Rumination and Impaired Cortisol RecoveryFollowing a Social Stressor in Adolescent Depression

Jeremy G. Stewart & Raegan Mazurka & Lea Bond &

Katherine E. Wynne-Edwards & Kate L. Harkness

# Springer Science+Business Media New York 2013

Abstract Response styles theory promotes rumination as acentral cognitive construct driving negative mood and depres-sion, and past research suggests that at least part of themechanism driving rumination’s depressogenic effect isthrough inhibiting the individual’s ability to shift attentionalfocus away from negative environmental stimuli. In the cur-rent study, we hypothesized that high trait rumination wouldbe associated with impaired recovery of the body’s biologicalresponse to psychological stress. In a community sample ofdepressed (n=31) and non-depressed (n=33) adolescents weassessed rumination and the more adaptive trait of distractionand problem-solving with the Children’s Response StylesQuestionnaire (CRSQ; Abela 2000), and diagnostic statuswas confirmed using the Child and Adolescent Schedule ofAffective Disorders and Schizophrenia (K-SADS; Kaufmanet al. Journal of the American Academy of Child andAdolescent Psychiatry 36:980–988, 1997). Participants com-pleted the Trier Social Stress Test (TSST; Kirschbaum et al.Neuropsychobiology 28:76–81, 1993), and the focus of ouranalyses was the change in salivary cortisol concentrationbetween peak cortisol output (25 min post-stressor) and asample taken during the “Recovery” period 65 minutes post-stressor. Consistent with the predictions of response styletheory, among the depressed adolescents only, high trait ru-mination was associated with delayed post-stressor cortisol

recovery, whereas high trait distraction and problem-solvingwas associated with more rapid recovery. In contrast, responsestyles were not associated with cortisol recovery in the non-depressed group. These findings implicate impaired post-stress cortisol recovery as a potential mechanism underlyingthe pathological effect of rumination on the development andmaintenance of Major Depressive Disorder (MDD).

Keywords Depression . Adolescence . Response styles .

HPA axis dysregulation . Stress

Rumination is a widely studied cognitive construct that hasemerged as central to the etiology, maintenance, and exac-erbation of Major Depressive Disorder (MDD). Accordingto Response Styles Theory (RST; Nolen-Hoeksema 1991),rumination is defined as a maladaptive cognitive processinvolving passive, repetitive thinking about the potentialcauses and consequences of one’s depressive symptoms,and is conceptualized as a stable trait that increases riskfor depressed mood. In contrast, distraction is defined asan adaptive alternative response to one’s depressive symp-toms and involves thoughts or activities that direct attentionaway from these symptoms (Nolen-Hoeksema et al. 2008)

RST proposes that the central maladaptive consequencesof rumination are a worsening and prolongation of negativemood and depression. Early experimental work with non-depressed college students has shown that laboratory-induced rumination is associated with increases in self-reported dysphoria over time, whereas distraction is associ-ated with decreases in dysphoria (e.g., Lyubomirsky et al.1998; Lyubomirsky and Nolen-Hoeksema 1993, 1995;Nolen-Hoeksema and Morrow 1993). Further, in both adultand adolescent clinical samples, high levels of ruminationprospectively predict new onsets of MDD, and are associat-ed with longer, more severe episodes of the illness (e.g.,

This research was by an operating grant from the Ontario MentalHealth Foundation (Harkness, Wynne-Edwards).

J. G. Stewart (*) :R. Mazurka :K. L. HarknessDepartment of Psychology, Queen’s University,Kingston, Ontario, Canada K7L 3N6e-mail: 2js4@queensu.ca

L. Bond :K. E. Wynne-EdwardsFaculty of Veterinary Medicine, University of Calgary,Calgary, Alberta, Canada

J Abnorm Child PsycholDOI 10.1007/s10802-013-9740-1

Ciesla and Roberts 2002; Just and Alloy 1997; Gibb et al.2012; Lara et al. 2000; Nolen-Hoeksema and Davis 1999;Nolen-Hoeksema et al. 1994; Robinson and Alloy 2003).

Researchers have proposed that the mechanism throughwhich rumination acts to worsen and prolong depressed moodis by inhibiting the individual’s ability to shift attentional focusaway from negative self-referential information in the environ-ment. For example, functional neuroimaging studies haveshown that high trait rumination, particularly in the contextof depression, is associated with sustained amygdala activity inresponse to negatively valenced stimuli (Ray et al. 2005;Siegle et al. 2002). Further, Johnson et al. (2009) found thatduring a laboratory-induced rumination task, participants withdepression showed lower activity than controls in areas of themedial prefrontal cortex (mPFC) that are associated with pos-itively valenced thought. Taken together, the above resultssuggest that dysphoric and depressed individuals with a ten-dency to ruminate selectively recruit neural regions responsiblefor processing negative self-relevant information, and inhibitregions responsible for processing positive information.

The above research has important implications for under-standing individual differences in recovery following stress.Young and Nolen-Hoeksema (2001) have suggested that aruminative self-focus in the face of stress may impair the body’snatural recovery process, and lead to amplification and prolon-gation of the stress response. The biological stress response isregulated by the hypothalamic pituitary adrenal (HPA) axis,which has the glucocorticoid hormone cortisol as its finaloutput. On average, the concentration of circulating cortisol inthe bloodstream reaches its peak 20–30 min after the onset ofan acute stressor. When individuals are exposed to laboratorystressors that involve uncontrollable, social-evaluative threat,circulating levels of cortisol fall towards baseline (i.e., recover)over the course of up to 60 min following the offset of thestressor through a process of negative feedback (Dickerson andKemeny 2004). Meta-analytic work in adults has documented asignificant prolongation of the biological response to laboratorystress challenge in patients with MDD versus non-depressedcontrols, as indicated by impaired recovery of cortisol levelspost-stressor (d=1.39, SE=0.35, n=4; Burke et al. 2005).Children and adolescents with MDD also show a dysregulatedHPA axis response to stress challenge as indexed by higherpeak cortisol (Luby et al. 2003; Rao et al. 2008) and highercortisol levels during recovery (Rao et al. 2008).

Given the research reviewed above relating high trait ru-mination to deficits in shifting attentional focus from negativeinformation, one would expect those high in rumination todisplay a prolongation of the biological stress response, asindexed by a slower return to baseline, and generally increasedcortisol concentration in the recovery phase. Presumably, thisrelation of rumination and prolonged HPA axis activation ismediated by increased amygdalar and decreased mPFC acti-vation, given the anatomical link between the HPA axis and

these cortico-limbic structures. Indeed, functional evidencesuggests that increased amygdalar and decreased mPFC acti-vation is associated with greater cortisol output (Kern et al.2008; Taylor et al. 2008). However, despite the shared neuralcorrelates of trait rumination and cortisol secretion, empiricalevidence for the impact of trait rumination on cortisol secre-tion in response to stress has been mixed.

Consistent with predictions, Zoccola et al. (2010) foundthat high trait rumination (measured using the rehearsalsubscale of the revised Emotion Control Questionnaire;Roger and Najarian 1989) was associated with heightenedcortisol reactivity and delayed recovery following a labora-tory stress challenge paradigm. In contrast, an earlier studyby the same group and using the same challenge paradigm,found that their sample of high ruminators (assessed withthe Response Styles Questionnaire [RSQ; Nolen-Hoeksemaand Morrow 1991]) showed a blunted pattern of cortisolresponse, and it was low ruminators who showed pro-nounced cortisol reactivity to the stressor (Zoccola et al.2008). Finally, Young and Nolen-Hoeksema (2001) reportedno difference between high and low ruminators, as assessedby the RSQ, in terms of their peak cortisol output or recov-ery in response to a social stress challenge.

