Community policing calls for a more selective approach for arresting suspects and influences officer decisions about who should be arrested and under what circumstances. The study anticipated that officers would become more instrumental in their arrest decisions and less bound by legal considerations. Data on 484 police-suspect nontraffic encounters were obtained from observations during ridealongs with Richmond patrol officers in 1992. The arrest/no arrest decision was regressed on a number of variables that represented certain legal and extralegal characteristics of the offense, victim, suspect, and officer. The overall probability for arrest for this sample was similar to that found in pre-community policing departments in previous research. As with previous analyses, situation factors rather than officer characteristics were the most powerful predictors of the arrest decision. Of these, legal considerations (offense seriousness, evidence strength, and victim's expressed preference for arrest) were the strongest predictors, and extralegal considerations (features relevant to the victim's and suspect's social status and threats to officer authority) failed to achieve either statistical or substantive significance. The more positively oriented an officer was to the concept of community policing, the less likely was an arrest. The study concluded that patterns of law enforcement under Richmond's community policing were not different from those reported in similar observational studies conducted in the 1960's and 1970's. An important exception was that officers in this study were governed more by legal considerations and less by extralegal factors than those in previous studies. 4 tables and 32 notes