Abstract
Although hedonic capacity is diminished during clinical depression, it is unclear whether that deficit constitutes a risk factor or persists after depression episodes remit. To examine these issues, adolescents with current/past major depression (probands; n = 218), never-depressed biological siblings of probands (n = 207), and emotionally well controls (n = 183) were exposed to several positively valenced probes. Across baseline and hedonic probe conditions, controls consistently reported higher levels of positive affect than high-risk siblings, and siblings reported higher levels of positive affect than probands (remitted and depressed probands’ reports were similar). Extent of positive affect across the protocol predicted adolescents’ self-reports of social support network and parental reports of offspring’s use of various adaptive mood repair responses in daily life. Attenuated hedonic responding among youths remitted from depression offers partial support for anhedonia as a trait, whereas its presence among never-depressed high-risk siblings argues for anhedonia as a potential diathesis for clinical depression.
Diminished ability to experience pleasure and joy (anhedonia) has been long recognized by clinicians as a cardinal feature of severe depression (e.g., Kraepelin, 1921); starting in 1980, its importance also was acknowledged by the official operational diagnostic criteria for depressive disorders (American Psychiatric Association, 1980). Usually joyous events, such as an unexpected present, reaching a coveted goal, being praised, watching a humorous show, or witnessing a beautiful sunset, for example, fail to elicit a sense of pleasure or happiness when a person is suffering from depression. However, it also has been posited that hedonic dysfunction, or a relatively low level of positive affectivity, is not merely a symptom of depression but also a trait that might predate the depressive syndrome, signal an elevated risk for it, or persist to some extent even after the depression has remitted (Hasler, Drevets, Manji, & Charney, 2004; Loas, 1996; Meehl, 1975). Recent research has underscored that depressed people’s reports of diminished enjoyment and lack of pleasure in everyday life is one facet of anhedonia, which also includes dysfunction in reward responsiveness and in related motivational, information processing, and decision making neural circuitries (see reviews by Pizzagalli, 2014; Treadway & Zald, 2011), and may identify a depression endophenotype (Hasler et al., 2004).
Although the nature and consequences of anhedonia have been extensively studied, basic questions remain about its stability across various phases of depression and its role as a risk factor for affective psychopathology. Furthermore, investigations of clinical samples have mostly involved adults, although depression in younger populations is a pressing problem (Kessler & Walters, 1998). All in all, however, numerous laboratory studies involving the presentation of positively toned stimuli have confirmed the presence of hedonic dysfunction during episodes of clinical depression. Compared with healthy controls, individuals suffering from depression (typically major depressive disorder, MDD) evidence blunted or less intense reactions to pleasant stimuli and potential or actual rewards (for reviews, see Bylsma, Morris, & Rottenberg, 2008; Eshel & Roiser, 2010; Zhang, Chang, Guo, Zhang, & Wang, 2013). For example, depressed adult patients usually report lower levels of happiness or enjoyment than do controls after viewing standardized positive visual prompts, such as pictures of pleasant scenes or smiling faces (Dunn, Dalgleish, Lawrence, Cusack, & Ogilvie, 2004; Sloan, Strauss, Quirk, & Sajatovic, 1997; Sloan, Strauss, & Wisner, 2001) or after reading highly positively toned scripts (Horner et al., 2014) and display reduced behavioral responses to monetary reward (Henriques & Davidson, 2000; Pizzagalli, Iosifescu, Hallett, Ratner, & Fava, 2009). Potential monetary reward also elicited lower levels of “joviality” among depressed college students than control peers (McFarland & Klein, 2009). In a similar vein, adolescents with various depression diagnoses were less likely to seek rewards during a gambling task than were healthy peers (Rawal, Collishaw, Thapar, & Rice, 2013). Although some studies have reported normative self-rated hedonic experience among adults with MDD (e.g., Osuch et al., 2009), a meta-analysis found that depressive episodes are associated with blunted positive affective reactions and impaired reward-related behaviors (Bylsma et al., 2008).
In contrast, fewer studies have examined individuals during periods of recovery from depression and the findings have been equivocal regarding whether, and to what extent, hedonic dysfunction persists after the depression episode remits. For example, college students remitted from depression and healthy control peers reported comparable levels of positive affect while anticipating monetary rewards (McFarland & Klein, 2009). Similarly, self-rated pleasure in response to chocolate failed to distinguish recovered depressed and never-depressed adults (McCabe, Cowen, & Harmer, 2009). Recovered depressed and control subjects also had similar behavioral responses to pictures of happy facial expressions (Kerestes et al., 2012). On the other hand, some studies did find adverse residual effects of depression on affective responding. Namely, self-ratings and clinical evaluations of middle-aged outpatients with remitted depression indicated that they had lower hedonic capacity than did community controls (Di Nicola et al., 2013). Likewise, in two separate studies, adults with remitted depression displayed blunted behavioral responses to monetary incentives and social praise, relative to healthy controls (Pechtel, Dutra, Goetz, & Pizzagalli, 2013). The persistence of hedonic dysfunction after remission of a depression episode supports the proposition that anhedonia has trait-like features, which has ramifications for prevention efforts. Thus, one goal of our study was to address this issue in a clinical sample of youths with prior histories of MDD.