There are likely several methodological factors that couldexplain the discrepant results reported above. Most impor-tantly, however, all of these studies employed analoguesamples of college undergraduates and none examined theindividuals most at risk for prolonged negative mood inresponse to ruminating — those with heightened levels ofdepression. Given that RST explicitly conceptualizes rumi-nation as a cognitive reaction to depressive symptoms, re-search to date has not provided an adequate test of thehypothesis that rumination is differentially associated withimpaired recovery of the biological stress response amongthose with and without depression.

The current study is the first to our knowledge to inves-tigate the relation of trait rumination to cortisol recoveryfrom stress in a sample of adolescents diagnosed with aunipolar depressive disorder according to Diagnostic andStatistical Manual of Mental Disorders (DSM-IV-TR;American Psychiatric Association 2000) criteria. By focus-ing on individuals at highest risk for the negative effects ofrumination, this will provide a more appropriate test of thehypothesis that rumination is related to HPA-axisdysregulation in MDD (Burke et al. 2005; Lopez-Duran etal. 2009). Adolescence is a key developmental period for theonset of depression (e.g., Kessler and Walters 1998) butresearch on the etiological mechanisms involved in initialonsets of depression in this group is lacking. Establishing arelation between rumination and impaired recovery of thebiological stress response in adolescence represents an im-portant first step to implicating rumination in the mainte-nance of depression over the lifespan.

J Abnorm Child Psychol

We will utilize the Trier Social Stress Test (TSST;Kirschbaum et al. 1993), which has distinct advantages inthe laboratory assessment of the stress response given its highdegree of social evaluative threat and uncontrollability, there-by ensuring biologically relevant responses (Dickerson andKemeny 2004). The current study is the first to our knowledgeusing an adolescent sample that examines cortisol recoveryfrom the TSST in a manner that controls for the inter-individual differences in peak cortisol output. Some studies(Rao et al. 2008; Zoccola et al. 2010) have found that groupsthat show higher cortisol concentration during recovery alsohave significantly higher peak cortisol values. Our study willaddress this confound by controlling for peak cortisol concen-tration, and thus examining the change in cortisol in responseto a stressor to index recovery.

Further, the current study is the first to examine the effectsof distraction on cortisol recovery following stress (Nolen-Hoeksema 1991). Studies on the neurobiological correlates ofemploying distraction to reduce negative affect have foundpatterns of activation that are opposite to those reported in thestudies on rumination. For instance, a recent study found thatparticipants displayed decreased amygdalar activationcoupled with increased activation in a large area of themPFC (including the dorsal Anterior Cingulate Cortex) asso-ciated with controlling attention to emotional stimuli and forsignaling the need for cognitive control, on trials on whichthey were distracted from a negatively valenced face using anunrelated stimulus (McRae et al. 2010). As was reviewedabove, the same neural correlates of rumination (i.e., increasedamygalar and decreased mPFC activation) are associated withgreater cortisol secretion in response to stress (Kern et al.2008; Taylor et al. 2008). In the current context, therefore,individuals with a tendency to distract may have more rapidcortisol recovery in the face of stress. By focusing on bothrumination and distraction we will clarify their respectiveroles in the pathology of depression.

Finally, the current study will extend previous research byalso examining the ratio of rumination to distraction, in addi-tion to rumination and distraction separately. This ratio indexesthe probability of engaging in rumination over distraction inresponse to low mood. Using this approach to response stylesassumes that both types of response styles exert independenteffects on cortisol recovery and that these effects can be com-bined in a linear fashion. In a sample of children, Abela et al.(2007) demonstrated that the ratio of rumination to distractionscores was more strongly related to change in depressive symp-toms over time than was either scale on its own.

For the purposes of this study, we define cortisol recoveryas the cortisol concentration 40 min following the offset of thestressor, controlling for the peak cortisol concentration col-lected immediately following the stressor. Therefore, lowercortisol recovery values indicate a more rapid return of circu-lating cortisol to baseline, whereas higher cortisol recovery

values indicate a slower return of circulating cortisol to base-line. We hypothesize that, among depressed adolescents only,higher self-reported trait rumination will be significantly as-sociated with higher cortisol recovery values (i.e., slowerrecovery). In contrast, we hypothesize that higher trait distrac-tion and problem-solving will be associated with lower corti-sol recovery values (i.e., more rapid recovery). Finally, againamong depressed adolescents only, we predict that a higherratio of rumination to distraction will be associated withhigher cortisol recovery values (i.e., slower recovery).

Method

Participants

The sample consisted of 64 adolescents (47 females) aged 12–18 years (M=15.30, SD=1.89) recruited from a mid-sizedcommunity in southeastern Ontario. These participants arepart of a larger ongoing study examining pathological stressprocesses in adolescent depression, and data including a sub-set of the current sample (n=37) have been published previ-ously (Harkness et al. 2011). Approximately 90 % (n=58) ofparticipants were of European ancestry, which is consistentwith the ethnic diversity of the population in the area.Demographic and clinical characteristics of the sample, strat-ified by diagnostic status, are presented in Table 1. Among thedepressed adolescents, 19 (61.3 %) were receiving some formof treatment, and 9 (29.0 %) were taking at least one psycho-tropic medication at the time of the study. Non-psychiatriccontrol participants were recruited from local high schools, aswell as community advertisements. Depressed adolescentswere referred from mental health practitioners in the commu-nity or were self-referred in response to advertisements.

The depressed group all met DSM-IV-TR (AmericanPsychiatric Association 2000) criteria for a current non-bipolar, non-psychotic mood disorder based on a structureddiagnostic interview (see below). Participants were exclud-ed if they met criteria for any of the following: a psychoticdisorder, bipolar disorder, substance dependence, conductdisorder, a developmental disability, or a medical disorderthat could cause depression. Adolescents in the non-psychiatric control group did not meet diagnostic criteriafor any current or past psychiatric illness.

Our initial sample included 106 adolescents. Twenty-eightindividuals were excluded based on our diagnostic exclusionarycriteria. Further, eight participants did not complete the measureof response styles, and six were missing one or more cortisolsamples, leaving 64 usable cases. The final sample did not differfrom excluded participants in terms of sex, !2(1, N=106) 0.56,p=0.45, age, t(104)=1.44, p=0.15, parental occupation status,t(104)=1.25, p=0.21, or ethnic distribution, !2 (3, N=105)=1.70, p=0.64.

J Abnorm Child Psychol

Measures

Child and Adolescent Schedule of Affective Disorders andSchizophrenia (K-SADS; Kaufman et al. 1997) We evaluatedthe presence of current and/or past DSM-IV-TR Axis Idiagnoses using the full K-SADS. Advanced graduate stu-dents in clinical psychology administered all interviewsafter being trained to gold-standard reliability status by thesecond author (see Grove et al. 1981). Trainees participatedin at least six interviews (three observing a trained inter-viewer, and three conducting the interview) and coded di-agnoses independently of their co-interviewer. Traineeswere required to have 100 % agreement with their trainedco-interviewer on diagnoses prior to conducting solo inter-views. Each interviewer was provided with ongoing clinicalsupervision, and conferenced each case, with the fifth authorfor the duration of the project.

