INTRODUCTION

The studies by Jones and her colleagues (e.g., Jones, 1965; Jones and Mussen, 1958) were among the first to examine behavioral effects of puberty not only as the result of biological events, for example, “raging hormones,” but rather as a biological event that occurs within a social context. Notably, the individual’s adjustment during and after puberty is in part associated with the values that friends, family, and society place upon different types and tempos of pubertal development. While much of the early work on these issues focused on personality characteristics, recent studies have demonstrated links between puberty and changes in adjustment during adolescence and beyond, especially for girls’ adjustment (see Graber, 2003 and Susman and Rogol, 2004, for recent reviews). Moreover, the original concept of timing of puberty as discussed by Jones and Mussen has been of increasing interest in the study of girls’ adjustment during adolescence. The goal of the present investigation is to test several hypotheses about the pathways through which puberty influences adjustment in girls during adolescence.

Hormone-Behavior Links

Models delineating the influence of puberty on adjustment identify several potential paths (Brooks-Gunn et al., 1994; Buchanan et al., 1992). The first hypothesis linking puberty to adjustment proposes direct effects of hormonal changes at puberty on increased moodiness, negative affect, and potentially poorer mental health. Of the few studies that have tested this hypothesis, supporting evidence is limited. Specifically, increases in depressive affect have been linked to functional hormonal levels (based on physiological influence on the reproductive system) as indexed by estradiol (Warren and Brooks-Gunn, 1989). Estradiol is well established as an indicator of gonadal maturation. In addition, Angold et al. (1999) found that elevated estradiol and testosterone levels were predictive of depression disorder in girls. Increases in aggressive affect have been linked to testosterone levels in boys and dehydroepiandrosterone (DHEA) or its sulfate (DHEAS) in girls (Olweus et al., 1988; Susman et al., 1987; Warren and Brooks-Gunn, 1989). DHEAS production increases with maturation of the adrenal gland resulting in adrenarche (Parker, 1999). Hence, hormone-behavior links, although limited, have been found for both gonadal and adrenal hormonal systems that change at puberty.

However, hormonal-based changes in mood would be expected to occur for all adolescents as they go through puberty and would not, in and of themselves, account for why some adolescents experience problems whereas others do not. A similar issue can be raised regarding links between adjustment and changes in the outward signs of pubertal development or pubertal status (the level of pubertal development at a point in time). It has been suggested that the observable physical changes of puberty may be experienced negatively within the social context and by the individual. Whereas some studies report limited associations between behavioral problems and pubertal status, such links are not pervasive (see Brooks-Gunn et al., 1994 for a review of this issue). As previous studies find few effects of status on adjustment and as status and estradiol levels were strongly associated in the present investigation, the present investigation focused on hormonal links to adjustment. Moreover, given that direct effects of models are not explanatory for understanding individual differences in adjustment, we examined models of hormone–affect associations that included indirect pathways.

We and others have suggested that hormonal changes at puberty may result in changes in emotional or mood arousal (Brooks-Gunn et al., 1994; Buchanan et al., 1992). Although arousal has been examined in terms of sexual feelings (e.g., Udry, 1988), our hypothesis is that physiological stimulation (i.e., not just sexual arousal) may be experienced by the adolescent as emotional turbulence or moodiness. Hormones that originate from either the hypothalamus–pituitary–gonadal (HPG) or hypothalamus–pituitary–adrenal (HPA) systems may be involved in such effects. In the present investigation, we sought to differentiate the HPA from the HPG effects via examination of DHEAS and estradiol, respectively.

Pubertal Timing and Adjustment

As noted by Jones and Mussen (1958; Jones, 1965), in contrast to levels of development (e.g., hormone levels or pubertal status), the timing of the pubertal transition may be most salient for differentiating who will go on to experience adjustment problems. Pubertal timing refers to going through puberty earlier, at about the same time, or later than one’s peers. In the case of psychological development, how individuals cope with and are influenced by puberty depends upon their emotional and cognitive skills when the transition occurs (Brooks-Gunn et al., 1985); hence, early maturation may result in poorer adjustment, especially for those who experience stressful life events but have not developed the skills needed to cope with such events. Timing effects via social paths or contexts are less related to the individual’s development than on a social referent which places a value on “normative” versus “nonnormative” development (e.g., Brooks-Gunn et al., 1985); in this case, early or late maturation is “deviant” from one’s peers. Evidence has amassed demonstrating that girls who mature earlier than their peers have poorer adjustment in several domains than other girls or boys (see Graber, 2003 for a recent review). In addition, recent studies have found that early-maturing girls have higher lifetime prevalence of major depressive disorder, conduct disorder, eating disorders, and suicide attempts in comparison to on-time and late maturing girls by the time they are in high school (Graber et al., 1997) as well as concurrent depression (Hayward et al., 1997).

