THE ANALYSIS COVERS 28,238 ARRESTS AND INTRODUCES CONTROLS FOR TYPE OF OFFENSE, AGE, SEX, AND RACIAL COMPOSITION OF NEIGHBORHOODS. NONWHITES ARE MORE LIKELY THAN WHITES TO BE ARRESTED UNDER CIRCUMSTANCES THAT WILL NOT CONSTITUTE SUFFICIENT GROUNDS FOR PROSECUTION; I.E., PROPORTIONATELY FEWER ARRESTS OF BLACKS THAN OF WHITES ARE UPHELD BY THE SUBSEQUENT ISSUANCE OF A WARRANT BY THE PROSECUTOR'S OFFICE. THE RACE OF THE OFFENDER IS A PARTICULARLY IMPORTANT FACTOR IN THE DECISION TO ARREST WITH LESS THAN SUFFICIENT EVIDENCE WHEN THE NONWHITE ARREST OCCURS IN A HEAVILY NONWHITE AREA. THE DATA ALSO SHOW THAT FEMALES ARE MORE LIKELY THAN MALES TO BE ARRESTED UNDER CONDITIONS IN WHICH A WARRANT IS NOT SUBSEQUENTLY ISSUED, PARTICULARLY FOR STEREOTYPICALLY 'FEMALE' CRIMES SUCH AS SHOPLIFTING. THE FINDINGS ARE SAID TO LEND SUPPORT TO THE LABELING AND CONFLICT PERSPECTIVES ON POLICE-CIVILIAN INTERACTIONS. THE FORMER SUGGESTS THAT EXISTING STEREOTYPES OF CRIMINAL OR POTENTIALLY CRIMINAL PERSONS INCLUDE A RACIAL COMPONENT. THE LATTER MAINTAINS THAT THOSE MOST LIKELY TO BE CRIMINALIZED ARE THOSE WITH RELATIVELY LITTLE POWER. SUPPORTING DATA AND A LIST OF REFERENCES ARE INCLUDED. (LKM)