Day-to-day friends’ victimization, aggression perpetration, and morning cortisol activity in late adolescents

Abstract

This study investigates bi-directional associations between adolescents’ daily experiences of victimization and aggression perpetration within friendships. We investigated a) across-day associations between victimization and aggression perpetration; b) morning cortisol activity as a moderator of cross-day victimization and aggression links; and c) potential sex differences in these patterns. For four consecutive days, 99 adolescents (M_age= 18.06, SD= 1.09, 46 females) reported whether they were victimized by or aggressive toward their friends. On three of these days, adolescents provided three morning saliva samples. Multilevel path analyses showed that across days, victimization and aggression were bi-directionally linked, but only for male adolescents. Additionally, for male adolescents, morning cortisol output (but not morning cortisol increase) moderated the association between victimization and next-day aggression; victimization predicted greater next-day aggression for boys with low, but not high, morning cortisol output. Findings implicate a physiological factor that may modify daily links between victimization and aggression in male adolescent friendships.

1. Introduction

Friendships are typically viewed as positive relationship contexts for adolescents. However, aggression can nevertheless occur between friends. In fact, studies among youth show that between 5% to 30% of youth report physical or relational victimization by close friends (Crick & Nelson, 2002; Leff, Waasdorp, & Paskewich, 2016; McDonald, Wang, Menzer, Rubin, & Booth-LaForce, 2011; Mishna, Wiener, & Pepler, 2008). One problematic outcome associated with peer or friend victimization (i.e., verbal, or physical aggression adolescents experience from a friend) is the potential escalation of aggression perpetration (i.e., aggression of the target adolescent toward friends) (Walters & Espelage, 2018). This, in turn, could increase adolescents’ risk for further victimization (Lam, Law, Chan, Zhang, & Wong, 2018).

Studies typically assess victimization-aggression links cross-sectionally or over the course of multiple months or years. However, it is unknown whether cycles of aggression play out from one day to the next. That is, does episodic (e.g., day-to-day) exposure to victimization prompt adolescents’ aggressive reactions? And, vice versa, does behaving aggressively one day render youth immediately more vulnerable to subsequent victimization? Although daily studies of peer victimization demonstrate that even single victimization incidents can increase adolescents’ emotional distress and school problems (Espinoza, Gonzales, & Fuligni, 2013; Lehman & Repetti, 2007; Nishina, 2012; Nishina & Juvonen, 2005), they have not explored whether victimization episodes are also accompanied by aggression perpetration. Understanding whether victimization and aggression are closely linked within adolescents’ daily lives has important intervention implications for identifying specific social conditions under which youth are at increased risk of being targeted or aggressing against close others.

In this study, we also consider whether daily changes in cortisol levels might amplify or dampen links between victimization and next-day aggression. Using daily data methodology, we adopted a within-person approach. Specifically, we tested two morning cortisol measures as plausible time-varying moderators: the cortisol awakening response (CAR) and the total morning cortisol output (AUCG). Both measures have been found to be sensitive to prior-day social stressors (Doane & Adam, 2010) and play a role in mobilizing bodily resources to handle upcoming daily demands (Adam, Hawkley, Kudielka, & Cacioppo, 2006; Fries, Dettenborn, & Kirschbaum, 2009). Methodologically, the daily diary approach captures within-person variability, minimizes recall bias, enhances ecological validity, and allows for inferences about the direction of effects (Bolger, Davis, & Rafaeli, 2003).

We focused the study on late adolescents, an understudied group with regard to friends’ aggression (Leff et al., 2016). However, during this time, aggression between friends still occurs (Bennett, Guran, Ramos, & Margolin, 2011). Aggression within late adolescents’ friendships may be especially disruptive, as reliance on friends for support increases while emotional dependence on parents decreases (Bagwell & Schmidt, 2011; Rubin, Wojslawowicz, Rose-Krasnor, Booth-LaForce, & Burgess, 2006).

1.1. Reinforcing links between friends’ victimization and aggression perpetration

According to the peer-socialization theory (Rose & Rudolph, 2006) victimized youth may learn and subsequently model aggressive behaviors toward peers. For example, cumulative peer victimization from childhood to early adulthood predicted global levels of aggression perpetration in adulthood (Logan-Greene, Nurius, Hooven, & Thompson, 2015). These links are especially important to understand given that youth who both are victimized and aggress against others often fare worse compared to nonaggressive victims (e.g., Juvonen, Graham, & Schuster, 2003; Leadbeater, 2006).

In turn, the sequential social process theory asserts that peer aggression increases youth’s risk for social rejection and subsequent victimization (Boivin & Hymel, 1997). Indeed, studies have documented associations between peer victimization and aggression perpetration both concurrently and across longer intervals (e.g., 6 months; one year) during early and mid- adolescence (Lam et al., 2018).

