AIMS: To examine Kreitman's preventive paradox of alcohol consumption and its revisions by Stockwell and colleagues and by Skog, with regard to alcohol-related social harm in Switzerland, and to shed light on the reporting of alcohol-related social harm in the low-volume drinking, non-binging subpopulation. The paper compares occurrence and severity of social harm in four subgroups defined by average consumption (volume) and binge drinking. Stage-of-change membership was used to further distinguish low-risk drinkers who might have changed their drinking patterns from those who had not. DESIGN, SETTING, AND PARTICIPANTS: Telephone interviews were conducted with 1256 current drinkers of a probabilistic two-stage sample of the general population of Switzerland. Moderate and hazardous mean consumption (volume) was defined by means of a quantity-frequency instrument. Daily average consumption of 20 g was set as the cut-off point for women, and 30 g for men. Binge drinking was defined as taking four or more drinks on an occasion for women, and five or more for men. Structural equation modelling was used to construct a severity scale of six alcohol-related consequences: work problems, accidents and problems with the police, with friends, with a partner or with the family. Explanatory factor analysis was used to assign drinkers to motivational stages of change. FINDINGS: Moderate drinkers in terms of volume reported more problems than hazardous drinkers, which confirms Kreitman's view. Binge drinkers reported more problems than non-binge drinkers, confirming the view of Stockwell and colleagues. Binge drinkers were more numerous in the moderate drinking group, which constituted the majority of drinkers, in accordance with Skog's view. Binge drinkers in the moderate-volume and hazardous-volume drinking groups did not differ significantly as to either severity or number of problems. Approximately 40% of moderate-volume, non-binge drinkers who reported alcohol-related social harm had already changed their consumption pattern, which indicated that reported harm was related to an earlier drinking pattern. CONCLUSIONS: As Skog has pointed out, the second-order preventive paradox of binge drinking reappeared, in that most binge drinkers were found to occur in the drinker group with low average consumption. Findings also indicate that, with respect to social harm, a preventive strategy aimed at the majority of the population, but on heavy-drinking occasions rather than on mean consumption, may be valuable.