Physiological Correlates of Affective Decision-Making in Anxiety and Depression

City University of New York (CUNY) CUNY Academic Works

All Dissertations, Theses, and Capstone Projects Dissertations, Theses, and Capstone Projects

9-2017

Louisa I. Thompson The Graduate Center, City University of New York

How does access to this work benefit ou?y Let us know!

More information about this work at: https://academicworks.cuny.edu/gc_etds/2254 Discover additional works at: https://academicworks.cuny.edu

This work is made publicly available by the City University of New York (CUNY). Contact: [email protected]

PHYSIOLOGICAL CORRELATES OF AFFECTIVE DECISION-MAKING IN ANXIETY AND DEPRESSION

LOUISA I. THOMPSON

A dissertation submitted to the Graduate Faculty in Psychology in partial fulfillment of the requirements for the degree of Doctor of Philosophy, The City University of New York.

2017

LOUISA I. THOMPSON

Physiological Correlates of Affective Decision-Making

In Anxiety and Depression

Louisa I. Thompson

This manuscript has been read and accepted for the Graduate Faculty in

Psychology in satisfaction of the dissertation requirement for the degree of

Doctor of Philosophy.

July 11, 2017 Sarit A. Golub, Ph.D., MPH Date Chair of Examining Committee

July 11, 2017 Richard Bodnar, Ph.D. Date Executive Officer

Supervisory Committee:

Douglas Mennin, Ph.D.

Justin Storbeck, Ph.D.

Jennifer Stewart, Ph.D.

Faith Gunning, Ph.D.

THE CITY UNIVERSITY OF NEW YORK

iii

ABSTRACT

Physiological Correlates of Affective Decision-Making in Anxiety and Depression

Louisa I. Thompson

Advisor: Sarit A. Golub, Ph.D.

Improving our understanding of cognitive and physiological profiles in anxiety and depression has the potential to reveal novel ways to target and improve treatments for these prevalent mental health conditions. The present study examined the impact of self-reported anxiety and depression symptoms on three established decision-making measures, the Iowa Gambling Task (IGT;

Bechara, Damasio, Damasio, & Anderson, 1994), Balloon Analogue Risk Task (BART; Lejuez et al., 2002), and Game of Dice Task (GDT; Brand et al., 2005), in a diverse sample of 100 college students (age 18 to 35). Physiological measures of tonic heart rate variability and galvanic skin response were obtained to better characterize autonomic flexibility and sympathetic reactivity, respectively, during decision-making performance. Interoceptive sensitivity, measured via a heart beat perception task (Schandry, 1981), was also examined as a potential moderator in the relationship between sympathetic reactivity and decision-making.

Consistent with the literature, BART performance was negatively associated with IGT performance, while GDT performance was positively associated with IGT performance.

Contrary to our hypotheses, physiological measures did not distinguish individuals who reported anxiety and/or depression from those who did not. Of the three tasks, only IGT performance was associated with sympathetic reactivity. Consistent with our hypotheses, anxiety and greater iv

sympathetic reactivity to losses in the task predicted better scores. Interoceptive sensitivity moderated the association between sympathetic reactivity and IGT performance, but only among those with anxiety, with better performance associated with a combination of lower interoceptive sensitivity and higher sympathetic reactivity. Low tonic HRV predicted worse IGT performance in depressed participants and worse GDT performance in anxious participants. These findings, though preliminary, have implications for treatment advances involving HRV biofeedback and interoceptive exposure. Our findings also highlight substantial differences between the IGT,

BART, and GDT in their associations with anxiety, depression, and physiological markers, for consideration in cross study comparisons and future research.

Table of Contents

Introduction………………………………………………………………………………..1

Literature Review……..…………………………………………………………………...2

The somatic marker hypothesis (SMH)…………………………………………...2

The Iowa Gambling Task (IGT)……..………………….....……………………...3

Applying the SMH to Anxiety and Depression………..……………………..…...4

Affective Decision-Making Performance in Anxiety………………...... ………5

Affective Decision-making Performance in Comorbid Anxiety and Depression...7

Skin Conductance and Affective Decision-Making Performance...... ………10

The SMH and IGT: Limitations and Future Directions………...... ……….…….11

Alternative Perspectives for Understanding Affective Decision-Making…….…13

The perseverative cognition hypothesis and autonomic flexibility…..….13

Interoceptive sensitivity……………………………………………...…..16

The Present Study…………...……….…………………………………………………..19

Aims………………………………………………………..…………………….19

Hypotheses…………………………………………..……...………………...….22

Method……………………………………...…………………..…………………..……24

Participants.………………………………………………………………………24

Procedure………...………………………………………………………………24

Neuropsychological Tasks………...……………………..………………….…...24

Physiological Measures………………………..………………….……………..28

Interoceptive Sensitivity Measure……………………………………….……….30

Anxiety and Depression Self-Report Measures…………………………….……31 vi

Additional Self-Report Measures…………………………………………….….32

Data Analysis…………………………………………………………………….33

Results……………..…………………………………………………………………..…33

Sample Size………………………………………………………………………33

Affective Variable Adjustments……………...…………………………………..34

Descriptives and Bivariate Comparisons…………………………...……………34

Analysis of IGT Performance………...………………………….………………35

Analysis of Game of Dice Task (GDT) Performance…………..………..………42

Analysis of Balloon Analogue Risk Task (BART) Performance...………….…..44

Affective Group Analysis …………...………………………………………..…47

Task Comparison Analysis.……………..………………………………..….…..47

Discussion……………………………………..…………………………………………48

IGT Findings...………………………………………………..……………….....49

GDT Findings…………………………………….….……...…………………...54

BART Findings………………………………………………………………...... 57

Task Comparisons………………………………………………………………..60

Implications…………….……………………………………………...…………62

Limitations……………………………...…………………………………...…...66

Conclusions……………………………………………………………………....69

Appendix A: Tables and Figures…………………………………..……………………..71

References …………………………...…………………………………………………...97

vii

List of Tables and Figures

Table 1. Sample demographic and affective characteristics…………………………...... 71

Table 2. Decision-making task variables………………………………….…………...... 72

Table 3. Spearman correlations: DM task, digit span, and affective variables…………..73

Table 4. Physiological and interoceptive sensitivity variable descriptives………...….....74

Table 5. Spearman correlations: physiological, affective, and digit span variables……..75

Table 6. One-way ANOVA results for variables predicting physiological outcomes…...76

Iowa Gambling Task Analyses:

Table 7. Independent regression: gender and affective variables…………………....…..77

Table 8. Independent regression: gender, depression, and LDSB variables………..…....78

Figure 1. Depression and working memory interaction……………………………….....79

Figure 2. Depression and working memory interaction using categorical LDSB…….....79

Table 9. Hierarchical regression: gender, depression, and dichotomized LDSB …….....80

Figure 3. Below average LDSB and depression predict worse performance……………80

Table 10. Hierarchical regression: tonic HRV interactions……………………………...81

Figures 4 and 5. Depression, tonic HRV, and working memory interactions………...... 81

Table 11. Independent regression: Mean gain and loss event SCR amplitude…………..82

Table 12. Independent regression: mean SCR amplitude interactions………...………...82

Figure 6. Greater SCR response to losses associated with better performance……….....83

Table 13. Independent regression: mean frequency of non-specific SCRs per minute.....83

Table 14. Hierarchical regression: HBPT accuracy and anxiety interaction………….....84

Figure 7. HBPT accuracy predicts worse IGT performance in anxiety……………...….84

Table 15. Independent regression: HBPT accuracy, SCR amplitude, and anxiety…...... 85 viii

Figure 8. HBPT accuracy, SCR amplitude, and anxiety interaction……………….…….85

Game of Dice Task Analyses:

Table 16. Independent regression: race and affective variables..……………………..…86

Table 17. Hierarchical regression: race and worry interaction…………………………..86

Table 18. Independent regression: race and working memory variables……….………..87

Table 19. Hierarchical regression: anxiety and dichotomous LDSB interaction……...…87

Figure 9. Anxiety and working memory interaction………………………..………..…..88

Table 20. Hierarchical regression: worry and tonic HRV interaction……………….…...89

Figure 10. Worry and HRV interaction………………...………………………………...89

Table 21. Independent regression: average SCR amplitude for gain and loss events…....90

Balloon Analogue Risk Task Analyses:

Table 22. Independent regression: gender and affective variables……………..………...90

Table 23. Independent regression: working memory variables……………………...…...91

Table 24. Hierarchical regression: anxiety and dichotomous LDSB interaction…………91

