Reinforcement sensitivity, depression and anxiety: A meta-analysis and meta-analytic structural equation model

Highlights

• Sensitivity to punishment positively predicts both depression and anxiety.

• Sensitivity to reward discriminates between them, negatively predicting depression.

• This pattern was observed even when directly controlling for comorbidity.

• Depression’s effect sizes are uniquely sensitive to clinical state.

• Depression’s effect sizes are also moderated by method of clinical assessment.

Abstract

Reinforcement Sensitivity Theory (RST) posits that individual differences in reward and punishment processing predict differences in cognition, behavior, and psychopathology. We performed a quantitative review of the relationships between reinforcement sensitivity, depression and anxiety, in two separate sets of analyses. First, we reviewed 204 studies that reported either correlations between reinforcement sensitivity and self-reported symptom severity or differences in reinforcement sensitivity between diagnosed and healthy participants, yielding 483 effect sizes. Both depression (Hedges’ g = .99) and anxiety (g = 1.21) were found to be high on punishment sensitivity. Reward sensitivity negatively predicted only depressive disorders (g = −.21). More severe clinical states (e.g., acute vs remission) predicted larger effect sizes for depression but not anxiety. Next, we reviewed an additional 39 studies that reported correlations between reinforcement sensitivity and both depression and anxiety, yielding 156 effect sizes. We then performed meta-analytic structural equation modeling to simultaneously estimate all covariances and control for comorbidity. Again we found punishment sensitivity to predict depression (β = .37) and anxiety (β = .35), with reward sensitivity only predicting depression (β = −.07). The transdiagnostic role of punishment sensitivity and the discriminatory role of reward sensitivity support a hierarchical approach to RST and psychopathology.

Introduction

Gray’s Reinforcement Sensitivity Theory (RST; Gray, 1970, Gray, 1987; Gray & McNaughton, 2000) posits that sensitivity to appetitive and aversive stimuli serves as a biological basis of human personality. According to the original theory, the Behavioral Approach System (BAS) governs processes related to appetitive stimuli, whereas the separate, independent Behavioral Inhibition System (BIS) governs responses to aversive stimuli (Corr, 2008). Individuals’ varied sensitivities in BAS and BIS functioning then lead to individual differences in reward processing, punishment processing, and personality. Those with a more sensitive BAS tend to show more cognitive styles and behaviors associated with reward promotion and the broad personality dimension of extraversion (Corr & McNaughton, 2008; Depue & Collins, 1999; Gray, 1987). A more sensitive BIS, on the other hand, impacts behaviors and psychological processes related to punishment and neuroticism (Gray, 1970; Smits & Boeck, 2006).

In 2000, RST was revised (Gray & McNaughton, 2000) to include three systems (Corr, 2008). The BAS system remained governing all appetitive processing, while aversive processing was ascribed to the newly named Fight/Flight/Freeze System (FFFS) instead of the original BIS. The BIS in the revised version was theorized to be activated when the goals of the BAS and FFFS came into approach-avoidance conflicts. New measures (e.g., Corr & Cooper, 2016; Jackson, 2009) have been developed with the purpose of differentiating the revised BIS from the FFFS (i.e., original BIS). In practice, they both correlate with measures of BIS developed under the original RST and much of the literature on RST, motivation and psychopathology continues to use the original terminology of BAS for appetitive sensitivity and BIS for aversive sensitivity (e.g., Bijttebier, Beck, Claes, & Vandereycken, 2009; Trew, 2011). The current meta-analysis, in reviewing this literature, uses the terminology of the original RST when referring to basic appetitive and aversive processes as well.

When appetitive and aversive sensitivities are dysregulated, they are theorized to be diatheses for subsequent psychopathology. BAS hyposensitivity is indicated by deficits in anticipatory pleasure, consummatory pleasure, and appetitive motivation (Alloy, Olino, Freed, & Nusslock, 2016; Olino, McMakin, & Forbes, 2018; Zald & Treadway, 2017). This anhedonic experience is a hallmark of dysthymia and major depressive disorder (American Psychiatric Association, 2013; Watson et al., 1995). Accordingly, numerous studies have found reduced BAS sensitivity to be associated with depression among clinical (e.g., DelDonno et al., 2015; Greenwood et al., 2013; Zaninotto et al., 2015) and nonclinical samples (e.g., Hundt, Nelson-Gray, Kimbrel, Mitchell, & Kwapil, 2007; Jarmolowicz et al., 2014; Peirson & Heuchert, 2001) – though not always (e.g., Carver & Johnson, 2009). Among depressed patients, BAS hyposensitivity was found to predict greater concurrent symptom severity and a worse outcome following treatment (Kasch, Rottenberg, Arnow, & Gotlib, 2002). Anxiety, on the other hand, is theorized to be weakly related to BAS sensitivity, if at all (Gray, 1987; Gray & McNaughton, 2000; McNaughton & Corr, 2004). Rather, anxiety is theorized to primarily occur as a function of aversive sensitivities, conceptualized as BIS hypersensitivity in the original RST (Gray, 1982, Gray, 1987). BIS hypersensitivity is associated with increased apprehension in the face of impending negative outcomes and more intense reactivity to punishment, which are experiences salient to anxiety disorders (Carver & White, 1994; Corr & McNaughton, 2008; Hollon, 2019; Leen-Feldner, Zvolensky, Feldner, & Lejuez, 2004). Later investigations, however, have found BIS levels to predict depression to a large degree as well (Eshel & Roiser, 2010; Hundt et al., 2007; Johnson, Turner, & Iwata, 2003; Pekka Jylhä & Isometsä, 2006).

