article in press - university of south floridauweb.cas.usf.edu/mood/docs/bylsma_inpress.pdfand...

Available online at www.sciencedirect.com

+ MODEL

CPR-00891; No of Pages 16

ARTICLE IN PRESS

Clinical Psychology Review xx (2007) xxx–xxx

A meta-analysis of emotional reactivity inmajor depressive disorder☆

Lauren M. Bylsma, Bethany H. Morris, Jonathan Rottenberg ⁎

Department of Psychology, University of South Florida, 4202 E. Fowler Ave, PCD4118G, Tampa, FL 33620, United States

Received 14 April 2007; received in revised form 30 September 2007; accepted 3 October 2007

Abstract

Three alternative views regarding how Major Depressive Disorder (MDD) alters emotional reactivity have been featured in theliterature: positive attenuation (reduced positive reactivity), negative potentiation (increased negative reactivity), and emotioncontext insensitivity (ECI; reduced positive and negative reactivity). Although empirical studies have accumulated on emotionalreactivity in MDD, this report is to our knowledge the first systematic quantitative review of this topic area. In omnibus analyses of19 laboratory studies comparing the emotional reactivity of healthy individuals to that of individuals with MDD, MDD wascharacterized by reduced emotional reactivity to both positively and negatively valenced stimuli, with the reduction larger forpositive stimuli (d=− .53) than for negative stimuli (d=− .25). Results were similar when 3 major emotion response systems (self-reported experience, expressive behavior, and peripheral physiology) were analyzed individually. The ECI view of emotionalreactivity in MDD is well supported by laboratory data. Implications for the understanding of emotions in MDD are discussed.© 2007 Elsevier Ltd. All rights reserved.

Keywords: Major depression; Emotion; Affect; Reactivity; Meta-analysis

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 01.1. Views of emotional reactivity in MDD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 01.2. Limitations of prior reviews. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 01.3. The present study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0

2. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 02.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 02.2. Rationale for emotion response domains covered in this review . . . . . . . . . . . . . . . . . . . . . . . . . 0

☆ The authors express their appreciation to Michelle Jones for her help in library research and to the members of the Mood and Emotion Laboratoryfor their comments. This research was supported by a University of South Florida New Researcher Grant awarded to Jonathan Rottenberg.⁎ Corresponding author. Tel.: +1 813 9746701; fax: +1 813 9744617.E-mail addresses: [email protected] (L.M. Bylsma), [email protected] (B.H. Morris), [email protected] (J. Rottenberg).

0272-7358/$ - see front matter © 2007 Elsevier Ltd. All rights reserved.doi:10.1016/j.cpr.2007.10.001

Please cite this article as: Bylsma, L. M., et al., A meta-analysis of emotional reactivity in major depressive disorder, Clinical PsychologyReview (2007), doi:10.1016/j.cpr.2007.10.001

mailto:[email protected]



http://dx.doi.org/10.1016/j.cpr.2007.10.001


2 L.M. Bylsma et al. / Clinical Psychology Review xx (2007) xxx–xxx

ARTICLE IN PRESS

2.3. Variables included in this review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 02.4. Literature search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 02.5. Study inclusion criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 02.6. Characteristics of selected studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 02.7. Computation of effect sizes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 02.8. Planned analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0

3. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 03.1. Omnibus analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 03.2. PER and NER in individual response systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0

3.2.1. PER. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 03.2.2. NER . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0

3.3. Heterogeneity and Moderator Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 03.4. File drawer analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0

4. Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 04.1. Limitations and future directions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 04.2. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0

References

1. Introduction

Major Depressive Disorder (MDD) is a debilitating condition that afflicts almost a sixth of the generalpopulation (Kessler, 2002). Reflecting the profound disturbance of affective function in MDD, it is classified as amood disorder by the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV; APA, 2000). DSM-IVdiagnostic criteria specify symptoms of at least two weeks duration that implicate deficient positive affect (e.g.,anhedonia), excessive negative affect (e.g., guilt, sadness), or both. When queried, patients who have beendiagnosed with MDD reliably report low positive affect and elevated negative affect on a variety of questionnaireand interview measures (Clark, Watson, & Mineka, 1994). Durable disturbance of mood is thus one of the mostsalient features of MDD.

Given that MDD is quintessentially a disorder of mood, one important question is how MDD influences emotionalreactions. Addressing this question requires a distinction between the constructs mood and emotion (e.g., Rottenberg &Gross, 2003). Moods have been defined as diffuse, slow-moving feeling states that are weakly tied to specific stimuli inthe environment (e.g., Watson, 2000). By contrast, emotions have been defined as quick-moving reactions that occurwhen an individual processes a meaningful stimulus (e.g., a threatening object). Emotional reactions are multi-componential and typically involve coordinated changes in several response systems (e.g., perception, feelings,expressive behavior, peripheral and central physiology; Ekman, 1992; Keltner & Gross, 1999). When mood andemotion are so distinguished, it becomes apparent that the various diagnostic criteria for MDD, such as pervasivesadness or anhedonia, indicate alterations in mood, but do not indicate alterations in emotion with correspondingspecificity.

Although the constructs are distinguishable, moods and emotions have generally been seen as interconnected, withmoods altering the probability of having specific emotions (e.g., Rosenberg, 1998). More specifically, moods arethought to potentiate like-valenced or matching emotions (e.g., irritable mood facilitates angry reactions, an anxiousmood facilitates panic, etc; Rottenberg, 2005). By extension, excessive negative mood in MDD would potentiatenegative emotional reactivity and/or a lack of positive mood would attenuate positive emotional reactivity. Indeed, theidea of mood-facilitation is one guiding source for three major views regarding emotional reactivity in MDD: (1)negative potentiation (2) positive attenuation and (3) emotion context insensitivity (ECI). We will briefly outline eachof these views before presenting a meta-analysis designed to determine which of them best fits the accumulated dataconcerning emotional reactivity in MDD.

Our main goal was to conduct a systematic meta-analytic review of the literature on emotional reactivity inMDD. We used rigorous literature search strategies (see below) to identify laboratory studies of emotional reactivityin MDD and estimate effect sizes. We report effect sizes from 19 laboratory studies that compared negative andpositive emotional reactivity between individuals with MDD and healthy participants. Analyses focused onomnibus group differences in positive and negative emotional reactivity, as well as the replicability of these group



3L.M. Bylsma et al. / Clinical Psychology Review xx (2007) xxx–xxx

ARTICLE IN PRESS

differences within three specific emotion response systems: self-report, behavioral, and peripheral physiologicalreactivity. To our knowledge, this study represents the first meta-analytic review of the literature on emotionalreactivity in MDD.

1.1. Views of emotional reactivity in MDD

Negative potentiation, the first view, holds that the pervasive negative mood states that are prevalent in MDDcontribute to potentiated emotional reactivity to negative emotional cues. Perhaps most relevant to the idea thatnegative moods and negative emotions are mutually reinforcing in MDD, cognitive theorists see negative moods asfacilitating negative cognitive processing which, in turn, results in negative cognitive interpretations that generatedysphoric reactions (e.g., Beck, 1976; Beck, Rush, Shaw, & Emery, 1979). Beck's schema model and relatedtheories of depression (e.g., Bower, 1981) conceptualize MDD in terms of cognitive structures, or schemas, whichare patterns in thinking that serve to negatively distort the processing of emotional stimuli (e.g., such as beliefs thatone is unlovable or a failure). Importantly, according to these theories, negative mood states prime, or activate, thesecognitive structures (Scher, Ingram, & Segal, 2005). Once activated, these structures precipitate negativeinterpretations which give rise to depressotypic emotional responses (e.g., crying spells) whenever schema-matching negative emotion stimuli are encountered, presumably potentiating reactivity to negative emotionalstimuli in MDD.

Positive attenuation, the second view, holds that individuals with MDD will have reduced reactivity in response topositive emotional stimuli. Because this hypothesis applies primarily to positive emotional stimuli, positiveattenuation is compatible with negative potentiation (i.e., individuals with MDD can exhibit both patternssimultaneously). The starting point for this hypothesis is the depressed persons' strong tendency to exhibit lowpositive mood. Indeed, anhedonia (the reduced ability to experience pleasure) is one of the cardinal symptoms ofMDD, and depressed individuals exhibit several other signs that are also indicative of deficient appetitive motivation(e.g., psychomotor retardation, fatigue, anorexia, apathy). Not surprisingly, several theorists have centered theiraccounts of abnormal emotional responding (e.g., abnormally low emotion response magnitude) in MDD on thisconstellation of motivational deficits (e.g., Clark et al., 1994; Depue & Iacono, 1989) often focusing on the centralnervous system substrates for deficient appetitive motivation (e.g., hypoactivation in the left frontal lobes; Henriques& Davidson, 1991).