A further critical issue to be resolved is whether hedonic dysfunction is present prior to the first episode of depressive illness and thus qualifies as a bona fide risk factor for it. The optimal approach to this topic is to establish the presence of hedonic impairment among not-yet-depressed individuals who are known to be at elevated risk for depression (usually by virtue of familial history) and then use follow-up to determine its predictive value for onset of clinical depression. However, studies of young, high-risk offspring (typically of depressed mothers) have yielded inconsistent results about the presence of hedonic impairment and the children have not been followed for a sufficiently long time to model risk factors for first onset depressive disorder. For example, in one study, evocative laboratory tasks elicited lower rates of positive affect displays among high-risk, 3-year-old offspring of depressed mothers than low-risk offspring (Durbin, Klein, Hayden, Buckley, & Moerk, 2005). A longitudinal developmental study also reported consistently lower levels of positive affect displays in the laboratory among high-risk, preschool, and young school-age children compared with low-risk offspring (Olino et al., 2011). However, in two other studies of preschool-age offspring, maternal depression histories were unrelated to children’s positive affect (Feng, Shaw, Skuban, & Lane, 2007; Olino, Klein, Dyson, Rose, & Durbin, 2010) and in a small sample of adolescents, those at high- versus low-risk reported comparable levels of positive affect in natural settings (Olino et al., 2014). Likewise, high- and low-risk 16- to 21-year-olds did not differ in ratings of pleasantness after tasting or viewing pictures of chocolate (McCabe, Woffindale, Harmer, & Cowen, 2012). The inconsistent findings could partly reflect that affect displays (along with behavioral regulation) across childhood are subject to notable developmental mediation and moderation (summarized in Olino et al., 2011), and such developmental effects probably varied across the studies. In the present investigation, we therefore focused on older high-risk subjects, namely, never-depressed adolescents whose brothers or sisters had histories of MDD. Depression in siblings is a robust risk factor for eventual depression in their yet unaffected brothers and sisters (e.g., Farmer et al., 2000), who have a cumulative probability of a first MDD episode of .42 by the time they are young adults (Ryan et al., 1992). However, with the exception of studies of twin pairs (e.g., Bogdan & Pizzagalli, 2009), to our knowledge, there are no laboratory investigations of positive affectivity involving young, high-risk siblings.
Hedonic capacity or positive affectivity has been shown to facilitate an array of important functions, including subjective well-being, attention allocation, and executive processes, as well as intra- and extrafamilial social interactions (e.g., Davis, Suveg, & Shaffer, 2015; Fredrickson, 2001; Grol, Koster, Bruyneel, & De Raedt, 2014; Smillie, Wilt, Kabbani, Garratt, & Revelle, 2015; Wadlinger & Isaacowitz, 2006; Yang & Yang, 2014). The positive affect system also facilitates approach behaviors focused on resources and therefore is associated with a range of adaptive consequences (e.g., Watson & Naragon, 2009). Conversely, attenuated hedonic capacity may have various adverse sequelae, including a disruption of social relationship and difficulties in maintaining social networks. Deficits in positive affectivity also may compromise emotion regulation repertoires (Kovacs & Lopez-Duran, 2010). For instance, in the presence of chronically low hedonic capacity, a person may be short on regulatory responses that involve positive affect (like recalling happy memories) as a way to lessen current distress. Thus, the final goal of the present study was to examine the functional significance of laboratory-based indices of positive affectivity.
In summary, there is compelling evidence that depression compromises hedonic capacity. However, questions have remained about whether such impairment predates depression among vulnerable people and thus operates as a true risk factor, and if hedonic dysfunction persists after depression. Furthermore, if empirical studies of affectivity are to have meaningful clinical implications, strengths or deficits identified in the laboratory should have functional correlates in daily life. We addressed these issues among adolescents with histories of childhood onset MDD who were either in remission or currently depressed (henceforth referred to as probands), never-depressed full biological (high-risk) siblings of probands, and healthy controls with no history of psychopathology. After exposure to a sampling of hedonic probes, namely, solvable puzzles, surprising receipt of a desired prize, and a film clip displaying slapstick comedy (presented in random order within a more extensive protocol), subjects repeatedly rated their own positive affect. Separately, youths also provided information about their social support networks, and parents independently reported whether offspring typically used positive affect-related regulatory responses to sadness in daily life.