In our sample, 31 (48 %) participants met diagnostic criteriafor a current depressive disorder: (1) Major DepressiveDisorder (n=18); (2), Dysthymia (n=5); (3) DepressiveDisorder Not Otherwise Specified (n=5), and (4) AdjustmentDisorder with Depressed Mood (n=3).1 Of these adolescents,17 (55%)met criteria for at least one comorbid Axis I disorder.Comorbid disorders included Social Phobia (n=6),Generalized Anxiety Disorder (n=5), Panic Disorder (n=2),Obsessive Compulsive Disorder (n=2), Post-Traumatic StressDisorder (n=2), Alcohol or Substance Abuse (n=7), EatingDisorder Not Otherwise Specified (n=1), and OppositionalDefiant Disorder (n=2).2

Basic demographic information was collected as part ofthe diagnostic interview. Parental occupation status wassubsequently coded by two independent judges on a 1- to7- point scale based on the Hollingshead Index of SocialPosition (Hollingshead 1975). Higher scores indicate a low-er social position. Any discrepancies between raters wereresolved by consensus, and the consensus ratings wereultimately used in analyses.

BeckDepression Inventory (BDI-II; Beck et al. 1996) Depressionseverity was assessed with the 21-item BDI-II. This measureis widely used in the assessment of depression in adoles-cents, and showed excellent internal consistence in oursample (α=0.95).

Mood and Anxiety Symptom Questionnaire (MASQ; Watsonand Clark 1991) The MASQ is a 90-item self-report ques-tionnaire designed to measure symptoms of depression andanxiety using a five-point Likert scale. The Anxious Arousal(AA) subscale is comprised of 17-items that measure symp-toms of somatic tension and hyperarousal that are hypothe-sized to be specifically associated with anxiety disorders,and not mood disorders. Watson et al. (1995) reported goodinternal consistency for the AA subscale across five separatesamples, with coefficient alpha values ranging from 0.86 to0.90 for subscale items. We used the AA scale to indexsymptoms of anxiety in our primary analyses.

Tanner Stages of Pubertal Maturation (Tanner 1962) In theTanner, participants chose between five illustrations ofbreast (females only) and pubic hair (males and females)development, as validated against physician assessment(Taylor et al. 2001). Scores on the scales range from 1 to5, with higher scores indicating later stages of development.None of the girls in the study had ever been pregnant.

Children’s Response Styles Questionnaire (CRSQ; Abela etal. 2000) The CRSQ was adapted from the RSQ (Nolen-Hoeksema and Morrow 1991) and assesses how often in-dividuals employ different cognitive strategies in responseto feeling sad. The CRSQ includes two subscales (Abela etal. 2007): a) Rumination (CRSQ-RUM) which includes 13items describing Ruminative Responses (e.g., “When I amsad, I think: ‘I’m ruining everything’”), and b) Distractionand Problem-Solving (CRSQ-DPS), which includes eightitems that redirect attention away from one’s mood (e.g.,“When I am sad, I do something fun with a friend”) oremploy strategies to overcome sadness (“When I am sad, Ithink of a way to make my problem better”). Regardingvalidity, previous research has demonstrated that the CRSQ-RUM is positively correlated with depressive symptoms inchildren and young adolescents, and that the CRSQ-DPS isnegatively correlated with depressive symptoms in youngadolescents (Abela et al. 2004, 2007). All items are rated ona 4-point scale from 1 (almost never) to 4 (almost always).In the current sample, the internal consistency for theCRSQ-RUM subscale was α=0.93, and for the CRSQ-DPS it was α=0.63. The two subscales were moderatelynegatively correlated (r=!0.39, p<0.01).

Procedure

The current study was conducted in compliance with ourhome institution’s Health Sciences Research Ethics Board.Adolescents participated in two 2.5-hour sessions separatedby 1 week, with both sessions occurring in the mid-afternoon with the same experimenter to avoid post-awakening elevations in cortisol. During the first session,

1 We re-ran our primary analyses using only adolescents who metcriteria for MDD and excluding those who met criteria for otherdepressive disorders. The results of these analyses did not differ fromthose using the full sample. Results of these analyses are availableupon request.2 As indicated in Table 1, nine adolescents were taking psychotropicmedication(s) at the time of the study. Given the documented effects ofcertain medications on cortisol function (Holsboer 2001), we re-ran allanalyses below excluding adolescents who were currently taking med-ication. The pattern of our results remained unchanged. Results areavailable upon request.

J Abnorm Child Psychol

written informed consent was obtained from all participants,and from parents or guardians of adolescents younger than18. At this time, participants were informed that they wouldbe asked to give a brief speech at the second session.Following consent, adolescents received the diagnostic in-terview and completed the questionnaires.

The second session involved the TSST (Kirschbaum et al.1993). Adolescents had been instructed not to eat or drink forat least one hour prior to their session. Upon arrival, partic-ipants were instructed to rest alone in the testing room for10 min, after which the experimenter collected the baselinesaliva sample (Sample A). Participants were then led to aseparate room and introduced to two research assistants whothey were told were members of a selection committee fromthe human resources department of a major department store.The participants were told to prepare a five-minute speechthat would serve as a job interview to be presented to thecommittee. They were told that the speech would be videorecorded and that an expert on the committee would beanalyzing their behavior. Participants were then broughtback to the original testing room and given 10 min to preparetheir speech, after which the experimenter collected a secondsaliva sample (Sample B). The adolescents then deliveredtheir speech to the committee, were asked several follow-upquestions, and performed an arithmetic task (i.e., serial

subtractions of 13 from 2,083, as quickly as possible withoutmaking mistakes). The stress task took approximately15 min, after which a third sample was collected (SampleC). Thus Sample C was collected 25 min after the onset ofthe stressor (i.e., being informed that one needs to prepare aspeech) and represented peak cortisol output. The experi-menter collected a fourth sample (Sample D) 40 min later,and a final sample 40 min after the fourth (Sample E).Following the TSST, participants were fully debriefed aboutthe purpose of the task and the nature of the deception.Participants were compensated $40 for their participation inthe study, and any self-referred participants who we identi-fied as suffering from a psychiatric condition were referredfor treatment.