As with the literature on hormone-behavior links, the influence of timing on adjustment is hypothesized to be the result of intermediary factors. For example, Caspi and Moffitt (1991) demonstrated that it was not simply early maturation that put girls at risk for behavioral problems in adolescence but the interaction of behavioral problems prior to entry into puberty and early timing. Such findings are supportive of a diathesis-stress model for depression (or adjustment problems more generally). In this model, individual differences in preexisting vulnerabilities are salient to explaining why some adolescents have more difficulty with normative adolescent transitions (see Graber, 2004 and Susman and Rogol, 2004 for recent reviews). Such vulnerabilities develop from genetic and environmental interactions that include atypical neurological functioning as seen in problems with hyperactivity, impulsivity or attention (Loeber et al., 1998). Individuals with such vulnerabilities are more likely to develop adjustment problems when they experience stressful events.

Notably, the experience of stressful life events increases during the early adolescent period. Stressful life events commensurate with changes in pubertal development have been linked to increases in depressive affect (Brooks-Gunn, 1991; Ge et al., 1994). In similar work, 2 groups (Petersen et al., 1991; Simmons and Blyth, 1987) found that going through puberty just before making a school change (a situation most often experienced by early-maturing girls and linked to stressful events) was predictive of having a depressive episode or problems over time. Hence, additional investigation of associations of puberty and stressful events is warranted.

The Current Investigation

Given that much of the existing literature on how puberty affects adjustment has been more supportive of indirect rather than direct pathways of effects, the present investigation specifically tested mediated pathways from puberty to adjustment. In particular, this study examined depressive affect and aggression as these are two of the more commonly studied aspects of adjustment in the puberty literature and both show increases in early adolescence. Pubertal timing, DHEAS, and estradiol levels have each been examined as predictors of these outcomes; these aspects of puberty are examined in the present investigation. In addition, we tested 3 potential mediators of puberty-adjustment pathways: emotional arousal, attention difficulties, and negative life events. Again, each of these has been identified as a predictor of adjustment problems and is potentially linked to puberty. To date, few studies have examined whether pubertal changes are predictive of arousal or attention difficulties in contrast with the literature on puberty and life events. Hence, this investigation sheds light on whether these factors are indicative of preexisting vulnerabilities or vulnerabilities that change with puberty.

METHOD

Participants and Procedure

Participants were 100 white girls who were between the ages of 10 and 14 (M=12.13, SD=.80) and in 5th, 6th, or 7th grade. Girls were from well-educated, middle- to upper-middle class families, and attended private schools in a major northeastern urban area. The sample was recruited from a larger cross-sectional study of girls’ development during adolescence. Of those families with girls who were 10–14 years of age, 67% agreed to participate in the substudy conducted at a hospital laboratory. Girls who participated in the laboratory visit did not differ from those who did not participate on demographic characteristics (e.g., maternal education, maternal employment, and social class), physical development, or adjustment (see Brooks-Gunn and Warren, 1989 and Warren and Brooks-Gunn, 1989 for more details).

At the laboratory visit, girls completed questionnaires about several aspects of their psychosocial functioning. In addition, physical examinations were conducted by a nurse practitioner or a physician to assess level of pubertal development. A blood sample was drawn for hormonal assays at this time. Blood samples were taken in the afternoon for all girls in order to reduce the effects of diurnal variation on assay results. Girls were paid for their participation.