We hypothesized that, regardless of whether depressed or in remission, probands will evidence lower hedonic capacity across all probes than controls. Based on Meehl’s (1975) model, we further hypothesized that high-risk, never-depressed siblings also will report consistently lower levels of positive affect than controls. We expected the differences across groups to be evident in terms of the absolute magnitudes of positive affect (raw scores of affect intensity) and the extent of reactivity to each probe (change scores). Finally, we hypothesized that the magnitude of positive affect elicited in the laboratory will have functional significance and predict the extent of youths’ social support networks (by self-report) and their use of mood repair responses that involve positive affect (by parental report).
Method
Subjects
Subjects included probands, whose histories of childhood onset MDD were established in a prior study (e.g., Baji et al., 2009; Tamás et al., 2007), never-depressed biological siblings of probands, and emotionally well controls with no history of major psychiatric disorders. Probands and siblings were a subset of a larger national sample in Hungary, enrolled in a molecular genetic study (the archival study) from approximately 1997 to 2006 (e.g., Burcescu et al., 2006; Dempster et al., 2009). Probands for the archival study were recruited through various child mental health facilities if they (a) had a current or recent Diagnostic and Statistical Manual of Mental Disorders (DSM-IV; American Psychiatric Association, 2000) diagnosis of depressive disorder, (b) were 7 to 14 years old at initial screen, (c) were not mentally retarded and had no major systemic medical disorder, (d) had at least one biological parent who could participate, and (e) had at least one full biological 7- to 18-year-old sibling (within ±3 years of age).
The current investigation of emotion reactivity and regulation enrolled 224 probands and 214 never-depressed siblings from the archival study, aged 11 to 18 years, who lived in the proximity of three research hubs. In this article, we report on n = 218 probands with MDD histories (6 with bipolar disorder were excluded) and 207 unaffected siblings of probands (7 siblings of bipolar probands were excluded) from a total of 297 families. A total of 102 families contributed a proband and one or more unaffected sibs, 116 families contributed only a proband (when a sibling was not available), and 79 families contributed one or more unaffected sibs (when a proband was not available). Probands were 17.0 years old on average (SD = 1.4), and 64.2% were boys; siblings were 15.9 years old (SD = 2.12) on average, and 44.9% were boys. Parental education served as an indicator of higher socioeconomic standing: 9.2% of proband mothers and 8.2% of fathers had more than secondary school education; for the partly overlapping sibling sample, rates of highly educated mothers and fathers were 17.8% and 11.9%, respectively.
Probands had the onset of their first MDD episode when they were 9.07 years old, on average (SD = 1.90 years). At enrollment in the current study, 56.0% had one MDD episode, 31.2% had 2 episodes, and 12.9% had 3 or more episodes. Furthermore, 32 probands (14.7%) were in a depressive episode, whereas 186 (85.3%) had recovered from their last MDD episode. Recovery was operationally defined as being essentially free of depressive symptoms for at least two consecutive months (Kovacs, Feinberg, Crouse-Novak, Paulauskas, & Finkelstein, 1984). Rates of comorbid psychiatric disorders were consistent with the literature (Birmaher et al., 1996): 39.0% of the probands had some type of anxiety disorder, and 36.24% had some behavioral disorder (e.g., attention-deficit/hyperactivity disorder, oppositional defiant disorder); in all, 70.2% had one or more comorbid major psychiatric disorder.
The 183 control subjects (age M = 16.1 years, SD = 2.13 years; 64.5% boys) were recruited for the present study to approximate the sex and age distribution of probands; they were identified in public elementary and secondary schools in the three cities in which most of the probands resided. Enrollment required lifetime history free of any major psychiatric or systemic medical disorder, having a parent willing to participate, and having at least one sibling (to approximate probands’ family units). For these youths, 48% of the mothers and 39% of the fathers had more than secondary school education.
As can be seen in Table 1, probands were approximately 1 year older than never-depressed siblings and controls, F(3, 565) = 15.31, p < .001,
Selected Demographic Characteristics and Baseline Self-Ratings of Affect
Note: Values are M ± SD unless otherwise indicated. CDI = Children’s Depression Inventory; PANAS = Positive and Negative Affect Schedule.
F or χ2, adjusted for family effect. b
p < .001. Group means with common subscripts did not differ significantly from one another in post hoc tests (Tukey, ps > .05).
The current research study was approved by the institutional review boards of the University of Pittsburgh and the Hungarian clinical research sites. Parents provided written informed consent, and depending on their ages, young subjects provided either assent or consent. All procedures, schedules, rating scales, and instruments used in this study were first developed in English, translated into Hungarian, and then back translated by bilingual child psychiatrists and clinical psychologists. An iterative procedure was used to resolve any discrepancies between original and back-translated versions of documents.
