بزاق تستوسترون و پرخاشگری، بزهکاری و تسلط اجتماعی در یک مطالعه طولی مبتنی بر جمعیت مردان نوجوان

Abstract Testosterone (T) has been found to have a stimulating effect on aggressive behavior in a wide range of vertebrate species. There is also some evidence of a positive relationship in humans, albeit less consistently. In the present study we investigated the relationship between T and aggression, dominance and delinquency over time, covering a period from early adolescence to adulthood. From a large population-based sample (n = 1.161) a subgroup of 96 boys was selected whose behavior had been assessed repeatedly by different informants from age 12 to 21 years, and who had provided multiple T samples over these years of assessment. On the whole, a decrease in aggressive and delinquent behavior was observed in a period in which T rises dramatically. Boys who developed a criminal record, had higher T levels at age 16. In addition, positive associations were observed between T and proactive and reactive aggression and self-reported delinquent behavior. Over the pubertal years different forms of aggressive and delinquent behavior were positively related to T, which may indicate that specific positive links are dependent on the social setting in which this relationship is assessed.

Introduction Testosterone (T), the most important male sex hormone, affects not only physical but also behavioral masculinization. For example, T has been found to increase aggressive behavior in a wide range of vertebrate species (Archer, 1988). Studies in male rodents show that competitive or intermale aggression increases at puberty, a time in which T levels dramatically rise. Also, administration of T results in an increase in aggression (Brain, 1979), whereas it is reduced by (chemical) castration (see Van Goozen et al., 1995 for results in humans). In human adults, T has been found to be related to delinquency, drug abuse (Dabbs and Morris, 1990) and criminal violence (Dabbs et al., 1995, Ehrenkranz et al., 1974, Kreuz and Rose, 1972 and Strong and Dabbs, 2000), as well as to conduct problems in childhood (Dabbs and Morris, 1990). However, Bain et al. (1987) found no difference in T between men charged for aggressive or non-aggressive crimes. It is clear, therefore, that in humans the evidence is at best suggestive of a positive relationship (Archer, 1991). Much less information exists about the relationship between T and aggressive behavior in children and adolescents. Some studies found a positive relationship between T and physical and verbal aggression (Olweus et al., 1980 and Olweus et al., 1988), persistent externalizing behavior (Maras et al., 2003), and age-graded norm-violating behaviors (Udry, 1990) in youngsters. Scerbo and Kolko (1994) studied pre- and early adolescent clinical cases and found that T was related to increased staff-rated aggression. Likewise, disruptive boys approaching puberty seem to have higher T levels as compared to normal controls (Chance et al., 2000). Sánchez-Martín et al. (2000) found a positive relation between T and the frequency of engaging in aggressive interactions in four-year-old boys. However, there are also quite a few studies that have found no relationship between aggression and T in youngsters (Constantino et al., 1993, Halpern et al., 1994, Inoff-Germain et al., 1988, Mattsson et al., 1980, Susman et al., 1987 and Van Goozen et al., 1998). One study found a relationship between T and disruptive behavior in girls, but not in boys (Granger et al., 2003). In another study (Schaal et al., 1996) boys with a history of high levels of physical aggression between ages 6 and 12 had lower T levels at age 13 than boys without such a history. These mixed findings in studies on children and adolescents may be due to important methodological differences between the studies. First, some studies used clinical samples of children referred for disruptive behavior disorders (Chance et al., 2000, Constantino et al., 1993, Scerbo and Kolko, 1994 and Van Goozen et al., 1998), while others used population-based samples (Olweus et al., 1980, Schaal et al., 1996 and Udry, 1990). Second, studies differ not only in sample size but also in the number of samples taken for T analysis. Third, different instruments have been used to investigate this relationship and therefore a mixture of information on various forms of aggressive or dominant behaviors (disruptive, assertive, or physically aggressive behaviors) has been collected (Tremblay et al., 1998). Studies that assess aggression have generally not taken into account different types of aggression, e.g., whether it is reactive or proactive in nature, and it could well be that different types of aggression have different relationships with T. Reactive and proactive aggression have been observed in children and adolescents (Brendgen et al., 2001, Dodge et al., 1997, Pulkkinen and Tremblay, 1992 and Vitaro et al., 1998). And finally, studies examining the T-aggression relationship in youngsters obviously differed in the ages of their participants, ranging from prepubertal to postpubertal, which could clearly affect the results. Schaal et al. (1996) found that boys who were persistently physically aggressive had lower T levels at age 13 than boys who were physically aggressive, but these same boys had higher T at age 16, and therefore a group by time interaction was observed (Tremblay et al., 1997). Studies in nonhuman primates also show a relatively strong association between testosterone and dominance (Mazur and Booth, 1998 and Paikoff and Brooks-Gunn, 1990). Social dominance may or may not involve aggressive behavior and it has been suggested that a more direct relationship exists between T and dominance (Albert et al., 1993 and Strong and Dabbs, 2000). This line of reasoning has been supported in humans (Schaal et al., 1996, Rowe et al., 2004 and Archer, 2006), in which it was shown that testosterone levels were associated with social success rather than with physical aggression. In the present study, we examined the relationship between testosterone, using multiple measurements of T in each year of assessment, and aggression, dominance, and delinquency, in a period covering early adolescence to adulthood. Puberty is a period in which T levels progressively rise from extremely low to mature levels, and it is also in most cultures a period of psychological development characterized by, among others, increases in antisocial and delinquent behavior (Moffitt, 1993 and Weisfeld and Berger, 1983). One would therefore expect that a longitudinal study from late childhood to adolescence and adulthood could provide crucial data on the influence of T on aggression. To this end, we based our study partly on data reported by Schaal et al. (1996), but extended the measurement period to adulthood, and used a larger number of behavioral assessments. A first goal was to examine whether, in line with a rise in T, different types of aggressive, dominant, or delinquent behavior increased from early adolescence into adulthood. Secondly, we investigated whether physical aggression, social dominance and/or delinquent behavior, as shown from childhood to adulthood, had a positive relationship with (changing) T levels. We expected to find an overall increase in aggressive, dominant, and delinquent behavior, together with a rise in T, in our assessment period. Moreover, when examining T levels in separate years, we expected to find positive relationships between T and physical aggression, social dominance, and/or delinquent behavior.

