This article reviews the literature and presents some unpublished data on the CNS effects of alcohol at doses not producing tolerance and dependence. The available evidence indicates that the effect of low doses of ethanol may qualitatively differ from those produced in animal models mimicking alcoholism. For example, rats exposed for two months to alcohol in drinking water at a concentration (3%) not inducing tolerance or dependence, as assessed by lack of withdrawal signs upon treatment suspension, appear to be less stressed in the two-way avoidance-learning tests. Accordingly, the treated rats perform better and learn faster than sucrose-fed controls, while this behavior is disrupted by high levels of ethanol intake. These initial observations suggest that discontinuity may exist between the effects of low and high doses of this substance and underscore the need to expand research on the effects of alcohol on the CNS to include the bottom end of the dose-response curve.