Emotion context insensitivity (ECI), the third view, holds that depressed individuals will exhibit reducedreactivity to all emotion cues, regardless of valence (Rottenberg, 2005, 2007). By this account, individuals withMDD should exhibit less reactivity to both positive and negative stimuli and events compared to healthyindividuals. ECI is derived from evolutionary accounts that describe depression as a defensive motivational statethat fosters environmental disengagement (Nesse, 2000). According to this view, depressed mood states evolved asinternal signals to bias organisms against action in adverse situations where continued activity might be potentiallybe dangerous or wasteful (e.g., famine). Thus, according to ECI, severe depressed mood states in MDD arepostulated to inhibit ongoing emotional reactivity. In sum, ECI makes similar predictions as the positiveattenuation view for positive stimuli, but makes opposite predictions as the negative potentiation view for negativestimuli.

1.2. Limitations of prior reviews

Prior generalizations about emotional reactivity in MDD have been based on narrative reviews, which can besusceptible to selection biases and are often unsystematic in their search strategies. Any review also faces thecomplexity of operationalizing the major constructs relevant to this topic area. For example, the depression constructcan be operationalized in different kinds of samples (e.g., analogue samples of dysphoric persons versus case-levelMDD), and even studies of MDDmay consider related conditions, such as dysthymia or minor depression, to be part ofthe depression construct. Thus, reviews may describe different effects of depression on emotional reactivity becausedifferent kinds of depressive phenomena are being considered. Likewise, complexity in the emotion construct alsocomplicates the task of review. Emotional reactivity can be operationalized in a large number of response systems,which often exhibit considerable independence from one another (e.g., self-reported experience, expressive behavior,cognition, peripheral nervous system responding, or neural activity, see Mauss, Levenson, Carter, Wilhelm, & Gross,




ARTICLE IN PRESS

2005). Prior reviews have not made a concerted effort to integrate data across self-report, behavioral, and physiologicaldomains (e.g., Forbes, Miller, Gohn, Fox, & Kovacs, 2005; Gehricke & Shapiro, 2000). Further, even within any givensystem of response, emotional reactivity can be computed using different metrics (e.g., level scores versus changescores), and commentators have generally not been explicit and/or systematic in defining their choice of reactivitymetric. Given these multiple challenges, perhaps it is not surprising that prior narrative reviews in this area of researchhave been inconclusive.

1.3. The present study

To overcome the limitations of prior reviews, we assessed the literature on emotional reactivity in MDD usingmeta-analysis — an important quantitative technique for systematically summarizing results across differentstudies which use different measurement techniques and arriving at stable estimates of effect sizes (Cohen, 1988).In contrast to narrative reviews, meta-analysis avoids reliance on statistical significance tests and provides precisequantitative estimates of the magnitude of effect sizes. For this meta-analysis, only studies that elicited emotion ina well-controlled manner and used DSM diagnostic criteria for MDD were included. Analyses focused on theextent to which MDD might increase or decrease positive and negative emotional reactivity relative to healthyindividuals.

2. Methods

2.1. Overview

A meta-analytic procedure was used to estimate the effect of MDD on emotional reactivity. Our major focus was on positiveemotional reactivity (PER; i.e., emotional reactivity to a positively valenced stimulus) and negative emotional reactivity (NER; i.e.,emotional reactivity to a negatively valenced stimulus). Emotional reactivity was defined as a positive (PER) or negative (NER)measured emotional response to a matching emotionally valenced stimulus that is reflected as a change from baseline affect. Wefocused on these two forms of emotional reactivity rather than specific emotions because (1) the major views of emotional reactivityin MDD concern PER and NER (2) valence has high theoretic importance in the emotion literature (e.g., Barrett, 2006), and (3)insufficient research exists to allow pooling of results for specific emotions.

2.2. Rationale for emotion response domains covered in this review

One general challenge in conducting a meta-analysis of PER and NER is the overwhelming number of responses that arepotentially relevant to emotion. Relatedly, there is a lack of consensus among affective scientists regarding just howmany emotionresponse systems there are, and how all of these response systems should be mapped on to the emotional reactivity construct (e.g.,posture, touch, perceptual changes, Rottenberg & Johnson, 2007). To make this problem more tractable, we made a judgment tofocus our review on three response systems that are important to emotional reactivity: (1) behavioral/expressive, (2) experiential/subjective, and (3) peripheral physiological responding in the autonomic nervous system. These three systems were prioritized forfour main reasons: First, these three systems have been historically important in guiding inquiry in emotion (e.g., Dolan, 2002;Ekman, 1992; Lang, Rice & Sternbach, 1972; Lang, 1988; Lazarus, 1991; Levenson, 1994). Second, there is reasonably goodagreement that these three systems do in fact index emotional reactivity. Third, there is reasonably good agreement on measuresthat can be extracted from these systems to index emotional reactivity (see below). Fourth, there is an adequate database ofempirical studies using metrics from these three systems in MDD to conduct a meta-analysis.

Although a focus on these three systems affords our review some breadth as well as some fidelity to the construct of emotion,this review is by no means all-inclusive of systems that are relevant to emotion. Perhaps most importantly, neurological andendocrine systems play a major role in emotional reactivity (e.g., LeDoux, 1998). However, we did not include neuroimagingmetrics (e.g., techniques such as fMRI, PET, and ERP), because existing reviews of neuroimaging and emotion indicate thatconsiderable disagreement remains about how to map emotional reactivity onto these techniques (e.g., Murphy, Nimmo-Smith, &Lawrence, 2003; Phan, Wager, Taylor, & Liberzon, 2002). More practically, an early inspection of potentially relevant studiesindicated that imaging studies of MDD did not routinely report statistics needed for computing effect sizes. Further, the technicalproblems of conducting meta-analytic analyses of neuroimaging data are considerable, as these are far from simple computationsof effect sizes, and, typically, effect sizes based on standardized mean differences cannot be calculated due to the complexity ofthe data, and because the focus is typically on activity location rather than effect size (Fox, Parsons, & Lancaster, 1998). Fox et al.(1998) have noted that meta-analytic procedures do not provide an adequate way of combining neuroimaging and non-neuroimaging measures into a single meta-analysis for both methodological and theoretical reasons. For these reasons, separate




ARTICLE IN PRESS

meta-analytic procedures must be used for neuroimaging data that have different theoretical interpretations. Likewise, we did notconsider the domain of cognition to be tractable for our purposes, even though cognition and emotion interact (e.g., Gray, 2004),and cognitive changes are among the known correlates of emotional reactions. An inspection of the literature suggests thatcognition is extraordinarily heterogeneous as a domain, and that there is little agreement or precedent concerning how metricsfrom the cognitive domain (e.g., memory bias, perceptual judgments, reaction times, etc) should be mapped onto the emotionalreactivity construct for meta-analysis.

2.3. Variables included in this review

Selection of measures of emotional reactivity within the three response systems covered in this review (self-reportedexperience, expressive behavioral, and peripheral physiological) was guided by prior literature demonstrating that each measurecould be interpreted as a valid index of emotional reactivity. Self-report measures of emotional responses included subjectivereports of emotional experience through various self-report measures such as the PANAS (Watson, Clark, & Tellegen, 1988), theSAM (Bradley & Lang, 1994), or discrete emotion measures. Behavioral measures included observed facial expression asmeasured through facial coding systems (e.g., FACS; Ekman, Matsumoto, & Friesen, 2005; Lang, Greenwald, Bradley, & Hamm,1993) or electromyographic (EMG) recordings of the face (e.g., Cacioppo, Martzke, Petty, & Tassinary, 1988; Cacioppo, Petty,Losch, & Kim, 1986; Lang et al., 1993), emotion-modulated eyeblink/startle responses (e.g., Bradley, Cuthbert, & Lang, 1999;Bradley & Lang, 2000), and approach/avoidance behavior (e.g., Carver, 2001; Henriques & Davidson, 2000). Physiologicalmeasures included measures of autonomic peripheral physiology, specifically including skin conductance responses (SCR; e.g.,Lang et al., 1993), heart-rate variability (HRV)/respiratory sinus arrhythmia (RSA) (e.g., Appelhans & Luecken, 2006; Butler,Wilhelm, & Gross, 2006), blood pressure (e.g., Melamed, 1987), respiration (e.g., Ritz, Thons, Fahrenkrug, & Dahme, 2005),heart rate (e.g., Lang et al., 1993), and finger pulse amplitude (FPA; e.g., Gross, 2002).

