The commonly held view among lay people and the still dominant view in the professional community is that drug involvement, as a problem for both user and society, does not begin until adolescence or perhaps a trifle earlier (Maddox & McCall, 1964; Kandel 1978). It is for this reason that the study of adolescent drug involvement has assumed such importance to workers concerned with etiology and with early intervention. Gateway phenomena of onset and of first problematic involvement with these substances are seen as taking place during this developmental time frame, so the period of youth assumes special importance (Braucht, 1982; Johnston, O’Malley & Bachman, 1984).

The present paper does not dispute this perspective, but it reviews recent evidence and theory which broadens the picture of acquisition considerably. The review suggests that the developmental frame of relevance for drug involvement needs to be extended backward in time to considerably earlier periods of childhood if our understanding of acquisition and involvement phenomena is to be complete. Also considered along the way are the role of interactional processes in contributing to these phenomena and the issue of continuity and discontinuity in developmental process. Last, discussion considers the design of intervention programs that might, early on, alleviate risk for later drug problems and related sequelae.

The paper is organized in four sections. The first section provides an overview of earlier studies of person-environment interactions as they relate to issues of the onset of drug use and of problem use in adolescence. Of special relevance here are the implied conceptual boundaries of “environment” utilized in this work. The next section reviews the small but growing literature on preadolescent populations, and deals with the issues of what individual and environmental factors during these earlier childhood years appear to be related to current and/or future drug involvement. Section Three examines issues relating to course and to developmental process as these help provide a framework of understanding within which to assimilate the studies of youth and earlier childhood. Of particular interest is the question of how risk factors may aggregate or disaggregate over developmental time, and the extent to which interactional factors facilitate or impede such processes. Section Four describes an early stage prevention-intervention program, the Michigan State University Prevention of Conduct Disorders Project, that has recently been implemented as a direct outgrowth of the work reviewed above. Discussion also focuses on the relevance of such a program to the larger issue of person-environment interaction over the life span, and the extent to which abusive drug involvement and related conduct problems may or may not be preventable phenomena.

An interactional perspective on behavior can be traced at least as far back as Kurt Lewin’s (1951) famous equation B= f(P,E) and Henry Murray’s (1938) conception that action is a result of the joint combination of needs (motives) and environmental presses, but it has only been within the past twenty years that significant attention has been paid to the details of the process. Within the arena of drug and alcohol studies, and most especially in the context of studies of drug use and onset, this era has produced major work on this problem. The most systematic of these studies have been produced by three research groups — at UCLA (e.g., Huba & Bentler, 1984), at Colorado (e.g., Jessor & Jessor, 1977), and at Columbia, involving Kandel and her colleagues (e.g., Kandel, Kessler & Margulies, 1978) — all of whom in different ways have found evidence for such interaction.

Working primarily with adolescent respondents, but in a few instances with parent and peer respondents as well, a number of across study correspondences are noted. Person variables — differently described as generalized deviant behavior (Jessor & Jessor, 1977), negative law abidance (Huba & Bentler, 1984), and delinquent involvement (Margulies, Kessler & Kandel, 1977; Kandel, Kessler & Margulies, 1978) — but all having to do with the potential for antisocial involvement, have all been linked to the onset of both alcohol and marijuana use. In fact, in one study of first graders, aggressiveness at T, was related to frequency of drug use as a teenager (Kellam & Brown, 1982).

At the attitudinal level, especially as related to onset, views reflective of an anticonformist, rebellious, prodeviancy position have similarly been linked to both onset and maintenance phenomena (cf. also Smith & Fogg, 1978; Bachman, Johnston & O’Malley, 1981). These findings, generated out of different theoretical superstructures, have alternately been considered to be evidence of lack of law abidance (Huba & Bentler), greater tolerance for deviance, and transition proneness (Jessor & Jessor), and evidence for anticipatory socialization via direct shaping effects from peers, and modelling effects from parents (the New York group).

Self-esteem, another personality variable concerned with central organizing tendencies, has been more equivocally related to onset and later involvement in drug use. Thus Kaplan and his co-workers systematically have shown associations between earlier low self-esteem and later drug involvement (e.g., Kaplan, Martin & Robbins, 1982). Others (e.g., Smith & Fogg, 1978) report an association to onset of marijuana use, and to the move to hard drug involvement (Huba & Bentler, 1984), but this association has not appeared in other work (Jessor & Jessor, 1978; Paton & Kandel, 1978).