Cortisol Collection and Hormone Determinations All salivasamples were collected in 5-ml polypropylene vials (RoseScientific Ltd, Edmonton, Alberta) by passive drool (Shirtcliffet al. 2001) between 3 pm and 5 pm to minimize the potentialfor circadian change during the trial (Groschl et al. 2003).Immediately after collection, all samples were placed in afreezer for short-term storage and eventually transported to asecure storage freezer (!20 °C). The resulting supernatant ineach sample was assayed for cortisol using the high sensitivityenzyme immunoassay designed for saliva (1–3002; Salimetrics

Table 1 Sample demographicand clinical characteristics bydiagnostic group

Tanner Tanner Stages of Puber-tal Maturation; CRSQ-RUM traitrumination; CRSQ-DPS traitdistraction and problem-solving;BDI Beck Depression Inventory;AA Anxious Arousal symptomsaPairwise t-tests were conductedusing Toothaker’s (1991) meth-od for combining between- andwithin-subjects error

Depressed(n=31)

Non-depressed(n=33)

Statistic(t or !2)

p

Sex (Female), n (%) 25 (80.6) 22 (66.7) 1.60 0.21

Age, M (SD) 16.00 (1.57) 14.64 (1.95) 3.07 0.003

Tanner, M (SD) 3.73 (3.42) 3.73 (2.43) 0.002 0.998

Parental Hollingshead Index, M (SD) 3.84 (1.73) 2.85 (1.64) 2.35 0.02

CRSQ_RUM, M (SD) 36.58 (8.08) 23.39 (8.28) 6.44 <0.001

CRSQ_DPS, M (SD) 14.48 (3.19) 18.94 (3.33) 5.34 <0.001

BDI, M (SD) 30.13 (9.67) 6.97 (6.11) 11.53 <0.001

AA, M (SD) 39.83 (13.49) 24.63 (7.09) 5.59 <0.001

Age at first onset, M (SD) 14.23 (2.14)

Depression History (Recurrent), n (%) 8 (25.8)

Current Treatment status

Medications Alone, n (%) 4 (12.9)

Therapy Alone, n (%) 10 (32.3)

Therapy + Medications, n (%) 5 (16.1)

No Treatment, n (%) 12 (38.7)

Comorbidity (yes), n (%) 17 (54.8)

Episode duration in months, M (SD) 17.90 (20.53)

Cortisola (ug/dL) M (SD)

Sample A 0.125 (0.011) 0.120 (0.010) 0.23 0.82

Sample B 0.119 (0.013) 0.132 (0.012) 0.60 0.55

Sample C 0.142 (0.023) 0.165 (0.022) 1.07 0.29

Sample D 0.122 (0.017) 0.137 (0.016) 0.70 0.49

Sample E 0.102 (0.011) 0.102 (0.010) 1.0 1.0

J Abnorm Child Psychol

LLC, State College, PA). To ensure that inter-assay variabilitydid not contribute to quantification error, all samples from oneindividual were placed on the same plate. Each sample wasquantified in duplicate at 25 μl, and duplicate high and lowcontrols were distributed across each plate to monitor precision.Following Berg and Wynne-Edwards (2001), samples that hada coefficient of variation of "15 % were repeated on anotherplate. The repeated samples were not used in the analyses, butwere only used to reject one of the original duplicates. In all, 32assay runs, in six batches, were conducted. The high control,measured at 1.071 μg/dL, had an intra-assay coefficient ofvariation of 3.6 % and an inter-assay coefficient of variationof 5.0 %. The low control, measured at 0.103 μg/dL had anintra-assay coefficient of variation of 6.2 % and an inter-assaycoefficient of variation of 9.8 %.

Data Analysis

Given the substantial positive skew for samples B, C and D(Skew=2.29–3.68), all cortisol samples were common loga-rithm transformed prior to analyses. There were no outliers(i.e., cases +/! 3SD from the mean) within the transformedvariables. We first conducted a 5 (time)!2 (group: control,depressed) repeated-measures Analysis of Covariance(RMANCOVA), using the five cortisol samples as thewithin-subjects factor. CRSQ-RUM, CRSQ-DPS and theirinteractions with a dummy-coded group variable (non-de-pressed vs. depressed) were entered as covariates. This omni-bus test allowed us to assess differences in the overall stressresponse curve (i.e., Sample A through Sample D) as a func-tion of group and response styles, and to evaluate whether itwas appropriate to conduct follow-up analyses targeting therecovery portion of the overall stress response. Any significantGroup X Time X response styles interactions were followedup by examining the two-way interactions within each group,and analyzing the polynomial trend of the stress response atvarying levels of the response styles variables. The results ofthe trend analyses, in conjunction with crude observation ofthe overall cortisol curves at varying levels of response style,were used to motivate more fine-grained statistical analysis ofthe recovery period.

Our follow-up analysis consisted of a hierarchical linearregression to examine the effects of diagnostic group, CRSQ-RUM, CRSQ-DPS and their interactions in predicting therecovery portion of the cortisol curve, and specifically thechange in participants’ cortisol concentration from peak(Sample C) to Sample D. We focused on Sample D because,on average, cortisol samples collected 41–60min post-stressorin tasks such as this TSST remain significantly elevated com-pared to pre-stressor levels, while also being substantiallylower than peak levels (Dickerson and Kemeny 2004)— thus,we deemed the change from Sample C to D the best index ofcortisol recovery because it was likely to contain the greatest

inter-individual variability in cortisol concentration valuesrelatively to peak concentation values. To analyze change incortisol, we inserted the common logarithm transformedSample C values in step 1 of our hierarchical regressionmodel. Step 2 included any demographic and/or clinical vari-ables that were significantly associated with the dependentvariable. In step 3, we entered the main effects of diagnosticgroup (0 = non-depressed, 1 = depressed), CRSQ-RUM andCRSQ-DPS. Finally, in step 4 we entered the Group X CRSQ-RUM and Group X CRSQ-DPS interaction effects — bothcontinuous variables were centered at their mean to create theinteraction effects. Two-way interaction effects were followedup by computing simple slopes for each diagnostic group(Aiken and West 1991).

Results

Descriptive Statistics

Relevant demographic and clinical characteristics for thesample, stratified by diagnostic status, are presented inTable 1. Adolescents in the depressed and non-depressedgroups did not significantly differ in terms of sex or Tannerscores. However, adolescents in the depressed group weresignificantly older, and had a lower parental occupationstatus, than controls. Not surprisingly, depressed adolescentsalso had significantly higher BDI-II, AA, and CRSQ-RUMscores, and lower CRSQ-DPS scores, compared to the non-depressed adolescents.

Further, there was no evidence that CRSQ-RUMor CRSQ-DPS were significantly associated with sex (t[62]=0.92,p=0.36; t[62]=0.15, p=0.88, respectively) or Tanner scores(r=!0.20, p=0.12; r=!0.01, p=0.94, respectively). However,CRSQ-RUM, but not CRSQ-DPS, was significantly correlat-ed with both older age (r=0.40, p=0.001; r=!0.18, p=0.16,respectively) and lower parental occupation status (r=0.29,p=0.02; r=!0.17, p=0.18, respectively). Finally, covaryingSample C, there was no evidence that Sample D values wereassociated with sex, F(1, 61)=0.11, p=0.74, Tanner scores,partial r=!0.06, p=0.65, medication status, F(1, 61)=0.82,p=0.37, BDI-II scores, partial r=!0.02, p=0.91, or AAscores, partial r=!0.19, p=0.15. However, higher Sample Dvalues were associated with younger age, partial r=!0.24,p=0.06, and lower parental occupation status (lowerHollingshead score), partial r=0.22, p=0.09, at trend levels.Therefore, we included age and parental occupation status inthe main models below.