Pubertal Development

As indicated either a physician or nurse practitioner rated each girls’ breast and pubic hair development using the Tanner stages. Tanner ratings are made on a 5-point scale ranging from 1 (no development) to 5 (adult development). Descriptive information on the girls’ pubertal development has been reported previously (Brooks-Gunn and Warren, 1989; Warren and Brooks-Gunn, 1989); pertinent information is repeated here. The percentage of girls in each stage of breast development was 20% in Stage 1, 35% in Stage 2, 28% in Stage 3, 15% in Stage 4, and 2% in Stage 5. The distribution was similar for pubic hair development with 17% in Stage 1, 25% in Stage 2, 25% in Stage 3, 23% in Stage 4, and 10% in Stage 5. All but 19 girls were premenarcheal; of these girls, nearly all had reached menarche within the previous 6 months (18 of 19) and did not have regular menstrual cycles.Footnote 1

Pubertal Timing

Pubertal timing was based on the Tanner ratings. The average of the ratings of breast and pubic hair development were calculated for each girl. Girls were classified as early, on-time, or late developers using norms for Tanner Stages by age from the National Health Examination Survey (Duke et al., 1982) which provides the most comparable national data to this sample (Brooks-Gunn and Warren, 1989). These norms for age were set such that girls in the upper and lower 20th percentiles of development were classified as early or late, respectively, with the middle 60% considered to be on-time in their development. Thirteen of the 100 girls were classified as early maturers based on this method. There was very little variation in Tanner Stage for the 10-year-old girls; most were Stages 1–2 of development. While these girls are not early maturers, it is impossible to tell whether they will be on-time or late maturers. Given that the majority of findings in the literature demonstrate negative effects for early maturing girls, we collapsed the late and on-time groups. Comparisons were made for early versus other girls. This approach allows for the inclusion of the 10-year-old girls in the analyses.

Hormonal Assessments

A blood sample was obtained from each girl in the midafternoon at the time of the laboratory visit. Because prior studies with this sample have found effects for estradiol (scored into categories) and DHEAS levels on depression and aggression, respectively, only these hormones are examined in order to follow up on the underlying pathways for prior effects. DHEAS (kits from Diagnostic Products) and Estradiol (kits from Serono Diagnostics) were measured from serum by RIA. For DHEAS, the cross-reactivity with other steroids was low and the sensitivity of the assay was 0.57 nmol/L. For estradiol, the intra- and interassay coefficients of variation ranged from 4 to 10% and 2.7 to 4.7%, respectively. (Descriptive information on the hormone levels in this sample of girls is available in Brooks-Gunn and Warren, 1989 and Warren and Brooks-Gunn, 1989.)

Estradiol was coded into 4 categories: 0–25, 25–50, 50–75, and greater than 75 pg/mL. These categories tapped global hormonal functioning and the actual effects of estradiol on the reproductive system. In prior studies, girls in hormonal Stages 2 and 3 had the highest levels of depressive affect; these are the girls experiencing the most rapid physiological changes. In contrast, girls in hormonal Stages 1 and 4 had lower depressive affect. Girls in hormonal Stage 1 are prepubertal and girls in hormonal Stage 4 are experiencing a leveling off of hormonal activity and are beginning adult physiological functioning.Footnote 2 Because the association between depressive affect and estradiol category was curvilinear, categories were collapsed to create a dichotomous variable: experiencing rapid change (Stages 2 and 3, n=36) versus not experiencing rapid change (Stages 1 and 4, n=59).

Table I. Descriptive Information and Correlations for Core Measures
Full size table

Mediators

Emotional Arousal

In order to tap emotional arousal, a scale was created from items on the Youth Self-Report (YSR, Achenbach and Edelbrock, 1986). The items were drawn predominantly from the Anxious–Obsessive subscale of the YSR. Construction of this scale was based on the larger sample of girls from the cross-sectional study for whom YSR data were available. A 6-item Arousal scale (α=.69) was created by summing the following items: “I cry a lot,” “I am nervous or tense,” “I am too fearful or anxious,” “I am self-conscious or easily embarrassed,” “My moods or feelings change suddenly,” and “I worry a lot.”

Attention Difficulties

A measure of attention difficulties was also created from the YSR. In this case, pubertal changes may result in arousal that affects concentration processes and therefore disturbs behavior. Again, construction of this scale was based on the larger sample of girls from the cross-sectional study for whom YSR data were available. A 5-item Attention scale (α=.60) was created by summing the following items: “I have trouble concentrating or paying attention,” “I can’t get my mind off certain thoughts,” “I have trouble sitting still,” “I feel confused or in a fog,” and “I daydream a lot.”

