There is consensus in the policing literature that demeanor and other extralegal variables help determine police decisions. A recent challenge to this consensus has been presented, however. Klinger (1994) has asserted that nearly all previous quantitative analyses of the effects of demeanor and other extralegal variables are fatally flawed, because they failed to limit demeanor to spoken words and failed to control for crime. Klinger concluded that all previous research is suspect until additional analyses of the data sets used in previous research and new observational research are presented. This research begins the first of these tasks by reporting additional analyses of data from three previously published papers based on the Midwest City Police-Citizen Encounters Study. With demeanor limited to spoken words and crime partially controlled, the reanalyzed data suggest that the effects of demeanor depend on how demeanor is represented and, to a lesser extent, model specification. Consequently, caution with respect to existing reports of the effects of demeanor and other extralegal variables remains necessary. In addition, carefully controlling for crime and limiting demeanor to spoken words many not be the only problems associated with efforts to assess the effects of demeanor. This research suggests that multiple representations of demeanor and more fully specified models may be important also. 3 tables and 50 references