Available online at www.sciencedirect.

com

Clinical Psychology Review 28 (2008) 676 691

A meta-analysis of emotional reactivity in major 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 the literature: positive attenuation (reduced positive reactivity), negative potentiation (increased negative reactivity), and emotion context insensitivity (ECI; reduced positive and negative reactivity). Although empirical studies have accumulated on emotional reactivity in MDD, this report is to our knowledge the first systematic quantitative review of this topic area. In omnibus analyses of 19 laboratory studies comparing the emotional reactivity of healthy individuals to that of individuals with MDD, MDD was characterized by reduced emotional reactivity to both positively and negatively valenced stimuli, with the reduction larger for positive stimuli (d = .53) than for negative stimuli (d = .25). Results were similar when 3 major emotion response systems (selfreported experience, expressive behavior, and peripheral physiology) were analyzed individually. The ECI view of emotional reactivity 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 . . . . . . . . . . . . . . . . . . . . . . . 1.1. Views of emotional reactivity in MDD . . . . . 1.2. Limitations of prior reviews . . . . . . . . . . . 1.3. The present study . . . . . . . . . . . . . . . . Methods . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Overview. . . . . . . . . . . . . . . . . . . . . 2.2. Rationale for emotion response domains covered . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . in this review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 677 678 678 679 679 679 679

2.

The authors express their appreciation to Michelle Jones for her help in library research and to the members of the Mood and Emotion Laboratory for 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: lbylsma@mail.usf.edu (L.M. Bylsma), bhmorris@mail.usf.edu (B.H. Morris), jrottenb@cas.usf.edu (J. Rottenberg). 0272-7358/$ - see front matter 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.cpr.2007.10.001

L.M. Bylsma et al. / Clinical Psychology Review 28 (2008) 676691

677

2.3. Variables included in this review . . . . . . . 2.4. Literature search . . . . . . . . . . . . . . . 2.5. Study inclusion criteria . . . . . . . . . . . . 2.6. Characteristics of selected studies . . . . . . . 2.7. Computation of effect sizes . . . . . . . . . . 2.8. Planned analyses . . . . . . . . . . . . . . . 3. Results . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Omnibus analyses . . . . . . . . . . . . . . . 3.2. PER and NER in individual response systems 3.2.1. PER . . . . . . . . . . . . . . . . . 3.2.2. NER . . . . . . . . . . . . . . . . . 3.3. Heterogeneity and Moderator Analyses . . . . 3.4. File drawer analyses . . . . . . . . . . . . . 4. Discussion . . . . . . . . . . . . . . . . . . . . . . 4.1. Limitations and future directions . . . . . . . 4.2. Conclusions . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

680 680 680 681 681 683 684 684 685 685 685 685 686 686 687 688 688

1. Introduction Major Depressive Disorder (MDD) is a debilitating condition that afflicts almost a sixth of the general population (Kessler, 2002). Reflecting the profound disturbance of affective function in MDD, it is classified as a mood disorder by the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV; APA, 2000). DSM-IV diagnostic 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 been diagnosed with MDD reliably report low positive affect and elevated negative affect on a variety of questionnaire and interview measures (Clark, Watson, & Mineka, 1994). Durable disturbance of mood is thus one of the most salient features of MDD. Given that MDD is quintessentially a disorder of mood, one important question is how MDD influences emotional reactions. 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 in the environment (e.g., Watson, 2000). By contrast, emotions have been defined as quick-moving reactions that occur when an individual processes a meaningful stimulus (e.g., a threatening object). Emotional reactions are multicomponential 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 and emotion are so distinguished, it becomes apparent that the various diagnostic criteria for MDD, such as pervasive sadness or anhedonia, indicate alterations in mood, but do not indicate alterations in emotion with corresponding specificity. Although the constructs are distinguishable, moods and emotions have generally been seen as interconnected, with moods altering the probability of having specific emotions (e.g., Rosenberg, 1998). More specifically, moods are thought to potentiate like-valenced or matching emotions (e.g., irritable mood facilitates angry reactions, an anxious mood facilitates panic, etc; Rottenberg, 2005). By extension, excessive negative mood in MDD would potentiate negative emotional reactivity and/or a lack of positive mood would attenuate positive emotional reactivity. Indeed, the idea 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 each of these views before presenting a meta-analysis designed to determine which of them best fits the accumulated data concerning emotional reactivity in MDD. Our main goal was to conduct a systematic meta-analytic review of the literature on emotional reactivity in MDD. We used rigorous literature search strategies (see below) to identify laboratory studies of emotional reactivity in MDD and estimate effect sizes. We report effect sizes from 19 laboratory studies that compared negative and positive emotional reactivity between individuals with MDD and healthy participants. Analyses focused on omnibus group differences in positive and negative emotional reactivity, as well as the replicability of these group differences within three specific emotion response systems: self-report, behavioral, and peripheral physiological reactivity. To our knowledge, this study represents the first meta-analytic review of the literature on emotional reactivity in MDD.