Negative Life Events

The total number of negative life events experienced in the past 6 months was assessed using Coddington’s (1972) scale, adapted for young adolescents. The adapted scale covered events in the family, school, and peer domains (Brooks-Gunn, 1991).

Adjustment Outcomes

Depressive affect and aggression were measured via the YSR (Achenbach and Edelbrock, 1986). Girls rated how much each item was true of themselves on a 0 (“not at all true of me”) to 2 (“very true or often true of me”) scale. Items for each scale were summed to form scale scores with high scores reflecting greater behavioral problems. The Depressive Withdrawal scale is an 11-item subscale (α=.65) that included items such as “I am unhappy, sad, or depressed,” and “I like to be alone.” The Aggression scale of the YSR is a 21-item subscale (α=.81) that included items such as “I argue a lot,” “I get in many fights,” and “I have a hot temper.”

Analysis Plan

The goal of this study was to examine mediated pathways from either hormonal levels or pubertal timing to depressive affect or aggression. The mediators were emotional arousal, attention difficulties, and negative life events. Mediation was tested using the guidelines delineated by Baron and Kenny (1986). Specifically, separate regression models tested the effect of the pubertal predictor on each mediator and the outcome variable. A full model was then tested with the pubertal predictor and all the mediators entered on the same step. For mediation to occur, the pubertal predictor must have a significant association with a mediator and the outcome variable, and, in the full model, the mediator must predict the outcome variable and the size of the association between the pubertal predictor and the outcome variable must be reduced. In all models, age was controlled for on a first step; this step is not shown in the figures.

RESULTS

Descriptive Analyses

Correlations among the variables and descriptive information (i.e., M and SD) are shown in Table I. Associations among pubertal indicators were weak. Estradiol category and timing were not associated as would be expected given that these measures were defined to tap different aspects of puberty (experience of rapid hormonal change versus timing relative to age, respectively). DHEAS demonstrated a weak association with each of these measures. Also, pubertal timing was not associated with age, in contrast to estradiol category and DHEAS which demonstrated positive, moderate associations with age.

Pubertal indicators demonstrated some associations with the outcomes of interest. Timing and estradiol category demonstrated weak associations (r=.27, p < .01) with depressive affect, but no association with aggression. In contrast, DHEAS was not associated with depressive affect, but showed a weak negative association with aggression. In addition, pubertal indicators demonstrated specific associations with the potential mediators. For example, estradiol category was moderately associated with negative life events such that being in the rapid change category was associated with a higher frequency of negative life events. DHEAS demonstrated a weak, negative association with attention difficulties that was only marginally significant. Finally, timing showed a weak association with emotional arousal.

The potential mediators of hormone or timing links to adjustment demonstrated consistent associations with the outcomes of interest (see Table I). Emotional arousal was positively and moderately correlated with depressive affect and aggression as well as with attention difficulties. Attention difficulties were also positively associated with depressive affect and aggression. Negative life events were positively associated with aggression.

For the 2 categorical predictor variables, ANCOVAs (covarying for age) were also conducted as the F-values, means, and standard deviations for the effects may be more illuminating than correlations. As expected estradiol category was associated with depressive affect [F(1, 95)=7.11, p < .01] with girls with rapidly changing estradiol experiencing higher levels of depressive affect than other girls (M=7.5, SD=3.28 and M=5.80, SD=2.93, respectively). A similar effect was found for negative life events F(1, 95)=9.34, p < .01) with girls in this group reporting a greater frequency of negative life events than other girls (M=2.29, SD=1.77 and M=1.31, SD=1.35, respectively).

For pubertal timing, effects were found for depressive affect and emotional arousal [F(1, 95)=7.09, p < .01, and F(1, 95)=5.23, p < .05, respectively). For depressive affect, early maturing girls had higher scores than other girls (M=8.67, SD=3.02 and M=6.14, SD=3.05, respectively), and higher emotional arousal (M=5.42, SD=2.23 and M=3.74, SD=2.39, respectively). Timing also demonstrated a trend for association with DHEAS (p < .10).