Clinical and diagnostic assessment
Caseness of probands was originally established in the archival study via standardized psychiatric diagnostic evaluations using a semistructured interview by trained interviewers. The information was gathered via the Interview Schedule for Children and Adolescents: Diagnostic version (ISCA-D). The ISCA-D, described previously (Baji et al., 2009; Kiss et al., 2007; Tamás et al., 2007), uses DSM-IV criteria (American Psychiatric Association, 2000). The clinician first interviews the parent about the child, and then separately the child about himself or herself, and then assigns an overall rating to each symptom. Those ratings, along with the child’s clinical history, determine whether DSM-IV criteria for any of the covered disorders are met. The diagnostic process also includes operational definitions of onset and recovery and psychiatric comorbidity (Kovacs, Feinberg, Crouse-Novak, Paulauskas, & Finkelstein, 1984; Kovacs, Feinberg, Crouse-Novak, Paulauskas, Pollock, & Finkelstein, 1984); diagnoses were always finalized by consensus among pairs of senior raters. Evaluations were conducted by child psychiatrists and psychologists trained in the ISCA-D, who had met predefined symptom rating goals on individually supervised interviews and reliability ratings against videotaped “gold-standard” diagnostic interviews. We have previously reported acceptable interrater reliability coefficients on ISCA-D symptom ratings (Tamás et al., 2007). A follow-up version of the ISCA-D, which queries about symptoms and adjustment since the prior evaluation, was used to establish the youth’s diagnostic status at the time of the current study. Sociodemographic information about the youths was obtained from the parents in accordance with a structured set of questions.
Experimental procedures
After the youth was familiarized with the experimental procedures and equipment, he or she was asked to rank order, from the least to the most preferred, 7 age-appropriate prizes (e.g., high-tech earphones, gift certificate to a multimedia store, combinations of snacks, all of comparable monetary value). The subject then was told that, at some point, the “computer will randomly decide” which prize he or she would receive. However, unbeknownst to the youth, he or she was to receive his or her most preferred prize. The protocol, which was introduced to subjects as an experiment to better understand how children and adolescents react to different tasks and stimuli, included probes of physiological and psychological reactions to affectively pleasant, unpleasant, and stressful stimuli/tasks, and took about 1 hr, with a 5- to 10-min midpoint break. To minimize potential order-of-task effects, subjects were randomized to predefined task sequences.
At the start of the protocol, the subject was seated in a comfortable chair in front of a table and was asked to provide affect ratings (see Positive Affect Self-Rating (PASR) during the protocol) for the first time (baseline). After that, he or she was connected to sensors for physiological recording (data not reported here) and then engaged in the specified series of tasks. In this article, we focus on affect self-ratings from baseline (pre-stimuli) and after three experimental probes that were designed to induce positive affect: presentation of solvable puzzles, provision of a preferred prize, and watching a funny film clip. The solvable puzzle and getting the prize were presented in the first 30 min of the protocol to 50% of the subjects and in the second half of the protocol to the other 50% of cases (interspersed with other tasks, not reported here). The happy film clip was the last stimulus for all cases.
Positive affect probes
The Happy Film Clip, one of three conditions used to induce positive affect, was a 186-second segment from Mr. Bean, a slapstick-style comedy, which has been used with adults (e.g., Joormann & Siemer, 2004; Rottenberg, Kasch, Gross, & Gotlib, 2002). Film clips also have been shown to elicit the desired valenced emotion in children (von Leupoldt et al., 2007). We nonetheless piloted the Mr. Bean clip with 240 Hungarian youth (53% girls; 62% of the sample aged 14 to 18 years, recruited from 6 schools in two cities). Youngsters rated each of 5 emotion words (from 0 to 8) after watching Mr. Bean, as well as neutral and sad film clips, presented in a randomized order. The Mr. Bean clip elicited robust reports of happiness (M = 6.45, SD = 1.77) but relatively low reports of nontarget emotions (e.g., anger; M = 1.35, SD = 1.15), as well as significantly higher ratings of happiness than the neutral, t(203) = 13.25, p < .001, dz = 0.93, and sad film clips, t(203) = 19.15, p < .001, dz = 1.34.
Solvable puzzles
Puzzles have been used as mood induction tools with youths (e.g., Berger, Miller, Seifer, Cares, & Lebourgeois, 2012; Cole et al., 2007; Hokoda & Fincham, 1995; Nolen-Hoeksema, Wolfson, Mumme, & Guskin, 1995). Our solvable puzzles were a series of computerized tasks in which the youth had to replicate a particular configuration of numbered tiles by sliding them around. The five puzzles (lasting for 60 s) were programmed to present a bit of a challenge but be relatively easy to complete. A training phase was provided.
Getting “a prize.”
As another probe of positive affectivity, youths were unexpectedly provided with their preferred prize. We pilot tested the desirability of 25 potential prizes (roughly equivalent in monetary value) with younger (9 to 13 years old) and older (14 to 18 years old) volunteer samples of convenience. The final list of 7 prizes (e.g., earphones, snacks) was based on the rankings these subjects provided.