Within Group Cortisol Output over Time

Common logarithm transformed cortisol concentrations(in μg/dL) at each TSST collection point were analyzed using

J Abnorm Child Psychol

a 5 (time)!2 (group: 0 = non-depressed, 1 = depressed)RMANCOVA, covarying the effects of age and parental oc-cupation status, and including the effects of CRSQ-RUM,CRSQ-DPS and their interactions with group. All continuousvariables were centered at their means. The 3-way interactionof time, group, and CRSQ-DPS did not reach statistical sig-nificance, nor did the 2-way interactions of CRSQ-DPS witheither time or group (all ps >0.39, "p

2 <0.07).However, there was a significant 3-way interaction of

time, group, and CRSQ-RUM, F(4, 224)=2.64, p=0.04,"p

2=0.05. The Time x CRSQ-RUM interaction was notsignificant for the non-depressed group (p=0.41, "p

2=0.01). In the depressed group, using an approach similar tosimple slopes for moderated regression (Aiken and West1991), CRSQ-RUM was centered at one standard deviationabove and below the mean, and these two new values werethen used to create interaction effects with time. We found asignificant quadratic effect of time at low levels (!1 SD) ofrumination, F(1, 56)=11.36, p=0.001, "p

2=0.17, and sig-nificant linear, F(1, 56)=8.40, p=0.005, "p

2=0.13, andquartic, F(1, 56)=10.94, p=0.002, "p

2=0.16, effects at highlevels of rumination (+1 SD). As displayed in Fig. 1, theseresults suggest a difference between depressed high and lowruminators during the recovery portion of the TSST, where-by high ruminators maintained relatively high levels ofcortisol while low ruminators recovered more rapidly. Inorder to statistically test these general trends in a morefine-grained manner, we conducted hierarchical linear re-gression analyses that targeted the recovery portion of thecurve exclusively.

Between Group Differences in Cortisol Recovery

Response Styles Sample C was entered alone in Step 1 ofour model, and accounted for a substantial portion of vari-ance in Sample D, R2=0.64, F(1, 62)=108.96, p<0.001. InStep 2 of our model, age and parental occupation statuscontributed a significant amount of unique variance to theprediction of Sample D, above the effect of Sample C,ΔR2=0.04, ΔF(2, 60)=3.19, p=0.048. The addition of the diag-nostic status and the response styles variables in Step 3 didnot significantly improve the model prediction (p=0.42), andnone of the individual parameter estimates were significant(all ps>0.12). However, consistent with hypotheses, theaddition of the group by response styles interaction effectsin Step 4 of the regression model resulted in a significantincrement in variance explained by our model, ΔR2=0.05,ΔF(2, 55)=5.55, p=0.006 (see Table 2). The Group xCRSQ-RUM interaction effect was significant in the finalmodel, whereas the Group x CRSQ-DPS interactionapproached significance. In total, the model accounted for74 % of the variance in Sample D scores.

Figure 2 presents the simple slopes for the Group x CRSQ-RUM interaction using the untransformed Sample D valuesfor clarity of interpretation. Among depressed adolescents,there was a significant positive association between CRSQ-RUM and Sample D, B=0.007, SE=0.003, t(55)=2.07,p=0.04, d=0.56. In contrast, among the non-depressed ado-lescents, the slope of the relationship between CRSQ-RUMand Sample D was negative and failed to reach significance,B=!0.006, SE=0.003, t(55)=!1.69, p=0.10, d=0.46.

0.07

0.09

0.11

0.13

0.15

0.17

0.19

0 20 40 60 80 100 120

Mea

n C

ortis

ol C

once

ntra

tion

(ug/

dL)

Time (minutes) Speech

Prep

A B

Rest Tasks Other Measures

Task Description & Meet Panel

ParticipantArrives

C D E

Non-Depressed

Depressed High CRSQ-RUM

Depressed Low CRSQ-RUM

Fig. 1 Estimated marginalmeans for untransformedcortisol concentrations at eachsample point for the non-depressed group (n=33), thedepressed group at high CRSQ-RUM (n=13), and thedepressed group at low CRSQ-RUM (n=18). Age, parentaloccupational status and CRSQ-DPS were centered at theirmeans when calculating theplotted marginal means. CRSQ-RUM trait rumination; CRSQ-DPS trait distraction andproblem-solving

J Abnorm Child Psychol

Although the Group x CRSQ-DPS did not meet statisticalsignificance, the relations between the CRSQ-DPS and SampleD in our models followed the opposite pattern reported for theGroup x CRSQ-RUM interaction. Among depressed adoles-cents, there was a significant negative association betweenCRSQ-DPS and Sample D, B=!0.022, SE=0.009, t(55)=2.39, p=0.02, d=0.64. In contrast, among the non-depressedadolescents, the slope of the relationship was not statisticallysignificant, B=0.001, SE=0.008, t(55)=0.12, p=0.91, d=0.03.

CRSQ-ratio Analyses Regression coefficients for the modelincluding the ratio of rumination to distraction are provided inTable 3. The total model accounted for 73 % of the variance.Once again, Sample C was entered in Step 1 of the model,followed by age and parental socioeconomic status in Step 2.The addition of diagnostic status and CRSQ-ratio in Step 3resulted in a nonsignificant increment in the variance inSample D explained by the predictors (p=0.23) and neitherof the terms were significant predictors (both ps >0.09). Thestep including the two-way interaction of group and responsestyle ratio resulted in a significant increment in variance

explained by our model, ΔR2=0.04, ΔF(1, 57)=8.32,p=0.006,. As displayed in Fig. 3, among depressed adoles-cents, the simple slope for CRSQ-ratio score was positive andstatistically significantly different from zero, B=0.108,SE=0.034, t(57)=3.16 p=0.003, d=0.84, whereas it failedto reach significance among non-depressed adolescents,B=!0.067, SE=0.049, t(57)=!1.37, p=0.18, d=0.36.

Discussion

The current study is the first to our knowledge to investigatethe association of two distinct response styles — depressiverumination and distraction/problem-solving— on biologicalrecovery from stress in a sample of clinically depressed andnon-depressed adolescents. Consistent with our hypotheses,among the depressed adolescents, trait rumination was as-sociated with higher post-stressor cortisol concentration,controlling for peak cortisol output, whereas distractionand problem-solving was associated with lower post-stressor cortisol concentration. In the non-depressed group,in contrast, there was no evidence for a relation of eitherresponse style to cortisol recovery. Thus, our results impli-cate HPA axis dysregulation — specifically, impaired corti-sol recovery — as a potential neurobiological mechanismthat may help to explain why higher levels of ruminationover distraction are so toxic for adolescents with depression.These adolescents are exposed to higher levels of cortisolover time, and as a result may suffer more of the cumulativebiological wear and tear— or “allostatic load” — associatedwith chronic exposure to a heightened neuroendocrine stressresponse (McEwen and Wingfield 2003). Heightenedallostatic load, and the changes to neural structure andfunction that it brings about over time, are associated withdepressive illness (McEwen 2003).