678

L.M. Bylsma et al. / Clinical Psychology Review 28 (2008) 676691

1.1. Views of emotional reactivity in MDD Negative potentiation, the first view, holds that the pervasive negative mood states that are prevalent in MDD contribute to potentiated emotional reactivity to negative emotional cues. Perhaps most relevant to the idea that negative moods and negative emotions are mutually reinforcing in MDD, cognitive theorists see negative moods as facilitating negative cognitive processing which, in turn, results in negative cognitive interpretations that generate dysphoric reactions (e.g., Beck, 1976; Beck, Rush, Shaw, & Emery, 1979). Beck's schema model and related theories of depression (e.g., Bower, 1981) conceptualize MDD in terms of cognitive structures, or schemas, which are patterns in thinking that serve to negatively distort the processing of emotional stimuli (e.g., such as beliefs that one is unlovable or a failure). Importantly, according to these theories, negative mood states prime, or activate, these cognitive structures (Scher, Ingram, & Segal, 2005). Once activated, these structures precipitate negative interpretations which give rise to depressotypic emotional responses (e.g., crying spells) whenever schema-matching negative emotion stimuli are encountered, presumably potentiating reactivity to negative emotional stimuli in MDD. Positive attenuation, the second view, holds that individuals with MDD will have reduced reactivity in response to positive emotional stimuli. Because this hypothesis applies primarily to positive emotional stimuli, positive attenuation is compatible with negative potentiation (i.e., individuals with MDD can exhibit both patterns simultaneously). The starting point for this hypothesis is the depressed persons' strong tendency to exhibit low positive mood. Indeed, anhedonia (the reduced ability to experience pleasure) is one of the cardinal symptoms of MDD, 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 their accounts of abnormal emotional responding (e.g., abnormally low emotion response magnitude) in MDD on this constellation of motivational deficits (e.g., Clark et al., 1994; Depue & Iacono, 1989) often focusing on the central nervous 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 reduced reactivity to all emotion cues, regardless of valence (Rottenberg, 2005, 2007). By this account, individuals with MDD should exhibit less reactivity to both positive and negative stimuli and events compared to healthy individuals. ECI is derived from evolutionary accounts that describe depression as a defensive motivational state that fosters environmental disengagement (Nesse, 2000). According to this view, depressed mood states evolved as internal signals to bias organisms against action in adverse situations where continued activity might be potentially be dangerous or wasteful (e.g., famine). Thus, according to ECI, severe depressed mood states in MDD are postulated to inhibit ongoing emotional reactivity. In sum, ECI makes similar predictions as the positive attenuation view for positive stimuli, but makes opposite predictions as the negative potentiation view for negative stimuli. 1.2. Limitations of prior reviews Prior generalizations about emotional reactivity in MDD have been based on narrative reviews, which can be susceptible to selection biases and are often unsystematic in their search strategies. Any review also faces the complexity of operationalizing the major constructs relevant to this topic area. For example, the depression construct can be operationalized in different kinds of samples (e.g., analogue samples of dysphoric persons versus case-level MDD), and even studies of MDD may consider related conditions, such as dysthymia or minor depression, to be part of the depression construct. Thus, reviews may describe different effects of depression on emotional reactivity because different kinds of depressive phenomena are being considered. Likewise, complexity in the emotion construct also complicates 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, 2005). Prior reviews have not made a concerted effort to integrate data across self-report, behavioral, and physiological domains (e.g., Forbes, Miller, Gohn, Fox, & Kovacs, 2005; Gehricke & Shapiro, 2000). Further, even within any given system of response, emotional reactivity can be computed using different metrics (e.g., level scores versus change scores), and commentators have generally not been explicit and/or systematic in defining their choice of reactivity metric. Given these multiple challenges, perhaps it is not surprising that prior narrative reviews in this area of research have been inconclusive.

L.M. Bylsma et al. / Clinical Psychology Review 28 (2008) 676691

679

1.3. The present study To overcome the limitations of prior reviews, we assessed the literature on emotional reactivity in MDD using meta-analysis an important quantitative technique for systematically summarizing results across different studies 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 precise quantitative estimates of the magnitude of effect sizes. For this meta-analysis, only studies that elicited emotion in a well-controlled manner and used DSM diagnostic criteria for MDD were included. Analyses focused on the extent to which MDD might increase or decrease positive and negative emotional reactivity relative to healthy individuals. 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 positive emotional 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. We focused on these two forms of emotional reactivity rather than specific emotions because (1) the major views of emotional reactivity in 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 are potentially relevant to emotion. Relatedly, there is a lack of consensus among affective scientists regarding just how many emotion response 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 to focus 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 for four 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 good agreement that these three systems do in fact index emotional reactivity. Third, there is reasonably good agreement on measures that can be extracted from these systems to index emotional reactivity (see below). Fourth, there is an adequate database of empirical 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 and endocrine systems play a major role in emotional reactivity (e.g., LeDoux, 1998). However, we did not include neuroimaging metrics (e.g., techniques such as fMRI, PET, and ERP), because existing reviews of neuroimaging and emotion indicate that considerable 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 studies indicated that imaging studies of MDD did not routinely report statistics needed for computing effect sizes. Further, the technical problems of conducting metaanalytic analyses of neuroimaging data are considerable, as these are far from simple computations of effect sizes, and, typically, effect sizes based on standardized mean differences cannot be calculated due to the complexity of the 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 meta-analytic procedures must be used for neuroimaging data that have different

