The study compared current gang members, gang-involved youth that were not currently members, and nongang youth on several dimensions, including sociodemographic characteristics, family background, school experiences, street experiences and exposure, emotional problems, and delinquency. Empirical models of gang delinquency were tested. It was hypothesized that the selection model would operate for the sociodemographic characteristics, family background, school experiences, street experience and exposure, and emotional problems; and the facilitation model would operate for delinquency. The first hypothesis suggests that gang members and gang-involved youth would differ from nongang youth and that gang members and gang-involved youth would not differ from one another. The second hypothesis suggests that gang members would engage in more delinquent behaviors than would gang-involved youth that would, in turn, engage in more delinquent behaviors than would nongang youth. Over 600 homeless and runaway adolescents in 4 Midwestern States were interviewed on the streets, in shelters, and in drop-in centers. Some evidence was found for the selection model in the following variables: family legal problems, having been suspended from school, age first on their own, deviant peers, and having made a suicide attempt. Evidence was found for the facilitation model for the externalizing behaviors scale and arrest record. There was no evidence found for the facilitation model for substance use, although the results supported the selection effects model. Youth gang members engaged in more deviant subsistence strategies than did noninvolved or involved youth. The findings and limitations of this study suggest directions for future research. This includes determining the relative timing of gang affiliation and running away or homelessness; asking the youth why they became involved with gangs; and providing a more complete picture of the youths’ broader social milieu. 1 table, 48 references