2.4. Literature search

This review covered published journal articles in the English language since 1975. Only published articles were reviewed forpossible inclusion, because published work generally has increased scientific rigor compared with unpublished work, and because itis readily available within the public domain for independent evaluation by others. Because null results may be more difficult topublish than significant findings (potentially leading to upward biases of effect sizes in the published literature), we included “file-drawer analyses” to provide an estimate of the number of studies with non-significant results that would be necessary to alter theinterpretation of the primary results (Rosenthal, 1991).

To generate potentially relevant articles, the first author used the PsychINFO and MEDLINE online database, entering thefollowing keywords: depress⁎ and emotion⁎ reactivity, depress⁎ and affect⁎ modulation, depress⁎ and affect⁎ regulation,depress⁎ and emotion⁎ functioning. The wildcard ⁎ was used to ensure that all forms of the keywords were searched. Afteridentifying potentially relevant articles, the reference sections were searched to identify additional articles which might meet ourinclusion criteria. As an additional check to ensure that no relevant articles were omitted, a research assistant also performed aparallel search.

2.5. Study inclusion criteria

Articles were required to meet the following inclusion criteria (a) a group of participants that qualified for a diagnosis of currentMDD using DSM criteria (b) a healthy control group, (c) appropriate statistics (e.g., means and standard deviations) that werereported in the published article or available from the author upon request (of nine articles that were missing data, the study authorswere able to supply it in six cases), (d) measured emotional reactivity with self-reported experience, expressive behavioral, orperipheral physiological indicators (e) a positively or negatively valenced emotion elicitation condition was included, along with(f) a neutral or baseline condition to allow computation of NER and PER (without a baseline of comparison responses to a valencedstimulus may reflect a general disposition or mood rather than reactivity to a given stimulus). Medication use by patients was notconsidered as an inclusion/exclusion criterion because virtually no studies in the entire literature used an un-medicated sample (oursearch yielded only four such studies).

2.6. Characteristics of selected studies

From an initial screening, a total of 35 potential articles were found. However, 6 of these did not include a control group, 4 did notinclude a neutral or baseline measure, 2 were not laboratory studies, and 3 did not include the necessary means and/or standarddeviations for inclusion in the meta-analysis and the authors were unable to be contacted. The final sample for inclusion in theanalysis consisted of 19 articles. All articles included measures of NER, and 14 also included measures of PER. Because many of thearticles includes measures of multiple response systems, overall there were 29 measures of NER and 24 measures of PER across




ARTICLE IN PRESS

response systems. The pooled sample size consisted of 465 participants with MDD and 452 control participants. See Table 1 for asummary of selected characteristics of studies included in the meta-analysis1.

The demographic characteristics of participants in the included studies appeared to be consistent with the known epidemiology ofMDD. The majority of the studies included more females than males in their samples, consistent with the prevalence rate ofdepression being higher in females (Kessler, 2002). Seven studies only included female participants, and one only included males.Most studies specified excluding participants with co-morbid psychosis, bipolar disorder, or substance abuse. However, co-morbidanxiety diagnoses were typically included (only two studies stated that they excluded for all anxiety disorders). Five studies specifiedincluding inpatient MDD samples, and 11 specified using outpatient samples. The majority of included studies did not reportethnicity or other participant demographic information. Only one study (Tsai, Pole, Levenson, & Munoz, 2003) used a minoritysample which consisted of Spanish-speaking Latinas, and one study used a British sample in London (Kaviani, Gray, Checkley,Raven, Wilson, & Kumari, 2004).

2.7. Computation of effect sizes

We conducted the analyses using the Hedges and Olkin approach (1985) which weights each study by the inverse of the variance,which is roughly proportional to the study's sample size. Effect sizes were computed using Cohen's d, a common measure of effectsize (Cohen, 1988). Effect sizes were calculated as (M1–M2)/SDp whereM1 is the control group mean,M2 is the MDD group mean,and SDp is the pooled standard deviation. This represents the standardized mean difference between groups. Standardizing thesemeans allowed us to account for scaling differences between different types of measures and pool data across studies. The meansused to compute the effect sizes were the average self-report, behavioral, or physiological emotional reactivity scores. If the study didnot report emotional reactivity scores directly, these scores were derived by subtracting neutral or baseline condition mean scoresfrom positive or negative response mean scores. In this case, the standard deviation was first pooled from the standard deviations ofthe positive and negative response mean scores and the neutral or baseline conditions, and then this pooled standard deviation wasused in the equation above to pool across groups in the calculation of the effect sizes. All effect sizes were calculated using theComprehensive Meta-analysis software package (Biostat, 2005).

Avariety of stimuli in the primary studies were used to elicit valenced emotional responses that have received some support fromprevious research as being valid measures of emotional reactivity, including emotional films or audio recordings describing anemotional situation (Rottenberg, Ray, & Gross, 2007), emotional pictures from the International Affective Picture System (IAPS;Lang, Bradley, & Cuthbert, 2005), facial expressions of emotion from the Facial Affective Booklet (FAB; Ekman, 1976, Ekman &

1 Studies excluded from the meta-analysis after the initial screening.No control group:

Carney, Hong, Brim, Plesons, and Clayton (1983)Greden, Price, Genero, Feinberg, and Levine (1984)Rottenberg, Wilhelm, Gross, and Gotlib (2002)Rottenberg, Salomon, Gross, and Gotlib (2005)Schwartz, Fair, Mandel, Salt, Mieske, and Klerman (1976)Schwartz, Fair, Salt, Mandel, and Klerman (1976)

No baseline or neutral measure:

Berenbaum and Oltmanns (1992)Katsikitis and Pilowsky (1991)Renneberg, Heyn, Gebhard, and Bachmann (2005)Tremeau et al. (2005)

Not laboratory studies (naturalistic):

Peeters, Nicolson, and Berkhof (2003)Peeters, Nicolson, Berkhof, Delespaul, and deVriews (2003)

Missing data necessary for computation of effect sizes:

Iacono, Lykken, Haroian, Peloquin, Valentine, and Tuason (1984)Lapierre and Butter (1980)Wexler, Levenson, Warrenburg, and Price (1994)Yecker, Borod, Brozgold, Martin, Alpert, and Welkowitz (1999).



Table 1Selected characteristics of studies included in the meta-analysis

Citation N Medicated Reactivity Emotional stimuli Baseline/neutral Reactivity measures

MDD/Control

Dawson, Schell, and Catania (1977) 20 /20 95% NER Stress tasks Rest, tones Physiological: HR and SCGreden, Genero, Price, Feinberg,

and Levine (1986)63 / 37 0% NER, PER Sad and happy imagery Rest Behavioral: facial EMG

Albus, Muller-Spahn, Ackenheil,and Engel (1987)

12 /63 0% NER Stress tasks Rest Physiological: HR, SC

Sigmon and Nelson-Gray (1992) 20 /20 Unknown NER, PER Audiotape of positive andnegative social interactions

Audiotape ofneutral social interactions

Self-report: DACL and affective ratings⁎

Physiological: SCPersad and Polivy (1993) 16 /16 Unknown NER Emotional faces from the FAB⁎ Neutral faces Self-report: reported sadness, fear to facesGuinjoan, Bernabo, and

Cardinali (1995)18 /18 0% NER Stress tasks Rest Physiological: HR, BP, SC

Sloan, Strauss, Quirk, &Sajatovic (1997)

24 /23 100% NER, PER IAPS pleasant andunpleasant pictures

IAPS neutral pictures Self-report: affective ratings of picturesBehavioral: facial expression (FACS)⁎

Allen, Trinder, & Brennan (1999) 14 /14 100% NER, PER IAPS pleasant andunpleasant pictures

IAPS neutral pictures Self-report: affective ratings of picturesBehavioral: emotion-modulated startle⁎

Gehricke & Shapiro (2000) 11 /11 27% NER, PER Sad and happy imagery Rest Self-report: DES⁎

Behavioral: facial EMGHenriques & Davidson (2000) 18 /15 0% NER, PER Reward, punishment Neutral condition Self-report: PANAS⁎

Behavioral: response to reward/punishmentSloan, Strauss, and Wisner (2001) 20 /20 100% NER, PER IAPS pleasant and

unpleasant picturesIAPS neutral pictures Self-report: affective ratings of pictures

Behavioral: facial expression (FACS)Rottenberg, Kasch, Gross, and Gotlib (2002) 72 /32 43% NER, PER Sad, fearful, and

amusing filmsNeutral film Self-report: reported sadness, amusement to films

Behavioral: facial expression (FACS)Physiological: SC, HR

Rottenberg et al. (2003) 25 /31 0% NER Sad film Neutral film Physiological: RSA, HR, respirationTsai et al. (2003) 11 /9 Unknown NER, PER Sad and amusing films Neutral film Self-report: emotion inventory