680

L.M. Bylsma et al. / Clinical Psychology Review 28 (2008) 676691

theoretical interpretations. Likewise, we did not consider 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 that cognition is extraordinarily heterogeneous as a domain, and that there is little agreement or precedent concerning how metrics from the cognitive domain (e.g., memory bias, perceptual judgments, reaction times, etc) should be mapped onto the emotional reactivity 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-reported experience, expressive behavioral, and peripheral physiological) was guided by prior literature demonstrating that each measure could be interpreted as a valid index of emotional reactivity. Self-report measures of emotional responses included subjective reports of emotional experience through various self-report measures such as the PANAS (Watson, Clark, & Tellegen, 1988), the SAM (Bradley & Lang, 1994), or discrete emotion measures. Behavioral measures included observed facial expression as measured 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), emotionmodulated eyeblink/startle responses (e.g., Bradley, Cuthbert, & Lang, 1999; Bradley & Lang, 2000), and approach/ avoidance behavior (e.g., Carver, 2001; Henriques & Davidson, 2000). Physiological measures 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 for possible inclusion, because published work generally has increased scientific rigor compared with unpublished work, and because it is readily available within the public domain for independent evaluation by others. Because null results may be more difficult to publish 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 the interpretation of the primary results (Rosenthal, 1991). To generate potentially relevant articles, the first author used the PsychINFO and MEDLINE online database, entering the following 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. After identifying potentially relevant articles, the reference sections were searched to identify additional articles which might meet our inclusion criteria. As an additional check to ensure that no relevant articles were omitted, a research assistant also performed a parallel 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 current MDD using DSM criteria (b) a healthy control group, (c) appropriate statistics (e.g., means and standard deviations) that were reported in the published article or available from the author upon request (of nine articles that were missing data, the study authors were able to supply it in six cases), (d) measured emotional reactivity with self-reported experience, expressive behavioral, or peripheral 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 valenced stimulus may reflect a general disposition or mood rather than reactivity to a given stimulus). Medication use by patients was not considered as an inclusion/exclusion criterion because virtually no studies in the entire literature used an un-medicated sample (our search yielded only four such studies).

L.M. Bylsma et al. / Clinical Psychology Review 28 (2008) 676691

681

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 not include a neutral or baseline measure, 2 were not laboratory studies, and 3 did not include the necessary means and/or standard deviations for inclusion in the meta-analysis and the authors were unable to be contacted. The final sample for inclusion in the analysis consisted of 19 articles. All articles included measures of NER, and 14 also included measures of PER. Because many of the articles includes measures of multiple response systems, overall there were 29 measures of NER and 24 measures of PER across response systems. The pooled sample size consisted of 465 participants with MDD and 452 control participants. See Table 1 for a summary 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 of MDD. The majority of the studies included more females than males in their samples, consistent with the prevalence rate of depression 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-morbid anxiety diagnoses were typically included (only two studies stated that they excluded for all anxiety disorders). Five studies specified including inpatient MDD samples, and 11 specified using outpatient samples. The majority of included studies did not report ethnicity or other participant demographic information. Only one study (Tsai, Pole, Levenson, & Munoz, 2003) used a minority sample 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

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).

682

Table 1 Selected characteristics of studies included in the meta-analysis Citation N MDD / Control Dawson, Schell, and Catania (1977) Greden, Genero, Price, Feinberg, and Levine (1986) Albus, Muller-Spahn, Ackenheil, and Engel (1987) Sigmon and Nelson-Gray (1992) Persad and Polivy (1993) Guinjoan, Bernabo, and Cardinali (1995) Sloan, Strauss, Quirk, & Sajatovic (1997) Allen, Trinder, & Brennan (1999) Gehricke & Shapiro (2000) Henriques & Davidson (2000) Sloan, Strauss, and Wisner (2001) Rottenberg, Kasch, Gross, and Gotlib (2002) 20 / 20 63 / 37 12 / 63 20 / 20 16 / 16 18 / 18 24 / 23 14 / 14 11 / 11 18 / 15 20 / 20 72 / 32 95% 0% 0% Unknown Unknown 0% 100% 100% 27% 0% 100% 43% NER NER, PER NER NER, PER NER NER NER, PER NER, PER NER, PER NER, PER NER, PER NER, PER Stress tasks Sad and happy imagery Stress tasks Audiotape of positive and negative social interactions Emotional faces from the FAB Stress tasks IAPS pleasant and unpleasant pictures IAPS pleasant and unpleasant pictures Sad and happy imagery Reward, punishment IAPS pleasant and unpleasant pictures Sad, fearful, and amusing films Sad film Sad and amusing films Rest, tones Rest Rest Audiotape of neutral social interactions Neutral faces Rest IAPS neutral pictures IAPS neutral pictures Rest Neutral condition IAPS neutral pictures Neutral film Physiological: HR and SC Behavioral: facial EMG Physiological: HR, SC Self-report: DACL and affective ratings Physiological: SC Self-report: reported sadness, fear to faces Physiological: HR, BP, SC Self-report: affective ratings of pictures Behavioral: facial expression (FACS) Self-report: affective ratings of pictures Behavioral: emotion-modulated startle Self-report: DES Behavioral: facial EMG Self-report: PANAS Behavioral: response to reward/punishment Self-report: affective ratings of pictures Behavioral: facial expression (FACS) Self-report: reported sadness, amusement to films Behavioral: facial expression (FACS) Physiological: SC, HR Physiological: RSA, HR, respiration Self-report: emotion inventory Behavioral: facial expression (FACS, smiles) Physiological: SC, HR, FPA, respiration Self-report: affecting ratings of pictures Behavioral: emotion-modulated startle Self-report: emotional response to pictures Self-report: affective ratings of films Behavioral: emotion-modulated startle Self-report: affective ratings of pictures Behavioral: emotion-modulated startle Self-report: emotional response to films Behavioral: facial EMG Physiological: HR, FPA, SC, respiration Medicated Reactivity Emotional stimuli Baseline/neutral Reactivity measures

