It uses log-linear analysis to allow simultaneous control for the influence of prior record, type of allegation, family type, sex, race, and county in analyzing data from police and court records in New Jersey. Consistent with the hypotheses, the findings indicate that racial bias is more apparent in police dispositions than in judicial decisions. Police bias is especially pronounced in the more urban of the two social settings studied, where minorities constitute a relatively high proportion of the population. However, this bias may be somewhat compensated for by the courts. Tabular data, 10 footnotes, and about 40 references are supplied. (Author abstract modified)