Behavioral: facial expression (FACS, smiles)Physiological: SC, HR, FPA, respiration

Dichter, Tomarken, Shelton,and Sutton (2004)


IAPS neutral pictures Self-report: affecting ratings of picturesBehavioral: emotion-modulated startle

Dunn, Dalgleish, Lawrence,Cusack, and Ogilvie (2004)


IAPS neutral pictures Self-report: emotional response to pictures

Kaviani et al. (2004) 22 /22 82% NER, PER Unpleasant andpleasant films

Neutral film Self-report: affective ratings of filmsBehavioral: emotion-modulated startle

Forbes et al. (2005) 38 /38 15% NER, PER IAPS pleasant andunpleasant pictures

IAPS neutral pictures Self-report: affective ratings of picturesBehavioral: emotion-modulated startle

Rottenberg, Gross, andGotlib (2005)

19 /22 32% NER, PER Sad and happy films Neutral film Self-report: emotional response to filmsBehavioral: facial EMGPhysiological: HR, FPA, SC, respiration

NER = negative emotional reactivity, PER = positive emotional reactivity, IAPS = international affective picture system, HR = heart rate, RSA = respiratory sinus arrhythmia, SC = skin conductance,BP = blood pressure, FPA = finger pulse amplitude, FACS = facial action coding system, EMG = electromyography, DACL = depression adjective checklist, FAB = facial affective booklet, PANAS =Positive and negative affect schedule, DES = differential emotions scale.⁎ These measures were not included in the meta-analysis due to missing means and/or standard deviations, which we were unable to obtain from the authors, or a neutral/baseline measure was notincluded to establish reactivity.

7L.M

.Bylsm

aet

al./Clinical

Psychology

Review

xx(2007)

xxx–xxx

ARTICLE

INPRESS

Please

citethis

articleas:

Bylsm

a,L.M.,et

al.,A

meta-analysis

ofem

otionalreactivity

inmajor

depressivedisorder,

Clinical

Psychology

Review

(2007),doi:10.1016/j.cpr.2007.10.001



ARTICLE IN PRESS

Oster, 1979), standardized laboratory stressors such as a loud noise or mental arithmetic tasks (Krantz, Manuck, &Wing, 1986), andthe presentation of rewards (Henriques, Glowacki, & Davidson, 1994). Most of the behavioral and physiological studies also usedself-report measures to validate whether the appropriate emotion was elicited. In order to ensure that emotional reactivity, defined asemotional reaction to a particular positively or negatively valenced stimulus, rather than a more general disposition of mood wasassessed, we defined PER and NER as a change from a baseline condition. Rest and response to a neutral stimulus (e.g., a neutralfilm) were both considered acceptable baselines. We felt this liberal definition of baseline was justified in light of disagreementconcerning what constitutes an optimal baseline condition for assessing the effects of emotional reactivity (for a discussion see, forexample, Rottenberg et al., 2007). The more liberal definition of baseline also allowed us to retain as many studies as possible in themeta-analysis. Nevertheless, when both a resting baseline and a neutral baseline were reported in the same study, we used the neutralcondition, because we believe there are drawbacks associated with using the resting baselines (e.g., rest may not be a representativestate of the organism, rest instructions may introduce unwanted variability, etc. For details, see Rottenberg et al., 2007).

When the authors of the studies included in the meta-analysis used more than one type of emotional reactivity measure for PER orNER those effect sizes were summarized by simply using the mean effect size for the omnibus analyses. For the analyses within eachdomain, averaging was only done if there was more than one emotional reactivity measure for PER or NER within a domain. Usingthe mean effect size is conservative, as it gives a lower estimate compared to overall composite variables (Rosenthal & Rubin, 1986).An alternate method would be to calculate the overall composite effect sizes. However, in order to make this computation, the inter-correlations between the multiple dependent measures would be needed and, in most studies, they were not reported.

2.8. Planned analyses

Primary omnibus analyses examined NER and PER across emotion response systems. Since activity in different emotionresponse systems are not always strongly inter-correlated (Lang, 1968; Mauss, Wilhelm, & Gross, 2004), and may even dissociate(e.g., Lang, 1988), we conducted secondary analyses within each emotion response domain (reported experience, behavior, andphysiology) to examine the consistency of emotional reactivity across domains.

The primary analyses employed a fixed effects model, which has high power and remains the predominant model in meta-analyses of psychological research (Hunter & Schmidt, 2000). However, fixed effects models have been criticized for inflating TypeI error, especially if there is heterogeneity in the data (Hunter & Schmidt, 2000). Due to the small number of studies available for thismeta-analysis and the unknown size of the expected effects, we chose to use the fixed effects model as the primary model. We alsoran tests of heterogeneity of the effect sizes using Cochran's heterogeneity statistic Q (Cochran, 1954). Since heterogeneity canpotentially inflate Type I error in a fixed effects approach, we planned to re-run the analyses using a random effects model, ifanalyses revealed significant heterogeneity in the data.

3. Results

3.1. Omnibus analyses

We first conducted omnibus analyses of positive and negative emotional reactivity using the fixed effects model. The analysis ofpositive emotional reactivity (PER) was significant (pb .0001) and revealed that PER was reduced in MDD compared to normalcontrols (see Fig. 1). The effect size for PER was d=− .53, a medium-sized effect by Cohen's (1988) conventions. Similarly, theomnibus analysis of negative emotional reactivity (NER) was also significant, (pb .0001) and revealed that NER was reduced inMDD compared to normal controls (see Fig. 1). The effect size for NER was d=− .25, corresponding to a small effect size. WhenPER and NER effect sizes were compared in a moderator analysis (with effect type PER versus NER coded as a moderator variable),a significant effect was obtained (Q=7.21, pb .01), reflecting that the PER effect was significantly larger than the NER effect,indicating that MDD individuals exhibited a more pronounced blunting of PER than of NER.

We conducted analyses of heterogeneity of both the PER and NER omnibus analyses to measure the variation around the meanweighted effect sizes. Significant heterogeneity was present for both PER (Q=111.80, pb .0001) and NER (Q=34.77, p=.01). Giventhe presence of heterogeneity in the magnitude of the effect sizes, we ran the planned follow-up random effects analysis for PER andNER. The random effects analyses, despite their reduced power, revealed similar results for both PER (d=− .59, p=.011) and NER(d=− .26, p=.008), with virtually identical effect sizes to those in the fixed effects analyses. Since the heterogeneity in omnibusanalysis suggests the possible presence of moderator variables, additional analysis of potential moderators is presented below.

3.2. PER and NER in individual response systems

3.2.1. PERTo examine the extent to which reduced PER generalized across different systems of emotional response, separate meta-analyses

were conducted on the self-report, behavioral, and physiological indicators of PER. For both self-report and behavioral measures ofPER, it was found that PER in MDD was reduced relative to controls. For self-report measures of PER, the effect size was d=− .703



Fig. 1. PER and NER across all domains. The MDD group exhibits reduced PER and NER compared to controls (pb .0001). Error bars represent 95%confidence intervals.


ARTICLE IN PRESS

( pb .0001), a large effect. This analysis included 10 studies with a sample size of 263 MDD individuals and 249 controls. Forbehavioral measures of PER, the effect was medium-sized (d=− .453, pb .0001). The analysis of behavioral measures included 10studies with 290 MDD individuals and 250 controls. Only 4 studies (with a total sample of 122 MDD individuals and 89 controls)included physiological measures of PER, and although the direction of effect was also towards a reduction of PER in MDD, none ofthese found a significant effect. When data from these 4 studies was pooled the overall analysis was non-significant (p=.293), with avery modest effect size for PER (d=− .151).

3.2.2. NERTo examine the extent to which reduced NER generalized across different systems of emotional response, separate meta-analyses

were conducted for self-report, behavioral, and physiological indicators of NER. For self-report measures of NER, the effect size wasmedium (d=− .359, pb .0001). This analysis included 11 studies with a total sample size of 279 MDD individuals and 264 controls.For behavioral measures, the effect size was small (d=− .054, and was not statistically significant, p=.544). For the analysis ofbehavioral measures, there were 9 studies, with a sample of 290 MDD individuals and 249 Controls. Finally, there was a mediumeffect for physiological measures of NER (d=− .223, pb .05). For the physiological measures, there were 8 studies, with a samplesize of 196 MDD individuals and 220 controls. In sum, similar to results for PER, when each emotional response system wasexamined separately individuals with MDD exhibited reduced NER compared to controls in two of the three systems.