Psychometric questionnaires
Adolescents completed the widely used 20-item Positive and Negative Affect Schedule (PANAS; Watson, Clark, & Tellegen, 1988), which we edited slightly for clarity with younger age groups. The factorial and concurrent validities of the PANAS were reported in the original article by Watson et al. (1988); further psychometric information, including on construct validity, has been reported by Crawford and Henry (2004), among others. PANAS items were rated (using a 1–5 Likert-type scale) for the past 2 weeks and were averaged to create the positive affect and negative affect scores. Level of depressive symptoms was quantified by the Children’s Depression Inventory–2 (CDI-2), a widely used self-rated questionnaire for youngsters, with documented concurrent, construct, and predictive validities (Kovacs & MHS Staff, 2011). The CDI also was completed to reflect the prior 2 weeks. We assessed adolescents’ social support networks via the 12-item Multidimensional Scale of Perceived Social Support (MSPSS; Zimet, Dahlem, Zimet, & Farley, 1988), which queries about friends, family, and others. MSPSS items are rated on a 7-point Likert-type scale and reflect both network size and support availability; the instrument has been used cross-culturally (e.g., Steese et al., 2006; Terzi-Unsal & Kapci, 2005; Zimet, Powell, Farley, Werkman, & Berkoff, 1990) and has good psychometric properties, including construct and concurrent validities (e.g., Kazarian & McCabe, 1991; Zimet et al., 1988; Zimet et al., 1990). Pubertal stage was determined via self-report on a questionnaire that uses drawings to depict stages of puberty along the lines established by Tanner (Morris & Udry, 1980). The validity of such self-rated pubertal staging against physicians’ assessments has been established (Duke, Litt, & Gross, 1980; Morris & Udry, 1980) as well as its correlation with chronological age (e.g., Angold, Costello, & Worthman, 1998).
Finally, parents completed a trait questionnaire of their offspring’s mood repair response repertoires (Feelings and My Child), which surveys a range of helpful (adaptive) and not helpful (maladaptive) regulatory responses to sadness and distress, and has good psychometric properties (Bylsma et al., in press; Gentzler, Santucci, Kovacs, & Fox, 2009). Using the six items that reflect adaptive responses involving the self-mobilization of positive affect (e.g., when my child feels sad, he/she tries to think of happier times in the past), we created a subscale. Total scores for this six-item scale ranged from 0 to 12, with higher scores indicating greater reliance on adaptive regulatory responses that mobilize positive affect to counter sadness. Based on n = 607 parental reports, the Cronbach’s alpha of .72 revealed acceptable internal consistency. The construct validity of this six-item questionnaire is supported by its significant correlation with youths’ self-ratings on the Positive Affect scale of the PANAS (Watson et al., 1988), r(397) = .22, p < .001, and their self-ratings on the Distraction subscale of the Children’s Response Style Questionnaire (Ziegert & Kistner, 2002), r(366) = .17, p < .01, which taps another dimension of adaptive emotion regulation. The scale’s discriminant validity is evidenced by the finding that depressed probands in this study (n = 32) had significantly lower scores than did a subset of controls (n = 32), matched for age and sex (Mdepressed = 2.97, SD = 2.12; Mcontrol = 5.53, SD = 1.88), t(62) = 5.12, p < .001, d = 1.28.
Positive Affect Self-Rating (PASR) during the protocol
At baseline, and after each hedonic probe, subjects indicated on a 0 (not at all) to 7 (very much) Likert-type scale the extent to which several discrete feeling states characterized them. The items included three manifestations of positive affect: happy, interested, and enthusiastic (along with items pertaining to negative affect). Because the three items were significantly intercorrelated at each measurement point (Cronbach’s alphas from .75 to .86), the scores were averaged at each rating, which yielded the Positive Affect Self-Rating (PASR) composite, used in statistical analyses.
Statistical analyses
Variables were compared across groups using ANOVAs, t tests, or χ2 tests, as appropriate. Mixed effects, repeated measures analysis of covariance (rANCOVA) was used to test for group (depressed proband, remitted proband, sibling, control) by task (baseline, prize, solvable puzzle, happy film) effects on PASR scores, while accounting for dependence between siblings. In these models, group was a between-subjects factor, experimental task was a within-subjects factor, age (in years), and sex were covariates, and family was a random effect. Nonsignificant covariate-by-task interactions were removed, and the best-fitting repeated measures’ covariance matrix was chosen (Gueorguieva & Krystal, 2004). The method of Kenward and Roger (Gomez, Schaalje, & Fellingham, 2005) was used to compute fixed effect denominator degrees of freedom, which, along with the corresponding F statistic, was used to compute the effect size, namely, the partial
Results
Characteristics of subject groups
Table 1 includes the mean and standard deviation affect ratings across subject groups at baseline. PANAS positive affect scores revealed a group effect, F(3, 548) = 12.09, p < .001,
Not surprisingly, probands reported higher levels of negative affect, in general (PANAS negative affect, Table 1), than did siblings or controls, F(3, 568) = 27.99, p < .001,
Manipulation check
Changes in mean PASR scores between baseline and each task indicated that the hedonic probes elicited the intended effect: getting a prize was associated with the greatest increase in positive affect (M = 0.42, SD = 1.02), t(607) = 10.22, p < .001, d = 0.41, followed by watching the happy film clip (M = 0.23, SD = 1.15), t(607) = 4.90, p < .001, d = 0.20. The solvable puzzles elicited trend-level increases in positive affect l (M = 0.08, SD = 1.06), t(607) = 1.78, p = .08, d = 0.07. Very few subjects endorsed absolutely no positive affect (composite score = 0) in response to the probes. Specifically, 3 depressed probands, 12 remitted probands, 3 siblings, and 2 controls reported no positive affect to one or more hedonic probes, with probands being overrepresented as nonresponders, χ2(3) = 11.15, p = .01, versus siblings: ORdepressed = 7.04, ORremitted = 4.69, ORcontrols = 0.75.