Results Behavioral and T changes in the pubertal period Fig. 1 and Fig. 2 show the patterns of mean T levels at ages 13, 16, and 21 (Fig. 1) and behavioral assessments from early adolescence to adulthood (Fig. 2). As expected, T levels increased significantly in this time period (F(2,94) = 160.4, P < 0.01). However, contrary to expectation, significant main effects of time were found reflecting an overall decrease in scores on teacher rated proactive- (F(3,66) = 4.6, P < 0.01) and reactive aggression (F(3,66) = 10.7, P < 0.01). With respect to self-reported delinquency we also found a significant curve linear effect of time (F(2,74) = 46.4, P < 0.01). Mean daily T at ages 13, 16 and 21 years. Fig. 1. Mean daily T at ages 13, 16 and 21 years. Figure options Changes in scores for proactive (- -▴- -) and reactive (- - -▵- - -) aggression ... Fig. 2. Changes in scores for proactive (- -▴- -) and reactive (- - -▵- - -) aggression and delinquent behavior (—▴—) from early adolescence to adulthood. Figure options T levels in aggressive, dominant, and delinquent subgroups Next, separate AN(C)OVAs were conducted to find out whether there was a relationship between T and aggressive, dominant, and/or delinquent behavior at ages 13, 16, and 21 years (see Table 4). Table 4. Differences in T between aggressive, dominant, and delinquent subgroups, assessed at ages 13, 16 and 21 years, using AN(C)OVAs Age 13 (ANCOVA) Age 16 (ANOVA) Age 21 (ANOVA) df F P df F P df F P Independent variables Toughness 1,83 0.59 0.45 1,94 0.06 0.81 1,94 0.04 0.84 Leadership 1,83 3.50 0.07 1,94 0.46 0.50 1,94 0.29 0.59 Proactive aggression 1,80 0.33 0.57 1,89 11.89 0.00 1,89 0.17 0.68 Reactive aggression 1,80 0.01 0.94 1,89 6.83 0.01 1,89 0.01 0.95 Aggression trajectory 2,82 0.11 0.89 2,93 2.16 0.12 2,93 1.16 0.32 Delinquency 1,83 0.09 0.76 1,91 0.34 0.56 1,91 11.11 0.00 Interactions Toughness × leadership 3,81 1.00 0.40 3,92 0.26 0.85 3,92 0.30 0.82 Table options At age 13, T correlated moderately strong with pubertal status (r =.30, P < 0.01), but at age 16, the size of the correlation was much lower (r = 0.12, ns). Pubertal status at age 13 was therefore entered as a covariate in the analyses for that particular year. When examining the differences between high and low aggression, dominance, or delinquency subgroups in T level at age 13, we did not find any significant differences (see Table 4). At age 16, boys with an official crime record were found to have significantly higher T levels compared to boys without such a record (Mann–Whitney U test; Z = −2.14, P = 0.03). Moreover, at age 16, T was higher in high proactive aggressive (HPA) boys than in low proactive aggressive (LPA) boys (Mean ± SD: LPA = 38.3 ± 19.0 pg/ml, HPA = 53.4 ± 22.6 pg/ml, F(1,89) = 11.89, P < 0.01), and higher in the high reactive aggressive (HRA) subgroup than in the low reactive aggressive (LRA) subgroup (Mean ± SD: LPA = 39.5 ± 19.3 pg/ml, HPA = 51.1 ± 23.1 pg/ml, F(1,89) = 6.83, P = 0.01). No T differences were found at age 16 between high and low dominant groups, high and low physically aggressive groups, groups with and without CD, or between self-reported delinquent subgroups (see Table 4). At age 21, high delinquent males (HD, n = 46) had higher T levels than low delinquent males (LD, n = 47; F(1,91) = 11.1, P < 0.01). No meaningful differences in T were found between the other aggressive, dominant, or delinquent subgroups. T trajectories Additional analyses were conducted to examine whether T trajectory groups differed in scores for aggressive, dominant, and/or delinquent behavior. We found that of the 5 boys who had a criminal record no one belonged to the LT group, one belonged to the MT group, and the remaining four boys belonged to the HT group (Pearson chi-square = 6.21, P = 0.05). Finally, a (linear) stepwise multiple regression analysis identified the behavioral assessment that best predicted T at age 21 when controlling for each of the other assessments. Although a high correlation was found between the mean reactive and proactive aggression scores (r = 0.79, P < 0.01) this score was not sufficiently high enough (the criterion being r > 0.90) to treat these variables as essentially the same. Both variables were therefore entered into the analysis. It turned out that the mean Z score on self-reported delinquency between ages 13 and 20 years best predicted T level at age 21, accounting for 5.5% of the variance (with Beta = 0.24, F(1,69) = 4.02, P = 0.05). Once the delinquency score had been entered into the equation, the other delinquency, dominance, or aggression variables did not add significantly to the prediction.