Some sex differences have been found in regard to aggression within the broader peer group and within friendships (Bennet et al., 2011) . Male, compared to female, adolescents are often found to report higher total levels of peer victimization and aggression perpetration (Card, Stucky, Sawalani, & Little, 2008; Cooley & Fite, 2016). In addition, some studies find stronger connections between peer victimization and aggression perpetration for male compared to female adolescents (Leadbeater, Boone, Sangster, & Mathieson, 2006).

In this study, we examine the victimization-aggression perpetration link in adolescents’ everyday lives. Specifically, we examined whether being victimized by friends on one day would increase adolescents’ aggression perpetration toward friends the following day, either in retaliation or as a way to displace anger. In turn, we also tested whether daily increases in aggression perpetration render adolescents more vulnerable to victimization by their friends.

1.2. Daily morning cortisol, peer victimization and aggression perpetration

To understand day-to-day links between victimization and aggression, it may be valuable to tune into adolescents’ daily physiological reactivity as a potential moderator of these links. Indeed, neurobiological models of aggression demonstrate how deficits in physiological regulation can contribute to aggressive behavior, particularly in the face of stress (see Alink et al., 2008; Murray‐Close, 2013a). These accounts conceptualize physiological activity as a stable, trait-like characteristic, often highlighting physiological under-activation as a hallmark of aggression perpetration (e.g., (Oberle et al., 2017). That is, aggression offers a means of up-regulating blunted physiological activity (Van Goozen, Fairchild, Snoek, & Harold, 2007). Physiological under-activation has also been documented as a biomarker of fearlessness and lack of inhibition (Raine, 2002). Alternatively, other researchers suggest that physiological over-activation might reflect high sensitivity to context or regulatory difficulties which can predispose youth to aggressive reactions to peer victimization (e.g., (Pitula, Murray‐Close, Banny, & Crick, 2015).

Cortisol, the end product of the hypothalamic-pituitary-adrenal HPA system, is particularly relevant to explore as a moderator of victimization-aggression perpetration links given its high sensitivity to social stressors. Indeed, a handful of recent studies tested the role of cortisol in moderating the between-individual associations between peer victimization and aggression (Kliewer, Dibble, Goodman, & Sullivan, 2012; Rudolph, Troop-Gordon, & Granger, 2010; Vaillancourt, Brittain, Haltigan, Ostrov, & Muir, 2018). For example, in a recent study among preschoolers, peer victimization was associated with higher aggression perpetration among boys who exhibited lower average of morning and afternoon cortisol levels across two days, but with lower aggression perpetration among boys with higher cortisol levels (Vaillancourt et al., 2018). Results for girls were non-significant. In other words, high levels of cortisol buffered victimization-aggression links for boys. However, in other work among pre-adolescents, heightened (but not blunted) pre-task cortisol levels predicted positive links between peer victimization and aggression perpetration (Rudolph et al., 2010). This divergent effect of cortisol might be due to several factors, especially the context of the cortisol assessment (i.e., reactivity versus basal levels), as has been previously suggested (Troop-Gordon & Erath, 2018).

Importantly, this past research has exclusively focused on between-person differences in cortisol (or in other physiological measures) in relation to the victimization-aggression links. However, cortisol indices show significant within-person variability across days (Rotenberg, McGrath, Roy-Gagnon, & Tu, 2012; Shirtcliff, Granger, Booth, & Johnson, 2005). It has been previously suggested that this within-person variability may influence adolescents’ likelihood to react aggressively to incidents of peer victimization (Murray‐Close, 2013b). However, to our knowledge, no prior studies have tested cortisol as it pertains specifically to victimization-aggression links in daily life.

Addressing this gap, this study explores how intraindividual (day-to-day) changes in cortisol levels may moderate the across-day link between incidents of victimization and subsequent aggression perpetration. We focused on two measures of the morning cortisol activity. The first is the cortisol awakening response (CAR), which refers to the typical morning increase in cortisol levels that peaks approximately 30 minutes after awakening (Law, Hucklebridge, Thorn, Evans, & Clow, 2013). The second is the total morning cortisol output (AUCG) which includes both the cortisol awakening levels and the cortisol increase (Pruessner, Kirschbaum, Meinlschmid, & Hellhammer, 2003). According to the anticipation hypothesis and recent supporting data (Bäumler et al., 2014), the CAR reflects a reactivation of memories from the previous day that prompts physiological preparedness to meet anticipated demands of the upcoming day (Fries et al., 2009). Some daily diary studies suggest lagged associations between social stressors and next-morning CAR (Doane & Adam, 2010; Sladek & Doane, 2015). For example, daily feelings of loneliness among high school students predicted a stronger next-day post-awakening cortisol increase (Doane & Adam, 2010). Importantly, this lagged effect was found for total morning cortisol production (AUCG) (Elder, Barclay, Wetherell, & Ellis, 2018; Zoccola, Dickerson, & Yim, 2011). Therefore, it is important to use the two measures of the cortisol morning activity, the CAR and the total morning cortisol output (AUCG) (Boggero, Hostinar, Haak, Murphy, & Segerstrom, 2017). This is especially relevant given recent findings, which suggest that cortisol at awakening is also sensitive to prior day stressors. For example, in a recent study, early adolescents secreted more cortisol at awakening following days they encountered peer problems (Bai, Robles, Reynolds, & Repetti, 2017).

