First published in 1963, Howard Becker’s Outsiders quickly became a reference work, a classic in the sociology of deviance, which discipline it helped to renew. The chapters on marijuana smokers gave rise to several ethnographic studies of illegal drug users and their “careers”, and Becker’s theses still resonate and provoke debate today. In an overview of French research on drug use, C. Faugeron and M. Kokoreff (1999) examine ethnographic studies, then epidemiological surveys, particularly those conducted in schools by means of self-administered questionnaires; they criticize the studies for being too “soft” and the surveys for not making stimulating use of sociological variables. It is worth remarking that these two distinct empirical approaches do not stand in the same relation to Becker’s analyses. Most ethnographic studies refer explicitly to Becker and claim to be working from the same perspective, whereas quantitative studies based on epidemiological surveys don’t even mention him. An examination of the international scientific literature confirms this, particularly the apparent incompatibility between the quantitative approach and that developed by Becker. It should be said that for his part Becker has often been critical of the statistical approach, deeming, for example, that multivariate analysis is ill-adapted to the study of deviance because it cannot grasp the sequentiality of a “career”; multivariate analysis in Becker’s view wrongly assumes a synchronic relation between a dependent variable (deviant behavior) and independent variables, all understood to act simultaneously (1985, ch. 2). For Becker, on the contrary, a variable that is well influential at a given stage in a career of deviance may not be at the next one (and vice versa). Be that as it may, statisticians do have modeling methods that make it possible to handle sequentiality (Gollac, 1997). But Becker doesn’t so much criticize the quantitative methods themselves as the use made of them in a certain school of American sociology (school qualified by Sorokin as “quantophrenic”). In Becker’s view, all research raises the problem of specific methods and interpretations, and sociologists must do their best to resolve these problems, being fully aware that each choice along the way may prove a source of error. He criticizes quantitativists for wanting to overstandardize procedures and research tools; this can lead them to commit the worst of errors, namely believing that their mathematical methods are “transparent”, induce no bias or arbitrariness, protect them from all risk of error (Becker, 1970, ch. 1). In this article I shall be looking at some of Becker’s analyses in Outsiders from a quantitative perspective, making use of a recent epidemiological survey. My approach will not involve a genre mix-up, however–quite the contrary. As Becker himself pointed out, what may seem highly antagonistic approaches often fuel each other; on the one hand, statistical data is precious in orienting and delimiting ethnographers’ research; on the other, we must be knowledgeable of the “softest”, least formalized methods if we wish to be able to interpret quantitative results. [1]

Becker opens Outsiders with an overview of existing definitions of deviance. After speedily dismissing statistical definitions as irrelevant, he moves on to criticize in some detail a widespread conception of deviance that follows the medical analogy : deviance has its source in the individual, which necessarily means the deviant has particular characteristics. In this conception, deviance is likened to a pathology, an “evil” that alienates the individual and whose “etiology” must be discovered (Becker, [1963] 1985, pp. 27-30 and p. 45). Such a conception is incompatible with a sequential approach in terms of career, since that approach assumes an actor with a certain degree of free will shaping his own trajectory through choices made in accordance with the circumstances he finds himself in. The medical metaphor presents the deviant as a diseased person in whom the disease acts, rather than as the actor of his own deviance. According to this “illness model” (Valleur, 1994), the addict in particular is the victim of a more or less irreversible, morbid or incurable process that develops independently of her will. In this perspective, the pathogenic agent, not the individual, has the leading role, and is understood to bring about the addict’s inevitable physical, moral, and social degeneration. For Becker it was important to assert that a cannabis smoker could very well be a “normal” individual who did not suffer from particular psychological problems. Above all he underlined the fact that marijuana does not cause physical dependence. This point, still regularly confirmed today (Zimmer and Morgan, 1997; Roques, 1999), is essential to his argument. In effect, absence of dependence makes it possible to reject the stereotype of the compulsive addict enslaved to his vice. [4] In methodological terms, no dependence justifies use of the career notion, and with it the rejection of determinist theories of deviance. [5] From this perspective it is no longer a matter of demonstrating why a given individual necessarily became a cannabis user, but how it was possible for him or her to become a user (Becker, 1994).

Epidemiologists today use multifactorial causal models; they no longer look for a unique determining cause for a given disease, but rather a cluster of causes that are neither necessary or sufficient; that is, “risk factors” for the disease. The multifactorial aspect requires the researcher to be attentive to confounding factors : the more variables there are, the more likely that some are statistically linked, meaning that the apparent influence of a given variable on the phenomenon may very well mediate the influence of another variable, then disappear as soon as this other variable is taken into account. In this case the first variable is the confound and must be closely controlled for. For example, “serious” suicide attempts are more frequent among cannabis “abusers”. Nonetheless, if we take sociodemographic profile and user’s personal history into account and then compare two individuals with the same profile and history, one a cannabis “abuser” and the other not, the risk that the second will seriously attempt suicide is only slightly lower than for the first (Beautrais et al., 1999). To neutralize confounds and have valid “otherthingsbeingequal” results to interpret, statisticians use analysis of variance methods. When a variable of interest is dichotomous (eg, whether or not to consume a substance) or, more generally, qualitative, they most often use logistic regression. [10] This method makes it possible to take numerous variables into account in a single model and measure the impact of each one on the dependent variable while controlling for the others.