Figure 11. Dichotomous LDSB and anxiety interaction……………………………..…...92

Table 25. Hierarchical regression: worry and dichotomous LDSB interaction……...…...92

Figure 12. Dichotomous LDSB and anxiety interaction………………………………….93

Table 26. Independent regression: gender, worry, anxiety, and tonic HRV……………...93

Table 27. Independent regression: mean SCR amplitude for gain and loss events….…...94

Table 28. ANOVA: physiological variables by gender and affect category……..……....94

Table 29. Independent regression: BART performance predicting IGT performance…...95

Table 30. Independent regression: GDT performance predicting IGT performance….....95

Table 31. Independent regression: BART performance predicting GDT performance.....96 ix

Physiological Correlates of Affective Decision-Making in

Anxiety and Depression

Anxiety disorders are a pervasive and growing problem in the U.S., with current estimates of lifetime prevalence at 29% (Kessler et al., 2005). Anxiety frequently co-occurs with other mental disorders, making it difficult to isolate and understand how it affects cognition

(Brown, Campbell, Lehman, Grisham, & Mancill, 2001). Furthermore, the taxonomy of anxiety disorders encompasses varying degrees of worry, central to generalized anxiety disorder (GAD), and anxious arousal, as seen most profoundly in panic disorder (PD). Approximately half of all individuals with anxiety disorders face a particularly debilitating prognosis involving the development of co-occurring depression associated with more severe and refractory symptomology (e.g., increased rates of suicidality and substance use), greater dysfunction of stress response systems (e.g., exaggerated adrenocorticotropic hormone response), and unique elevations in amygdala activity and disrupted frontolimbic connectivity relative to either disorder alone (Belzer & Schneier, 2005; Young, Abelson, & Cameron, 2004; Canu et al., 2015; Mineka,

Watson, & Clark, 1998).

Understanding the mechanisms of affective decision-making in anxiety and depression can inform prevention and treatment efforts. The ability to integrate emotional information (e.g., reward and risk) in decision-making is critical for our daily functioning and mental health.

Understanding how anxiety, depression, and anxious depression differentially impact decision- making can help us decipher the cognitive-behavioral processes contributing to the common progression from anxiety to anxious-depression and the development of treatment refractory symptoms and comorbid substance use problems (Hettema, Prescott, & Kendler, 2003; Kessler et al., 1996; Parker et al., 1999). In turn, findings could aid therapists and patients in addressing 1

the role of decision-making in behaviors (e.g., avoidance) and contexts (e.g., social events) characteristic of anxiety and depression. Despite these promising applications, it remains difficult to piece together a consistent picture from the existing literature of how different pathological mood states may moderate pre-existing affective decision-making processes.

LITERATURE REVIEW

The Somatic Marker Hypothesis

The somatic marker hypothesis (SMH) is a strongly supported neurobiological model of affective decision-making that can be applied to understanding how anxiety and depression impact decision-making. Somatic markers are learned associations between reinforcing stimuli and the rudimentary physiological affective responses that they evoke. Relevant markers can then be reactivated and summated in future decision-making to efficiently provide input based on prior experience (Damasio, 1994).

The central autonomic network (CAN) is critical to this process, and consists of both peripheral networks of the autonomic nervous system (ANS; e.g., vagus nerve) and brain structures (e.g., prefrontal cortex and insula) that allow for bidirectional feedback necessary for the generation of somatic markers. The ANS also helps to regulate emotion by rapidly adjusting physiological arousal in response to changes in the environment (Thayer, Friedman, &

Borkovec, 1996). Cortical top-down inhibition works to keep sympathetic fight and flight responses in check, while peripheral physiological feedback provides cues about emotional experience to inform cognitive processes, such as decision-making (Bechara, Damasio, &

Damasio, 2000; Paulus, 2007).

Reactivation of somatic markers is proposed to occur through one of two pathways, which have been termed the ‘body loop’ and the ‘as if body loop.’ The body loop involves 2

emotional arousal activated through physiological changes in the body that are communicated back to the brain via autonomic feedback pathways, particularly the vagus nerve and spinal cord.

The as if body loop bypasses the physiological correlates of emotional experience and instead relies more efficiently on cognitive representations of those previously experienced states

(Damasio, Everitt, & Bishop, 1996). The ventral medial prefrontal cortex (VMPFC) is considered to play a central regulatory role in the summation and communication of cognitive and physiological cues during decision-making (Bechara, Damasio, Damasio, and Anderson

1994; Bechara, Tranel, and Damasio, 2000).

Although the engagement of somatic affective cues in decision-making may be largely unconscious, the experience has been described in terms of feelings of “gut instinct” and

“hunches,” which are critical components of the decision-making process even as rational, conscious considerations are being made. Among individuals with good mental health and cognitive functioning, this system makes decision-making guided by emotional learning more efficient, in so far as one does not have to undergo a slow conscious consideration of prior experience in order to arrive at an informed choice (Bechara et al., 2000; Gray, 2004).

The Iowa Gambling Task (IGT)

The IGT is an affective decision-making task originally created to test the tenants of the

SMH. The IGT was first used with a sample of patients with VMPFC lesions in order to evaluate the role of this brain area in affective decision-making (Bechara et al., 1994). These individuals not only did poorly on the IGT, but they also lacked autonomic skin conductance responses

(SCRs) to emotional stimuli during performance, which was interpreted as a failure to integrate physiological affective cues in the decision-making process. In the general population as well, the IGT is sensitive to normal variability in affective decision-making, and better performance is 3

associated with stronger autonomic responses (Carter & Pasqualini, 2004). Since its introduction, the IGT has been used extensively in clinical populations, and has identified disrupted decision- making in samples ranging from patients with substance abuse and gambling addiction to those with fibromyalgia or chronic pain (Apkarian et al., 2004; Bolla et al., 2003; Golub, Starks,

Kowalczyk, Thompson, & Parsons, 2012; Linnet, Møller, Peterson, Gjedde, & Doudet, 2011;

Verdejo-García, Lopez-Torrecillas, Calandre, Delgado-Rodriguez, & Bechara, 2009).

Applying the SMH to Anxiety and Depression

The influence of somatic markers also has the potential to be detrimental, biasing decision-making toward suboptimal choices, particularly if some aspect of the stimuli or environment is misread (Paulus & Yu, 2012). This kind of misguided affective decision-making may be more common in anxiety and mood disorders for a number of reasons. The over- reactivity of physiological stress responses biases attention, and the subsequent generation of somatic markers, toward potential environmental threats (Engelmann, Meyer, Fehr, & Ruff,

2015; Paulus & Yu, 2012). For example, individuals with GAD or elevated state anxiety show attention and memory bias for threat words on emotional Stroop and dot probe tasks (Friedman,

Thayer, & Borkovec, 2000; Mathews & MacLeod, 1994, 1986; Nelson, Purdon, Quigley,

Carriere, & Smilek, 2015). As a result, decision-making may become increasingly risk-avoidant, which can be adaptive in some situations, but maladaptive in others (Engelmann et al., 2015;

Mueller, Nguyen, Ray & Borkovec, 2010). Indeed, avoidance (e.g., of social situations, medical tests) plays an important role in the functional impairment of anxiety. However, there is also evidence that chronically stressed and anxious individuals are more likely to make poor decisions when faced with immediate rewards, particularly when cortisol is chronically elevated

(Epel, Lapidus, McEwen, & Brownell, 2001; Piazza & Le Moal, 1997; van den Bos, Harteveld, 4

& Stoop, 2009), when choosing the reward facilitates avoidance of uncertainty (Luhmann,

Ishida, & Hajcak, 2011), or when the reward can alleviate current distress (Tice, Bratslavsky, &

Baumeister, 2001). For example, health risk behaviors (e.g., smoking) often relapse under stress

(Shiffman & Waters, 2004).

In contrast, depression may impair the processing of rewards and the ability to discriminate between reward and punishment information (Knutson, Bhanji, Cooney, Atlas, &

Gotlib, 2008). This impairment has been associated with reduced autonomic responding to positive emotional stimuli, such as cash incentives for accurate task performance (McFarland &

Klein, 2009). Such deficits can make it more difficult for depressed individuals to learn reward and punishment contingencies during decision-making, and may explain why they sometimes take longer to avoid poor choices on decision-making tasks (Cella, Dymond, & Cooper, 2010;

Must et al., 2006).