To our knowledge, no prior study has examined how within-adolescent changes in cortisol might modify day-to-day associations between victimization and aggression perpetration. However, an increasing number of studies support the notion that both heightened CAR and AUCG may actually promote coping with upcoming daily tasks (Adam et al., 2006; Hoyt, Zeiders, Ehrlich, & Adam, 2016). For example, higher total morning cortisol output predicted more stamina later that day among older adults (Adam et al., 2006). Pioneering laboratory studies, albeit also among adults, suggest that beyond a surge of energy, a heightened cortisol morning increase also predicts improved cognitive flexibility (Law, Evans, Thorn, Hucklebridge, & Clow, 2015). Particularly relevant to the prediction of aggression is a prior finding that stronger cortisol morning increases predicted better inhibition of unwanted responses in a standardized laboratory task later that day (Shi et al., 2018). Elsewhere, higher mean cortisol increase level across 3 days was linked to the use of adaptive problem-solving strategies among young adults (Gilbert, Mineka, Zinbarg, Craske, & Adam, 2017). In addition, a heightened CAR might have a favorable effect on mood later that day. Though it was not found directly with the CAR, hourly upshots of cortisol have been found to predict better mood the next hour (Hoyt et al., 2016). Together, these studies provide evidence that a heightened CAR may serve a regulatory function in daily life, particularly in the face of social stressors.

1.2.1. Sex as a moderator

Sex might further moderate the effect of cortisol on daily victimization-aggression links. Male compared to female adolescents have often been found to show lower morning cortisol increase and total output (Hollanders, van der Voorn, Rotteveel, & Finken, 2017). In addition, associations between low morning cortisol and higher total aggression have been sometimes found for male adolescents only (Dietrich et al., 2013; Shirtcliff et al., 2005; Sondeijker et al., 2008). In addition, low cortisol levels have been found to potentiate the victimization-aggression link among preschool boys but not girls, as we mentioned above (Vaillancourt et al., 2018).

1.3. The present study

Our two main goals were first to test bi-directional across-day links between victimization by friends and aggression perpetration against friends and, second, to explore whether daily measures of the CAR would modify the across-day link between victimization on one day and aggression perpetration on the next in adolescents’ friendships. Although we expected that links between victimization and aggression would be bidirectional, we also hypothesized that a stronger CAR following victimization would buffer adolescents from next-day aggression perpetration (i.e., CAR as facilitator of self-regulation; Shi et al., 2018).

Next, we hypothesized that morning cortisol measures (i.e., CAR and AUCG) would moderate the link between victimization and next-day aggression perpetration (HO2): Specifically, we hypothesized that stronger morning cortisol activity would buffer the effect of victimization on next-day aggression perpetration among friends. To fully capture daily aggressive incidents, we probed for a wide range of behaviors, including verbal (e.g., insults) and physical (e.g., push) behaviors.

Finally, we tested sex as a moderator of each model given the previously found sex differences in mean levels of the daily constructs and the links among them, which we described above (e.g., Dietrich et al., 2013; Leadbeater et al., 2006; Vaillancourt et al., 2018). Nonetheless, there is currently not enough in the literature to base specific hypotheses on how sex might moderate the daily links.

2. Methods

2.1. Overview

The present study uses data from the fifth wave of a longitudinal study of adolescent adjustment (Author/s, masked). Eligibility for the larger study required that both parents were living with the child for 3 or more years at study entry, all three family members were able to complete the procedures in English, and the child was either age 9 or 10 (Cohort 1) or in middle school for families joining the project at a later wave (Cohort 2). All study procedures were approved by the university IRB. We obtained parental consent and adolescent consent or assent, depending upon the youth’s age.

2.2. Participants

Adolescents who completed daily diary data and provided saliva samples for diurnal cortisol assays were included in the current study. Of the 131 adolescents who participated in some portion of Wave 5 procedures, 32 did not complete saliva procedures, resulting in a sample of 99 adolescents (46 female), who were 18.06 years old on average (SD = 1.09, min= 14.92, max= 21.05, with the majority of participants (95%) younger than 20 years of age). Reflecting the metropolitan community from which they were drawn, 32.3% of adolescents self-reported their ethnicity as Hispanic/Latino. Self-identified race was 5.1% Asian/Pacific Islander, 20.1% Black/African American, 37.4% Caucasian, 29.3% multi-racial, and 8.1% other. Median household income was $80,000 (M = $91,828, SD = $58,884); 18.0% of families reported incomes below $40,000; 33.7% reported $40,001-$80,000; 24.7% reported $80,001-$120,000; and 23.6% reported > $120,000.