[13]

As explained, Becker developed a sequential approach to marijuana use that involves two axes : gradual learning about the product’s effects and moving from initiation to occasional use to regular use. Becker wanted to break with what he saw as an over-“psychologizing” research tradition; for him marijuana use is not determined by psychological characteristics or motivations because motivations themselves are modified with use. Experimentation is due above all to curiosity, and the sought-after effects do not become a motivation until the individual has learned to bring them on, perceive them, and enjoy them. It is “learning” to use that produces motivation, rather than the other way around. Later qualitative studies have confirmed the existence of a learning process for cannabis (Green and Miller, 1975; Hirsch et al., 1998; Aquatias et al., 1999) as well as other substances, particularly heroin (Duprez and Kokoreff, 2000). Closed-ended-question questionnaires are probably not a well-adapted means of observing this process. We should note, however, that pupil surveys confirm the primacy of curiosity as a motive for experimentation : this was the reason most often cited by Paris high school students (Peretti and Leselbaum, 1999) as well as middle school and high school students in the ESPAD survey for whom cannabis was the first drug ever taken (asked about their reasons for taking cannabis, 82% chose “I was curious”). Moreover, since cannabis is most often smoked in the form of user-rolled cigarettes, learning how to use it is facilitated by having smoked tobacco. It is hardly surprising that in the ESPAD study, almost all the respondents who had already smoked cannabis had also already smoked a cigarette (98%). Tobacco is not so much a “risk factor” as a near-necessary condition of cannabis use. In the learning process described by Becker, peer group in which a cannabis smoker was initiated plays an important role. Quantitative surveys produce an indication of peers’ role, in that the presence of cannabis users among teenager’s immediate family and friends is more marked for regular consumers than experimenters (Peretti and Leselbaum, 1999).

According to Becker a starter must first and foremost get beyond the norms condemning marijuana as a dangerous, toxic substance that causes the individual to lose control of himself. The individual’s opinion on whether marijuana is dangerous will change with use : if he moves from the occasional to the regular phase, he will revise that opinion, specifically by deeming the sub-stance does not engender dependence. One user Becker questioned compared marijuana to alcohol–to underline the advantages of the former. Another respondent, to convince himself that he was not dependent, that he was in control of his consumption, decided to stop smoking for a week; afterward, reassured, he began again. Yet another, sensitive to ideas presenting drug use as a sign of psychological and moral weakness, ultimately concluded that because he was aware of this risk he must still have it under control. In sum, to start, continue, or increase consumption, a marijuana smoker has to “neutralize” the stereotypes condemning his practice by convincing himself that they reflect the opinion of people who don’t know anything about the matter, and substituting a more favorable interpretation, based on his own experience and that of his peers.

It should be specified from the outset that the ESPAD survey is crosssectional whereas Becker’s perspective is longitudinal, following the trajectory of a given individual over time to apprehend the successive stages of the process leading, or not, to regular cannabis use. In sum, his approach is diachronic. The ESPAD survey in contrast consists of data collected on a single date; it observes cannabis consumption among teenage respondents at that instant and only that instant; it is synchronic. To reconcile the two approaches I will hypothesize that the various use levels indicated by respondents correspond to different phases in the process Becker analyzed. In other words, the ESPAD survey offers no more than a cross-section of the process but nonetheless can be assumed to provide information on the various stages in that, at the moment of the survey, the different respondents were at different stages (no stage at all or barely initiated for some, intermediate or “last” for others). I will assume here that comparison of individuals who have identified themselves as being at levels that are clearly distinct from one another will enable us to understand why a given individual moves from one level to another.

Table I is to be read by columns; for example, 44.5% of abstainers are boys and 55.5% girls. All variable combinations presented are significant at a highly satisfactory statistical threshold (except stating that it’s easy to procure cannabis in a café or discotheque). The column entitled “at least once” aggregates all the following columns and thus makes it possible to specify the results we would have obtained if we hadn’t distinguished between different consumption levels. On the basis of the table we may state the following :

Let us first consider stated opinions on people who take one or another of the drugs and on how dangerous the different substances and using them are. All the combinations presented in Table III are significant at a highly satisfactory statistical level. Table III reads thus : 28.4% abstainers say they are “not against” people who try cannabis once or twice.