L.M. Bylsma et al. / Clinical Psychology Review 28 (2008) 676691

Rottenberg et al. (2003) Tsai et al. (2003)

25 / 31 11 / 9

0% Unknown

NER NER, PER

Neutral film Neutral film

Dichter, Tomarken, Shelton, and Sutton (2004) Dunn, Dalgleish, Lawrence, Cusack, and Ogilvie (2004) Kaviani et al. (2004) Forbes et al. (2005) Rottenberg, Gross, and Gotlib (2005)

17 / 16 25 / 25 22 / 22 38 / 38 19 / 22

100% 100% 82% 15% 32%

NER, PER NER, PER NER, PER NER, PER NER, PER

IAPS pleasant and unpleasant pictures IAPS pleasant and unpleasant pictures Unpleasant and pleasant films IAPS pleasant and unpleasant pictures Sad and happy films

IAPS neutral pictures IAPS neutral pictures Neutral film IAPS neutral pictures Neutral film

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 not included to establish reactivity.

L.M. Bylsma et al. / Clinical Psychology Review 28 (2008) 676691

683

d, a common measure of effect size (Cohen, 1988). Effect sizes were calculated as (M1M2)/SDp where M1 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 these means allowed us to account for scaling differences between different types of measures and pool data across studies. The means used to compute the effect sizes were the average self-report, behavioral, or physiological emotional reactivity scores. If the study did not report emotional reactivity scores directly, these scores were derived by subtracting neutral or baseline condition mean scores from positive or negative response mean scores. In this case, the standard deviation was first pooled from the standard deviations of the positive and negative response mean scores and the neutral or baseline conditions, and then this pooled standard deviation was used in the equation above to pool across groups in the calculation of the effect sizes. All effect sizes were calculated using the Comprehensive Meta-analysis software package (Biostat, 2005). A variety of stimuli in the primary studies were used to elicit valenced emotional responses that have received some support from previous research as being valid measures of emotional reactivity, including emotional films or audio recordings describing an emotional 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 & Oster, 1979), standardized laboratory stressors such as a loud noise or mental arithmetic tasks (Krantz, Manuck, & Wing, 1986), and the presentation of rewards (Henriques, Glowacki, & Davidson, 1994). Most of the behavioral and physiological studies also used self-report measures to validate whether the appropriate emotion was elicited. In order to ensure that emotional reactivity, defined as emotional reaction to a particular positively or negatively valenced stimulus, rather than a more general disposition of mood was assessed, we defined PER and NER as a change from a baseline condition. Rest and response to a neutral stimulus (e.g., a neutral film) were both considered acceptable baselines. We felt this liberal definition of baseline was justified in light of disagreement concerning what constitutes an optimal baseline condition for assessing the effects of emotional reactivity (for a discussion see, for example, Rottenberg et al., 2007). The more liberal definition of baseline also allowed us to retain as many studies as possible in the meta-analysis. Nevertheless, when both a resting baseline and a neutral baseline were reported in the same study, we used the neutral condition, because we believe there are drawbacks associated with using the resting baselines (e.g., rest may not be a representative state 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 or NER those effect sizes were summarized by simply using the mean effect size for the omnibus analyses. For the analyses within each domain, averaging was only done if there was more than one emotional reactivity measure for PER or NER within a domain. Using the 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 emotion response 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, and physiology) 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 Type I error, especially if there is heterogeneity in the data (Hunter & Schmidt, 2000). Due to the small number of studies available for this meta-analysis and the unknown size of the expected effects, we chose to use the fixed effects model as the primary model. We also ran tests of heterogeneity of the effect sizes using Cochran's heterogeneity statistic Q (Cochran, 1954). Since heterogeneity can potentially inflate Type I error in a fixed effects approach, we planned to re-run the analyses using a random effects model, if analyses revealed significant heterogeneity in the data.

684

L.M. Bylsma et al. / Clinical Psychology Review 28 (2008) 676691

3. Results 3.1. Omnibus analyses We first conducted omnibus analyses of positive and negative emotional reactivity using the fixed effects model. The analysis of positive emotional reactivity (PER) was significant (p b .0001) and revealed that PER was reduced in MDD compared to normal controls (see Fig. 1). The effect size for PER was d = .53, a medium-sized effect by Cohen's (1988) conventions. Similarly, the omnibus analysis of negative emotional reactivity (NER) was also significant, (p b .0001) and revealed that NER was reduced in MDD compared to normal controls (see Fig. 1). The effect size for NER was d = .25, corresponding to a small effect size. When PER 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, p b .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.