Affective Decision-Making Performance in Anxiety

A number of studies have used the IGT and other behavioral affective decision-making measures to investigate how anxiety and depression impact emotional learning and decision- making performance. Anxiety and worry are often hypothesized to predict better IGT performance, but significant associations are not always found, as demonstrated by two recent studies looking at worry, either as a trait or as the primary feature of GAD. In the first, a college sample completed the GAD-Q-IV (Newman et al., 2002) to measure GAD symptoms, and then completed both the IGT and a variant version of the task designed to capture sensitivity to punishment by requiring acceptance of large short-term losses in order to achieve long-term gain

(Mueller et al., 2010). Individuals with significant GAD symptoms were found to perform better on both versions of the task relative to controls, suggesting that their learning was driven by 5

greater attention to future adverse outcomes in decision-making, and not simply a desire to avoid immediate losses.

In the second study, which examined trait worry using the Penn State Worry

Questionnaire (PSWQ; Meyer, Miller, Metzger, & Borkovec, 1990) in an undergraduate sample, the authors created the IGT-P, a version of the task designed to capture the avoidance typically associated with worry by allowing individuals to skip or “pass” on trials of the task. Although the number of passes taken was negatively associated with performance, it did not mediate a relationship between worry and task performance. In fact, worry was neither associated with task performance, nor with the number of passes taken (Drost, Spinhoven, Kruijt, & Van der Does,

2014).

Differences between the two self-report measures (i.e., trait worry vs. a broader assessment of GAD symptomology) may have contributed to the variable results between these two studies. In addition, demand effects (e.g., worry about being negatively judged by the experimenter) may have contributed to high worry participants’ not wanting to use the pass option in the second study, and it is also likely that participants were naturally engaged in completing the IGT, which in and of itself is not particularly anxiety provoking. These contrasting findings bring into question whether individuals with anxiety may demonstrate variable affective decision-making across contexts and emotional states.

Luhmann and colleagues (2011) sought to address this question in a study examining the impact of uncertainty on affective decision-making ability in individuals with anxiety. They argued that because situations involving extended periods of uncertainty are particularly uncomfortable for individuals with anxiety, they should be motivated to avoid such outcomes in their decision-making. They developed a decision-making task with choices for both immediate 6

small rewards and delayed rewards that were often larger but required an uncertain period of waiting. In a college sample, higher intolerance of uncertainty among individuals with elevated trait anxiety (STAI-T; Spielberger, 1983) was associated with greater selection of less valuable immediate rewards relative to larger delayed rewards. This finding suggests that anxiety may be detrimental to decision-making in situations involving extended periods of uncertainty (Luhmann et al., 2011).

Other contextual factors such as acute stress may also impact affective decision-making ability, particularly among anxious individuals. Preston and colleagues (2007) examined the effects of anticipatory stress (i.e., public speech manipulation) on IGT performance among college students. Participants completed the Positive and Negative Affect Scale (PANAS;

Watson, Clark, & Tellegen, 1988) and State Trait Anxiety Index – Trait and State Scales (STAI-

TS; Spielberger, 1983). Participants in the experimental anticipatory stress condition were slower to learn the IGT, but overall performed similarly to controls. However, during the last few blocks of the task, females in the experimental condition performed significantly better than controls, while the opposite was true for males, suggesting that gender may interact with anxiety to impact decision-making in the context of stress (Preston, Buchanan, Stansfield, & Bechara, 2007).

Affective Decision-Making Performance in Comorbid Anxiety and Depression

One likely source of variability in this area of research is the limited amount of attention given to comorbid mood disorders. Relatively few studies have controlled for depression or examined both anxiety and depressive symptoms in the same sample. One study used the

Balloon Analogue Risk Task (BART; Lejuez et al., 2002) to examine risk-aversion in a college sample. Controlling for negative affect on the PANAS, trait anxiety (STAI-T) and worry

(PSWQ) both predicted risk-aversion (i.e., lower game winnings) during task performance 7

(Maner et al., 2007). The researchers also found that anxious individuals (identified via the Mini

International Neuropsychiatric Interview; Lecrubier et al., 1997) endorsed greater risk aversion on a self-report measure than those diagnosed with depression and non-distressed controls. These findings suggest that several different dimensions of anxiety predict risk aversion; either on the

BART, a task that requires some risk-taking for optimal performance, or on a self-report measure. However, despite attention to negative affect and depression, this study did not address comorbidity beyond controlling for negative affect on the PANAS.

Research on affective decision-making in clinical samples has more thoroughly addressed comorbidity between anxiety and depression. Findings from clinical samples generally confirm that anxious individuals are risk averse, while suggesting that those with depression or comorbid anxious-depression may have decision-making deficits. Several studies with clinical samples have made use of behavioral decision-making tasks other than the IGT. Kaplan and colleagues

(2006) recruited non-medicated individuals with PD, both with and without comorbid major depressive disorder (MDD), to complete the Cambridge Gambling Task (CGT; Rogers et al.,

1999), a measure of risky decision-making with reward and loss contingencies that are more transparent than those in the IGT. There were no differences in performance between control and

PD only groups, but individuals with comorbid PD and MDD took longer to make decisions

(consistent with findings for depression) and showed attentional bias toward negative stimuli in the CGT (consistent with findings for anxiety). As anxiety has often been associated with risk aversion, and the Cambridge Gambling Task rewards a risk adverse approach to decision making, it is consistent that participants with PD and without comorbid depression were found to perform well on this task.

In another study, no significant neurocognitive differences were observed between adolescent patients (inpatient and outpatient, medicated and non-medicated) with comorbid GAD and MDD versus those with pure MDD alone (Han et al., 2012). Interestingly however, a significant gender by group interaction was observed for IGT performance, such that females with MDD did better than female controls, but MDD males did worse than male controls. Of note, previous research suggests that females do better on the IGT during adolescence (Hooper,

Luciana, Conklin, & Yarger, 2004), but this pattern seems to change in adulthood, with males typically outperforming females on the IGT, and showing differential lateralization of brain activity during the task (Bolla, Eldreth, Matochik, & Cadet, 2004). In the context of previously discussed findings (de Visser et al., 2010; Lighthall, Mather, & Gorlick, 2009), the Han et al.

(2012) study provides further support for the existence of differential interactions between emotional distress and decision-making as a function of gender.

Findings from research with depression-only samples have generally shown deficits in decision-making. Knutson et al. (2008) used a monetary incentive delay task created for functional magnetic resonance imaging to examine neural responses to monetary incentives as participants try to win and avoid losing virtual money. Relative to controls, they found impaired discrimination between gain vs. non-gain outcomes, and a reduced medial PFC response to gains in their sample of non-medicated, depressed adults without comorbid diagnoses. This finding in a unique sample provides support for the idea that, unlike purely anxious individuals who may have a heightened sensitivity to threat and adverse outcomes, depressed individuals may have deficits in decision-making due to reduced sensitivity to reward and threat. Additional clinical studies have generally supported this conclusion as well (Cella, Dymond, & Cooper, 2010; Must et al., 2006; W. Zhang, Chang, Guo, K. Zhang, & Wang, 2013). 9

SCRs and Affective Decision-Making Performance

Galvanic skin responses (GSR), also known as SCRs, are common methods for characterizing the autonomic physiological feedback that is theorized to play a central role emotional learning and affective decision-making processes. GSR measures real-time changes in sweat secretion that reflect fluctuations in sympathetic nervous system arousal. Event specific

SCRs can typically be observed when individuals experience or anticipate an emotionally salient event (e.g., a gain or loss) and larger SCRs, especially in anticipation of losses, have been associated with better performance on the IGT (Carter & Pasqualini, 2004).

Werner and colleagues (2009) found that higher trait anxiety (STAI-T) was associated with better IGT performance in a college sample. They also looked at self-reported emotion regulation with the scales for emotional experience. Using GSR physiological recording during task performance, they found that trait anxiety was associated with larger anticipatory SCRs, while the experience of emotion regulation was associated with smaller anticipatory SCRs

(Werner, Duschek, & Schandry, 2009). This finding raises the possibility that individuals with anxiety may develop greater anticipatory SCRs due to poor emotion regulation, which could in turn actually serve as a mechanism for improved task performance.

