Where Is the Syndrome? Examining Co-Occurrence Among Multiple Problem Behaviors in Adolescence Teena Willoughby, Heather Chalmers, and Michael A. Busseri Brock University and Youth Lifestyle Choices–Community Research Alliance The authors examined co-occurrence among a wide range of adolescent problem behaviors: alcohol, smoking, marijuana, hard drugs, sexual activity, major and minor delinquency, direct and indirect aggression, and gambling. Using a large self-report survey of high school students, confirmatory factor analysis was used to test the problem syndrome model proposed by problem behavior theory. A 3-factor model provided better overall fit than did a single problem syndrome factor model. Contingency table analyses were used to examine the co-occurrence of problem behaviors at different levels of involvement within individuals, as well as relative risk ratios. Analyses offered modest support for a limited problem syndrome encompassing the report of high-risk involvement with alcohol, minor delinquency, direct aggression, and, to a lesser extent, marijuana. For each problem behavior, the majority of adolescents did not report high-risk involvement, and only a minority reported any involvement with multiple behaviors. The research literature is replete with studies that examine multiple problem behaviors in adolescence. Much of this work has used the concept of a “problem behavior syndrome” (Jessor & Jessor, 1977) to explain intercorrelations among behaviors, includ- ing smoking, alcohol use, marijuana use, and sexual activity. A tendency toward deviance or unconventionality has been offered as the cause for this syndrome (Donovan & Jessor, 1985). Initial support came from studies in which a single principal component (in exploratory factor analysis) accounted for covariation among problem behavior measures (Donovan & Jessor, 1985; Donovan, Jessor, & Costa, 1988). Results from these exploratory models were considered as strong support for a problem behavior syn- drome. Proponents have suggested that adolescents reporting one type of problem behavior show a range of problem behaviors (Barnes, 1984; Barnes & Welte, 1986), all of which can be subsumed under a single problem behavior construct (e.g., Ary et al., 1999; Zhang, Welte, & Wieczorek, 2002). In subsequent work, structural equation modeling and confir- matory factor analysis have been used to test relations among measures of problem behaviors. The structure that the syndrome factor usually takes is one of two forms; in both cases, a latent problem syndrome is modeled as the ultimate cause of the indi- vidual behaviors. In the basic model, behavioral indicators (usually single-item measures) are modeled as loading onto a single latent syndrome factor (e.g., Farrell, Danish, & Howard, 1992; New- comb & McGee, 1991; Osgood, Johnston, O’Malley, & Bachman, 1988; White, 1992). In other studies, problem behavior indicators have been modeled as loading onto one of several latent problem behavior factors (e.g., a Soft Drug factor or a Delinquency factor) and these latent first-order factors are then modeled as loading onto a higher order problem syndrome (e.g., Barrera, Biglan, Ary, & Li, 2001; Bui, Ellickson, & Bell, 2000; Newcomb & Felix-Ortiz, 1992). Studies comparing the relative fit of these two types of structural models find support for the higher order factor model compared with a single-factor model (Cooper, Wood, Orcutt, & Albino, 2003; Dembo et al., 1992; Farrell, Kung, White, & Valois, 2000; Garnefski & Diekstra, 1997; Gillmore et al., 1991; Grube & Morgan, 1990; Mason & Windle, 2002; McGee & Newcombe, 1992; Mitchell & O’Nell, 1998; Osgood et al., 1988; Resnicow, Ross-Gaddy, & Vaughan, 1995; Tildesley, Hops, Ary, & Andrews, 1995; Vingilis & Adlaf, 1990; Williams & Ayers, 1996; Zhang et al., 2002). In general, results from a wide array of studies provide evidence for a general problem behavior syndrome (but see Benda, 1999; Gillmore et al., 1991; Grube & Morgan, 1990; Tildesley et al., 1995; White & Labouvie, 1994; Williams & Ayers, 1996). How- ever, although structural models typically provide an adequate description of the covariation among problem behaviors, intercor- relations typically range from .10 to .40. Furthermore, because the amount of explained variation in individual behaviors is rarely more than 50% (and typically averages much less across behaviors in a given study), much of the meaningful variation is not ac- counted for by the syndrome model. Thus, in addition to being part of a syndrome, problem behaviors can be considered as unique phenomena (Allen, Leadbeater, & Aber, 1994; Dembo et al., 1992; Farrell et al., 2000; Garnefski & Diekstra, 1997; Grube & Morgan, 1990; Mason & Windle, 2002; Mitchell & O’Nell, 1998; New- combe & McGee, 1989; Osgood et al., 1988; Resnicow et al., 1995; White, 1992; White, Loeber, Stouthamer-Loeber, & Far- rington, 1999) and adolescents involved in one type of problem behavior will not necessarily be involved in other problem behav- iors (Farrell et al., 1992; Gillmore et al., 1991; Tildesley et al., 1995; White & Labouvie, 1994). Teena Willoughby, Heather Chalmers, and Michael A. Busseri, Depart- ment of Child and Youth Studies, Brock University, St. Catharines, On- tario, Canada, and Youth Lifestyle Choices–Community Research Alli- ance, St. Catharines, Ontario, Canada. This research was supported by the Social Sciences and Humanities Research Council of Canada and by Human Resources and Development Canada. Correspondence concerning this article should be addressed to Teena Willoughby, Department of Child and Youth Studies, Brock University, 500 Glenridge Avenue, St. Catharines, Ontario L2S 3A1, Canada. E-mail: twilloug@brocku.ca Journal of Consulting and Clinical Psychology Copyright 2004 by the American Psychological Association 2004, Vol. 72, No. 6, 1022–1037 0022-006X/04/$12.00 DOI: 10.1037/0022-006X.72.6.1022 1022