Importantly, most of the depressed adolescents in oursample were very early in their course of depression (75 %

Table 2 Final regression model estimating recovery (sample D) fromdiagnostic group, response styles, and their interactions

Variable B (SE) t p d

Sample C 0.840 (0.074) 11.32 < 0.001 3.05

Age 0.025 (0.012) 2.10 0.04 0.57

Parental Hollingshead !0.017 (0.012) !1.37 0.18 0.37

Diagnostic group !0.059 (0.054) !1.10 0.28 0.30

CRSQ-RUM !0.006 (0.003) !1.69 0.10 0.46

CRSQ-DPS 0.001 (0.008) 0.12 0.91 0.03

Group X CRSQ-RUM 0.013 (0.005) 2.67 0.01 0.72

Group X CRSQ-DPS -0.023 (0.013) !1.74 0.09 0.47

CRSQ-RUM trait rumination; CRSQ-DPS trait distraction and problem-solving

Fig. 2 Predicted Sample Dvalues from CRSQ-RUM,controlling for Sample C,CRSQ-DPS, age, and parentaloccupational status computedusing simple slopes, fordepressed and non-depressedgroups. Low = 1 StandardDeviation below the mean;High = 1 Standard Deviationabove the mean

J Abnorm Child Psychol

were on their first episode). Therefore, the current resultssuggest that the relation of rumination to impaired HPA axisrecovery from stress may not require long-term, repeatedexposure to depressive illness and any resultant conse-quences (e.g., personality ‘scars’ from depressive episodes;Zeiss and Lewinsohn 1988). Indeed, high trait ruminatorsmay increase vulnerability to impaired HPA axis recoveryfrom stress even prior to the first onset of depression,particularly in the presence of other depression vulnerabilitymarkers. For instance, studies have demonstrated an associ-ation between maternal depression and disruptions in HPAaxis functioning, as indexed by elevated early morningcortisol concentration (Mannie et al. 2007) and dexametha-sone non-suppression (Young et al. 2006).

Importantly, our pattern of results held when responsestyles were considered independently and also when theywere combined in a ratio of rumination to distraction andproblem-solving. Indeed, substituting a single response styleratio in place of the two separate response style variablesresulted in only a 1 % reduction in explained variance (74 %versus 73 %) in the overall model. These findings lendsupport to Abela and colleagues’ (2007) contention that itis the relative use of these two strategies that has the

strongest implications for depression, and suggest specifi-cally that the extent to which depressed adolescents employa ruminative response style over a more adaptive distractionand problem-solving response style may be particularlyrelevant to biological recovery from a stressor.

Response styles theory proposes rumination and distrac-tion as adaptive and maladaptive responses to depressedmood, respectively, and proposes that they should havecorrespondingly opposing relations to depression symptoms(see Bagby and Parker 2001; Nolen-Hoeksema et al. 1994).However, the evidence supporting this conceptualization ismixed, with some studies finding a paradoxical positivecorrelation between rumination and distraction (e.g.,Schmaling et al. 2002), and others showing that distractionis unrelated to depression (e.g., Just and Alloy 1997; Nolen-Hoeksema et al. 1993). Our study provided support for theRST by showing, first, that rumination and distraction andproblem-solving were significantly negatively correlated,and, second, that the two constructs were associated withopposite patterns of cortisol recovery following a stressor.Given that disruptions in the HPA axis system are prospec-tively associated with chronicity (Goodyer et al. 2003) andrecurrences (Adam et al. 2010) in adolescent MDD, ourresults provide corroborative, neurobiological support forthe adaptive nature of distraction and problem-solving overrumination, and contributes to clarifying the roles of theseresponse styles in the pathology of depression.

The precise mechanisms mediating the relation of rumina-tion to impaired cortisol recovery following stress were not thefocus of the current study. Nevertheless, our results are con-sistent with neural mechanisms of rumination and distraction.Specifically, high rumination has been associated with a fail-ure to disengage attention from negative stimuli, mediated atleast in part by decreased PFC inhibition of amygdala activity(e.g., Siegle et al. 2002; Johnson et al. 2009). Conversely, hightrait distraction has been associated with greater PFC inhibi-tion of amygdalar activity and thus, enhanced ability to shift

Table 3 Final regression model estimating recovery from diagnosticgroup, response styles ratio, and their interaction

Variable B (SE) t p d

Sample C 0.844 (0.073) 11.62 <0.001 3.08

Age 0.022 (0.011) 1.99 0.05 0.53

Parental Hollingshead !0.019 (0.012) !1.60 0.12 0.42

Diagnostic Group !0.028 (0.054) !0.51 0.61 0.14

CRSQ-ratio !0.067 (0.049) !1.37 0.18 0.36

Group X CRSQ-ratio 0.175 (0.061) 2.88 0.01 0.76

CRSQ-ratio the ratio of trait rumination to trait distraction and prob-lem-solving

Fig. 3 Predicted Sample Dvalues from CRSQ-Ratio,controlling for Sample C, age,and parental occupation status,computed using simple slopes,for depressed and non-depressed by groups. CRSQ-Ratio the ratio of traitrumination to trait distractionand problem-solving; Low = 1Standard Deviation below themean; High = 1 StandardDeviation above the mean

J Abnorm Child Psychol

away from negative stimuli. For example, when psychiatrical-ly healthy women were instructed to cognitively reappraise anegative emotional experience, activation in several PFC re-gions (i.e., medial, dorsolateral, and ventrolateral) was asso-ciated with subsequent reduction in amygdala activity (Goldinet al. 2008). This corticolimbic circuit has direct implicationson functioning of the HPA axis. For example, Taylor et al.(2008) found that activation of the amygdala in response tothreatening faces was associated with heightened cortisolreactivity to the TSST, whereas Kern et al. (2008) found thatactivation of the medial prefrontal cortex was associated withdampened cortisol reactivity to a laboratory stressor. In thecontext of our study, then, high levels of rumination overdistraction may be associated with a failure to disengage fromstress, thereby prolonging the cortisol recovery period.

Our study had a number of strengths, including the use of astructured diagnostic interview to rigorously characterize oursample clinically, an ecologically valid stress induction pro-cedure, and a psychometrically strong and relevant measure ofrumination and distraction. The RSQ assesses ruminationfocused on “symptoms of distress and on the possible conse-quences and causes of these symptoms” (Nolen-Hoeksema etal. 2008, p. 400). Previous studies using college studentsamples have failed to find a relation between ruminationassessed with the RSQ and cortisol recovery following theTSST (Young and Nolen-Hoeksema 2001) or have reportedan overall blunted pattern of cortisol secretion in individualshigh, versus low, on a measure of trait rumination, in responseto an evaluative stressor (Zoccola et al. 2008). In contrast,studies in non-clinical undergraduate samples that focus onstress-focused rumination, using measures such as theEmotion Control Questionnaire (Roger and Najarian 1989),have found support for a relation between rumination anddelayed cortisol recovery (Zoccola et al. 2010). These resultssuggest that symptom-focused versus stress-focused rumina-tion are differentially relevant to stress recovery for individ-uals high versus low in depressive symptoms, respectively,and that these differences should be taken into considerationwhen choosing appropriate constructs for study.

The current findings need to be considered in light of thefollowing limitations. First, our study employed a modestsample size and the results should be replicated in a largersample. However, our interpretations are based on mediumto large effects that were consistent with our a priori hy-potheses, indicating that these findings are likely robust.Second, consistent with the TSST protocol, cortisol wasonly assessed at five points. In their meta-analysis,Dickerson and Kemeny (2004) found that individuals’ peakcortisol concentration was found in a range of time between20 and 40 min post-stressor. Thus, it is possible that highruminators may have reacted to a later portion of our stress-or, leading them to peak later than low ruminators, and at atime that was not captured by our assessment points. Future

research can address this issue by assessing cortisol concen-tration at a higher frequency, both during the reactivity andrecovery phases, than the TSST protocol calls for. Third, wedid not collect data on non-psychotropic medication use,including the use of oral contraceptive hormones, whichmay have affected cortisol functioning in our sample.Fourth, we did not confirm self-reported alcohol/substanceabuse by using urine screenings for these substances.Finally, the Distraction and Problem-Solving subscale usedin our analyses possessed modest internal consistency in thissample. This finding is consistent with Abela and colleagues(2007) and points to the need for further psychometricscrutiny of the CRSQ.

