Male mice of the BALB/c strain were given a solution of 15% ethanol as their only source of fluid during either 24 or 48 weeks. They were submitted to a sequential alternation (SA) task in a T-maze (6 successive trials). It was found that 48 but not 24 weeks of alcohol administration lead to a deficit as compared to pair-fed or tap-water controls. Whereas experimental mice performed as well as controls on the first 3 choices, they exhibited a gradual decrease in the SA rate on subsequent trials. We suggest that this deficit might result from an exaggerated vulnerability to proactive interference (PI). In order to further test this hypothesis, a second experiment investigated whether a between-trials variation of context of the maze would increase performance. It was found that the SA rate improved as soon as the variation was provided (5th trial). We suggest that the deficit of experimental mice results from an impairment of retrieval processes. A neuroanatomical study was conducted to quantify cell losses resulting from 8, 24 or 48 weeks of ethanol treatment in the mammillary bodies (MM) or the hippocampus (HPC). At the time of appearance of the deficit, MM exhibited a -32% cellular loss, whereas this was only -18% in the HPC. This result emphasises the importance of MM lesion in memory deficits resulting from long-term alcohol consumption.