In contrast, Miu and colleagues also found that individuals with higher trait anxiety

(STAI-T – Romanian version) had larger anticipatory SCRs during task performance, but these individuals performed worse on the IGT than individuals without anxiety (Miu, Heilman, &

Houser, 2008). There are several critical differences between these studies that may explain why their results differ. First, it should be noted that Miu and colleagues comprised their groups of anxious and non-anxious individuals by selecting participants who scored either greater than or less than on standard deviation from the sample mean on the STAI-T, while Warner and 10

colleagues used the STAI-T as a continuous measure. It is also potentially significant that

Werner and colleagues used a computerized version of the IGT (most common), while Miu and colleagues used the original hand administered version of the task. Again returning to the possibility of contextual variance, it is possible that anxious individuals may experience greater performance and social anxiety when completing the task administered in person compared to the task administered on the computer. If replicated, the finding by Miu and colleagues could provide a closer representation of how anxiety impacts decision-making in social situations that are potentially more anxiety provoking. A comparable study examining depression has not yet been conducted; however, Tchanturia and colleagues found that individuals with anorexia, the majority of whom had comorbid clinical depression, demonstrated lower SCRs to IGT losses and performed worse in comparison to significantly less depressed control and anorexia recovery groups (Tchanturia et al., 2007).

The SMH and IGT: Limitations and Future Directions

The majority of research on the SMH still comes from the IGT, and some aspects of the model require further investigation. The specificity of the task continues to be questioned

(Buelow & Suhr, 2009; Dunn, Dalgleish, & Lawrence, 2006). Several studies have found decision-making impairments across patients with a variety of different lesions, including areas such as the amygdala (Bechara, Damasio, Damasio, & Lee, 1999; Manes et al., 2002; Fellows &

Farah, 2005). More recent imaging studies suggest that while the VMPFC is the primary frontal region activated during task learning, other areas are consistently involved at various points during performance. For example, choice selection from disadvantageous decks seems to elicit activation in the medial frontal gyrus, lateral orbitofrontal cortex, and insula, and activity in these

areas is positively associated with performance (Lawrence, Jollant, O’Daly, Zelaya, & Phillips,

2009).

Also concerning is the paucity of strong causal evidence for the role of physiological feedback in IGT performance (Dunn et al., 2006). Although autonomic SCRs have been associated with both good performance on the task and activity in cortical areas important for attention and emotion, including the VMPFC (Critchley, Elliott, Mathias, & Dolan, 2000), there has been controversy surrounding what role these SCRs actually play in performance. Some have posited that anticipatory SCRs relate more to the magnitude of win and loss amounts rather than the valence (i.e., win or loss) of any given card selection, and suggest that this indicates that anticipatory SCRs do not play a role in the process of learning the long-term consequences associated with each deck (Tomb, Hauser, Deldin, & Caramazza, 2002).

Looking beyond evidence from research measuring SCRs, individuals with spinal cord damage have been shown to perform comparatively to controls, which is not inconsistent with the SMH, but suggests that the contribution of afferent spinal pathways may be minimal relative to that of the vagus and other cranial nerves (North & O’Carroll, 2001). In a unique study of epileptic patients implanted with vagus nerve stimulators, Martin and colleagues (2004) found that receiving vagal stimulation during the IGT improved performance, suggesting that the vagus nerve is indeed involved in the relay of somatic signals (Martin, Denburg, Tranel, Granner, &

Bechara, 2004). Yet surprisingly, individuals with a disease causing extensive peripheral denervation of autonomic neurons have been able to perform better than controls on the IGT

(Heims, Critchley, Dolan, Mathias, & Cipolotti, 2004). This suggests that autonomic feedback may not be critical for task performance either, and leaves only the somatosensory information traveling via spinothalamic neurons of the spinal cord as a possible source of somatic feedback 12

in this particular sample. Taken together, these findings suggest that if physiological feedback is indeed critical for affective decision-making, it must occur via diffuse and flexible networks encompassing both autonomic and somatosensory pathways.

Looking beyond the IGT, future investigations of the SMH could benefit from exploring different task paradigms and other behavioral approaches to facilitate a more well rounded examination of its principles. For example, in unique study Batson and colleges provided individuals with “false” physiological feedback during exposure to different value threats and were able to show that later decisions favored whichever value had received the stronger false feedback during exposure to the earlier threat (Batson, Engel, & Fridell, 1999).

Alternative Perspectives for Understanding Affective Decision-Making

The perseverative cognition hypothesis (PCH) and autonomic flexibility. Another framework for understanding how emotional pathology impacts decision-making is the PCH, which suggests that a pathological pattern of cognitive perseveration, intrusive repetitive thoughts common in both anxiety (i.e., worry) and depression (i.e., rumination), is related to physiological disruptions in self-regulatory systems (Brosschot, Gerin, and Thayer, 2006).

Chronic cognitive perseveration disrupts the tonic inhibitory function of the PFC, allowing the reactive disinhibition of stress responses, as well as persistent hypervigilance and attentional bias. In turn, additional components of the CAN become disrupted, leading to rigid patterns physiological reactivity. As discussed in the context of the SMH, the CAN integrates information from the sympathetic and parasympathetic nervous systems and is critical for neural control of physiological states and the regulation of emotion (Thayer & Lane, 2000).

The vagus nerve is an important component of the CAN responsible for the parasympathetic inhibitory control of the heart, which works in contraposition to sympathetic 13

processes in order to make respiration sensitive adjustments in heart rate that are important for homeostasis. Heart rate variability (HRV) can thus be used as a measure of the impact of regulatory parasympathetic influences on heart rate, reflecting autonomic flexibility, and ultimately the capacity for emotion regulation (Allen, Chambers, & Towers, 2007; Porges,

2007). Research examining HRV indices of autonomic flexibility distinguish between tonic (i.e., resting) HRV, and phasic HRV, which is stimulus or event specific. While resting HRV is widely utilized and easy to measure, phasic HRV analysis can be used to examine changes in autonomic responding across contexts, and provides a more accurate measure of autonomic contributions at specific time points, such as during feedback in a behavioral decision-making task or exposure to threat stimuli (Ruiz-Padial, Vila, & Thayer, 2011; Thayer, Friedman,

Borkovec, Johnsen, & Molina, 2000).

CAN dysfunction has been associated with depression and anxiety. Low resting HRV, or reduced vagal adjustment in heart rate, is thought to represent dysfunction in the negative feedback mechanisms of the CAN that support flexible and efficient self-regulation (Friedman,

2007). Reduced resting HRV as an indicator of autonomic rigidity has been implicated in the etiology of poor emotion regulation, anxiety, and antisocial behavior in children and adolescents

(Porges, 1992; Mezzacapa et al. 1997; for a review see Beauchaine, Gatzke-Kopp, & Mead,

2007). Consistent with evidence of hypersensitivity to bodily changes and increased physiological arousal in anxiety, several studies have shown that HRV is disrupted in adult anxiety (Thayer, Friedman, & Borkovec, 1996). Reduced resting HRV has specifically been associated with GAD (Thayer et al., 1996) and PD (Friedman, 2007), representing disparate ends of the spectrum of anxiety pathology. Preliminary evidence suggests that such disruptions may be more closely related to thought processes associated with worry, a cognitive avoidance 14

strategy, than to those associated with rumination, which is characterized by repetitive negative thoughts about the past and more strongly associated depression (Aldao, Mennin, &

McLaughlin, 2013). Aldao and colleagues found that worry predicted reduced phasic HRV during exposure to happiness and fear inducing film clips, while rumination did not predict HRV across any film clips (happy, sad, or fearful). This finding is consistent with evidence that trait rumination is not associated with resting HRV (Key, Campbell, Bacon, & Gerin, 2008). While worry can be reactive and closely tied with reactive physiological changes, rumination may be less relevant to phasic changes in HRV in response to emotional videos, as it reflects more of a maladaptive type of self-reflection.

In contrast, others have shown that phasic HRV may be reduced in response to stress in depressed individuals (Hughes & Stoney, 2000) and those with higher state rumination (Key et al., 2008). Reduced HRV has been associated with depression in several other studies (Nahshoni et al., 2004; Udupa et al., 2007), and has been negatively related to depression severity (Agelink,

Boz, Ullrich, & Andrich, 2002; Agelink et al., 2001). Associations between depression and reduced HRV have also been extensively reported in research on individuals with cardiovascular disease (Carney et al., 1995; Stein et al., 2000; Carney et al., 2000).

Although affective decision-making ability per se has not been explored within the framework of the PCH, there is considerable overlap in the neural and physiological feedback loops implicated in the SMH and the PCH. Despite this, there has been little reported use of

HRV as a physiological method that could be useful for testing the SMH and characterizing autonomic flexibility as it relates to affective decision-making ability in healthy populations or in individuals with anxiety and depression.