In summary, the current study provides compelling evi-dence in a clinically diagnosed sample of adolescents thathigh trait rumination relative to distraction/problem-solvingis significantly associated with later cortisol recovery fol-lowing stress. This relation was only present among adoles-cents with depression. As such, these results extendprevious literature on the relation between response stylesand cortisol recovery to a clinical sample. It is hoped that theresults of this study will stimulate further research integrat-ing neurobiological and cognitive factors to understandingthe etiology of MDD during adolescence — a key develop-mental period that has been relatively understudied. Thismay lead to a more fine-grained understanding of howbiological and cognitive factors work in concert with devel-opmental factors (e.g., hormones, normative intrapersonaland social stressors) to predict the course of MDD fromearly adolescence over the lifespan.

Acknowledgements Funding for this study was provided by a grantfrom the Ontario Mental Health Foundation (OMHF), awarded to Dr.Kate Harkness and Dr. KatherineWynne-Edwards. Additionally financialsupport was provided by a donation from Bombardier Inc. The authorsgreatly acknowledge Eric Bulmash, Alexandra Sutherland, NazaninAlavi, Cherie La Rocque and Dustin Washburn who performed thediagnostic interviews and Lindsey Lytle, Shannon Coyle, JordanTheriault, and Monica Haberl for their help with data coding andmanagement.

Conflict of Interest The authors declare that they have no conflict ofinterest.

References

Abela, J. R. Z., Rochon, A., & Vanderbilt, E. (2000). The children'sresponse styles questionnaire. Unpublished questionnaire.

Abela, J. R. Z., Vanderbilt, E., & Rochon, A. (2004). A test of theintegration of the response styles and social support theories ofdepression in third and seventh grade children. Journal of Socialand Clinical Psychology, 23, 653–674.

Abela, J. R. Z., Aydin, C. M., & Auerbach, R. P. (2007). Responses todepression in children: reconceptualizing the relation among re-sponse styles. Journal of Abnormal Child Psychology, 35, 913–927.

J Abnorm Child Psychol

Adam, E. K., Doane, L. D., Zinbarg, R. E., Mineka, S., Craske, M. G.,& Griffith, J. W. (2010). Prospective prediction of major depres-sive disorder from cortisol awakening responses in adolescence.Psychoneuroendocrinology, 35, 921–931.

Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing andinterpreting interactions. Newbury Park: Sage.

American Psychiatric Association. (2000). Diagnostic and statisticalmanual of mental disorders (4th ed., text rev.). Washington, DC:Author.

Bagby, R. M., & Parker, J. D. A. (2001). Relation of rumination anddistraction with neuroticism and extraversion in a sample ofpatients with major depression. Cognitive Therapy andResearch, 25, 91–102.

Beck, A. T., Steer, R. A., & Brown, G. K. (1996). Beck DepressionInventory manual (2nd ed.). San Antonio: Psychological Corporation.

Berg, S. J., & Wynne-Edwards, K. E. (2001). Changes in testosterone,cortisol and estradiol levels in men becoming fathers.Mayo ClinicProceedings, 76, 582–592.

Burke, H. M., Davis, M. C., Otte, C., & Mohr, D. C. (2005).Depression and cortisol responses to psychological stress: ameta-analysis. Psychoneuroendocrinology, 30, 846–856.

Ciesla, J. A., & Roberts, J. E. (2002). Self-directed thought andresponse to treatment for depression: a preliminary investigation.Journal of Cognitive Psychotherapy, 16, 435–453.

Dickerson, S. S., & Kemeny, M. E. (2004). Acute stressors and cortisolresponses: a theoretical integration and synthesis of laboratoryresearch. Psychological Bulletin, 130, 355–391.

Gibb, B. E., Grassia, M., Stone, L. B., Uhrlass, D. J., & McGeary, J. E.(2012). Brooding rumination and risk for depressive disorders inchildren of depressed mothers. Journal of Abnormal ChildPsychology, 40, 317–326.

Goldin, P. R., McRae, K., Ramel, W., & Gross, J. J. (2008). The neuralbases of emotion regulation: reappraisal and suppression of neg-ative emotion. Biological Psychiatry, 63, 577–586.

Goodyer, I. M., Herbert, J., & Tamplin, A. (2003). Psychoendocrineantecedents of persistent first-episode major depression in adoles-cents: a community-based longitudinal inquiry. PsychologicalMedicine, 33, 601–610.

Groschl, M., Rauh, M., & Dorr, H. G. (2003). Circadian rhythm ofsalivary cortisol, 17 alpha-hydroxyprogesterone and progesteronein healthy children. Clinical Chemistry, 49, 1688–1691.

Grove, W. M., Andreasen, N. C., McDonald-Scott, P., Keller, M. B., &Shapiro, R. W. (1981). Reliability studies of psychiatric diagnosis:theory and practice. Archives of General Psychiatry, 38, 408–413.

Harkness, K. L., Stewart, J. G., & Wynne-Edwards, K. E. (2011).Cortisol reactivity to social stress in adolescents: role of depres-sion severity and child maltreatment. Psychoneuroendocrinology,36, 173–181.

Hollingshead, A. B. (1975). Four factor index of social position.Unpublished manuscript. New Haven, CT: Yale University.

Holsboer, F. (2001). Stress, hypercortisolism, and corticosteroid re-ceptors in depression: implications for therapy. Journal ofAffective Disorders, 62, 77–91.

Johnson, M. K., Nolen-Hoeksema, S., Mitchell, K. J., & Levin, Y.(2009). Medial cortex activity, self-reflection and depression.Social Cognitive and Affective Neuroscience, 4, 313–327.

Just, N., & Alloy, L. B. (1997). The response styles theory of depres-sion: tests and an extension of the theory. Journal of AbnormalPsychology, 106, 221–229.

Kaufman, J., Birmaher, B., Brent, D., Rao, U., Flynn, C., Moreci, P., et al.(1997). Schedule for Affective Disorders and Schizophrenia forSchool-Age Children — Present and Lifetime Version (K-SADS-PL): initial reliability and validity data. Journal of the AmericanAcademy of Child and Adolescent Psychiatry, 36, 980–988.

Kern, S., Oakes, T. R., Stone, C. K., McAuliff, E. M., Kirschbaum, C.,& Davidson, R. J. (2008). Glucose metabolic changes in the

prefrontal cortex are associated with HPA axis response to apsychosocial stressor. Psychoneuroendocrinology, 33, 517–529.

Kessler, R. C., & Walters, E. E. (1998). Epidemiology of DSM-III-Rmajor depression and minor depression among adolescents andyoung adults in the national comorbidity survey. Depression andAnxiety, 7, 3–14.

Kirschbaum, C., Pirke, K. M., & Hellhammer, D. H. (1993). The triersocial stress test— a tool for investigating psychobiological stressresponses in laboratory setting. Neuropsychobiology, 28, 76–81.

Lara, M. E., Klein, D. N., & Kasch, K. L. (2000). Psychosocial pre-dictors of the short-term course and outcome of major depression:a longitudinal study of a nonclinical sample with recent-onsetepisodes. Journal of Abnormal Psychology, 109, 644–650.