As a further manipulation check, we examined PASR ratings subsequent to other tasks interspersed throughout the protocol and found that they were consistently associated with decreases in positive affect. For example, relative to levels of positive affect at baseline, PASR scores declined after a physical challenge, face exposed to ice (M = –0.31, SD = 0.83), t(307) = −6.55, p < .001, d = −0.37, a task involving unsolvable puzzles (M = –0.77, SD = 1.09), t(607) = −17.48, p < .001, d = −0.71, and watching a sad film clip (M = –0.87, SD = 1.10), t(607) = −19.55, p < .001, d = −0.79. Thus, PASR scores reflect changing levels of positive affect as a function of experimental task.
Patterns of positive affect experience
The mean PASR scores are graphically portrayed in Figure 1. Consistent with our first hypothesis, a repeated measures ANCOVA (with age and sex as covariates) yielded effects for task, F(3, 606) = 12.87, p < .001,

Self-rated positive affect composite score across baseline and three hedonic probes. Effects are from the full factorial model of group and time, adjusting for age and sex fixed effects, family random effect, and within-subject unstructured correlation. The probes were presented in randomized order across cases.
To examine positive affect reactivity (change from baseline to each task), PASR scores were modeled via rANCOVA, with group (four levels) as the independent variable, and baseline PASR rating, age, and sex as covariates. As expected, there was a main effect of task, F(2, 607) = 11.43, p < .001,
To examine whether the group differences in positive affectivity were due to residual depression symptoms, we repeated the analyses with CDI scores added to the models. Controlling for depression symptoms, the group effect remained significant in the rANCOVA of PASR scores, F(3, 570) = 15.34, p < .001,
We also considered whether the results were affected by pubertal stage, comorbid anxiety and externalizing disorders, and the interaction of sex by group. However, neither pubertal stage, comorbid anxiety disorder, nor the sex-by-group interaction term emerged as statistically significant in the repeated measures ANCOVA of the raw PASR scores or the reactivity scores. Comorbid externalizing disorder was significant in the rANCOVA of PASR scores, F(1, 607) = 10.34, p < .01,
Hedonic responding in the laboratory and daily functioning
As hypothesized, lower levels of positive affect in response to experimental probes predicted the extent of adolescents’ perceived social support networks and parental reports of the adolescents’ utilization of mood repair responses that mobilize positive affect. Adjusting for age, sex, and family effect in a random effects model, perceived social support was predicted by average level of positive mood across the hedonic probes (b = 0.08, SE = 0.02), F(1, 596) = 10.73, p < .01,
Discussion
We found that currently and previously depressed adolescent probands, as well as their never-depressed high-risk adolescent siblings, reported diminished hedonic responses to a variety of laboratory probes of positive affect, as compared with emotionally healthy control peers. In fact, the three groups of youths differed in a linear fashion in levels of positive affect at both baseline and across the hedonic probes: controls reported the highest levels of positive affect, followed (in decreasing order) by high-risk siblings, and then by probands. Remitted and depressed probands reported comparable levels of positive affect. The results were similar when we examined changes in affect ratings, which are believed to reflect affective flexibility (Bylsma et al., 2008). Thus, personal history of clinical depression and familial history of depression in adolescence both appear to diminish the intensity of subjective positive affect and attenuate the magnitude of that affective response. Importantly, the experimental probes modeled some daily life situations that presumably trigger positive emotion and mood, such as receiving a desired object, solving a mild challenge, and watching a funny movie. The fact that probands and high-risk siblings exhibited some degree of hedonic dysfunction in response to all probes suggests that they may not be able to capitalize on the varied opportunities in daily life to experience positive affect. Crucially, across-group differences in positive affect persisted when we controlled for levels of current depression symptoms. Moreover, the extent of reported positive affect in the laboratory had real-world functional significance; it predicted adolescents’ perceived social support networks and their habitual use of adaptive mood repair responses that involve positive affect.