ARTICLE IN PRESS

3.3. Heterogeneity and Moderator Analyses

Omnibus analyses presented earlier indicated that effect sizes for both PER and NER were heterogeneous. Given that all threedistinct response systems were analyzed together, this was not completely surprising. To help isolate the sources of heterogeneity, wealso ran tests of heterogeneity within each emotion response system. For NER, the effect sizes in the physiological analyses hadsignificant heterogeneity (Q=21.16, pb .01), self-report effect sizes had marginally significant heterogeneity (Q=17.19, p=.05) andbehavioral (Q=8.67, p=.37) were not heterogeneous. For PER, behavioral (Q=150.46, pb .0001) and self-report (Q=86.70,pb .0001) but not physiological analyses (Q=0.39, p=.82) indicated significant heterogeneity. Thus, the only response systemwhich exhibited heterogeneity with any consistency was self-reported emotion, which may be due to the diverse measures includedwithin this domain of emotional responding. Heterogeneity can also indicate the presence of outliers in the data. The PER behavioralmeasures demonstrated most notable degree of heterogeneity. From a visual inspection of the plot of the PER behavioral effectssizes, it would appear that were two outliers present. Specifically, all of the effect sizes fell within the narrow range of − .22 to +.10except for two outliers (+0.64 and −3.95).

The presence of moderator variables, which systematically influence the effect sizes, can also cause heterogeneity. We originallyintended to run additional moderator analyses designed to identify variables that might account for systematic variation in PER andNER effect sizes (e.g., medications, depression severity, co-morbidity). Unfortunately, these analyses were either impossible due toinadequate information from primary studies and/or were underpowered (see Discussion below).2 The small subset of studiesavailable for each response system likewise precluded adequately powered analyses of potential moderator variables within eachsystem.3

3.4. File drawer analyses

Finally, since our review considered only published sources, we conducted file-drawer analyses to assess the reliability of theomnibus effects in the face of additional unpublished null results (Rosenthal, 1991). For the probability for the omnibus analysis ofPER to become non-significant ( pN .05), there would have to be over 90 unpublished studies with non-significant results hiddenaway in the file-drawer. Similarly, for the omnibus analysis of NER to become non-significant, over 40 unpublished studies withnon-significant results would be needed.

4. Discussion

MDD is well-established as a disorder of mood, but theories and prior narrative reviews have disagreed about howMDD influences ongoing emotional reactivity. This meta-analysis represents the first quantitative review of the MDDemotional reactivity literature. The major results indicate that MDD involves consistent reductions in both PER andNER. These results have implications for each of the three major views of emotional reactivity in MDD. First, as NERwas found to be reduced rather than increased, the negative potentiation view was not supported. Second, the reductionof PER in MDD, and the fact that the PER effect was larger than the NER effect both provided good support for thepositive attenuation view. Third, the ECI view appears to offer the most parsimonious overall account of these databecause (1) the pattern of decreased emotion reactivity in MDD was not restricted to PER, and (2) the ECI viewuniquely predicts reduced NER.

2 For example, we attempted to examine whether depression severity exerted a systematic effect on emotion reactivity in MDD; however, thestudies included in the meta-analysis inconsistently reported severity measures. For the measures that reported BDI scores for the MDD group,we correlated these scores with the NER and PER effect sizes. No significant correlation was found for either NER (r= .156, p= .647) or PER(r=− .452, p= .190); however, this may be due to the limited number of studies providing BDI data (11).3 Given that depression is defined diagnostically by tonic mood disturbances, we conducted additional exploratory analyses that examined the

extent to which PER and NER differences between studies in self-reported, behavioral, and physiological reactivity might be explained by tonic,baseline differences in each corresponding system of measured response. In self report of affect at baseline, as expected, MDD individuals onaverage reported less positive affect (d=−0.865, pb .0001) and more negative affect (d=1.454, pb .0001) than controls. However, correlationalmoderator analyses did not find a significant relationship between the magnitude of positive affect baseline differences and self-report indices ofPER (r=.26, p=.49). Likewise, no significant relationship was found between the magnitude of negative affect baseline differences and self-reportindices of NER (r=.48, p=.34). For behavioral measures, the groups did not differ at baseline in positive (d=.005, p=.968) or negative (d=− .154,p=.188) behavioral indices. For physiological indices, MDD individuals exhibited higher activation at baseline compared to controls (d=.263,pb .05). Too few studies were available to examine the association of baseline physiology with physiological indices of PER. However, themagnitude of the difference in baseline activation was not correlated with physiological indices of NER (r=− .352, p=.439). Although the lownumber of included studies limited the power of these analyses, there did not appear to be a strong relationship between tonic differences in responseand observed differences in PER and NER.




ARTICLE IN PRESS

Lending additional credence to the ECI view, these reductions in PER and NER were robust in several respects.First, both effects were highly statistically reliable. Second, file-drawer analyses indicated that the present results foroverall NER and PER would hold even in the face of a substantial number of unpublished null results. And third, thereductions in PER and NER appeared to generalize across different systems of emotional response, and were notrestricted to a single response system.

These results also have more general implications for the study of how moods and emotions interact. Althoughmoods and emotions are highly familiar constructs that have generated extensive research, it is not yet clear how thesetwo kinds of affective processes are related. Intuitively, it makes sense that emotional reactions are stronger when theyare congruent with a preexisting mood, an idea reinforced by contemporary emotion theory. Yet there are surprisinglyfew strong empirical demonstrations that moods actually facilitate emotional reactivity to mood-congruent stimuli. Inthis respect, the analyses presented here appear to provide a notable exception to the idea of mood-facilitation.Excessive negative mood in MDD should potentiate NER by the logic of mood-facilitation, yet this was clearly not thecase in the present data. One hypothesis to help make sense of this paradox is that moods may have non-linear effectson emotional reactivity, with mood-facilitation holding for mild and moderate depressed mood states but not for severedepressed mood states. Although this hypothesis waits further testing, the idea that milder depression potentiates NERis consistent with a number of “analog” studies of dysphoric samples that have obtained data consistent with thisprediction (Golin, Hartman, Klatt, Munz, &Wolfgang, 1977; Lewinsohn, Lobitz, &Wilson, 1973). Careful assessmentof emotional reactivity in samples that contain individuals with the full range depressed mood will be critical for testingfor non-linear effects, an analysis that may enrich an already active debate about whether depression is best conceivedas a continuum or a discrete category (e.g., Beach & Amir, 2003).

4.1. Limitations and future directions

While this study adds to our knowledge about emotions in MDD and has implications for the understanding ofnormal mood variation, four important limitations should be noted. First, to estimate effect sizes for PER and NER, thisreview prioritized the controlled measurement of emotion in laboratory settings. It is an open question as to whetherthese laboratory findings generalize to naturalistic settings. The research database that examines emotion in everydaylife adults among with MDD is currently limited; thus studies that use emotion experience sampling techniques (e.g.,Peeters, Berkhof, Delespaul, Rottenberg, & Nicolson, 2006) or informant ratings of emotion drawn from ecologicallyimportant contexts (e.g., spousal relations) represent important directions for future work.

Second, there was significant heterogeneity in the omnibus analyses of NER and PER as well as consistentheterogeneity in the self-report system when individual response systems were analyzed, but the proximal source(s) ofthis heterogeneity were difficult to isolate. Second, we were not able to identify moderators that might explainsystematic variation in PER and NER effect sizes across studies, either because the relevant information in the primarystudies was incomplete or absent and/or because the analyses of candidate moderators were underpowered. This pointmerits emphasis.

For example, it is possible that the antidepressant medications commonly taken by individuals with MDD couldinfluence the magnitude of PER and NER (Tomarken, Shelton, & Hollon, 2007). The studies in this meta-analysis weredisparate in their practices with respect to medications. Very few studies used a non-medicated or an all-medicatedsample, and the all-medicated samples were often taking a variety of medications, which typically were not reported insufficient detail to analyze in any meaningful way.

Another possible moderator that could not be addressed by the present meta-analysis was co-morbidpsychopathology in the MDD individuals. Most studies did not exclude for anxiety or personality disorder co-morbidity, and did not report emotional reactivity in “pure MDD” separately. In fact, only two studies excluded for alltypes of anxiety disorders in MDD individuals, which are commonly co-morbid with MDD (DSM-IV; APA, 2000). Theeffect of co-morbid disorders on PER and NER in MDD is not well understood, but there are indications in theliterature that anxiety levels and forms of co-morbidity such as substance use (Zvolensky & Schmidt, 2003) mayinfluence the magnitude of observed effects on emotional reactivity in MDD. For example, one study (Kaviani et al.,2004) found that within MDD individuals, anxiety levels as measured by self-report measures were associated withincreased reactivity as measured by EMG for both positively and negatively valenced film clips. In sum, the presenceof normative co-morbidity in MDD samples precludes any strong claim that blunted emotional reactivity is specific toMDD. To advance such a claim, it would be useful to have carefully conducted comparisons of emotional reactivity




ARTICLE IN PRESS

between pure samples of MDD participants, co-morbid MDD participants and participants with other psychiatricconditions.

