All four hypotheses tested were supported. In the first hypothesis, co-offending was most common at onset and decreases thereafter in most careers. The test of hypothesis 2 concerning the relationship of co-offending and criminal experience resulted in the discovery of two types of offenders or delinquent careers: high-activity offenders with eight or more recorded offenses in their career, and low-activity offenders, with seven or fewer offenses. Consistent with hypotheses 3 and 4 the prevalence of co-offending decreases with the age of the offender, except during childhood, and varies by the type of offense. The hypotheses tested were derived from Reiss's (1986, 1988) taxonomic theory of co-offending using police-reported data on the delinquent careers and co-offending of 55,336 Canadian offenders. The police-reported data was used on many young offenders and co-offenders to investigate the role of co-offending in the development of the delinquent career and to test hypotheses derived from Reiss's theory. The four hypotheses tested include 1) there are two types of delinquents who are distinguished both by their levels of criminal activity and by the role that co-offending plays in their delinquent careers; 2) co-offending is most common at the onset and early stages of the delinquent career; 3) the tendency to co-offend, rather than to offend alone, decreases with the offender's age; and 4) the tendency to co-offend, rather than to offend alone, varies with the type of offense. Figures, tables, references, and appendix