Objective: The present study was designed to assess whether the serving practices of a sample of bartenders in an American university town would vary as a function of the number of behavioral cues of intoxication displayed by apparently real patrons (who were actually experimental confederates). Method: We trained two male and three female confederates to role-play three different conditions (three ambiguous signs, eight apparently obvious signs, or no signs of intoxication) and assessed whether they would be served a second beer after "accidentally" knocking over a first beer. Results: Service of another beer was denied in only 10% of episodes when a confederate displayed no signs or three signs indicative of intoxication; however, confederates were denied service during 50% of episodes when multiple, obvious signs were displayed. Furthermore, neither the type of the establishment, gender of the bartender or gender of the confederate was associated with likelihood of being served the second drink. Conclusion: Although some bartenders are attentive to multiple behavioral signs of intoxication, we recommend additional research on the most effective combinations of server training and legal measures to reduce further the proportion of intoxicated customers served alcohol, both for their own benefit and the public health. (Contains 1 table.)