Testosterone, Social Class, and Antisocial Behavior in a Sample of 4,462 Men

Two hypotheses have been offered to explain the relation between testosterone and antisocial behavior in delinquent and criminal populations. One is that testosterone leads directly to antisocial behavior. The other is that a constellation of dominance, competitiveness, and sensation seeking associated with testosterone leads to either antisocial or prosocial behavior, depending upon an individual's resotirces and background. Analysis of archival data from 4,462 U.S. military veterans supported the first hypothesis: Jj^st^sterone was correlated with a variety of antisocial behaviors among all individuals. However, socioeconomicstali4sJ^SES)_2rovedto^ be a moderatmg variable, with weaker testosterone^ havior relationshipj_among^ high SESTuhJecis. Testosterone has begjTjelated to huma7i"antisocial behavior, but primarily in delinquent and cnminal populations. It hasjgen found that prisoiLmmates-high-^. in testosterone are morg_likely-4Q-ha3K£_ committed v io len t^ imes (Dabbs, FfaHyTCarr, & BeschTl^STrDabbs, Ruback, Frady, Hopper, & Sgoutas, 1988; Ehrenkranz, Bliss, & Sheard, 1974), engaged in criminal activity at an early age (Kreuz & Rose, 1972), been more dominant in prison and violated prison rules more often (Dabbs et al., 1987; Ehrenkranz et al.., 1974), and been judged more severely by parole boards (Dabbs et al., 1988; Kreuz & Rose, 1972). Testosterone in adolescent males has been associated with dominance, leading especially lo delinquency when they feel unfairly restricted (Olweus, 1983) or prematurely try to engage in activities (especially involving sex, drugs, alcohol, and tobacco) usually reserved for older persons (Udry, 1989). Focussing on criminals and delinCorrespondence and reprint requests to James M. Dabbs, Jr., Depanment of Psychology, Georgia State University, Atlanta, GA 30303. quents, however, may obscure more positive relationships. JDaitzman agd ZiickeEmanJj980)^uggest that testostergne is related to a general sensationsee^kin£j£iidenc^SensaTion seeking ca have positive or negative effects, pending on one's social background and resources. Individuals low in socioeconomic status (SES) often find the most exciting things to do are illegal, while high SES individuals can do things that are both exciting and socially acceptable—driving fast cars instead of stealing them, arguing instead of fighting, playing college football instead of assaulting. Data from a large sample of normal adults could help clarify the relationship between testosterone and behavior. We obtained such data from a study of 4,462 male U.S. military veterans. These men, half of whom had served in Vietnam during 1965-1971, were studied beginning in 1985 to assess long-term effects of the Vietnam military experience (Centers for Disease Control, 1988). Their mean age was 37 years, and they were representative ofthe U.S. population in race, education, income, and occupation. All had been enlisted men. They were interviewed by telephone and underwent extensive medical, psychological, neuropsychological, and laboratory examinations at the Lovelace Medical Foundation (Albuquerque, NM). Laboratory examinations included assays of testosterone in serum samples collected in the morning before eating; the coefficient of variation (CV) among quality control samples was less than 10%. Subjects completed the Minnesota Multiphasic Personality Inventory and were interviewed using the Diagnostic Interview Schedule (DIS), which provides information on DSM-III psychiatric diagnoses (American Psychiatric Association, 1980). The DIS was modified to include additional items on antisocial behavior, The present paper examines relationships between subjects' testosterone levels and reports of their own behavior. Following the epidemiologieal approach used by the Centers for Disease Control (CDC) in the original study, we treated testosterone as a risk factor that might increase the likelihood of antisocial behavior. This approach dichotomizes continuous variables and thus loses some information, but it has the advantage of focussing attention directly on the behavior of extreme groups. We classified subjects as high in testosterone if they were in the upper 10% ofthe distribution and as nonnal if they were in the remainder of the distribution. We then computed ratios to contrast the risk (or likelihood) of various behaviors in the high testosterone group with the risk of these behaviors in the normal testosterone group. For example, a risk ratio of 2.0 would indicate a given behavior was twice as probable in the high as in the normal testosterone group. A risk ratio is considered statistically significant (p < .05) when its 95% confidence interval excludes 1.0. Risk ratios were adjusted (SAS Institute, 1985) for subjects' age, race (white or nonwhite), and place of service (Vietnam or other), variables that might plausibly affect the testosterone-behavior relationship. The findings are summarized in Table 1. The behaviors in Table 1 are described in footnotes to the table; they are as defmed in the original CDC data set. These behaviors occurred at times ranging from childhood to the recent past. Individuals higher in testosterone more often reported having trouble with parents, teachers, and classmates; being assaultive toward other adults; going AWOL in the military; and using hard drugs, marijuana, and alcohol. They also reported having more sexual partners. The overall picture is one of delinquency, substance abuse, and a general tendency toward excessive behavior. We then examined SES, to see if it moderated the testosterone-behavior relationship. There was relatively little social class information in the data set, and we used subjects' education and current VOL. L NO. 3. MAY 1990 Copyright © 1990 American Psychological Society 209 PSYCHOLOGICAL SCIENCE 'estostcronc and antisocial behavior Table 1. l\-rccnui,iic of normal and high testosterone subjects reporting high levels of various hchaviois