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

L.M. Bylsma et al. / Clinical Psychology Review 28 (2008) 676691

685

We conducted analyses of heterogeneity of both the PER and NER omnibus analyses to measure the variation around the mean weighted effect sizes. Significant heterogeneity was present for both PER (Q = 111.80, p b .0001) and NER (Q = 34.77, p = .01). Given the presence of heterogeneity in the magnitude of the effect sizes, we ran the planned follow-up random effects analysis for PER and NER. 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 omnibus analysis 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. PER To 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 of PER, it was found that PER in MDD was reduced relative to controls. For self-report measures of PER, the effect size was d = .703 ( p b .0001), a large effect. This analysis included 10 studies with a sample size of 263 MDD individuals and 249 controls. For behavioral measures of PER, the effect was medium-sized (d = .453, p b .0001). The analysis of behavioral measures included 10 studies 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 of these found a significant effect. When data from these 4 studies was pooled the overall analysis was non-significant (p = .293), with a very modest effect size for PER (d = .151). 3.2.2. NER To 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 selfreport measures of NER, the effect size was medium (d = .359, p b .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 of behavioral measures, there were 9 studies, with a sample of 290 MDD individuals and 249 Controls. Finally, there was a medium effect for physiological measures of NER (d = .223, p b .05). For the physiological measures, there were 8 studies, with a sample size of 196 MDD individuals and 220 controls. In sum, similar to results for PER, when each emotional response system was examined separately individuals with MDD exhibited reduced NER compared to controls in two of the three systems. 3.3. Heterogeneity and Moderator Analyses Omnibus analyses presented earlier indicated that effect sizes for both PER and NER were heterogeneous. Given that all three distinct response systems were analyzed together, this was not completely surprising. To help isolate the sources of heterogeneity, we also ran tests of heterogeneity within each emotion response system. For NER, the effect sizes in the physiological analyses had significant heterogeneity (Q = 21.16, p b .01), self-report effect sizes had marginally significant heterogeneity (Q = 17.19, p = .05) and behavioral (Q = 8.67, p = .37) were not heterogeneous. For PER, behavioral (Q = 150.46, p b .0001) and self-report (Q = 86.70, p b .0001) but not physiological analyses (Q = 0.39, p = .82) indicated significant heterogeneity. Thus, the only response system which exhibited heterogeneity with any consistency was self-reported emotion, which may be due to the diverse measures included within this domain of emotional responding. Heterogeneity can also indicate the presence of outliers in the data. The PER behavioral measures demonstrated most notable degree of heterogeneity. From a visual inspection of the plot of the PER behavioral effects sizes, it would appear that were two outliers present. Specifically, all of the effect sizes fell within the narrow range of .22 to +.10 except for two outliers (+ 0.64 and 3.95). The presence of moderator variables, which systematically influence the effect sizes, can also cause heterogeneity. We originally intended to run additional moderator analyses designed to identify variables that might account for systematic variation in PER and NER effect sizes (e.g., medications, depression severity, co-morbidity). Unfortunately, these analyses were either impossible due to inadequate information from primary studies and/or were underpowered

686

L.M. Bylsma et al. / Clinical Psychology Review 28 (2008) 676691

(see Discussion below).2 The small subset of studies available for each response system likewise precluded adequately powered analyses of potential moderator variables within each system.3 3.4. File drawer analyses Finally, since our review considered only published sources, we conducted file-drawer analyses to assess the reliability of the omnibus effects in the face of additional unpublished null results (Rosenthal, 1991). For the probability for the omnibus analysis of PER to become non-significant ( p N .05), there would have to be over 90 unpublished studies with non-significant results hidden away in the file-drawer. Similarly, for the omnibus analysis of NER to become non-significant, over 40 unpublished studies with non-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 how MDD influences ongoing emotional reactivity. This meta-analysis represents the first quantitative review of the MDD emotional reactivity literature. The major results indicate that MDD involves consistent reductions in both PER and NER. These results have implications for each of the three major views of emotional reactivity in MDD. First, as NER was found to be reduced rather than increased, the negative potentiation view was not supported. Second, the reduction of PER in MDD, and the fact that the PER effect was larger than the NER effect both provided good support for the positive attenuation view. Third, the ECI view appears to offer the most parsimonious overall account of these data because (1) the pattern of decreased emotion reactivity in MDD was not restricted to PER, and (2) the ECI view uniquely predicts reduced NER. 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 for overall NER and PER would hold even in the face of a substantial number of unpublished null results. And third, the reductions in PER and NER appeared to generalize across different systems of emotional response, and were not restricted to a single response system. These results also have more general implications for the study of how moods and emotions interact. Although moods and emotions are highly familiar constructs that have generated extensive research, it is not yet clear how these two kinds of affective processes are related. Intuitively, it makes sense that emotional reactions are stronger when they are congruent with a preexisting mood, an idea reinforced by contemporary emotion theory. Yet there are surprisingly few strong empirical demonstrations that moods actually facilitate emotional reactivity to mood-congruent stimuli. In this 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 the case in the present data. One hypothesis to help make sense of this paradox is that moods may have non-linear effects on emotional reactivity, with mood-facilitation holding for mild and moderate depressed mood states but not for severe depressed mood states. Although this hypothesis waits further testing, the idea that milder depression potentiates NER
For example, we attempted to examine whether depression severity exerted a systematic effect on emotion reactivity in MDD; however, the studies 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 on average reported less positive affect (d = 0.865, p b .0001) and more negative affect (d = 1.454, p b .0001) than controls. However, correlational moderator analyses did not find a significant relationship between the magnitude of positive affect baseline differences and self-report indices of PER (r = .26, p = .49). Likewise, no significant relationship was found between the magnitude of negative affect baseline differences and self-report indices 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, p b .05). Too few studies were available to examine the association of baseline physiology with physiological indices of PER. However, the magnitude of the difference in baseline activation was not correlated with physiological indices of NER (r = .352, p = .439). Although the low number of included studies limited the power of these analyses, there did not appear to be a strong relationship between tonic differences in response and observed differences in PER and NER.
2