Tests of Mediation

Figure 1 shows the path models predicting depressive affect. Figure 1(a) shows the model testing mediated pathways of estradiol category to depressive affect; Fig. 1(b) shows the model from hormonal arousal (DHEAS) to depressive affect; and Fig. 1(c) shows the mediated model from early timing to depressive affect. The β coefficients are shown in the figures. Specifically, on the left side of the models, the β values were calculated separately for each pathway for the predictor to depressive affect and to each potential mediator. The β coefficients on the right side of the models (shown in bold type) were calculated with all variables (predictors and mediators) entered simultaneously into the model. Hence, there are 2 β coefficients for the pubertal predictor to the outcome.

Fig. 1.
figure 1

Path models for each pubertal predictor to depressive affect. On the left side of the models, the β values were calculated separately for each pathway for the predictor to depressive affect and to each potential mediator. The β coefficients on the right side of the models (shown in bold type) were calculated with all variables (predictors and mediators) entered simultaneously into the model; there are 2 β coefficients for the pubertal predictor to the outcome.

Full size image

As can be seen in Fig. 1(a), estradiol category was associated with depressive affect and one of the mediators, negative life events. However, life events were not predictive of depressive affect and no mediation is demonstrated in this model. DHEAS had no association with depressive affect; the model is included in Fig. 1(b) for comparison. In contrast, the model for pathways from pubertal timing to depression shown in Fig. 1(c) demonstrated mediation. There is evidence in this model for mediation of the effect of timing on depressive affect via emotional arousal. As per specifications of Baron and Kenny (1986), the coefficient of early timing predicting depressive affect is reduced and is no longer significant when the mediators are included in the model (i.e., β is reduced from 0.26, p < .05, to 0.14, NS). In addition, early timing predicted higher emotional arousal which subsequently predicted increased depressive affect. Only the emotional arousal pathway demonstrated mediation.

Figure 2 shows the path models for each pubertal predictor to aggression. Interestingly, in contrast with the model for depressive affect, the model for pathways from estradiol category to aggression (shown in Fig. 2(a)) demonstrates mediation. There is evidence in this model for mediation of the effect of estradiol category on aggression via negative life events (i.e., β for estradiol category to aggression is reduced from 0.22, p < .05, to 0.08, NS, and pathway via negative life events is significant). In this model, rapidly changing estradiol levels were associated with reporting more negative life events which was predictive of increased aggression. A virtually identical model for mediated effects can be seen in Fig. 2(b) for DHEAS (i.e., β is reduced from −0.20, p < .10, to 0.07, NS). Even though the direct association of DHEAS with aggression was only marginally significant, the simple correlation was significant and it does appear that a trend for mediation occurs; the size of the coefficients suggests that this is not a strong initial association. The main difference between the models for estradiol category and DHEAS was that DHEAS demonstrated a negative association with negative life events and aggression. In addition, both the estradiol category and DHEAS models suggest that attention difficulties may also be a mediator of the association with aggression; with more power such an effect may be significant. Finally, pubertal timing had no association with aggression; the model is included in Fig. 2(c) for comparison.

Fig. 2.
figure 2

Path models for each pubertal predictor to aggression. On the left side of the models, the β values were calculated separately for each pathway for the predictor to aggression and to each potential mediator. The β coefficients on the right side of the models (shown in bold type) were calculated with all variables (predictors and mediators) entered simultaneously into the model; there are 2 β coefficients for the pubertal predictor to the outcome.

Full size image

Follow-up Analyses

The analysis of mediated pathways demonstrated some hypothesized findings. For depression, the effect of early timing on depression was mediated by emotional arousal. For aggression, it was expected that DHEAS would be associated most likely via an emotion or attention pathway. Rather both DHEAS and estradiol category demonstrated links to depression with more evidence of a pathway for negative life events. Thus, we conducted additional analyses to test for moderating effects rather than mediation to help explain some of the findings. The first set focused on whether or not pubertal timing interacted with hormonal activity. In this sense, were the hormonal and timing effects independent or were there hormonal associations with timing? Perhaps some early maturers have higher “hormonal” arousal (i.e., DHEAS) which leads to emotional arousal, depressive affect, or aggression.