Moreover, heterogeneity of MDD samples with respect to symptom severity, and demographic characteristics arealso potential factors that influences the magnitude of PER and NER. Samples of MDD individuals in this meta-analysis were drawn from a variety of sources and included inpatients, outpatients, and untreated individuals recruitedfrom the general population. Preliminary moderator analyses designed to examine the relationship between depressionseverity and emotion reactivity found no relationship (see footnote 1), but because of incomplete data, this analysis didnot provide a strong test. Other work has suggested that the severity of MDD may have an effect on emotion reactivity(Rottenberg, Kasch et al., 2002). Finally, it is also possible that emotional reactivity in MDD varies as a function ofdepression subtype (e.g., atypical versus melancholic MDD). Unfortunately, the majority of the studies did not specifyinto which subtype individuals in their MDD samples fell. Finally, the sparse reporting of race, ethnicity, andsocioeconomic status, precluded analyses of these factors despite the fact that these demographic factors may beimportant in modifying emotional reactivity (e.g., Gallo, Bogart, Vranceanu, & Matthews, 2005).

In light of these considerations, we strongly recommend that researchers be more conscientious in their reporting onpossible moderators of emotional reactivity in MDD, including medications, co-morbid anxiety, gender, demographicvariables, depression severity, depression subtypes. Importantly, despite the multiple factors which might influencePER and NER, and the well known heterogeneity of depression, this meta-analysis found reliable effects for NER andPER. In fact, the effect sizes for NER and PER appear to be comparable in magnitude to those found in other meta-analyses in related domains (e.g., Matt, Vazquez, & Campbell, 1992; Mor & Winquist, 2002).

Finally, as noted earlier, a fourth limitation of this meta-analysis is that it did not include every response system thatis relevant to the emotion construct. Thus, one important avenue for future work will be to examine the generalizabilityof the findings presented here to other response systems. Interestingly, a recent meta-analysis of a substantial literatureon endocrine responses to stress in MDD (which focused on the hormone cortisol) found reduced reactivity in MDD(Burke, Davis, Otto, & Mohr, 2005), which is consistent with the ECI view. Evidence from neural systems involved inemotional responding in MDD is not easily summarized in narrative form, but the possibility should be noted thatfindings may be more mixed than our meta-analytic results. For example, Deldin, Keller, Gergen, and Miller (2000)used ERP to measure encoding of emotional stimuli in individuals with MDD and controls and found that controlsshowed enhanced P300 during encoding and reduced P300 during recognition of positive stimuli, which the authorsinterpreted as a response bias for positive information. Using fMRI, a blunted response in the amygdala to facialexpressions of fear (relative to neutral) has been observed in both depressed adults (Drevets, 2001) and depressedchildren (Thomas et al., 2001), which might be interpreted in terms of ECI. However, consistent with the negativepotentiation view, Siegle, Steinhauer, Thase, Stenger, and Carter (2002) found that MDD individuals exhibited asimilar initial amygdala response but greater sustained amygdala activity in response to negative words. In this samestudy, and underlining the complexity of this domain, the MDD individuals exhibited decreased reactivity indorsolateral prefrontal cortex (DLPFC), compared to controls for the same negative words. Relatedly, Canli, Sivers,Thomason, Whitfield-Gabrieli, Gabrieli, and Gotlib (2004) found that MDD individuals had increased reactivity tonegative words in the inferior parietal lobule (IPL) but decreased reactivity in the superior temporal gyrus (STG) andthe cerebellum. The high complexity of these neuroimaging findings again underlines our caution in not assuming thegeneralizability of these results to other systems of emotional response.

4.2. Conclusions

Clinical scientists have increasingly recognized the importance of emotion to understanding psychopathology(Kring & Bachorowski, 1999; Rottenberg & Johnson, 2007). Given this wide interest, the field has been surprisinglyslow to undertake quantitative reviews of emotional reactivity in Axis-I disorders. This report, as the first quantitativereview of emotion reactivity in major depression, begins to address this gap by suggesting that changes in emotionalreactivity are a reliable correlate of MDD. In addition to augmenting the clinical description of MDD as a disorder ofemotion, these analyses have implications for interventions designed to treat and prevent MDD. For example, to theextent that diminished emotional reactivity is a key affective deficit in MDD, it may be the case that individuals whostrongly exhibit this deficit will have a more pernicious course of disorder (e.g., Rottenberg, Wilhelm et al., 2002). Thislogic would also support the development of psychologically-based and pharmacological treatment techniques tobolster MDD patients' appropriate reactivity to both positive and negative emotional stimuli. In fact, given the




ARTICLE IN PRESS

centrality of affective disturbance in MDD, developing an accurate, system-by-system account of how MDD altersemotional reactivity, including testing hypothesized proximal mechanisms (e.g., individual differences in cognition orbiological functioning), is a sine qua non for developing more effective, targeted treatments for this disorder. Wesubmit this meta-analysis as a modest first step towards this goal.

References⁎

*Albus, M., Muller-Spahn, F., Ackenheil, M., & Engel, R. R. (1987). Psychiatric patients and their response to experimental stress. Stress Medicine,3, 233−238.

*Allen, N. B., Trinder, J., & Brennan, C. (1999). Affective startle modulation in clinical depression: Preliminary findings. Biological Psychiatry, 46,542−550.

American Psychiatric Association (2000). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author.Appelhans, B. M., & Luecken, L. J. (2006). Heart rate variability as an index of regulated emotional responding. Review of General Psychology, 10,

229−240.Barrett, L. F. (2006). Valence is a basic building block of emotional life. Journal of Research in Personality, 40, 35−55.Beach, S. R., & Amir, N. (2003). Is depression taxonic, dimensional, or both? Journal of Abnormal Psychology, 112, 228−236.Beck, A. T. (1976). Cognitive therapy and the emotional disorders. Madison, CT: International Universities Press.Beck, A. T., Rush, J., Shaw, B. F., & Emery, G. (1979). Cognitive therapy of depression. New York: The Guilford Press.Berenbaum, H., & Oltmanns, T. F. (1992). Emotional experience and expression in schizophrenia and depression. Journal of Abnormal Psychology,

101, 37−44.Biostat. (2005). Comprehensive meta-analysis [Computer software].Bradley, M. M., Cuthbert, B. N., & Lang, P. J. (1999). Affect and the startle reflex. In M. E. Dawson A. Schell & A. Boehmelt (Eds.), Startle

modification: implications for neuroscience, cognitive science and clinical science (pp. 242−276). Stanford, CA: Stanford University Press.Bradley, M. M., & Lang, P. J. (1994). Measuring emotion: The Self-Assessment Manikin and the semantic differential. Journal of Behavior Therapy

and Experimental Psychiatry, 25, 49−59.Bradley, M. M., & Lang, P. J. (2000). Measuring emotion: Behavior, feeling, and physiology. In R. D. Lane & L. Nadel (Eds.), Cognitive

neuroscience of emotion (pp. 106−128). New York: Oxford University Press.Butler, E. A., Wilhelm, F. H., & Gross, J. J. (2006). Respiratory sinus arrhythmia, emotion, and emotion regulation during social interaction.

Psychophysiology, 43, 612−622.Bower, G. H. (1981). Mood and memory. American Psychologist, 36, 129−148.Burke, H. M., Davis, M. C., Otto, C., & Mohr, D. C. (2005). Depression and cortisol responses to psychological stress: A meta-analysis. Psycho-

neuroendocrinology, 30, 846−856.Cacioppo, J. T., Martzke, J. S., Petty, R. E., & Tassinary, L. G. (1988). Specific forms of facial EMG response index emotions during an interview:

From Darwin to the continuous flow hypothesis of affect-laden information processing. Journal of Personality and Social Psychology, 54,592−604.

Cacioppo, J. T., Petty, R. E., Losch, M. E., & Kim, H. S. (1986). Electromyographic activity over facial muscle regions can differentiate the valenceand intensity of affective reactions. Journal of Personality and Social Psychology, 50, 260−268.

Canli, T., Sivers, H., Thomason, M. E., Whitfield-Gabrieli, S., Gabrieli, J. D., & Gotlib, I. H. (2004). Brain activation to emotional words in depressedvs. healthy subjects. NeuroReport, 15, 2585−2588.

Carney, R. M., Hong, B. A., Brim, J., Plesons, D., & Clayton, P. (1983). Facial electromyography and reactivity in depression. The Journal ofNervous and Mental Disease, 171, 312−313.

