VII. Testing Social Learning Theory
Although full empirical tests of all of the dimensions of Akers’s social learning theory did not emerge in the literature until the late 1970s, early research, such as Sutherland’s (1937) qualitative study of professional theft and Cressey’s (1953) well-known research on apprehended embezzlers, provided preliminary support for differential association (e.g., also offering support for social learning). Following these seminal studies, research now spanning more than five decades has continued to demonstrate varying levels of support for the various components of social learning theory, and the evidence is rather robust (see Akers & Jensen, 2006). There are far too many studies to make an attempt to list or discuss each individually; therefore, the following discussion is limited to noting the findings of some of the most recognizable and comprehensive tests of Akers’s social learning theory performed by Akers and his associates.
Akers has tested his own theory with a number of scholars over the years across a variety of samples and on a range of behaviors from minor deviance to serious criminal behavior, and this research can best be summarized in terms of four projects: (1) the Boys Town study, (2) the Iowa study, (3) the elderly drinking study, and (4) the rape and sexual coercion study. The first of these projects, and by far the most well-known and cited, is the Boys Town study (for a review, see Akers & Jensen, 2006). This research project involved primary collection of survey data from approximately 3,000 students in Grades 7 through 12 in eight communities in the Midwest. The majority of the survey questions focused on adolescent substance use and abuse, but it was also the first survey that included questions that permitted Akers and his associates (Akers et al., 1979) to fully test the four components of social learning theory.
The results of the studies relying on the Boys Town data provided overwhelming support for Akers’s social learning theory, including each of its four main sets of variables of differential association, definitions, differential reinforcement, and imitation. The multivariate results indicated that greater than half of the total variance in the frequency of drinking alcohol (R2 = .54) and more than two thirds of the variance in marijuana use (R2 = .68) were explained by the social learning variables. The social learning variables also affected the probability that the adolescent who began to use substances would move on to more serious involvement in drugs and alcohol. Not only did the social learning variables yield a large cumulative effect on explaining substance use, but also each of the four elements exerted a substantial independent effect on the dependent variable (with the exception of imitation). The more modest results found for the effect of imitation on substance use was not surprising considering the hypothesized interrelationships among the social learning variables. Also, imitation is expected to play a more important role in initiating use (first use) versus having a strong effect on the frequency or maintenance of use. Lanza-Kaduce, Akers, Krohn, and Radosevich (1984) also demonstrated that the social learning variables were significantly correlated with the termination of alcohol, marijuana, and hard drug use, with cessation being related to a preponderance of non-using associations, aversive drug experiences, negative social sanctions, exposure to abstinence models, and definitions unfavorable to continued use of each of these substances.
The second research project, the Iowa study, was a 5-year longitudinal examination of smoking among junior and senior high school students in Muscatine, Iowa (for a review, see Akers & Jensen, 2006). Spear and Akers (1988) provided the initial test of social learning theory on the first wave (year) of the Iowa data in an attempt to replicate the findings of the Boys Town study. The results of the cross-sectional analysis revealed nearly identical results among the youth in the Iowa study as was previously found in the Boys Town study. Once again, the social learning variables explained over half of the variance in self-reported smoking, and each of the social learning variables had a rather strong independent effect on the outcome (with the exception of imitation). Additional evidence provided by Akers (1998) illustrated the substantial influence of the adolescents’ parents and peers on their behavior. When neither of the parents or friends smoked, there was a very high probability that the adolescent abstained from smoking, and virtually none of these youth reported being regular smokers. In contrast, when the adolescent’s parents and peers smoked, more than 3 out of every 4 of these youth reported having smoked, and nearly half reported being regular smokers.
The longitudinal analysis of the Iowa data also provided support for social learning theory. Path models constructed using the first 3 years of data indicated that the direct and indirect effects of the social learning variables explained approximately 3% of the variance in predicting who would be a smoker in Year 3 if that individual had not reported being a smoker in either of the 2 prior years. Although this evidence was relatively weak, stronger results were found for the ability of the social learning variables to predict the continuation and the cessation of smoking by the third year (approximately 41% explained variance; Krohn, Skinner, Massey, & Akers, 1985). Akers and Lee (1996) also provided longitudinal support for the social learning variables’ capacity to predict the frequency of smoking using the complete 5 years of data from the Iowa study and revealed some reciprocal effects for smoking behavior on the social learning variables.
The third project was a 4-year longitudinal study of the frequency of alcohol use and problem drinking among a large sample of elderly respondents in four communities in Florida and New Jersey (for a review, see Akers & Jensen, 2006). Similar to the results of the Boys Town and Iowa studies, which examined substance use among adolescents, the multivariate results in this study of elderly individuals also demonstrated significant effects for the social learning variables as predictors of the frequency of alcohol use and problem drinking. The social learning process accounted for more than 50% of the explained variance in self-reported elderly alcohol use/abuse.
The last project by Akers and his associates reviewed here is a study of rape and sexual coercion among two samples of college men (Boeringer, Shehan, & Akers, 1991). The findings in these studies also mirrored the results of the previous studies by Akers and his associates, with the social learning variables exerting moderate to strong effects on self-reported use of nonphysical coercion in sex in addition to predicting rape and rape proclivity (i.e., the readiness to rape). Although Akers and his associates have continued to test social learning theory to various degrees using dependent variables such as adolescent alcohol and drug use (Hwang & Akers, 2003), cross-national homicide rates (Akers & Jensen, 2006), and even terrorism (Akers & Silverman, 2004), the findings from the classic studies just reviewed clearly identify the strength of the empirical status of social learning theory.