The finding of hedonic dysfunction among our remitted probands is in line with reports of attenuated positive affectivity among adult patients recovered from depression (e.g., Di Nicola et al., 2013), and extends the literature to clinical samples of adolescents. The fact that these adolescents, although no longer depressed, still experienced lower levels of positive affect than controls did does support the thesis that hedonic dysfunction has trait-like features. Given reports of impaired mood repair in these very same subjects (Bylsma et al., 2015; Kovacs et al., 2015), remitted major depression in youths appears to be associated with various dysfunctions in affective processing.
Never-depressed siblings reported higher levels of positive affect than probands did, but yet failed to match the hedonic responses of control peers. Thus, the attenuated positive affectivity that tends to characterize young offspring at elevated risk for depression (e.g., Dietz et al., 2008; Durbin et al., 2005; McMakin, Burkhouse, et al., 2011) also is detectable in older youths who were identified via a different definition of familial risk. Given that high-risk siblings were also found to display impairment in the tone and accessibility of positive autobiographical memories (Begovic et al., 2015) and that high-risk adolescent offspring evidenced attenuated positive affectivity during mother-child interactions (Dietz et al., 2008; McMakin, Burkhouse, et al., 2011), familial risk for depression may involve multiple forms of hedonic dysfunction. Furthermore, the finding of lowered levels of positive affectivity among never-depressed youths at elevated risk for eventual depression supports the proposition that impaired hedonic capacity is a plausible risk factor or diathesis for clinical depression (Meehl, 1975; Watson & Naragon, 2009). Although the actual risk of an episode of clinical depression posed by reduced hedonic capacity in our sibling sample is yet to be documented, low positive affectivity in other samples of youths has predicted subsequent depression symptoms (Lonigan, Phillips, & Hooe, 2003; Wetter & Hankin, 2009). And low reward seeking (defined via behavioral indices) reportedly prognosticated onset of depressive disorder among high-risk adolescents (Rawal et al., 2013). Thus, there is accumulating evidence that hedonic impairment may serve as a diathesis for clinical depression.
Attenuated hedonic capacity may increase the risk of a depression episode through various psychosocial and physiological routes (e.g., Dockray & Steptoe, 2010; Forbes & Dahl, 2012). For instance, based on our findings, low hedonic capacity may contribute to depression risk by adversely affecting social support networks. Social support networks, in turn, play an important role in buffering the effects of stress events that may trigger depression. Our findings also suggest the possibility that lower levels of positive affect may compromise the availability of certain mood repair responses, such as the recall of positive memories to attenuate sadness (Isen, 1985; Josephson, Singer, & Salovey, 1996). In turn, a constricted mood repair response repertoire may make it more difficult to manage the affective consequences of depressogenic triggers (Kovacs & Lopez-Duran, 2010).
Because both the ecological validity and practical implications of laboratory probes of affect have been questioned, it is particularly notable that we documented various extralaboratory functional correlates of experimentally induced positive affect. Youths who reported higher levels of positive affect in the laboratory had better social support networks and were more likely to deploy adaptive mood repair responses in daily life that mobilize positive affect. These findings are in line with the facilitating role of positive affect in various areas of functioning (Fredrickson, 2001) and were based on reports provided independently by youths and parents, thereby increasing confidence in the results.
The present study has various strengths, including several large samples, its extension of the traditional high-risk study design to juvenile siblings, thorough diagnostic evaluation of the subjects across the years, and the use of multiple hedonic probes and their randomized presentation. Furthermore, the probes included a sampling of stimuli/tasks that might be encountered in daily life, increasing the validity of the results. As well, some of the findings may have practical implications. For example, receipt of a preferred prize was the most robust hedonic probe, suggesting that personal relevance is an important dimension to be considered in protocols using positive affect probes.
Along with similar work by others in this area (reviewed earlier), our findings complement a corpus of investigations that point to blunted neural responses to positive stimuli or reward related tasks in high risk compared with low risk young offspring (Kujawa, Hajcak, Torpey, Kim, & Klein, 2012; Monk et al., 2008), adolescents and young adults (Foti, Kotov, Klein, & Hajcak, 2011; Gotlib et al., 2010; McCabe et al., 2012), and adult offspring and older adult siblings (Macoveanu et al., 2013). A recent study also found that blunted neural response to reward predicted onset of the first major depression episode among adolescent girls at variable familial risk for depression (Bress, Foti, Kotov, Klein, & Hajcak, 2013). However, neural, subjective, and behavioral dimensions of hedonic abnormalities have not yet been compellingly linked within and across studies.