We tested for selective attrition by comparing the 32 participants who were eligible to provide saliva but did not do so to the 99 who did participate. No significant differences emerged with respect to adolescents’ race/ethnicity, annual family income, parental level of education, or adolescents’ mean levels of peer aggression in the past year.

2.3. Procedures

During the in-lab visit, adolescents completed measures of past-year victimization and aggression within friendships. At the end of laboratory procedures, participants completed day 1 of the home data collection by reporting on the preceding day via a daily Qualtrics-based online survey. Participants were then instructed to fill out one survey at the end of each of 9 additional consecutive days. Out of the 10 days of data collection, participants chose 3 consecutive weekdays to collect saliva samples, allowing for at least one day of questionnaires to precede the first day of saliva collection. On each of the 3 days of saliva collection, participants provided 5 saliva samples to measure diurnal cortisol patterns. To capture the CAR, we used 3 saliva samples that were collected (a) upon awakening; (b) 20 minutes after awakening; and (c) 40 minutes after awakening.

We based our choice for the CAR timing on the expert recommendations at the time of data collection (Chida & Steptoe, 2009). Alarms were set as reminders; the first alarm was before 8 am. Saliva samples were collected using oral swabs that were frozen immediately after use. Participants were asked not to consume caffeine or alcohol for the 24 hours before saliva sampling and not to eat, drink, exercise, or brush their teeth 30 minutes before providing samples. In the lab, the samples were frozen at −20 degrees Celsius and then shipped on dry ice for assay at Salimetrics, LLC (State College, PA). The last sample of each day was also assayed for cotinine, a byproduct of nicotine (Granger et al. 2007).

2.4. Compliance

Compliance with the daily surveys and cortisol sampling was relatively high; we received 849 of the 990 possible daily reports (86%). Of the 891 cortisol samples that we aimed to collect (9 for each of 99 participants), 30 samples could not be assayed due to insufficient saliva quantity or missed sample collection (nine awakening samples, eight second samples, and 13 third samples). We dropped 3 additional morning samples which were more than 3 SD above the cortisol sample mean level for that time of day. We statistically accounted for the 33 missing samples (3.7%) in the analyses using full information maximum likelihood estimation (FIML; Schafer & Graham, 2002). Time of saliva collection was reported through a brief questionnaire, which participants completed during the 2-minute saliva sample. To further verify compliance with the timing of saliva collection, we used Medication Event Monitoring System (MEMS) caps that recorded when the saliva collection vial was opened. This additional procedure, completed with a sub-sample of 25 participants, showed relatively high correspondence between MEMs cap-recorded sampling times and participants’ reported times (mean difference = 5.15 minutes, median difference = 1.93 for the morning samples and mean difference = 11.33 minutes, median difference = 1.11 for the last saliva sample).

2.5. Measures

2.5.1. Cortisol indices

We used the three morning cortisol samples to calculate the CAR as the area under the curve with respect to cortisol increase above the awakening levels (CAR AUCI) and the total cortisol morning output as the area under curve with respect to ground (AUCG) (Pruessner et al., 2003). Morning cortisol indices were normally distributed and did not require log-transformation.

2.5.2. Daily items

Daily friend victimization.

The daily survey included 16 items, adapted from the How Friends Treat Each Other scale (HFTEO, Bennett et al. 2011), assessing whether adolescents were victimized by their friends. Items assessed verbal (e.g., “Today a friend swore or cursed at me”), physical (e.g., “Today a friend slapped me or pulled my hair”), relational (e.g., “Today a friend talked about me behind my back”) and electronic (e.g., “Today a friend sent a mean or hurtful text message to me”) aggression on a 3-point scale ranging from 0 (not at all) to 2 (a lot). We averaged scores across all items. Internal consistency of the measure, accounting for nested structure of the data (Shrout & Lane, 2012) was .78.

Daily friend aggression.

To assess aggression perpetration we used 16 items, analogous to those used for daily aggression from friends (e.g., “Today I talked about a friend behind his/her back” (see above). Internal consistency of the measure, accounting for nested structure of the data was .73.

2.5.3. Covariates

Based on recent guidelines (Stalder et al., 2016), we included the first cortisol sample taken upon awakening in models involving the CAR. Age was tested as a covariate due to putative associations with the CAR (Romeo, 2013) and daily aggression (e.g., Rose & Rudolph, 2006). In addition, we tested the possible effect of cotinine levels, hours of sleep, time of morning awakening, and the use of medications, which may potentially influence daily cortisol patterns. Including these covariates in the models did not change the significance, magnitude, or direction of the findings and thus there were subsequently dropped.