Interoceptive sensitivity. Interoceptive sensitivity is the degree to which one is able to sense subtle bodily changes and internal states, and includes sensations such as heart rate, pain, temperature, and muscle tension. Interoceptive representations of physiological condition are a function of the insula, a brain area that is an important part of the ANS network (Craig, 2003).

The anterior insula is thought to be important for subjective feeling states more generally, including awareness of self and context (e.g., environmental, social), and conveys this information to cortical areas important for decision-making and other executive functions

(Paulus & Stein, 2010).

Interoceptive sensitivity has most commonly been measured with a simple heartbeat perception task (HBPT; Schandry, 1981). Although a number of variations on this task have been developed over the years, the primary method involves having the participant concentrate on the sensation of their heartbeat and count the number of beats during a series of time intervals (Dunn et al., 2010; Ehlers & Breuer, 1992). The number of heartbeats counted by the participant is then compared to the electrocardiogram (ECG) recording of their actual heartbeat during the same time period to determine accuracy. Significant variability in performance within the general population has been well documented (Yates, Jones, Marie & Hogben, 1985; Katkin, Cestaro, &

Weitkunat, 1991). Good performance on the task is associated with activity in the right insula and dorsal cingulate cortex (Pollatos, Kirsch, & Schandry, 2005). The anterior cingulate may serve as a link between visceroperception and emotional processing (Cameron 2001, 2002).

Consistent with the idea that heart rate and blood pressure are critical aspects of emotional reactivity (James, 1884), good heartbeat perceivers report experiencing more intense emotions in response to emotional stimuli relative to poor perceivers (Wiens, Mezzacappa, & Katkin, 2000;

Pollatos, Gramann, & Schandry, 2007). 16

Interoceptive sensitivity has been shown to predict affective decision-making ability, and may also moderate brain activity during IGT performance. In a sample of healthy adults, Werner and colleagues (2013) found that interoceptive sensitivity was positively associated with activity in the right anterior insula. In turn, anterior insula activity was associated with decision-making performance only in individuals with greater interoceptive sensitivity (Werner et al., 2013).

Dunn and colleagues (2009) have also found that bodily responses to reward and loss influence intuitive ability during decision-making to a greater extent among individuals with higher interoceptive sensitivity. They have argued that, additional physiological feedback during decision-making could be advantageous or disadvantageous depending on the situation and the valence of physiological cues, and therefore, greater interoceptive sensitivity should not be equated with better decision-making (Dunn, Billotti, Murphy, & Dalgleish, 2009). This interpretation contributes to our understanding of how interoceptive sensitivity functions in anxious individuals.

Anxious individuals show elevated interoceptive sensitivity on self-reports and heart-beat perception measures. For example, individuals with social anxiety have been shown to have greater heartbeat perception ability during baseline and speech anticipation conditions relative to controls (Stevens et al., 2011). Despite their greater sensitivity to somatic states, people with anxiety are also more likely to interpret these states inaccurately. Individuals with PD, in particular, show a tendency to interpret bodily sensations as distressing (Ehlers & Breuer, 1995).

Conversely, during interoceptive tasks, depressed individuals perform poorly relative to controls and show bilateral reductions in dorsal medial insula activity, with the degree of hypoactivity associated with depression severity (Furman, Waugh, Bhattacharjee, Thompson, &

Gotlib, 2013; Avery et al., 2013). These findings are consistent with the evidence of reduced physiological reactivity in individuals with depression (McFarland & Klein, 2009).

The contrasting findings on interoceptive sensitivity in anxiety and depression bring into question how interoceptive processes may be influenced by the co-occurrence of these two disorders. Thus far, only a few studies have investigated interoceptive sensitivity in individuals with anxious-depression (Dunn et al., 2010; Pollatos, Traut-Mattausch, & Schandry, 2009).

While both confirmed a negative association between depression and interoceptive sensitivity, their findings regarding anxious-depression were inconsistent and additional research will be needed in order to develop reliable conclusions.

The incorporation of interoceptive sensitivity measures in the study of affective decision- making ability in anxiety and depression has the potential to further characterize and clarify the role of physiological input in decision-making across these disorders and their co-occurrence. To our knowledge, only two studies have been conducted in this vein. Wölk and colleagues found that greater interoceptive sensitivity predicted poor IGT performance among individuals with PD

(adult inpatients), but predicted good IGT performance among age matched controls (Wölk et al.,

2013). This finding confirms previous indications that anxiety is associated with improved interoceptive ability, but shows that this association may be detrimental to decision-making, in at least some types of anxiety. Furman and colleagues found that among women with MDD

(without comorbid anxiety disorders), self-reported decision-making difficulty on the Structured

Clinical Interview for DSM-IV was associated with poorer heartbeat perception, suggesting that reduced interoceptive sensitivity may be related to decision making difficulties in MDD (Furman et al., 2013).

PRESENT STUDY

Overall, the literature reviewed suggests that regulatory physiological feedback loops play an important role in the cognitive and emotional pathology of anxiety and depression, but additional research is needed to clarify whether, and through what mechanisms, anxiety and depression may differentially impact affective decision-making. The present study aimed to bring greater coherence and generalizability to prior neuropsychological findings in this area of research by examining potential physiological and contextual factors contributing to differences in affective decision-making observed in anxiety and depression.

The first major aim was to evaluate anxiety- and depression-associated differences in affective decision-making using methods that facilitate comparisons with, and address the limitations of, prior studies. As discussed above, prior research in this area has largely focused on either anxiety or depression, without examining or controlling for the high rates of comorbidity between these two conditions. Consideration of both conditions seems prudent, however, given that profiles of affective decision-making differ when comparing these studies, with higher rates of impaired decision-making observed in association with depression, and risk adverse decision-making often associated with anxiety. To address this limitation, we assessed dimensional and comorbid characterizations of anxiety and depression through the use of continuous self-report measures of anxiety (i.e., Beck Anxiety Inventory; BAI), worry (i.e.,

PSWQ), and depression (i.e., Beck Depression Inventory – 2nd Edition; BDI-II) in the prediction of task performance. Analyses involving participants with comorbid anxious-depression were largely exploratory given the paucity of prior research addressing decision-making in this population.

An additional limitation to the generalizability of prior research has been the number of tasks that have been used to study affective decision-making, and the limited degree to which their similarities and differences have been characterized within the same study or sample.

Although the IGT continues to be used frequently, it is not without its limitations, and other tasks may be more suitable for addressing certain research questions. While having an array of tasks is beneficial for developing more nuanced and flexible models of affective decision-making, this progress has been somewhat encumbered by uncertainties about how some of the measures relate to each other, and subsequently about the extent to which findings from one might be generalizable to another. The present study compared performance across three established computer tasks assessing affective decision-making (i.e., the IGT, Game of Dice Task: GDT, and

BART), which have generally been used independently, in order to establish their associations with one another and their relative sensitivities to anxiety and depression. By comparing these three tasks within the same sample, we aimed to better characterize how anxiety and depression interact with specific contextual factors in decision-making, such as uncertainty and appropriateness of risk taking, which vary across the three tasks.

The second major aim of the present study was to determine whether potential differences in ANS responding associated with anxiety and depression would predict affective decision- making performance. As discussed above, the integration of physiological processes in decision- making is a central tenant of the SMH, and has previously been characterized in research on affective decision-making, particularly in studies using the IGT. Using Biopac MP36-R equipment to measure GSR, we aimed to clarify whether gain- and loss-related SCRs, which have previously been associated with affective decision-making performance, vary in their associations with task performance among individuals with anxiety versus depression. 20

Additionally, we sought to determine whether the established association between SCR amplitude and decision-making performance on the IGT could also be observed using the BART and the GDT.

An additional physiological aim of this study was to examine the potential role of autonomic flexibility, operationalized as HRV, in affective decision-making performance.

Autonomic flexibility is a self-regulatory process of the CAN, which may be disrupted by chronic stress, perseverative thought patterns, and depression. Though low autonomic flexibility has previously been associated with anxiety and depression, few studies have examined its potential effects on affective decision-making. Toward expanding on the GSR-based characterization of ANS involvement in affective decision-making, we strove to clarify whether low tonic HRV differentially predicts decision-making ability in individuals with depression or anxiety.