L.M. Bylsma et al. / Clinical Psychology Review 28 (2008) 676691

687

is consistent with a number of analog studies of dysphoric samples that have obtained data consistent with this prediction (Golin, Hartman, Klatt, Munz, & Wolfgang, 1977; Lewinsohn, Lobitz, & Wilson, 1973). Careful assessment of emotional reactivity in samples that contain individuals with the full range depressed mood will be critical for testing for non-linear effects, an analysis that may enrich an already active debate about whether depression is best conceived as 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 of normal mood variation, four important limitations should be noted. First, to estimate effect sizes for PER and NER, this review prioritized the controlled measurement of emotion in laboratory settings. It is an open question as to whether these laboratory findings generalize to naturalistic settings. The research database that examines emotion in everyday life 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 ecologically important 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 consistent heterogeneity in the self-report system when individual response systems were analyzed, but the proximal source(s) of this heterogeneity were difficult to isolate. Second, we were not able to identify moderators that might explain systematic variation in PER and NER effect sizes across studies, either because the relevant information in the primary studies was incomplete or absent and/or because the analyses of candidate moderators were underpowered. This point merits emphasis. For example, it is possible that the antidepressant medications commonly taken by individuals with MDD could influence the magnitude of PER and NER (Tomarken, Shelton, & Hollon, 2007). The studies in this meta-analysis were disparate in their practices with respect to medications. Very few studies used a non-medicated or an all-medicated sample, and the all-medicated samples were often taking a variety of medications, which typically were not reported in sufficient detail to analyze in any meaningful way. Another possible moderator that could not be addressed by the present meta-analysis was co-morbid psychopathology in the MDD individuals. Most studies did not exclude for anxiety or personality disorder comorbidity, and did not report emotional reactivity in pure MDD separately. In fact, only two studies excluded for all types of anxiety disorders in MDD individuals, which are commonly co-morbid with MDD (DSM-IV; APA, 2000). The effect of co-morbid disorders on PER and NER in MDD is not well understood, but there are indications in the literature that anxiety levels and forms of co-morbidity such as substance use (Zvolensky & Schmidt, 2003) may influence 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 with increased reactivity as measured by EMG for both positively and negatively valenced film clips. In sum, the presence of normative co-morbidity in MDD samples precludes any strong claim that blunted emotional reactivity is specific to MDD. To advance such a claim, it would be useful to have carefully conducted comparisons of emotional reactivity between pure samples of MDD participants, co-morbid MDD participants and participants with other psychiatric conditions. Moreover, heterogeneity of MDD samples with respect to symptom severity, and demographic characteristics are also potential factors that influences the magnitude of PER and NER. Samples of MDD individuals in this metaanalysis were drawn from a variety of sources and included inpatients, outpatients, and untreated individuals recruited from the general population. Preliminary moderator analyses designed to examine the relationship between depression severity and emotion reactivity found no relationship (see footnote 1), but because of incomplete data, this analysis did not 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 of depression subtype (e.g., atypical versus melancholic MDD). Unfortunately, the majority of the studies did not specify into which subtype individuals in their MDD samples fell. Finally, the sparse reporting of race, ethnicity, and socioeconomic status, precluded analyses of these factors despite the fact that these demographic factors may be important 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 on possible moderators of emotional reactivity in MDD, including medications, co-morbid anxiety, gender, demographic