More generally, person characteristics in the Lewinian equation are conceived as one facet of the independent variable network, that are interactively engaged with aspects of the environment. But from the standpoint of psychology as a discipline, such characteristics are also aspects of the dependent variable network insofar as the behavior system is very much a part of the person system.¹ Thus a number of research groups have, in the language of this review, been interested in person-environment interactions as they predict (other) person characteristics. In this context drug involvement has been construed as one aspect of a cluster of activities, involving earlier and greater involvement in delinquent acts, in aggressive and assaultive behavior, less church attendance, and more generally poorer performance in prosocial activities such as school achievement (Donovan & Jessor, 1985; Huba & Bentler. 1984; Kaplan et al., 1982; Kandel, 1978; Pandina & White, 1984). Insofar as this conceptualization is correct, it raises the question about whether the phenomenon being investigated fits a traditional cause-effect model which includes action upon dependent variables. An alternative, possibly more accurate representation of what is taking place is the trans-active model, involving numerous feedback loops and an iterative process among multiple sources of influence. We shall return to this point later in the paper.

The environment is a large arena, and how it has been divided up and conceptualized by different research groups shows substantial variation in the organizing schemas guiding the research. Such variation also leads to differences in the selection (and measurement) of the potentially vast ecological domain. Nonetheless all of these researchers have attended to aspects of the sociocultural system, the family system (i.e., family of origin), and the peer system.

The pathway of effect from influence at the sociocultural system level to impact upon the person is in the strictest sense, a one way street (cf. Zucker, 1979), and in that respect is not capable of producing person-environment interactions. The manner in which the social structure concretely interacts with the individual is by way of the community, the neighborhood, the job and the educational system. In these respects sociocultural influences need to be regarded as opportunity structures within which the interactive relationships play themselves out (cf. Jessor, Graves, Hanson & Jessor, 1968). It is undoubtedly for these very reasons that investigators of sociodemographic characteristics as related to drug involvement, including both alcohol and marijuana use and abuse, have found low order to negligible relationships between social status and involvement (Jessor & Jessor, 1977; Kandel et al., 1978; Rachel, Maisto, Guess & Hubbard, 1982). In contrast, those aspects of one’s demography that impinge upon life-style (e.g., what sex one is, whether one lives in an urban as compared to a rural area, what region of the country one lives in, what school one attends) consistently are moderately related to patterns of onset, use, and to particular drug of involvement (Johnston, O’Malley & Bachman, 1984).

At a life-style influencing level closer to home, in this case the social structure of the high school, one study found that group membership in different cliques in the school was significantly related to differences in consumption patterns (Riester & Zucker, 1968). Members of the two highest prestige cliques (the Collegiates and the Leathers) had higher alcohol consumption patterns than did members of less prestigious cliques, or nonmembers. While this effect might appear attributable solely to peer influences, it is more correctly a social structural effect that extends beyond the immediate peer group. Cliques retain identities and norms over periods of time that outlive the membership of any particular set of peers or individual participants.

The possibility that socialization effects by parents influence the adolescent’s early drug involvement has frequently been considered to be one plausible hypothesis to account for onset and early use patterns, and the three major longitudinal studies already referred to, have not surprisingly, at least tentatively looked at this possibility.

The Columbia study, in particular, is useful in examining the extent of parent contribution because data have been assembled on adolescents, mothers and fathers, rather than simply utilizing perceptions of parent behavior generated by the adolescent.² The collection of parallel data from friends of the primary respondents also allowed for the evaluation of comparative parent versus peer effects. The pattern of contributory effects is a complex one (Margulies, Kessler & Kandel, 1977; Kandel, Kessler & Margulies, 1978). With regard to initiation into hard liquor use, parent effects, and most especially the modelling of the parent’s own pattern of distilled spirit use, are positively and significantly related to earlier onset. However, this influencing network provides a small proportion of the contributory variance. Evaluations of differences in strength of effect between parents, peers, and other sources of influence showed that the parental influences were fourth in line, after child’s own prior drug involvement in cigarettes and beer/wine, the child’s earlier pattern of delinquent behavior involvement, and peer influences.

To underscore the fact that there is no simple answer to this developmentally complex issue, the Margulies et al. (1977) analyses demonstrate that the influencing structures shift in some spheres but not in others as one takes different cross-sections of developmental time. Thus, (a) parent effects remain constant in amount of contributory variance to distilled spirit onset, but cohorts that begin this use pattern at older ages increasingly show friends becoming more influential as the cohorts progres...