The third and final core aim of this study was to determine whether interoceptive sensitivity (i.e., awareness of bodily sensations) moderates the relationship between physiological reactivity and decision-making performance. A substantial body of literature has developed on the topic of interoception and interoceptive sensitivity in anxiety and depression, and generally suggests that interoceptive sensitivity may be increased in anxiety and reduced in depression. While only a few investigations have begun to look at the potential role of interoceptive sensitivity in affective decision-making, preliminary findings suggest that it may aid in determining how strongly physiological feedback influences the decision-making process.

By examining interoceptive sensitivity alongside our other measures we aimed to provide additional information about how specific profiles of decision-making in anxiety (i.e., risk aversion) and depression (i.e., slower, impaired) are developed. By operationalizing 21

interoceptive sensitivity as HBPT accuracy, and we investigated whether interoceptive sensitivity interacts with GSR variables to explain additional variance in affective decision- making performance among individuals with anxiety and/or depression.

There were four main hypotheses related to the aims of this study. First, we hypothesized that anxiety would be associated with increased sympathetic reactivity, and that the reverse would be true of depression. Since a substantial body of prior research indicates that autonomic arousal is chronically elevated in anxious individuals and may be somewhat muted in individuals with depression, we expected to replicate these findings.

Second, we hypothesized that greater physiological reactivity, as indicated by increased frequency and/or amplitude of SCRs, would predict better performance on the IGT and GDT, and worse performance on the BART. Because emotional learning should depend upon physiological arousal, we expected that greater reactivity would facilitate learning on tasks that require a risk-averse approach to decision-making. However, because the BART is uniquely designed to require moderate risk-taking and tolerance of uncertainty, we expected that heightened physiological reactivity in this context might deter individuals from earning money in the game.

Third, we hypothesized that interoceptive sensitivity would moderate the relationship between physiological reactivity and decision-making task performance. Because interoceptive sensitivity reflects the extent to which information about physiological processes is conveyed back to the brain, it may help to gauge how strongly decision-making is informed by physiological reactivity. Physiological reactivity to reward and loss has been shown to influence decision-making ability to a greater extent among individuals with better heartbeat perception, but it is not yet known how this relationship may play out in anxious and depressed individuals 22

who typically have higher and lower interoceptive sensitivity, respectively (Paulus & Stein,

2010; Werner et al., 2013).

Fourth, we hypothesized that autonomic flexibility, as indicated by tonic HRV, would be reduced in both anxiety and depression. We aimed to replicate this finding from the literature and predicted that autonomic flexibility would be associated with decision-making ability; however, there was too little conclusive evidence about this relationship in the literature to hypothesize the direction of a potential association. Prior research suggests that anxious and depressed individuals have reduced tonic HRV, reflecting difficulties with adjusting physiological arousal in response to changes in in the environment (Friedman & Thayer, 1998; Gorman & Sloan,

2000), but how this relates to decision-making is less clear. Reduced HRV in individuals with anxiety may give them an advantage on tasks like the IGT where the generation of somatic markers to avoid risk is beneficial, but support for this explanation is mixed (Miu et al., 2008;

Mueller et al., 2010; Werner et al., 2009).

In testing these hypotheses, the present study contributes to the affective decision-making literature by synthesizing relevant theoretical frameworks and bridging a range of neuropsychological and physiological methodologies, in order to clarify their associations and relative utilities in assessing affective decision-making within the same sample. Our findings provide an improved understanding of the flexibility and appropriateness of three different decision-making tasks, and characterize less studied mechanisms (e.g., autonomic flexibility and interoception) of affective decision-making that can be targeted for anxiety and depression treatments.

METHOD

Participants

Participants were 104 adult undergraduate students recruited from the Introductory

Psychology course participant pool, as well as the general student population at Hunter College.

In order to be eligible, participants had be at least 18 years of age. Recruiting on a large public college campus in an urban setting provided a diverse sample of participants (e.g., 78% non- white). Undergraduate students are an important population for this research because they are at increased risk for anxiety and depression (Eisenberg, Gollust, Golberstein, & Hefner, 2007).

Furthermore, emerging adulthood (i.e., ages 18-29) is a critical time when affective cues are especially likely to impact decision-making (Brown, Tolou-Shams, Lescano, & Lourie, 2006).

Procedure

After providing informed consent, participants were connected to the physiological equipment and baseline readings were established while they watched an ocean nature video for three minutes. They next completed a HBPT to measure interoceptive sensitivity. The three decision-making tasks were then completed on the computer, followed by a brief verbal working memory task administered by the experimenter. Finally, participants completed a survey about their emotional health and cognition, also on the computer. The study duration ranged from 60 to

90 minutes, and participants were compensated with Psychology 1000 course research credits.

Neuropsychological Tasks

IGT. The IGT is a computerized card game test of judgment and decision-making specifically designed to test the tenets of the SMH (Bechara et al., 1994). Participants are told to win as much virtual money as possible by selecting cards from four available decks, with each selection resulting in either a gain or a gain and a loss of money. Good performance requires 24

learning to select from decks that produce smaller wins in the short-term but ultimately yield the highest gains (i.e., good decks). This approach has been associated with stronger physiological autonomic responding (e.g., GSR) to emotional stimuli in the game, which is thought to represent the generation of somatic markers as anticipatory cues of reward and loss (Carter &

Pasqualini, 2004). In contrast, continual selection from bad decks initially results in high rewards, but ultimately delivers even larger losses, and is thought to reflect insensitivity to future consequences and a failure to incorporate physiological affective feedback in decision-making.

The IGT is organized so that participants are generally unable to consciously predict its reward/loss schedules; however, most individuals gradually learn to select from the good decks and perform well over the course of the task’s 100 trials. In order to examine possible differences in the type of decision-making assessed during early versus later trials of the IGT, performance on the first 60 trials (i.e., first 3 blocks; decisions under ambiguity) is sometimes examined separate from performance on the last 40 trials (i.e., last 2 blocks; decisions under risk). Previous studies have noted that the point at which the different reward/loss contingencies of the task become more apparent is subject to individual variability; however, in most cases this learning process takes at least 40 trials, and can be expected to have occurred by the last 40 trials (Brand,

Recknor, Grabenhorst, & Bechara, 2007b; Noël, Bechara, Dan, Hanak, & Verbanck, 2007).

For the present study we used a version of the IGT based on the original task (Bechara et al., 1994) and adapted for presentation in Millisecond Inquist 3 SoftwareTM. This version of the task was used in order to program event-specific markers that signal to physiological recording software whenever a participant wins or loses money. Raw total scores were computed for each of the five 20-trial blocks of the task and reflect the number of bad deck selections subtracted

from the number of good deck selections. For the purposes of the present study, IGT total score across all blocks (adjusted for losses) was used as the primary IGT dependent variable.

BART. The BART is a well-established measure of risky decision-making designed to be sensitive to impulsivity and capture real-word risk behavior (Lejuez et al., 2002). Participants are asked to earn money by pumping up 30 balloons, each representing a new round of the game.

Money accumulates with each pump, as does the risk of the balloon popping, which results in a loss of all earnings for that round of the game. Raw scores are calculated for the total dollar earned amount, the total number of pumps (across non-pop trials), and the number of pops over the course of the task. Unlike the IGT, good performance on the BART requires some willingness to take risks and therefore should be particularly sensitive to the risk-aversion common in anxiety disorders (Hunt, Hopko, Lejuez, & Robinson, 2005). The BART was used to address our research question of whether increased physiological reactivity associated with anxiety may be beneficial in some decision-making contexts (e.g., if optimal outcomes require attention to future outcomes) but not others (e.g., if optimal outcomes require risk). Further, it was used to parse whether the deficits in discriminating reward and punishment cues in depression contribute to risk behavior, or simply to difficulties with making decisions due to slower learning.

Like the IGT, the version of the BART used in the present study was programed using

Millisecond Inquist 3 SoftwareTM in order to send event markers to our physiological recording software to indicate whenever the participant popped a balloon or successfully completes a round of the game.

GDT. The GDT is an affective decision-making task that provides explicit rules for gains and losses, such that decisions regarding risk can be made with certainty (Brand et al., 2005). 26

Participants are tasked with predicting how a single dice will land, with more money at stake for single outcome bets versus multiple outcome bets (i.e., the dice will land on any 1/1, 1/2, 1/3 or

1/4 different numbers). There are 18 trials of the task, and raw scores are computed for the total amount earned and the total number of safe dice selections minus risky dice selections. We compared performance on the GDT with performance on the IGT and BART, we in order to capture whether intolerance of uncertainty, a common feature of anxiety, potentially disrupts decision-making (Heilman, Crisan, Houser, Miclea, & Miu, 2010). The GDT task was programed using Inquist 3 SoftwareTM and event marker signals were created to indicate each win and loss.

