Data were obtained from the NSCR (Netherlands Institute for the Study of Crime and Law Enforcement) School Project, which has collected detailed information on juvenile delinquency, peer involvement, and social networks. The study's focus was on whether a person's best and regular friends differed in their level of delinquency, whether the relationship between respondent and peer delinquency differed between best and regular friends, how the level of respondent delinquency differed for different combinations of regular and best friends' delinquency, and whether indirect and direct measurements of peer delinquency resulted in different findings about the relationship between respondent and peer delinquency. Best friends differ from regular friends in the meaning of the friendship for the individual, the activities shared through the friendship, and the types of conversations. The study used two methods for measuring peer delinquency: the conventional method of asking respondents about their peers' behavior; and the social network method, through which peers report about themselves. The study findings indicate that respondents can be best and regular friends who differ in their degree of delinquency and that the link between respondent and peer delinquency does not differ much between regular friends and best friends. Apparently both types of peers influence the level of respondents' delinquency. Measures based on the direct network method resulted in higher estimates of peer delinquency but in lower estimates of the link between respondent and peer delinquency. 4 tables, 35 references, and appended cross-tabulation of indirect measurements of delinquencies