Across more than four decades of previous related research, social scientists have reported that demeanor of the citizen and other extralegal factors shape police actions such as arrest. Recently, however, Klinger (1994) has asserted that all previous research in this area is suspect, because it failed to control for crime and failed to limit demeanor to legally permissible words and displays of hostility. The current research further probes this issue by using previously unpublished data. The previously unexamined police-citizen encounters were from the Midwest City Police-Citizen Encounters Study (Sykes and Brent 1983; Sykes and Clark 1975; also see Sherman 1980). These encounters involved drunk driving. During a 15-month period, beginning in June 1970, a group of seven observers accompanied police patrol officers in a large midwestern city. With prior permission but without prior notice as to specific precinct and cruisers, observers rode in randomly selected patrol cars for 365 8-hour shifts and observed approximately 2,000 police-citizen encounters. Of the 365 8-hour shifts, 79 (21.6 percent), or slightly more than 1 in 5, involved police contact with at least one drunk driver. A wide variety of data were encoded by the observers, either as an encounter occurred or immediately after the encounter. The dependent variable was the arrest decision made by the police officer who handled the drunk-driving encounter. The study concluded that demeanor mattered when it was limited to legally permissible words and displays of hostility and when crime was partially controlled, although the effects of demeanor varied with how it was represented. Other extralegal variables, especially race and class, also mattered. This analysis, therefore, provides support for the findings of research over four decades, i.e., that demeanor and other extralegal variables shape police actions and arrest decisions. 1 table, 5 notes, and 60 references