Recent approaches (particularly those focused on neural mechanism) have emphasized demarcating the anticipatory/motivational and consummatory (i.e., wanting versus liking) phases of hedonic processing (Treadway & Zald, 2011; Whitton, Treadway, & Pizzagalli, 2015). The present study did not address hedonic function during task anticipation but instead focused on the experiential or “liking” aspect of pleasurable experience. Contrary to some work indicating no impairment during the consummatory phase of hedonic processing among depression-prone adults (reviewed by Treadway & Zald, 2011; Whitton et al., 2015) and similar findings in a small sample of high-risk adolescents (Olino et al., 2014), but in line with various reports we cited earlier, probands and high-risk youths in our study did evidence “consummatory” hedonic impairment. Such discrepant results probably reflect important methodological differences across studies, including varying definitions of the outcome and the type of experimental hedonic probe that was used.
The findings of this study should be evaluated in the context of several limitations. First, it is possible that we failed to find significant differences between depressed (n = 32) and remitted probands (n = 186) in positive affectivity because of low statistical power. Furthermore, although there is compelling evidence that positive affectivity is strongly heritable (e.g., Watson & Naragon, 2009), our study was not designed to disentangle the influence of familial depression versus low familial positive affectivity on the results. Finally, our study focused on subjective experience as indexed by self-report, and did not consider other response systems.
Given the accumulating findings and the ongoing groundswell of interest regarding hedonic dysfunction in depression, what questions should be prioritized and what studies should be conducted as the logical next steps? One basic question still to be conclusively answered is whether hedonic dysfunction in the context of depression is a trait or a state. Although the best interpretation of extant findings (based on cross-sectional comparisons of depressed, remitted, and control subjects) is that hedonic dysfunction has both trait and state components, a definitive answer can only be based on longitudinal studies that repeatedly test the same high- and low-risk individuals across varying mood states. Is hedonic dysfunction a bona fide risk factor that precedes and predicts first onset of clinical depression? Based on accumulating evidence, the answer probably is yes, although further supportive findings are needed. However, the relevant literature reflects group differences across samples at high versus low risk for depression, whereas hedonic capacity is clearly an individual differences variable, impairment in which is only one route to clinical depression (e.g., Meehl, 1975). Thus, the challenge is to identify those who exhibit this characteristic and then intervene to remediate the presumed dysfunction.
How can individuals with low hedonic capacity best be identified? On the one hand, there is consensus that (regardless of how it is assessed) the construct of hedonia has various facets beyond experiential or consummatory pleasure and positive affect. The complexity is well reflected by the numerous studies of reward processing, which have underscored the additional importance of anticipation, stimulus salience, as well as related motivational, decision making, and learning processes (e.g., Treadway & Zald, 2011; Whitton et al., 2015). On the other hand, the practical ramifications of this broad literature for the identification of vulnerable cases are yet to be established. For example, it is unclear how the various facets of hedonic functioning relate to positive affect (Treadway & Zald, 2011), which has long been considered a fundamental dimension of pleasure and joy. Consequently, it is unclear whether early identification of high-risk cases should be based on attenuated responses to positive stimuli, impaired incentive motivation, problems in reward-related learning, some other index of performance, or combinations of these variables. Importantly, little is known about how these facets “play out” in everyday life and which facet is most predictive of functional impairment or clinical course.
Several of these issues could be addressed in samples of interest by monitoring hedonic related responses and processes in daily life (along the lines of ecological momentary assessment designs). Such studies could explore the relative importance of affective, cognitive, motivational, and information processing mechanisms and their interactions in hedonic responding, as well as the roles of hedonic stimulus access and salience in situ, and how contextual features may attenuate or facilitate hedonic dysfunction. The resultant information could yield integrated guidelines about which parameters best identify hedonically challenged individuals and also inform existing approaches to remediate anhedonia in depressed people (e.g., McMakin, Siegle, & Shirk, 2011).
Finally, another critical topic is the developmental origins of anhedonia in depression-prone youths. For example, although affective processes appear to be most salient in the consummatory (liking) phases of hedonic experience, cognitive processes appear to predominate in the motivational/anticipatory phases; this is important because components of affective and cognitive processing (and associated neural mechanisms) have different developmental trajectories. Thus, it would be helpful to know how early in development children at high risk for depression evidence atypical cognitive (e.g., motivational) versus the affective (e.g., experiential) facets of hedonic processing, which type of impairment is most likely to derail subsequent development, and how these processes relate to early life stress, which has been implicated in eventual affective abnormalities (e.g., Pechtel & Pizzagalli, 2011). Information about the developmental course of dysfunctional hedonic processes, and the conditions under which they persist, could yield new approaches to intervention with depression-prone youths in the search to lower the risk of clinical depression among them.
Footnotes
Declaration of Conflicting Interests
The Children’s Depression Inventory is published by Multi-Health Systems, Inc., from which Maria Kovacs receives royalties.
Funding
This research was supported by Grants MH-084938 and MH-056193 from the National Institute of Mental Health, Rockville, MD, USA, and Grant NN-85285 from the Hungarian Scientific Research Fund, A, Budapest, Hungary.
