short-term and long-term spatial learning rely on distinct neurobiological mechanisms.