Carver, C. S. (2001). Affect and the functional bases of behavior: On the dimensional structure of affective experience. Personality and SocialPsychology Review, 5, 345−356.

Clark, L. A., Watson, D., & Mineka, S. (1994). Temperament, personality, and the mood and anxiety disorders. Journal of Abnormal Psychology.Special Issue: Personality and Psychopathology, 103, 103−116.

Cochran, W. (1954). The combination of estimates from different experiments. Biometrics, 10, 101−129.Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.*Dawson, M. E., Schell, A. M., & Catania, J. J. (1977). Autonomic correlates of depression and clinical improvement following electroconvulsive

shock therapy. Psychophysiology, 14, 569−577.Deldin, P. J., Keller, J., Gergen, J. A., & Miller, G. A. (2000). Right-posterior face processing anomaly in depression. Journal of Abnormal

Psychology, 109, 116−121.Depue, R. A., & Iacono, W. G. (1989). Neurobehavioral aspects of affective disorders. Annual Review of Psychology, 40, 457−492.*Dichter, G., Tomarken, A., Shelton, R., & Sutton, S. (2004). Early- and late-onset startle modulation in unipolar depression. Psychophysiology, 41,

433−440.Dolan, R. J. (2002). Emotion, cognition and behavior. Science, 298, 1191−1194.Drevets, W. C. (2001). Neuroimaging and neuropathological studies of depression: implications for the cognitive-emotional features of mood

disorders. Current Directions in Neurobiology, 2001, 240−249.

⁎ Studies included in the meta-analyses.




ARTICLE IN PRESS

*Dunn, B. D., Dalgleish, T., Lawrence, A. D., Cusack, R., & Ogilvie, A. (2004). Categorical and dimensional reports of experienced affect toemotion-inducing pictures in depression. Journal of Abnormal Psychology, 113, 654−660.

Ekman, P. (1976). Pictures of facial affect. Palo Alto, CA: Consulting Psychologists Press.Ekman, P. (1992). An argument for basic emotions. Cognition and Emotion, 6, 169−200.Ekman, P., Matsumoto, D., & Friesen, W. V. (2005). Facial expression in affective disorders. In P. Ekman & E. L. Rosenberg (Eds.), What the face

reveals: basic and applied studies of spontaneous expression using the facial action coding system (FACS) (pp. 429−440). (2nd edition) New York:Oxford University Press.

Ekman, P., & Oster, H. (1979). Facial expressions of emotion. Annual Review of Psychology, 30, 527−554.*Forbes, E. E., Miller, A., Gohn, J., Fox, N., & Kovacs, M. (2005). Affect-modulated startle in adults with childhood onset depression: relations to

bipolar course and number of lifetime depressive episodes. Psychiatry Research, 134, 11−25.Fox, P. T., Parsons, L. M., & Lancaster, J. L. (1998). Beyond the single study: Function/location metanalysis in cognitive neuroimaging. Current

Opinion in Neurobiology, 8, 178−187.Gallo, L. C., Bogart, L. M., Vranceanu, A., & Matthews, K. A. (2005). Socioeconomic status, resources, psychological experiences, and emotional

responses: A test of the reserve capacity model. Journal of Personality and Social Psychology, 88, 386−399.*Gehricke, J. G., & Shapiro, D. (2000). Reduced facial expression and social context in major depression: Discrepancies between facial muscle

activity and self reported emotion. Psychiatry Research, 95, 157−167.Golin, S., Hartman, S. A., Klatt, E. N., Munz, K., & Wolfgang, G. L. (1977). Effects of self-esteem manipulation on arousal and reactions to sad

models in depressed and nondepressed college students. Journal of Abnormal Psychology, 86, 435−439.Gray, J. R. (2004). Integration of emotion and cognitive control. Current Directions in Psychological Science, 13, 46−48.*Greden, J. F., Genero, N., Price, L., Feinberg, M., & Levine, S. (1986). Facial electromyography in depression. Archives of General Psychiatry, 43,

269−274.Greden, J. F., Price, L., Genero, N., Feinberg, M., & Levine, S. (1984). Facial EMG activity levels predict treatment outcome in depression.

Psychiatry Research, 13, 345−352.Gross, J. J. (2002). Emotion regulation: Affective, cognitive, and social consequences. Psychophysiology, 39, 281−291.*Guinjoan, S. M., Bernabo, J. L., & Cardinali, D. P. (1995). Cardiovascular tests of autonomic function and sympathetic skin responses in patients

with major depression. Journal of Neurology, Neurosurgery and Psychiatry, 59, 299−302.Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis. San Diego, CA: Academic Press.*Henriques, J. B., & Davidson, R. J. (1991). Left frontal hypoactivation in depression. Journal of Abnormal Psychology, 100, 535−545.Henriques, J. B., & Davidson, R. J. (2000). Decreased responsiveness to reward in depression. Cognition and Emotion, 14, 711−724.Henriques, J. B., Glowacki, J. M., & Davidson, R. J. (1994). Reward fails to alter response bias in depression. Journal of Abnormal Psychology, 103,

460−466.Hunter, J. E., & Schmidt, F. L. (2000). Fixed effects vs. random effects meta-analysis models: Implications for cumulative research knowledge.

International Journal of Selection and Assessment, 8, 275−292.Iacono,W. G., Lykken, D. T., Haroian, K. P., Peloquin, L. J., Valentine, R. H., & Tuason, V. B. (1984). Electrodermal activity in euthymic patients with

affective disorders: One-year retest stability and the effects of stimulus intensity and significance. Journal of Abnormal Psychology, 93, 304−311.Katsikitis, B. A., & Pilowsky, I. (1991). A controlled quantitative study of facial expression in Parkinson's disease and depression. Journal of

Nervous and Mental Disease, 179, 683−688.*Kaviani, H., Gray, J. A., Checkley, S. A., Raven, P. W., Wilson, G. D., & Kumari, V. (2004). Affective modulation of the startle response in

depression: Influence of the severity of depression, anhedonia, and anxiety. Journal of Affective Disorders, 83, 21−31.Keltner, D., & Gross, J. J. (1999). Functional accounts of emotions. Cognition and Emotion, 13, 575−599.Kessler, R. C. (2002). Epidemiology of depression. In I. H. Gotlib & C. L. Hammen (Eds.), Handbook of depression (pp. 23−42). New York:

Guilford Press.Krantz, D. S., Manuck, S. B., & Wing, R. R. (1986). Psychological stressors and task variables as elicitors of reactivity. In K. A. Matthews S. M.

Weiss T. Detre T. M. Dembroski B. Falkner S. B. Manuck & R. B. Williams (Eds.), Handbook of stress reactivity and cardiovascular disease(pp. 85−107). New York: Wiley-Interscience.

Kring, A. M., & Bachorowski, J. A. (1999). Emotion and psychopathology. Cognition and Emotion, 13, 575−600.Lang, P. J. (1968). Fear reduction and fear behavior: Problems in treating a construct. In J. M. Shlien (Ed.), Research in Psychotherapy, Vol. 3

(pp. 90–103). Washington, DC: American Psychological Association.Lang, P. J. (1988). What are the data of emotion? In V. Hamilton G. H. Bower & N. H. Frijda (Eds.), Cognitive perspectives on emotion and

motivation (pp. 173−191). New York: Kluwer Academic/Plenum Press.Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (2005). International affective picture system (IAPS): Affective ratings of pictures and instruction

manual. Technical Report A-6. Gainesville, FL: University of Florida.Lang, P. J., Greenwald, M. K., Bradley, M. M., & Hamm, A. O. (1993). Looking at pictures: affective, facial, visceral, and behavioral reactions.

Psychophysiology, 30, 261−273.Lang, P. J., Rice, D. G., & Sternbach, R. A. (1972). The psychophysiology of emotion. In N. S. Greenfield & R. A. Sternbach (Eds.), Handbook of

psychophysiology (pp. 623−643). New York: Holt, Rinehart, & Winston.Lapierre, Y. D., & Butter, H. J. (1980). Agitated and retarded depression: A clinical psychophysiological evaluation. Neuropsychobiology, 6,

217−223.Lazarus, R. S. (1991). Emotion and adaptation. London: Oxford University Press.LeDoux, J. (1998). The emotional brain: the mysterious underpinnings of emotional life. New York: Simon & Schuster.Levenson, R. W. (1994). Human emotions: A functional view. In P. Ekman & R. J. Davidson (Eds.), The nature of emotion: Fundamental questions

(pp. 123−126). New York: Oxford University Press.




ARTICLE IN PRESS

Lewinsohn, P. M., Lobitz, W. C., & Wilson, S. (1973). “Sensitivity” of depressed individuals to aversive stimuli. Journal of Abnormal Psychology,81, 259−263.