For these analyses, the interaction of DHEAS and pubertal timing was examined. The upper third of the distribution of DHEAS was used as a cut-off for high “hormonal” arousal. As expected given the prior analyses, there was a main effect for timing, but not for DHEAS. Interestingly, the interaction of DHEAS and timing was a significant predictor of depressive affect [F(1, 95)=4.96, p < .05]. Follow-up tests indicated that girls with high DHEAS and early maturation had significantly higher depressive affect than other girls (M=10.00, SD=2.94). Other girls had similar levels of depressive affect: low DHEAS arousal girls who were not early maturers had depressive affect levels of 6.46 (SD=3.26); low DHEAS arousal girls who were also early maturers had mean levels of 6.80 (SD=2.77); and girls with high DHEAS arousal who were not early maturers had mean levels of 5.40, (SD=2.42). In addition, a trend for an interaction effect of DHEAS arousal and timing was found for emotional arousal (p < .07). Means for emotional arousal followed the same pattern as for depressive affect; that is, high DHEAS girls who were early maturers had the highest reports of emotional arousal (M=6.4) and all other girls had similar reports of emotional arousal (M ranging from 3.5 to 3.9). In a similar analysis, the interaction of estradiol category with timing on depression was not significant.

Finally, in order to explore the aggression pathways more fully, the same analyses with hormonal arousal (high vs. low-DHEAS groups) and the interaction with timing were conducted. No effects were found. Given that negative life events were significant in the mediation model, its role as a moderator was tested. However, inclusion of the interaction of estradiol category and negative life events was not a significant predictor of aggression.

DISCUSSION

As indicated, our goal was to move beyond the demonstration of main effects and try to understand some of the pathways that underlie potential effects of puberty on adjustment. Indications that adrenal activity, early pubertal timing, and rapid pubertal hormone increases may have separate pathways to depressive affect or aggression are intriguing and highlight the potential for more intensive investigation of gonadal and adrenal processes in combination with psychosocial development.

For depressive affect, we had hypothesized that estradiol category, tapping rapid change in hormones, would lead to emotional arousal which would lead to depression. However, there was no support for this pathway in our analyses. At the same time, such a pathway existed for early maturation and depressive affect. Why do early maturing girls have higher emotional arousal scores, which mediated the maturation–depression link? The subsequent analyses that demonstrated an interaction between DHEAS and timing are relevant to this question. Not all early maturing girls experienced high emotional arousal and depression. Those girls with high levels of DHEAS and early maturation were those who had the highest emotional arousal and depressive affect scores. Notably, while the early maturers did in fact have higher estradiol levels than other girls, DHEAS did not differ substantially by timing. Moreover, high DHEAS, alone, did not predict emotional arousal or depressive affect. Thus, the interaction of DHEAS and timing suggests several possibilities. DHEAS may be an indicator of a preexisting physiological disposition that is exacerbated or activated by the social and psychological demands placed on early maturing girls. In contrast, some early maturing girls who are experiencing difficulties coping with their maturation may have elevated DHEAS as a product of this stress. Because the present analyses were cross-sectional, the difference between these two possibilities cannot be determined.

These findings are consistent with those of Goodyer and his colleagues (Goodyer et al., 2000), who have found that high levels of DHEAS were predictive of onset of depressive disorder over a 1-year period. It should be noted that Goodyer’s sample was postpubertal and timing effects were not examined in their study. From their analyses, Goodyer and his colleagues felt that the effects of DHEAS were more distal to depression whereas factors such as recent stressful life events or negative mood were more proximal predictors of depression. Such an interpretation would fit with our own data regarding emotional arousal. Future research would need to start with younger girls and follow them longitudinally in order to determine how these associations are established. Inclusion of other adrenal steroids such as cortisol in the study designs would also help elucidate how physiological arousal or reactivity interacts with pubertal development.

Furthermore, because of the cross-sectional approach, this investigation presents only one direction of effects. Bidirectional associations cannot be tested. Recent studies have found that social and psychological factors such as lower quality of family relationships and father absence predict earlier maturation in girls (e.g., Ellis, 2004; Graber et al., 1995). As we have suggested previously (e.g., Graber, 2003), factors that predict earlier maturation may also predict subsequent psychopathology. Unfortunately, most of these studies have not collected hormonal data. Thus, hormonal mechanisms that may predict timing and may subsequently lead to different emotional arousal and subsequent depression are needed to delineate these pathways more clearly (see Ellis, 2004, for a recent review).