2.6. Data analysis approach

We conducted multilevel path analyses with daily observations nested within participants, estimating random intercepts and fixed slopes. To examine effects across time, we used the lagging approach described by Bolger and colleagues (2003). For the moderation analysis, we created two level 1 product terms (i.e., victimization × CAR; victimization × AUCG). Within-person predictors (and moderators) were group-mean centered consistent with recommendations by Enders and Tofighi (2007). Missing data were statistically accounted for using full information maximum likelihood estimation (FIML; Schafer & Graham, 2002) in Mplus Version 7 (Muthén & Muthén, 2012). To help account for the non-normality of aggression perpetration, we used robust standard errors. Multiple group analyses were used to test for moderating effects of sex (Yuan & Chan, 2016): We individually constrained paths in each model to be invariant between sexes. The one degree of freedom nested chi-square difference test was used to determine if the relevant parameters were significantly different between groups. First, all parameters were free; then we equated each of the paths across sexes. If a significant increase in fit was found, that model was then selected and subsequent constraints were made using that model. When the difference in model fit was not significant, we present the more parsimonious model.

3. Results

3.1. Descriptive statistics and preliminary analyses

Over the 4 days of data collection included in the study (3 days of self-reported victimization and aggression, which overlapped with the cortisol assessment and the prior day), 52 adolescents (50.0 % of female participants; 58.5% of the male participants) reported at least one incident of friends’ aggression (i.e., being victimized by friends) and 51 reported at least one incident of aggression perpetration towards friends (36.9 % of the females; 56.6% of the males). Incidents of victimization by friends were reported on 107 (27.4%) of total days and aggression perpetration toward friends was reported on 139 (35.4%) of the days. Male, compared to female, adolescents reported significantly more daily incidents of victimization, and more incidents of aggression perpetration [t (97) = −2.12, p = .04, t (97) = −2.08, p = .04; respectively]. In addition, intraclass correlations were .26 for the CAR, .57 for the AUCG, .56 for friends’ victimization and .57 for aggression perpetration. These ICCs suggest meaningful within-person variability in our main constructs of interest and highlight the importance of modeling their daily associations.

Table 1 presents means and standard deviations for study measures and correlations for male and female adolescents. Though the morning cortisol measures were higher for females, these differences were non-significant for either the CAR or AUCG, [t (97) = 1.18, p = .24, t (97) = .98, p = .33; respectively]. Key correlational findings are the strong same-day associations between victimization and aggression perpetration for both sexes, the negative association between the AUCG and same-day aggression perpetration for male adolescents, but not females, and the lack of significant within-person associations between the cortisol measures and experiences of victimization later that day. Finally, we did not find differences in study variables (i.e., daily victimization, aggression and cortisol measures) between the 13 participants who used medications versus the 86 who did not (all p values >. 17).

Table 1.

Correlations between all variables for male (above diagonal) and female adolescents (below diagonal)

	1	2	3	4	5	6	7	8
1. Victimizationa	-	.75***	−.10	.04	.10	.29*	.09	−.37*
2. Aggressiona	.51*	-	−.19	−.13*	.09	.27*	−.44*	−.26
3. CARa	−.19	−.01	-	.53***	−.32*	−.11	.05	−.08
4. AUCGa	.10	−.05	.58***	-	.63***	−.06	.10*	−.13
5. Awakening cortisola	.09	.06	−.38***	.51***	-	.02	.12	−.08
6. Time of first cortisola	.10	.08	.08	.02	−.01	-	.00	.22
7. Cotinine levelsa	−.48**	.09	.15	−.02	.18	.01	-	−.05
8. Age	−.28	−.25	−.001	−.32*	−.24	.15	−.03	-
	M(SD)	M(SD)	M(SD)	M(SD)	M(SD)	M(SD)	M(SD)	M(SD)
Total sample	.04(0.10)	.05(0.10)	.05(0.10)	.31(.21)	.13(.30)	7.39(0.69)	4.76(21.62)	18.06(1.09)
Female	.04(.12)	.05(.11)	.07 (.19)	.35(.25)	.43(.44)	7.33 (.08)	1.59 (6.50)	18.13 (1.12)
Male	.05(.06)	.06(.08)	.03 (.13)	.29(.17)	.37(.23)	7.41 (.95)	7.81 (28.62)	17.96 (1.05)

Note: N = 99. CAR = cortisol increase above awakening levels. AUCG = total morning cortisol production measured as the area under the curve with respect to ground. Cortisol indices reported in μg/dl.

