Aims: The aim of this study was to investigate performance on neuropsychological tasks and tasks measuring different aspects of attention in a representative sample of sober alcoholics. Methods: The study followed a between-groups design whereby sober alcoholics were compared to a matched non-alcoholic control group. The alcoholics were recruited from a six-week residential rehabilitation unit for addicts. A total of 98 alcoholics (64 males) and 30 non-alcoholic controls (22 males) participated in the study. The alcoholics were assessed on four standard neuropsychological tasks and three attention tasks from cognitive experimental literature targeted towards measuring attention. Results: In comparison to a non-alcoholic control group the alcoholics were significantly impaired on all neuropsychological tasks, the divided attention task and the Stroop task. However, a normal pattern of performance for the alcoholics appears in the Eriksen task, a measure of selective attention. Alcoholics also showed elevated levels of negative affect at treatment intake. Conclusion: It is concluded that a representative sample of alcoholics show specific deficits of attention as opposed to a general decline of attention at treatment intake. Thus, sober alcoholics appear to be as efficient as controls at selecting on the basis of location. Nonetheless when they are required to select on the basis of semantic information, as in the Stroop task, or required to respond to two independent sources of information, as in the divided attention task, they are at a deficit. This study has added to previous research by highlighting the specific attentional processes impaired in alcoholics. (C) 2004 Elsevier Ireland Ltd. All rights reserved.