688

L.M. Bylsma et al. / Clinical Psychology Review 28 (2008) 676691

variables, depression severity, depression subtypes. Importantly, despite the multiple factors which might influence PER and NER, and the well known heterogeneity of depression, this meta-analysis found reliable effects for NER and PER. In fact, the effect sizes for NER and PER appear to be comparable in magnitude to those found in other metaanalyses 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 that is relevant to the emotion construct. Thus, one important avenue for future work will be to examine the generalizability of the findings presented here to other response systems. Interestingly, a recent meta-analysis of a substantial literature on 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 in emotional responding in MDD is not easily summarized in narrative form, but the possibility should be noted that findings 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 controls showed enhanced P300 during encoding and reduced P300 during recognition of positive stimuli, which the authors interpreted as a response bias for positive information. Using fMRI, a blunted response in the amygdala to facial expressions of fear (relative to neutral) has been observed in both depressed adults (Drevets, 2001) and depressed children (Thomas et al., 2001), which might be interpreted in terms of ECI. However, consistent with the negative potentiation view, Siegle, Steinhauer, Thase, Stenger, and Carter (2002) found that MDD individuals exhibited a similar initial amygdala response but greater sustained amygdala activity in response to negative words. In this same study, and underlining the complexity of this domain, the MDD individuals exhibited decreased reactivity in dorsolateral 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 to negative words in the inferior parietal lobule (IPL) but decreased reactivity in the superior temporal gyrus (STG) and the cerebellum. The high complexity of these neuroimaging findings again underlines our caution in not assuming the generalizability 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 surprisingly slow to undertake quantitative reviews of emotional reactivity in Axis-I disorders. This report, as the first quantitative review of emotion reactivity in major depression, begins to address this gap by suggesting that changes in emotional reactivity are a reliable correlate of MDD. In addition to augmenting the clinical description of MDD as a disorder of emotion, these analyses have implications for interventions designed to treat and prevent MDD. For example, to the extent that diminished emotional reactivity is a key affective deficit in MDD, it may be the case that individuals who strongly exhibit this deficit will have a more pernicious course of disorder (e.g., Rottenberg, Wilhelm et al., 2002). This logic would also support the development of psychologically-based and pharmacological treatment techniques to bolster MDD patients' appropriate reactivity to both positive and negative emotional stimuli. In fact, given the centrality of affective disturbance in MDD, developing an accurate, system-by-system account of how MDD alters emotional reactivity, including testing hypothesized proximal mechanisms (e.g., individual differences in cognition or biological functioning), is a sine qua non for developing more effective, targeted treatments for this disorder. We submit 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, 233238. *Allen, N. B., Trinder, J., & Brennan, C. (1999). Affective startle modulation in clinical depression: Preliminary findings. Biological Psychiatry, 46, 542550. 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, 229240. Studies included in the meta-analyses.

L.M. Bylsma et al. / Clinical Psychology Review 28 (2008) 676691

689

Barrett, L. F. (2006). Valence is a basic building block of emotional life. Journal of Research in Personality, 40, 3555. Beach, S. R., & Amir, N. (2003). Is depression taxonic, dimensional, or both? Journal of Abnormal Psychology, 112, 228236. 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, 3744. 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. 242276). 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, 4959. 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. 106128). 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, 612622. Bower, G. H. (1981). Mood and memory. American Psychologist, 36, 129148. Burke, H. M., Davis, M. C., Otto, C., & Mohr, D. C. (2005). Depression and cortisol responses to psychological stress: A meta-analysis. Psychoneuroendocrinology, 30, 846856. 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, 592604. Cacioppo, J. T., Petty, R. E., Losch, M. E., & Kim, H. S. (1986). Electromyographic activity over facial muscle regions can differentiate the valence and intensity of affective reactions. Journal of Personality and Social Psychology, 50, 260268. Canli, T., Sivers, H., Thomason, M. E., Whitfield-Gabrieli, S., Gabrieli, J. D., & Gotlib, I. H. (2004). Brain activation to emotional words in depressed vs. healthy subjects. NeuroReport, 15, 25852588. Carney, R. M., Hong, B. A., Brim, J., Plesons, D., & Clayton, P. (1983). Facial electromyography and reactivity in depression. The Journal of Nervous and Mental Disease, 171, 312313. Carver, C. S. (2001). Affect and the functional bases of behavior: On the dimensional structure of affective experience. Personality and Social Psychology Review, 5, 345356. 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, 103116. Cochran, W. (1954). The combination of estimates from different experiments. Biometrics, 10, 101129. 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, 569577. Deldin, P. J., Keller, J., Gergen, J. A., & Miller, G. A. (2000). Right-posterior face processing anomaly in depression. Journal of Abnormal Psychology, 109, 116121. Depue, R. A., & Iacono, W. G. (1989). Neurobehavioral aspects of affective disorders. Annual Review of Psychology, 40, 457492. *Dichter, G., Tomarken, A., Shelton, R., & Sutton, S. (2004). Early- and late-onset startle modulation in unipolar depression. Psychophysiology, 41, 433440. Dolan, R. J. (2002). Emotion, cognition and behavior. Science, 298, 11911194. Drevets, W. C. (2001). Neuroimaging and neuropathological studies of depression: implications for the cognitive-emotional features of mood disorders. Current Directions in Neurobiology, 2001, 240249. *Dunn, B. D., Dalgleish, T., Lawrence, A. D., Cusack, R., & Ogilvie, A. (2004). Categorical and dimensional reports of experienced affect to emotion-inducing pictures in depression. Journal of Abnormal Psychology, 113, 654660. 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, 169200. 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. 429440). (2nd edition) New York: Oxford University Press. Ekman, P., & Oster, H. (1979). Facial expressions of emotion. Annual Review of Psychology, 30, 527554. *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, 1125. 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, 178187. 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, 386399. *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, 157167. 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, 435439. Gray, J. R. (2004). Integration of emotion and cognitive control. Current Directions in Psychological Science, 13, 4648.

