The study analyzed data collected in 1977, using 611 police-citizen encounters observed by civilian observers riding on patrol shifts. Results indicate that police arrest decisions are influenced by the victims' race and by the race of female suspects. Police may behave differently in poorer neighborhoods, making more arrests if the suspects are discovered in lower socioeconomic areas. Police may ascribe more pejorative traits to these suspects, heightening their propensity to arrest. This affects all residents of poor neighborhoods -- white and black. Although little evidence of racial bias against suspects was found, differential police response to victims is evident. Police are more likely to arrest in situations involving white complainants, while denying black complainants in similar situations the protection of law. Differential responsiveness appears strongest with property crimes and encounters in which complainants want offenders arrested. In both situations, police are more responsive to white victims. Police may exhibit bias because they view black complainants as less deserving of legal protection, because they are unsympathetic to minority groups generally and blacks specifically, because they view lower class residents as more deserving of arrest, and because they see whites as high status and thus more credible. Police racial bias can be offset by prosecutors refusing to fully prosecute such cases. A total of 41 footnotes and data tables are supplied.