Matt, G. E., Vazquez, C., & Campbell, W. K. (1992). Mood-congruent recall of affectively toned stimuli: A meta-analytic review. ClinicalPsychology Review, 12, 227−255.

Mauss, I. B., Levenson, R. W., Carter, L., Wilhelm, F. H., & Gross, J. J. (2005). The tie that binds? Coherence among emotion experience, behavior,and physiology. Emotion, 5, 175−190.

Mauss, I. B., Wilhelm, F. H., & Gross, J. J. (2004). Is there less to social anxiety than meets the eye? Emotion experience, expression, and bodilyresponding. Cognition and Emotion, 18, 631−662.

Melamed, S. (1987). Emotional reactivity and elevated blood pressure. Psychosomatic Medicine, 49, 217−225.Mor, N., & Winquist, J. (2002). Self-focused attention and negative affect: A meta-analysis. Psychological Bulletin, 128, 638−662.Murphy, F. C., Nimmo-Smith, I., & Lawrence, A. D. (2003). Functional neuroanatomy of emotions: A meta-analysis. Cognitive, Affective and

Behavioral Neuroscience, 3, 207−233.Nesse, R. M. (2000). Is depression an adaptation? Archives of General Psychiatry, 57, 14−20.Peeters, F., Berkhof, J., Delespaul, P., Rottenberg, J., & Nicolson, N. A. (2006). Diurnal mood variation in major depressive disorder. Emotion, 6,

383−391.Peeters, F., Nicolson, N. A., & Berkhof, J. (2003). Cortisol responses to daily events in major depressive disorder. Psychosomatic Medicine, 65,

836−841.Peeters, F., Nicolson, N. A., Berkhof, J., Delespaul, P., & deVriews, M. (2003). Effects of daily events on mood states in major depressive disorder.

Journal of Abnormal Psychology, 112, 203−211.*Persad, S. M., & Polivy, J. (1993). Differences between depressed and nondepressed individuals in the recognition of and response to facial

emotional cues. Journal of Abnormal Psychology, 102, 358−368.Phan, K. L., Wager, T., Taylor, S. F., & Liberzon, I. (2002). Functional neuroanatomy of emotion: A meta-analysis of emotion activation studies in

PET and fMRI. NeuroImage, 16, 331−348.Renneberg, B., Heyn, K., Gebhard, R., & Bachmann, S. (2005). Facial expression of emotions in borderline personality disorder and depression.

Journal of Behavior Therapy and Experimental Psychiatry, 36, 183−196.Ritz, T., Thons, M., Fahrenkrug, S., & Dahme, B. (2005). Airways, respiration, and respiratory sinus arrhythmia during picture viewing. Psycho-

physiology, 42, 568−578.Rosenberg, E. L. (1998). Levels of analysis and the organization of affect. Review of General Psychology, 2, 247−270.Rosenthal, R. (1991). Meta-analytic procedures for social research. Thousand Oaks, CA: Sage.Rosenthal, R., & Rubin, D. B. (1986). Meta-analytic procedures for combining studies with multiple effects sizes. Psychological Bulletin, 99,

400−406.Rottenberg, J. (2005). Mood and emotion in major depression. Current Directions in Psychological Science, 14, 167−170.Rottenberg, J. (2007). Major depressive disorder: Emerging evidence for emotion context insensitivity. In J. Rottenberg & S. L. Johnson (Eds.),

Emotion and psychopathology: Bridging affective and clinical science. Washington DC: American Psychological Society.Rottenberg, J., & Gross, J. J. (2003). When emotion goes wrong: realizing the promise of affective science. Clinical Psychology: Science and

Practice, 10, 227−232.*Rottenberg, J., Gross, J. J., & Gotlib, I. H. (2005). Emotion context insensitivity in major depressive disorder. Journal of Abnormal Psychology, 114,

627−639.Rottenberg, J. & Johnson, S. L. (Eds.). (2007). Emotion and psychopathology: Bridging affective and clinical science. Washington, D.C.: APA

Books.*Rottenberg, J., Kasch, K. L., Gross, J. J., & Gotlib, I. H. (2002). Sadness and amusement reactivity differentially predict concurrent and prospective

functioning in major depressive disorder. Emotion, 2, 135−146.Rottenberg, J., Ray, R. D., & Gross, J. J. (2007). Emotion elicitation using films. In J. A. Coan & J. J. B. Allen (Eds.), The handbook of emotion

elicitation and assessment. London: Oxford University Press.Rottenberg, J., Salomon, K., Gross, J., & Gotlib, I. H. (2005). Vagal withdrawal to a sad film predicts subsequent recovery from depression. Psy-

chophysiology, 42, 277−281.Rottenberg, J., Wilhelm, F. H., Gross, J. J., & Gotlib, I. H. (2002). Respiratory sinus arrhythmia as a predictor of outcome in major depressive

disorder. Journal of Affective Disorders, 71, 265−272.*Rottenberg, J., Wilhelm, F. H., Gross, J. J., & Gotlib, I. H. (2003). Vagal rebound during resolution of tearful crying among depressed and

nondepressed individuals. Psychophysiology, 40, 1−6.Scher, C. D., Ingram, R. F., & Segal, Z. V. (2005). Cognitive reactivity and vulnerability: Empirical evaluation of construct activation and cognitive

diatheses in unipolar depression. Clinical Psychology Review, 25, 487−510.Schwartz, G. E., Fair, P. L., Mandel, M. R., Salt, P., Mieske, M., & Klerman, G. L. (1976). Facial electromyography in the assessment of improvement

in depression. Psychosomatic Medicine, 40, 355−360.Schwartz, G. E., Fair, P. L., Salt, P., Mandel, M. R., & Klerman, G. L. (1976). Facial expression and imagery in depression: An electromyographic

study. Psychosomatic Medicine, 38, 337−347.Siegle, G. J., Steinhauer, S. R., Thase, M. E., Stenger, V. A., & Carter, C. S. (2002). Can't shake that feeling: fMRI assessment of sustained amygdala

activity in response to emotional information in depressed individuals. Biological Psychiatry, 51, 693−707.*Sigmon, S., & Nelson-Gray, R. (1992). Sensitivity to aversive events in depression: Antecedent, concomitant, or consequent? Journal of

Psychopathology and Behavioral Assessment, 14, 225−246.*Sloan, D. M., Strauss, M. E., Quirk, S. W., & Sajatovic, M. (1997). Subjective and expressive emotional responses in depression. Journal of

Affective Disorders, 46, 135−141.




ARTICLE IN PRESS

*Sloan, D. M., Strauss, M. E., & Wisner, K. L. (2001). Diminished response to pleasant stimuli by depressed women. Journal of AbnormalPsychology, 110, 488−493.

Thomas, K. M., Dahl, R. E., Ryan, N. D., Neal, D., Birmaher, B., Eccard, C. H., et al. (2001). Amygdala response to faces in anxious and depressedchildren. Archives of General Psychiatry, 58, 1057−1063.

Tomarken, A. J., Shelton, R. C., & Hollon, S. D. (2007). Affective science as a framework for understanding the mechanisms and effects ofantidepressant medications. In J. Rottenberg & S. L. Johnson (Eds.), Emotion and psychopathology: Bridging affective and clinical science.Washington DC: American Psychological Society.

Tremeau, F., Malaspina, D., Duval, F., Correa, H., Hager-Budny, M., Coin-Barious, L., et al. (2005). Facial expressiveness in patients withschizophrenia compared to depressed patients and nonpatient comparison subjects. American Journal of Psychiatry, 162, 92−101.

*Tsai, J. L., Pole, N., Levenson, R. W., & Munoz, R. F. (2003). The effects of depression on the emotional responses of Spanish speaking Latinas.Cultural Diversity and Ethnic Minority Psychology, 9, 49−63.

Watson, D. (2000). Mood and temperament. New York: The Guilford Press.Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales.

Journal of Personality and Social Psychology, 54, 1063−1070.Wexler, B. E., Levenson, L., Warrenburg, S., & Price, L. (1994). Decreased perceptual sensitivity to emotion-evoking stimuli in depression.

Psychiatry Research, 51, 127−138.Yecker, S., Borod, J. C., Brozgold, A., Martin, C., Alpert, M., & Welkowitz, J. (1999). Lateralization of facial emotional expression in schizophrenic

and depressed patients. Journal of Neuropsychiatry and Clinical Neurosciences, 11, 370−379.Zvolensky, M. J., & Schmidt, N. B. (2003). Panic disorder and smoking. Clinical Psychology: Science and Practice, 10, 29−51.



article in press - university of south floridauweb.cas.usf.edu/mood/docs/bylsma_inpress.pdfand...

Documents