Wechsler Adult Intelligence Scale-Fourth Edition (WAIS-IV) Digit Span Subtest.

Working memory is often implicated in complex decision-making, with numerous studies indicating associations between performance on working memory and decision-making tasks such as the IGT (Bechara & Martin, 2004; Hinson, Jameson, & Whitney, 2003). Furthermore, working memory contributes to emotion regulation and is a component of executive functioning frequently disrupted in anxiety (Schmeichel, Volokhov, & Demaree, 2008; Shackman et al.,

2006). In order to consider these relationships and take related variance in decision-making into account, the present study included the administration of the Wechsler Adult Intelligence Scale

(WAIS-IV) Digit Span subtest forward and backward trials (Wechsler, 2008). For this brief task, participants were asked to repeat strings of numbers in either forward or backward order relative to how they were read by the experimenter. The backward portion of the task assesses working- memory by requiring mental manipulation of the number sequence. The present study used the

Longest Digit Span Backwards (LDSB) score as our primary variable for assessment of working memory. Longest Digit Span Forward (LDSF), a measure of basic auditory attention, was also 27

examined to rule the possibility of LDSB effects arising due to differences in basic auditory attention versus working memory. The WAIS-IV Digit Span subtest is a widely used and well- validated measure of working memory capacity, and has the additional advantage of being well normed (Conway et al., 2005). All analyses examining working memory for this study were exploratory in nature, with the aim of characterizing potential differences in sensitivity to working memory between the three decision-making tasks, and to capture any variability in working memory that might be associated with depression or anxiety and thereby affect task performance.

Furthermore, the Digit Span Subtest is often used as an embedded effort measure in neuropsychological assessment. The reliable digit span (RDS) is the most widely researched symptom validity measure, with cut off scores of ≤ 7 commonly used to indicate poor effort in normal adult populations (Schroeder, Twumasi-Ankrah, Baade, & Marshall, 2012). The present study used this metric to screen for low effort among study participants. Individuals with scores of ≤ 7 were flagged for data inspection.

Physiological Measures

Physiological data was gathered from participants throughout each decision-making task in order to capture the autonomic correlates of emotional responding as they related to performance. We selected two different physiological measures (i.e., tonic HRV and GSR) widely used in psychophysiology and decision-making research (Allen et al., 2007; Figner &

Murphy, 2011; Navqi & Bechara, 2006). Physiological data was collected using Biopac MP36-R recording equipment. Data was recorded at a rate of 1000 Hz (1000 samples/second) using

AcqKnowledge 4.0 software. All tasks were programed to communicate events of interest (e.g.,

money lost) to the AcqKnowledge software to be marked in the physiological data for event related analysis.

GSR. GSR is a measure of subtle changes in sweat secretion on the fingertips indicating physiological arousal driven by the sympathetic branch of the ANS. Event-specific SCRs are temporary increases in the amplitude of the GSR signal that indicate greater sympathetic arousal in response to a stimulus. For the present study, we defined event-specific SCRs with a threshold of 0.03 uhmo/µS, based on the sensitivity of our recording methods and current directions in the literature (Braithwaite, Watson, Jones, & Rowe, 2013; Figner & Murphy, 2011). Electrodes were placed on the distal flanges of the non-dominant hand’s index and middle fingers following preparation by wiping the skin with water and a paper towel. SCRs were captured before and after negative and positive feedback (e.g., win or loss events) in the tasks to track sympathetic arousal during performance. For each task, we examined the overall frequency of event-specific and non-specific SCRs, as well as their amplitudes, using AcqKnowledge 4.0 software. GSR has been used extensively with the IGT to demonstrate decision-making deficits in clinical samples

(Bechara et al., 1994; Werner et al., 2009). The generation of SCRs during decision-making is associated with activity in brain areas important for emotion, attention, and decision-making

(Critchley et al., 2000).

HRV. HRV is variability is the autonomic control of heart rate measured by sampling the inter-beat interval (i.e., time between R-spikes) in ECG heart rate data. HRV is a common index of autonomic flexibility and the capacity for emotion regulation (Allen et al., 2007; Thayer et al.,

1996: Thayer & Lane, 2000). In the present study, HRV was measured during each task and at baseline. Tonic (i.e., resting) HRV during the three-minute baseline was used for the final analyses to broadly characterize variability in autonomic flexibility among participants. Data 29

were collected via recording from electrodes placed: 1) above the right clavicle, 2) on the left side of the abdomen above the hip, and 3) on the inside of the left forearm. Each participant’s interbeat interval (IBI) time series was hand corrected, and HRV metrics were calculated using

QRSTool and CMET software (Allen et al., 2007).

Two metrics of HRV were selected based on prior research (Allen et al., 2007; Williams et al., 2015). The PNN50 metric represents the proportion of IBIs in consecutive heart rate data that differ by more than 50 milliseconds and is a time-domain measure of HRV that may be more sensitive to respiration-linked parasympathetic (i.e., vagal) influence. The root mean square differences (RMSSD) is another commonly reported time-domain measure of vagally-mediated

HRV. Both PNN50 and RMSSD have been reported as highly correlated with high frequency

(HF) power HRV, which suggests that they are both primarily vagally mediated (Williams et al.,

2015; Ellis, Sollers, Edelstein, & Thayer, 2008). RMSSD and PNN50 have been shown to load onto the same factor and have a correlation of .89 at rest in a normal college sample (Hibbert,

Weinberg, & Klonsky, 2012). Though both metrics were highly correlated and are reported in our descriptives tables, RMSSD was ultimately chosen for our final analyses. The published guidelines for the Task Force of The European Society of Cardiology and The North American

Society of Pacing and Electrophysiology (Malik et al., 1996) state that RMSSD is preferable to

PNN50 because it has better statistical properties, and suggest that it may provide better results when the recording period is relatively brief. Thus, this metric seemed most appropriate for our data given our short (i.e., three minute) baseline period. RMSSD values were log transformed based on literature guidelines (Nunan, Sanderrock, & Brodie, 2010; Hovland et al., 2012;

Williams et al., 2015).

Interoceptive Sensitivity Measure 30

The HBPT is a common measure of interoceptive sensitivity that involves comparing the participant’s count of self-detected heartbeats to their ECG recoding to determine self-detection accuracy (Schandry, 1981). Participants completed three different heartbeat perception trials with durations of 35s, 25s, and 45s followed by three control trials (23s, 56s, and 40s), during which they were asked to estimate the duration of the trial in seconds, and finally another three heartbeat perception trials lasting 35s, 25s, and 45s (Dunn, Dalgleish, Ogilvie, & Lawrence,

2007; Ehlers & Breuer, 1992; Schandry, 1981). Participants were instructed to not use strategies such as taking their pulse or holding their breath, and were asked at the end of each trial to rate their confidence in the accuracy of their estimates. Perception accuracy was calculated using the following equation: ([|actual – estimated number of heart beats| ÷ actual] x 100) (Dunn et al.,

2010a).

Anxiety and Depression Self-Report Measures

Participants completed the PSWQ (Meyer et al., 1990), the BAI (Beck & Steer, 1993), and the BDI-II (Beck et al., 1996). These measures were selected because of evidence in the literature that they capture clinically valid and reliable features of anxiety (e.g., worry and anxious arousal) and depression (e.g., mood and somatic complaints) (Beck, Guth, Steer, & Ball,

1997; Creamer, Foran, & Bell, 1995; Dozois, Dobson, & Ahnberg, 1998; Molina & Borkovec,

1994). Each of these scales has the advantage of having normed clinical cut-offs for the detection of clinically significant anxiety and depression; however, for our primary analyses these measures were used as continuous variables. Evaluating anxiety and depression on a spectrum that includes milder symptomology increases the relevance of findings to a broader range of individuals, and is consistent with the recent movement in clinical psychology to adjust

diagnostic criteria to reflect the continuous, rather than dichotomous, nature of mental illness symptomology.

Additional Self-Report Measures

All participants completed survey measures to assess demographics (e.g., age, gender, education, and race), mental health, personality characteristics, ability to regulate emotions, and approaches to decision-making from the participant’s perspective. Race was assessed using categories (i.e., American Indian or Alaska Native, Asian, Black or African American, Native

Hawaiian or Other Pacific Islander, and White) consistent with current NIH guidelines.