^aMultilevel correlations are calculated for associations involving daily-level variables.

^†p<.10;

^*p < .05;

^**p < .01;

^***p < .001

Associations between victimization and next-day CAR were non-significant for both female (r =.15, p = .62) and male participants (r =.11, p = .10). The association between victimization and next-day AUCG was non-significant (r = −.03, p = .88) for females but positive and significant for males (r =.14, p = .04).

3.2.1. Victimization and next-day aggression perpetration

Cross-lagged panel models were used to simultaneously test the bidirectional associations between variables (i.e., daily friends’ victimization predicts next-day aggression perpetration and vice versa) while statistically adjusting for associations at the same time point and for autoregressive effects.

Figure 1 presents results for the entire sample, and for males and females separately. Results showed a strong positive association between victimization and perpetration within the same day. We equated the same-day links between victimization and aggression across sex and this did not significantly decrease model fit, [for day t: Δχ2(1) = 43, p < .001, for day t+1: Δχ2(1) = 32, p < .001]. Victimization on one day did not predict next-day aggression, whereas aggression perpetration did predict next-day victimization levels. However, multiple group analyses showed that sex moderated the paths from daily victimization to next-day aggression and from daily aggression perpetration to next-day victimization, ([Δχ2(1) = 25.66, p < .001, Δχ2(1) = 83, p < .001; respectively)]; both cross-day associations were significant for male but not female adolescents. The autoregressive paths from victimization to next-day victimization and from aggression to next-day aggression were non-significant. Sex did not moderate the autoregressive paths, [Δχ2(1) = 3.4, p = .07, Δχ2(1) = 2.3, p = .13; respectively].

Figure 1.

Models of bidirectional same and across day associations between friends’ victimization and aggression. Same-day paths are equated across sexes.

3.2.2. Daily morning cortisol activity as a moderator of the friends’ victimization next-day aggression perpetration link

Table 2 presents results of final models testing interactive effects of daily friends’ victimization and cortisol measures on next-day aggression perpetration for the whole sample and final models for males and females. Multilevel path analyses did not show that either of the cortisol measures significantly moderated the link between victimization to next-day aggression perpetration for the whole sample. However, sex significantly moderated the victimization x AUCG interaction, which was significant for males but not for females, Δχ2(1) = 7.33, p < .001. Sex did not moderate the paths from the AUCG, Δχ2(1) = 8.34, p < .001, or daily aggression perpetration to next-day aggression perpetration, Δχ2(1) = 1.51, p = .22, which were equated across sexes.

Table 2.

The combined effect of cortisol morning output indices and friend victimization on aggression

			Aggression (t+1)
	Total Sample		Females		Males		Δχ2
Variable	ß	SE	ß	SE	ß	SE
Moderator: AUCG (t +1)
Intercept	.07*	.03	.07	.05	.08*	.03
Victimization (t)	.37*	.08	−.10	.36	.41***	.05	14.9***
AUCG (t+1)	−.06	.07	−.10	.09	−.10	.11	8.3***
Victimization (t) * AUCG (t+1)	−.12	.07	−.10	.11	−.15*	.06	7.3***
Covariates
Aggression (t)	−.11	.03	−.11	.13	−.11	.12	1.5
Moderator: CAR (t +1)
Intercept	.01	.02	.01	.03	.02	.03
Victimization (t)	.08	.10	−.34*	.12	.19*	.08	6.62***
CAR (t+1)	−.01	.02	−.03	.03	−.12	.10	3.51†
CAR (t+1) *Victimization (t)	−.03	.04	−.06	.04	.03	.04	5.18*
Covariates
Aggression (t)	−.18*	.08	−.22*	.19	−.22*	.16	0.65
Cortisol at awakening (t+1)	.03	.10	.11	.11	−.11	.07	4.7*

Note. Results are presented for the entire sample and final multiple group models for sex. ß = standardized coefficient; SE = standard errors. AUCG= cortisol total output computed as the area under the curve with respect to ground. CAR= cortisol awakening response computed as the area under the curve with respect to increase. In both models, paths from the cortisol measure to next-day aggression perpetration and the autoregressive associations were equated across sexes in final models.

^†p<.10,

^*p < .05.

^**p < .01.

^***p < .001.

As shown in Figure 2, we decomposed the interaction for males through follow-up simple slopes analyses testing the associations between victimization and next-day aggression perpetration at males’ low (−1SD) average (M) and high (+1SD) levels of the AUCG. High and average AUCG levels buffered the victimization-aggression perpetration link for males such that victimization by friends no longer predict next-day aggression perpetration towards friends, [b= .03, SE= .14, p = .86, b= .17, SE= .15, p = .26; respectively]. However, the association between victimization and next-day aggression perpetration remained significant when adolescents exhibited blunted AUCG levels (b= .33, SE= .16, p = .04).