Findings for aggression are perhaps a bit more difficult to explain. As indicated, pubertal timing was not associated with aggression in these analyses despite expectations based on the literature linking early maturation and conduct disorder in girls (e.g., Graber et al., 1997). Of course, in the present investigation, delinquency or conduct disorder, per se, were not examined.Footnote 3 Caspi and his colleagues (Caspi et al., 1993) examined the effect of pubertal timing on externalizing behaviors in young adolescents who attended mixed-gender and single-gender schools in New Zealand. Early-maturing girls in mixed-gender schools had higher levels of behavior problems than other girls; however, early-maturing girls in single-sex schools did not have elevated behavior problems. Our sample was drawn predominantly from private girls’ schools; although comparisons of single versus mixed gender contexts were not made for this study, the absence of an effect of timing on aggression is consistent with the interaction of timing and school context demonstrated previously by Caspi and colleagues (Caspi et al., 1993).

Aggression was associated with hormonal measures and some mediated pathways were identified. Mediated associations with aggression via negative life events, and possibly via attention difficulties, were found for both estradiol category and DHEAS. As indicated, negative life events were included in the models because of the existing literature linking puberty and life events. Hence, the partial mediation of estradiol category associations with aggression via negative life events is consistent with prior studies that found that girls experience more negative events as they go through puberty and that experiencing more events during this time is linked to depression (e.g., Brooks-Gunn, 1991; Ge et al., 1994). Hypotheses that DHEAS would predict negative life events have less support than hypotheses of moderating effects; that is, that DHEAS would interact with negative life events to predict aggression. However, moderated pathways were not significant for DHEAS and life events vis-à-vis aggression. Again, interpretation of the findings may be limited by the cross-sectional analysis of the data. Negative life events may be altering DHEAS along with influencing aggression. In contrast, the nonsignificant pathway for attention difficulties would have been more consistent with the literature on predictors of aggression in boys (Loeber and Keenan, 1994). Clearly, future studies need to examine similar models for girls and boys in order to understand where pathways are consistent versus unique by gender.

One other finding merits some discussion. Specifically, DHEAS moderated the effect of timing on depressive affect such that high DHEAS interacted with early timing to predict higher depressive affect. High DHEAS levels have been found in depressed adults, although findings with adolescents have been more mixed (see Wolf and Kirschbaum, 1999, for a recent review). In contrast, in the present investigation, DHEAS demonstrated negative associations with aggression and attention difficulties. A negative correlation between DHEAS and aggression has been found by others (e.g., Susman et al., 1987). The different directions of the associations with DHEAS and depressive affect and aggression may be indicative of curvilinear effects. McBurnett and colleagues (McBurnett et al., 2000) find higher aggression in boys with persistently low cortisol over 2 assessments. Puberty was not examined in their study and findings are based on a small sample of boys. While cortisol and DHEAS are both adrenal steroids, their interconnection is not fully understood. However, the findings for low-HPA activity and the present findings for DHEAS are consistent. Also, as indicated, the present investigation was consistent with findings of Goodyer and colleagues (Goodyer et al., 2000) linking high DHEAS and depression. It is certainly possible that physiological effects of DHEAS differ at the high versus low ends of the distribution. It is also possible that contextual and pubertal factors are important in understanding these alternate pathways.

This investigation has several limitations, many of which have already been identified and discussed throughout the paper (e.g., cross-sectional design, sample characteristics). An additional limitation of the study is that we did not have a specific measure of emotional arousal or attention difficulties, and hence, we created scales from the YSR that we felt tapped these constructs. We have a limited ability to examine the validity of these scales. It should be noted that the scale constructed to tap emotional arousal demonstrated a strong association with the depression scale. However, our constructed scales were not uniformly correlated with other scales from the YSR and demonstrated unique associations with puberty and the adjustment outcomes. More comprehensive assessments of arousal and attention are needed.

Our goal was to move beyond demonstrating effects to identifying potential pathways. In some aspects, we have been successful in this endeavor. In particular, given the recent convergence of findings in the literature that early maturation is associated with serious psychopathology in girls, identifying factors that place these girls at risk is particularly important. Physiological vulnerabilities as tapped by high DHEAS or emotion regulation vulnerabilities were identified in the present investigation as particularly salient to explaining links between timing and adjustment in girls. At the same time, these and our other findings serve to illuminate additional potential pathways among these constructs that merit fuller examination. The task of including multiple hormonal pathways, multiple contextual factors, multiple outcomes, and both genders will probably be achieved across several investigations that examine pieces of the puzzle rather than a single definitive study.