690

L.M. Bylsma et al. / Clinical Psychology Review 28 (2008) 676691

*Greden, J. F., Genero, N., Price, L., Feinberg, M., & Levine, S. (1986). Facial electromyography in depression. Archives of General Psychiatry, 43, 269274. Greden, J. F., Price, L., Genero, N., Feinberg, M., & Levine, S. (1984). Facial EMG activity levels predict treatment outcome in depression. Psychiatry Research, 13, 345352. Gross, J. J. (2002). Emotion regulation: Affective, cognitive, and social consequences. Psychophysiology, 39, 281291. *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, 299302. 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, 535545. Henriques, J. B., & Davidson, R. J. (2000). Decreased responsiveness to reward in depression. Cognition and Emotion, 14, 711724. Henriques, J. B., Glowacki, J. M., & Davidson, R. J. (1994). Reward fails to alter response bias in depression. Journal of Abnormal Psychology, 103, 460466. 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, 275292. 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, 304311. 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, 683688. *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, 2131. Keltner, D., & Gross, J. J. (1999). Functional accounts of emotions. Cognition and Emotion, 13, 575599. Kessler, R. C. (2002). Epidemiology of depression. In I. H. Gotlib & C. L. Hammen (Eds.), Handbook of depression (pp. 2342). 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. 85107). New York: Wiley-Interscience. Kring, A. M., & Bachorowski, J. A. (1999). Emotion and psychopathology. Cognition and Emotion, 13, 575600. 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. 90103). 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. 173191). 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, 261273. 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. 623643). New York: Holt, Rinehart, & Winston. Lapierre, Y. D., & Butter, H. J. (1980). Agitated and retarded depression: A clinical psychophysiological evaluation. Neuropsychobiology, 6, 217223. 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. 123126). New York: Oxford University Press. Lewinsohn, P. M., Lobitz, W. C., & Wilson, S. (1973). Sensitivity of depressed individuals to aversive stimuli. Journal of Abnormal Psychology, 81, 259263. Matt, G. E., Vazquez, C., & Campbell, W. K. (1992). Mood-congruent recall of affectively toned stimuli: A meta-analytic review. Clinical Psychology Review, 12, 227255. 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, 175190. Mauss, I. B., Wilhelm, F. H., & Gross, J. J. (2004). Is there less to social anxiety than meets the eye? Emotion experience, expression, and bodily responding. Cognition and Emotion, 18, 631662. Melamed, S. (1987). Emotional reactivity and elevated blood pressure. Psychosomatic Medicine, 49, 217225. Mor, N., & Winquist, J. (2002). Self-focused attention and negative affect: A meta-analysis. Psychological Bulletin, 128, 638662. Murphy, F. C., Nimmo-Smith, I., & Lawrence, A. D. (2003). Functional neuroanatomy of emotions: A meta-analysis. Cognitive, Affective and Behavioral Neuroscience, 3, 207233. Nesse, R. M. (2000). Is depression an adaptation? Archives of General Psychiatry, 57, 1420. Peeters, F., Berkhof, J., Delespaul, P., Rottenberg, J., & Nicolson, N. A. (2006). Diurnal mood variation in major depressive disorder. Emotion, 6, 383391. Peeters, F., Nicolson, N. A., & Berkhof, J. (2003). Cortisol responses to daily events in major depressive disorder. Psychosomatic Medicine, 65, 836841. 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, 203211.

L.M. Bylsma et al. / Clinical Psychology Review 28 (2008) 676691

691

*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, 358368. 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, 331348. 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, 183196. Ritz, T., Thons, M., Fahrenkrug, S., & Dahme, B. (2005). Airways, respiration, and respiratory sinus arrhythmia during picture viewing. Psychophysiology, 42, 568578. Rosenberg, E. L. (1998). Levels of analysis and the organization of affect. Review of General Psychology, 2, 247270. 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, 400406. Rottenberg, J. (2005). Mood and emotion in major depression. Current Directions in Psychological Science, 14, 167170. 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, 227232. *Rottenberg, J., Gross, J. J., & Gotlib, I. H. (2005). Emotion context insensitivity in major depressive disorder. Journal of Abnormal Psychology, 114, 627639. 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, 135146. 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. Psychophysiology, 42, 277281. 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, 265272. *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, 16. 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, 487510. 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, 355360. 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, 337347. 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, 693707. *Sigmon, S., & Nelson-Gray, R. (1992). Sensitivity to aversive events in depression: Antecedent, concomitant, or consequent? Journal of Psychopathology and Behavioral Assessment, 14, 225246. *Sloan, D. M., Strauss, M. E., Quirk, S. W., & Sajatovic, M. (1997). Subjective and expressive emotional responses in depression. Journal of Affective Disorders, 46, 135141. *Sloan, D. M., Strauss, M. E., & Wisner, K. L. (2001). Diminished response to pleasant stimuli by depressed women. Journal of Abnormal Psychology, 110, 488493. 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 depressed children. Archives of General Psychiatry, 58, 10571063. Tomarken, A. J., Shelton, R. C., & Hollon, S. D. (2007). Affective science as a framework for understanding the mechanisms and effects of antidepressant 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 with schizophrenia compared to depressed patients and nonpatient comparison subjects. American Journal of Psychiatry, 162, 92101. *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, 4963. 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, 10631070. Wexler, B. E., Levenson, L., Warrenburg, S., & Price, L. (1994). Decreased perceptual sensitivity to emotion-evoking stimuli in depression. Psychiatry Research, 51, 127138. 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, 370379. Zvolensky, M. J., & Schmidt, N. B. (2003). Panic disorder and smoking. Clinical Psychology: Science and Practice, 10, 2951.