Figure 2.

AUCG as a moderator between daily friends’ victimization and next-day aggression perpetration for female and male adolescents. Note: Coefficients represent simple slopes. b = unstandardized coefficients. Paths from AUCG to next-day aggression perpetration and the autoregressive associations for aggression perpetration are equated across sexes.

*p < .05. **p < .01. ***p < .001.

For the model testing CAR as a moderator of victimization-aggression links, sex significantly moderated victimization x CAR interaction effect on aggression, Δχ2(1) = 5.18, p = .02, which was nevertheless non-significant for both sexes. Sex did not moderate the paths from the CAR, Δχ2(1) = 3.51 p =.06, or daily aggression perpetration to next-day aggression perpetration, Δχ2(1) = 0.65, p = .42, which were equated across sexes.

Given the contribution of the cortisol awakening levels to the morning cortisol output AUCG, we also tested the moderating effect of the first (awakening) cortisol level on the victimization-aggression links, finding non-significant results for the whole sample, Β=.02, SE=.02, p=.18, and for females, Β=.04, SE=.03, p=.18, and males, Β=.05, SE=.04, p=.89, separately.

4. Discussion

We explored whether daily incidents of victimization and aggression perpetration within friendships are bi-directionally linked across days. We tested cortisol morning activity as a potential moderator of the across-day effect of daily victimization on next-day aggression perpetration. In partial support of Hypothesis 1, victimization and perpetration were bidirectionally linked across days for male but not female adolescents. However, within a day, victimization and aggression perpetration were positively linked for the whole sample. Also, for male adolescents only, we found that a higher than usual AUCG buffered the link between victimization and next-day aggression perpetration. Unexpectedly, CAR did not moderate the victimization-perpetration link.

The finding that male, but not female, adolescents increase their aggression toward friends the day after being victimized by friends might reflect sex differences in relation to youths’ social goals (De Goede, Branje, & Meeus, 2009). Female adolescents may prioritize efforts to repair relationships with the aggressive friend or to affiliate with other close friends as a way to strengthen their social connections (Crick & Zahn–Waxler, 2003). This is in line with the notion that women, when under stress, are more likely to ‘tend and befriend’ rather than ‘fight or flight’ (Taylor, 2006), which has been shown in young adults (Nickels, Kubicki, & Maestripieri, 2017). Male adolescents, on the other hand, might be more concerned about potential threats to their social status and use aggression to stabilize their social dominance or to seek revenge (Rose & Asher, 1999). Some studies, which were conducted over longer period of time, have not found sex differences in the effect of total peer victimization on total aggression perpetration (Lam et al., 2018). It is plausible that for girls, it takes repeated experiences of victimization to induce aggression perpetration, whereas for boys, aggressive reactions to victimization are more immediate. This might be explained by accumulating findings which suggest reduced sensitivity to punishment among boys compared to girls (Cross, Copping, & Campbell, 2011). However, for both sexes, within-day victimization and aggression were positively and significantly linked. Thus, perhaps female adolescents do react aggressively to victimization, but these reactions do not linger to the next day.

In addition, our results indicate that increased aggression contributes to further victimization by friends. This finding supports prior findings that showed the effect of aggression on further victimization across months (Lam et al., 2018) and highlights the immediate social costs of aggressive responses. Consistent with some prior findings (Leadbeater et al., 2006), we found this effect only for boys. This might be driven by sex differences in the type of aggression perpetrated. Physical aggression, which is more typical for male compared to female adolescents (Karriker-Jaffe, Foshee, Ennett, & Suchindran, 2008), has been found to predict more peer victimization (Leadbeater et al., 2006). Because our study was limited to four days (overlapping with cortisol assessment) and aggression and victimization were already relatively low-frequency experiences, we could not explore our hypotheses separately for types of aggression. Future studies using daily victimization and aggression perpetration across more days are needed to test the effect of different types of aggression on the victimization-aggression perpetration links.

With regard to our second hypothesis, as we expected, increases in AUCG buffered the link between victimization and next-day aggression perpetration. Based on the boost hypothesis (Adam et al., 2006), this mitigating effect of the AUCG on aggression might be explained by the favorable impact of high morning cortisol on daily mood. Specifically, increases in cortisol were found previously to improve mood (Hoyt et al., 2016). Thus, it is plausible that a heightened AUCG helps to reduce anger and irritability, which have been found in the past to relate to aggression perpetration towards peers (Walters & Espelage, 2018). Also, the negative effect of the AUCG on same-day aggression might be due to its catalyzing effect on inhibitory control later that day (Shi et al., 2018). These favorable effects on metacognitive functioning have been suggested to result from intertwined activity of the morning cortisol activity and the prefrontal cortex, which is beyond the scope of this study (Shields, Bonner, & Moons, 2015).