Lopez-Duran, N. L., Kovacs, M., & George, C. J. (2009). Hypothalamic-pituitary-adrenal axis dysregulation in depressed children and adoles-cents: a meta-analysis. Psychoneuroendocrinology, 34, 1272–1283.

Luby, J. L., Heffelfinger, A., Mrakotsky, C., Brown, K., Hessler, M., &Spitznagel, E. (2003). Alterations in stress cortisol reactivity indepressed preschoolers relative to psychiatric and no-disorder com-parison groups. Archives of General Psychiatry, 60, 1248–1255.

Lyubomirsky, S., & Nolen-Hoeksema, S. (1993). Self-perpetuatingproperties of dysphoric rumination. Journal of Personality andSocial Psychology, 65, 339–349.

Lyubomirsky, S., & Nolen-Hoeksema, S. (1995). Effects of self-focusedrumination on negative thinking and interpersonal problem solving.Journal of Personality and Social Psychology, 69, 176–190.

Lyubomirsky, S., Caldwell, N. D., & Nolen-Hoeksema, S. (1998).Effects of ruminative and distracting responses to depressed moodon retrieval of autobiographical memories. Journal of Personalityand Social Psychology, 75, 166–177.

Mannie, Z. N., Harmer, C. J., & Cowen, P. J. (2007). Increased wakingsalivary cortisol levels in young people at familial risk of depres-sion. The American Journal of Psychiatry, 164, 617–621.

McEwen, B. S. (2003). Mood disorders and allostatic load. BiologicalPsychiatry, 54, 200–207.

McEwen, B. S., & Wingfield, J. C. (2003). The concept of allostasis inbiology and biomedicine. Hormones and Behavior, 43, 2–15.

McRae, K., Hughes, B., Chopra, S., Gabrieli, J. D. E., Gross, J. J., &Ochsner, K. N. (2010). The neural bases of distraction andreappraisal. Journal of Cognitive Neuroscience, 22, 248–262.

Nolen-Hoeksema, S. (1991). Responses to depression and their effectson the duration of depressive episodes. Journal of AbnormalPsychology, 100, 569–582.

Nolen-Hoeksema, S., & Davis, C. G. (1999). “Thanks for sharingthat”: ruminators and their social support networks. Journal ofPersonality and Social Psychology, 77, 801–814.

Nolen-Hoeksema, S., & Morrow, J. (1991). A prospective study ofdepression and post traumatic stress symptoms following a natu-ral disaster: the 1989 Loma Prieta earthquake. Journal ofPersonality and Social Psychology, 61, 115–121.

Nolen-Hoeksema, S., & Morrow, J. (1993). Effects of rumination anddistraction on naturally occurring depressed mood following loss.Cognition and Emotion, 7, 561–570.

Nolen-Hoeksema, S., Morrow, J., & Fredrickson, B. L. (1993).Response styles and the duration of episodes of depressed mood.Journal of Abnormal Psychology, 102, 20–28.

Nolen-Hoeksema, S., Parker, L., & Larson, J. (1994). Ruminativecoping with depressed mood following loss. Journal ofPersonality and Social Psychology, 67, 92–104.

Nolen-Hoeksema, S., Wisco, B. E., & Lyubomirsky, S. (2008).Rethinking rumination. Perspectives on Psychological Science,3, 400–424.

Rao, U., Hammen, C., Ortiz, L. R., Chen, L.-A., & Poland, R. E.(2008). Effects of early and recent adverse experiences on adrenalresponse to psychosocial stress in depressed adolescents.Biological Psychiatry, 64, 521–526.

J Abnorm Child Psychol

Ray, R. D., Ochsner, K. N., Cooper, J. C., Robertson, E. R., Gabrieli, J. D.E., & Gross, J. J. (2005). Individual differences in trait ruminationand the neural systems supporting cognitive reappraisal. Cognitive,Affective, & Behavioral Neuroscience, 5, 156–168.

Robinson, L. A., & Alloy, L. B. (2003). Negative cognitive styles andstress-reactive rumination interact to predict depression: a pro-spective study. Cognitive Therapy and Research, 27, 275–292.

Roger, D., & Najarian, D. (1989). The construction and validation of anew scale for measuring emotional control. Personality andIndividual Differences, 10, 845–853.

Schmaling, K. B., Dimidjian, S., Katon, W., & Sullivan, M. (2002).Response styles among patients with minor depression and dysthy-mia in primary care. Journal of Abnormal Psychology, 111, 350–356.

Shirtcliff, E. A., Granger, D. A., Schwartz, E. B., & Curran, M. J. (2001).Use of salivary biomarkers in biobehavioral research: cotton-basedsample collection methods can interfere with salivary immunoassayresults. Psychoneuroendocrinology, 26, 165–173.

Siegle, G. J., Steinhauer, S. R., Thase, M. E., Stenger, A., & Carter, C. S.(2002). Can’t shake that feeling: event-related fMRI assessment ofsustained amygdala activity in response to emotional information indepressed individuals. Biological Psychiatry, 51, 693–707.

Tanner, J. M. (1962). Growth and adolescence (2nd ed.). Springfield:Thomas.

Taylor, S. J., Whincup, P. H., Hindmarsh, P. C., Lampe, F., Odoki, K., &Cook, D. G. (2001). Performance of a new pubertal self-assessmentquestionnaire: a preliminary study. Paediatric and PerinatalEpidemiology, 15, 88–94.

Taylor, S. E., Burkland, L. J., Eisenberger, N. I., Lehman, B. J.,Hilmert, C. J., & Lieberman, M. D. (2008). Neural bases of

moderation of cortisol stress responses by psychosocialresources. Journal of Personality and Social Psychology,95, 197–211.

Toothaker, L. E. (1991). Multiple comparisons for researchers.Newbury Park: Sage Publications.

Watson, D. & Clark, L. A. (1991). The Mood and Anxiety SymptomQuestionnaire. Unpublished manuscript, Southern MethodistUniversity, Dallas, TX.

Watson, D., Clark, L. A., Weber, K., Assenheimer, J. S., Strauss, M. E.,& McCormick, R. A. (1995). Testing a tripartite model: II.Exploring the symptom structure of anxiety and depression instudent, adult and patient samples. Journal of AbnormalPsychology, 104, 15–25.

Young, E. A., & Nolen-Hoeksema, S. (2001). Effect of ruminations on thesaliva cortisol response to a social stressor.Psychoneuroendocrinology,26, 319–329.

Young, E. A., Vazquez, D., Jiang, H., & Pfeffer, C. R. (2006). Salivacortisol and response to dexamethasone in children of depressedparents. Biological Psychiatry, 60, 831–836.

Zeiss, A. M., & Lewinsohn, P. M. (1988). Enduring deficits afterremission of depression: a test of the scar hypothesis. BehaviourResearch and Therapy, 26, 151–158.

Zoccola, P. M., Dickerson, S. S., & Zaldivar, F. P. (2008). Ruminationand cortisol responses to laboratory stressors. PsychosomaticMedicine, 70, 661–667.

Zoccola, P. M., Quas, J. A., & Yim, I. S. (2010). Salivary cortisolresponses to a psychosocial laboratory stressor and later verbalrecall of the stressor: the role of trait and state rumination. Stress,13, 435–443.

J Abnorm Child Psychol