Thus, it appears that the BIS predicts both depression and anxiety while the BAS uniquely predicts depression (Bijttebier et al., 2009; Khan, Jacobson, Gardner, Prescott, & Kendler, 2005). This pattern has been linked to similar hierarchical models of psychopathology such as the tripartite model, where general distress is present across depression and anxiety disorders, but deficits in positive affect are unique to depression (Clark & Watson, 1991; Lahey, Krueger, Rathouz, Waldman, & Zald, 2017; Zinbarg & Yoon, 2008).

The multidisciplinary literature on the relationship between reinforcement sensitivity dysregulation and mood disorders has been summarized in a number of influential theoretical frameworks and models. The BAS dysregulation model highlights the unique role that reward sensitivity plays in the development and maintenance of depression and bipolar disorder (Alloy et al., 2016; Stange et al., 2013). The joint system hypothesis suggests that internalizing pathology is most closely predicted by a confluence of hyposensitive BAS and hypersensitive BIS (Corr, 2001; Eddington, Majestic, & Silvia, 2012; Hundt et al., 2007). Furthermore, RST has provided a theoretical basis for disorder classification models related to depression and anxiety in particular (Nusslock, Abramson, Harmon-Jones, Alloy, & Hogan, 2007; Zinbarg & Yoon, 2008), and mental illness in general (Bijttebier et al., 2009). RST has even served key roles in explaining how individual differences in general may impact the etiology (Kimbrel, 2008; Trew, 2011), severity (e.g., Brown, Chorpita, & Barlow, 1998; Cloninger, Bayon, & Svrakic, 1998) and classification (e.g., Caspi et al., 2014; Clark & Watson, 1991; Eaton et al., 2013; Krueger & Markon, 2006; Watson, 2009) of mental disorders. The National Institute of Mental Health has since included reward and punishment processing among its Research Domain Criteria (RDoC; Insel et al., 2010), further emphasizing the central role that RST plays in basic clinical science.

The current theoretical work, however, is limited by its reliance on narrative reviews, which often under-represent nonsignificant effects, unpublished effects, or effects that were secondary to the study at hand (Easterbrook, Gopalan, Berlin, & Matthews, 1991; Rosenthal & DiMatteo, 2001; Sterne, Gavaghan, & Egger, 2000). This biased source selection may have theoretical implications. Constructs such as positive emotionality (Watson & Naragon-Gainey, 2010) and extraversion (Krueger, Caspi, Moffitt, Silva, & McGee, 1996) have been hypothesized to differentiate depression from anxiety. However, when subjected to the scrutiny of meta-analyses, neither construct conclusively discriminated between these disorders (Khazanov & Ruscio, 2016; Kotov, Gamez, Schmidt, & Watson, 2010). Similarly, the relationship between BAS hyposensitivity and anhedonic depression has been highlighted in narrative reviews of findings (e.g., Alloy et al., 2016; Zinbarg & Yoon, 2008) derived from biological (e.g., DelDonno et al., 2015), behavioral (e.g., Treadway, Bossaller, Shelton, & Zald, 2012), and self-report data (e.g., Kircanski, Mazur, & Gotlib, 2013), with some studies based upon quite sizeable samples (e.g., Johnson et al., 2003). However, it is also possible that the unique relationship between BAS sensitivity and anhedonia may not withstand the scrutiny of meta-analysis, as was the case of positive emotionality and extraversion.

Attempts to discriminate between depression and anxiety are further complicated by their high comorbidity (Watson, 2009). Depression and anxiety are noted risk factors for each other (Jacobson & Newman, 2017), with overlapping symptom criteria (Borsboom & Cramer, 2013), some shared basic mechanisms (Nolen-Hoeksema & Watkins, 2011; Ruscio, Seitchik, Gentes, Jones, & Hallion, 2011), and are especially comorbid among more severe clinical samples (Kessler, Chiu, Demler, & Walters, 2005). To understand each disorder’s unique relationship with reinforcement sensitivity, it is necessary to control for their shared variance. Only meta-analytic tools enable that, either by indirectly projecting shared covariance (e.g., Khazanov & Ruscio, 2016) or by using structural equation modeling to directly control for it (M. W.-L. Cheung & Chan, 2005; Jak, 2015). Thus, a systematic quantitative review is needed to adequately estimate the unique relationships between BAS, BIS, depression and anxiety, after controlling for the disorders’ high rates of comorbidity.

Furthermore, a meta-analysis of the relationship between reinforcement sensitivity, depression and anxiety may answer the calls for identifying potential moderators to the relationship (see Bijttebier et al., 2009). One moderator of note is clinical state. Indeed, differences have been found between acute depressed patients and those in remission in measures of both reinforcement sensitivity (Pinto-Meza et al., 2006) and temperament (Nery et al., 2009; Takahashi et al., 2013). This was particularly true for punishment sensitivity (Hansenne & Bianchi, 2009). To the best of our knowledge, however, the moderating role of clinical state on otherwise stable traits is largely neglected in the meta-analytic literature (Kotov et al., 2010; cf. Zaninotto et al., 2016), including the literature on reinforcement sensitivity.

In sum, numerous multimodal lines of research have found that dysregulated reinforcement sensitivity plays a critical role in the development and maintenance of depression and anxiety. Summaries of these findings, however, have been limited to narrative reviews. Without a quantitative summary of the literature, the range of observed effect sizes remains unknown, central hypotheses laid out in RST remain untested, and potential moderators remain unexamined. These questions impact the nature of reinforcement sensitivity’s relationship with depression and anxiety, and warrant the performance of a meta-analysis.