Similar to sex differences found in the behavioral response to daily victimization (hypothesis 1), AUCG buffered the victimization-aggression perpetration link only for boys. This sex effect is also consistent with several prior studies, (Rudolph et al., 2010; Vaillancourt et al., 2018) and might indicate that physiological underarousal predisposes males but not females to aggression in the context of peer victimization. It is plausible that low AUCG is related differently to emotional activation across sexes. For example, daily increases in positive mood were related to lower total daily cortisol levels among females but not males (Polk, Cohen, Doyle, Skoner, & Kirschbaum, 2005).

On a more biological level, the inverse relations between cortisol and aggression have been previously found to be intertwined with diminished activity of other physiological systems, particularly testosterone (Platje et al., 2015) and sympathetic nervous system activity (Gordis, Granger, Susman, & Trickett, 2006) to predict lower aggression. This explanation can also help to explain why we found a moderation effect for the AUCG only for male adolescents. Research shows that male adolescents generally have higher morning testosterone levels (Harden et al., 2016; Marceau, Ruttle, Shirtcliff, Essex, & Susman, 2015; Marceau, Ruttle, Shirtcliff, Hastings, et al., 2015). Thus, perhaps higher testosterone levels among male adolescents drive more aggressive behaviors in reaction to the prior day experiences of victimization. This explanation is consistent with prior research which found positive associations between higher testosterone levels and externalizing symptoms among male youth (Maras et al., 2003). However, it is also possible that higher daily rates of victimization and aggression among male compared to female participants might statistically enable us to detect effects for male compared to female adolescents.

Only the AUCG and not the CAR was a significant moderator of daily victimization-aggression links. In contrast to the CAR, the AUCG also includes the cortisol awakening levels. As other researchers have suggested before, it is plausible that the catalyzing effect of prior day stressors already starts pre-awakening to produce increased cortisol levels the next day (Adam et al., 2006; Bai et al., 2017). The fact that the AUCG and not the CAR predicted less aggression later that day points to the contribution of total morning cortisol levels for same-day behaviors. As it has been suggested before, perhaps total morning cortisol levels exert a stronger effect on metabolic processes which provide the body available energy to cope (Adam et al., 2006). Though a small number of studies have reported non-significant effects for the cortisol increase (Adam et al., 2006; Zoccola et al., 2011), this finding is unexpected and requires further investigation. However, the non-significant moderation effect of the initial cortisol awakening levels on the victimization-aggression link suggests that the combination of the cortisol morning increase (CAR) and the awakening levels contribute the buffering daily buffering effect.

4.1. Strengths, limitations, and future studies

There are several limitations to this study that should be noted. First, we tested our hypotheses across only four days, overlapping with the three days of cortisol sampling. This limits our ability to capture the range of victimization and perpetration experiences for individuals. In addition, we could not account for the effect of different types of victimization on aggression perpetration and the HPA axis. This is important given that divergent links have sometimes been found between HPA measures and relational versus physical aggression (Murray-Close, Han, Cicchetti, Crick, & Rogosch, 2008; Pitula et al., 2015). Moreover, we do not know whether the target’s aggression was directed toward the friend who previously was aggressive (i.e., retaliation), or to other non-aggressive friends (displacement of aggression). In addition, we did not assess the extent to which participants were upset by the victimization, which is likely to affect their reaction (Oberle et al., 2017). Also, daily victimization and aggression perpetration were based on self-reports, which might lead to underreporting of its occurrence (Hymel & Swearer, 2015). In addition, we did not assess how the quality of friendship between the adolescent and perpetrating friend impacted our findings. This is relevant to include given prior research showing associations between global levels of peer aggression and friendship quality (Kawabata, Crick, & Hamaguchi, 2010). Finally, we did not have objective data on sleep duration and quality which have been found to influence the CAR (Sladek & Doane, 2015; Zeiders, Doane, & Adam, 2011) or validated awakening times, which are recommended in order to assess compliance with the saliva collection (Stalder et al., 2016). Furthermore, future research is needed to test the role of sleep duration and quality on the victimization- aggression perpetration link, given its potential influence on self-regulation capacities (Gruber & Cassoff, 2014). Even so, using adolescents’ day-to-day self -reports on incidents of victimization and aggression perpetration in combination with collection of cortisol across three days provide a unique opportunity to test the role of cortisol in time-linked behavioral responses to social adversities.

4.2. Conclusion

This study shows how cycles of victimization and aggression in friendship contexts are maintained on a daily level and identifies a physiological marker that disrupts this pattern. This new knowledge may help to identify points of intervention for preventing these reciprocal links. In addition, male but not female adolescents showed a bio-behavioral sensitivity to daily peer victimization. This highlights the need to further consider sex-specific processes contributing to the maintenance of victimization and aggression links in daily life.