Participants were also given the option to enter their own descriptors in an “other” category, and the most frequent entries in this section were for individuals of Middle Eastern descent. Specific measures included: a modified version of the College Chronic Life Stress Survey (Towbes &

Cohen, 1995; CCLS), the Behavioral Inhibition and Activation scales (Carver & White 1994;

BIS/BAS), the Barratt Impulsivity Scale (Patton & Stanford, 1995; BIS), the Emotion Regulation

Questionnaire (Gross & John, 2003; ERQ), a modified version of the The Risk Behavior Scale

(Weber, Blais, & Betz, 2002; RBS), and the General Decision-Making Style Scale (Scott &

Bruce 1995; GDMS). The BAI was designed to minimize overlaping symptomology between anxiety and depression scales. In the literature, correlations reported between the BAI and BDI range from 0.4 to 0.7, depending on the sample (Stulz et al., 2010, Beck et al., 1996). The CCLS was modified for relevance to city-dwelling college students, the majority of whom live off campus with family or friends and do not drive. The RBS was modified to to increase its relevance for college students living in New York City during the current decade (e.g., the item

“chasing a tornado or hurricane by car to take dramatic photos” was removed and the item

“illegally downloading music, tv shows, or movies” was added). 32

Data Analysis

To assess relationships between anxiety, depression and physiological reactivity (i.e.,

GSR), Spearman’s rank correlations were calculated with continuous variables for the frequency

(i.e., number) and strength (i.e., amplitude) of SCRs and participant’s scores on affective measures (i.e., BAI, BDI-II, and PSWQ). The same approach was used to assess whether autonomic flexibility (i.e., tonic HRV) was negatively associated with anxiety and depression.

To determine whether physiological reactivity predicts decision-making performance, multiple regression analyses were conducted for each task, with GSR variables and continuous measures of distress as predictors of performance. To assess whether interoceptive sensitivity moderates the relationship between physiological reactivity and task performance, we added HBPT ability as an additional factor in the regression models described above and tested for interactions with

GSR variables. To explore whether autonomic flexibility predicted task performance, multiple regression analyses were conducted with continuous variables for tonic HRV and each distress measure predicting performance on each task. Bivariate comparisons were made to explore potential differences between anxiety, depression, and anxious-depression in terms of physiological measures, task performance, and self-report covariates. Exploratory analyses were conducted to control for potential variability in decision-making performance related to attention and working memory capacity (i.e., WAIS-IV Digit Span Task). Toward characterizing potential differences in sensitivity to working memory between the three decision-making tasks, multiple regression analyses were conducted for each task with LDSF and LDSB as predictors.

RESULTS

Sample Size 33

A total of 104 participants completed the study; however, the final sample size was 100, as some of the survey data required for evaluating demographic, anxiety, and depression characteristics was missing (n = 1) or invalid (n = 3) due to participants’ rushed/indiscriminate responding (i.e., selecting the same response item for all questions). During data cleaning and analysis, seven participants were identified who needed to be excluded from all analyses. Two were found to have only partial data for each of the decision-making tasks due to software problems. Another two failed to meet the cut off for adequate effort on an embedded validity measure (i.e., reliable digit span performance), and a final three were excluded for additional issues (i.e., use of the internet during the appointment, skipping trials of the BART, and stated cultural bias against gambling that prevented full task engagement). Thus, our starting analytic sample consisted of 93 participants.

Affective Variable Adjustments

Internal consistency was acceptable for all affective self-report measures (BDI-II: α =.89;

BAI: α = .91, PSWQ = .91). All three measures showed some departure from normality. The

BAI showed more pronounced skewness (1.49) and kurtosis (3.52) than the BDI-II (skewness =

0.76; kurtosis = -0.10) and the PSWQ (skewness = -0.42; kurtosis = -0.59). Analyses involving continuous BAI scores were thus based on log transformed data (skewness= 0.35; kurtosis= -

0.15). Specifically, a Box-Cox procedure was used to produce λ ranging from -2.00 to 1.00, and we determined that the optimal power transformation (after BAI was anchored at 1.0) would be achieved using λ = 0.5, as this yielded the lowest combined skew and kurtosis. Descriptives, including correlations, show both the transformed and non-transformed BAI variables for comparison.

Descriptives and Bivariate Comparisons 34

Descriptive statistics for demographics and affective characteristics are presented in

Table 1, and include values for both the full (N = 100) and analytic sample (N = 93). Table 2 shows descriptive statistics for the continuous measures of decision-making task performance, working memory, and affective variables. Nonparametric (i.e. Spearman’s rank) correlations between primary variables of interest are shown in Table 3. Anxiety, worry, and depression measures were all positively associated with each other. Anxiety was positively associated with performance on the GDT only, while depression was negatively associated with the effort measure of reliable digit span (RDS) score only. Worry was negatively associated with both

LDSF and RDS scores. IGT performance was positively associated with GDT performance and

LSDF score. None of the variables were associated with BART performance. Physiological and interoceptive sensitivity variables of interest are described in Table 4. For these measures, sample size varied due to missing data and artifacts, and the adjusted sample sizes are noted below each variable name. Log transformations were performed based on literature guidelines; both raw and transformed scores shown for comparison. Spearman’s rank correlations between affective variables, working memory, interoceptive sensitivity (HPBT accuracy), and physiological variables are shown in Table 5. There were no associations between affective and physiological variables. Working memory performance (LDSB) was positively associated with tonic HRV. Finally, affective variables, working memory, and interoceptive sensitivity were entered as factors in a series of ANOVAs predicting the physiological variables; these results are shown in Table 6. There were no significant findings.

Analysis of IGT Performance

Outliers in our dependent variable, IGT performance (i.e., total score with gains adjusted for losses), were identified as any values three or more standard deviations above the mean; 35

based on this criterion, a total of two cases were excluded. Additionally, our main effects models were screened for multivariate outliers using the Cook’s distance cut off criteria (i.e., > 4/sample size), leverage cut off criteria (i.e., ≥ 1), and large standardized residuals (i.e., ≥ 3) to identify overly influential cases. Based on this process, two additional cases were excluded from this section of the analysis, creating a final analytic sample size of 89.

Demographic variables. We first conducted exploratory main effects analyses examining demographic variables (i.e., age, gender, and race) to determine which factors might account for significant variance in IGT performance. Gender, but not race or age, significantly predicted IGT total score, such that men (M = 0.4, SD = 1.1) performed significantly better than women (M = -0.2, SD = 0.9; R2 = 0.10, F(1, 87) = 9.6, p < .01).

Regression models examining gender and affect. Independent, multivariate linear regressions predicting IGT performance were conducted with gender and continuous measures of depression, worry, and anxiety. Results of each individual affective model are presented in Table

7, as well as an adjusted model that includes all three affective variables simultaneously. Each model was adjusted for gender, given the significant bivariate association between gender and

IGT performance. Controlling for gender, depression predicted significant variance in IGT performance (Table 7, model 1), such that greater depression was associated with worse IGT performance. The parameters associated with worry or anxiety were not significant predictors of

IGT performance (Table 7, models 2 and 3). With all three affective variables in the same model, the parameter associated with depression remained fairly stable (β = -.23 and β = -.21, respectively); however, power was diminished such that only the main effect of gender remained significant (Table 7, adjusted model). Interactions between the continuous affective variables were also conducted and were not significant (results not shown). A hierarchical regression 36

model adding a gender by depression interaction did not explain any additional variance in IGT performance after the main effects for gender and depression (data not shown).

Exploratory regression models examining working memory. WAIS-IV Digit Span subtest variables, including auditory attention (i.e., LDSF) and working memory (i.e., LDSB), were entered independently into independent linear regression models predicting IGT performance. Controlling for gender, neither of these digit span variables were significant predictors of IGT performance (Table 8, models 1 and 2).

To determine whether working memory might moderate the relationship between affective distress and IGT performance, we kept LDSB in the model and checked for an interaction with depression. In this adjusted model, a main effect of LDSB emerged, and the interaction between depression and LDSB was also significant (Table 8, interaction model), such that greater working memory was associated with better performance among those reporting higher levels of depression, and worse performance in those with lower levels of depression. To confirm that this effect was driven by working memory, and not simple auditory attention, we checked for an interaction between LDSF and depression, which was not significant (R2 change

= 0.00, F change (1, 84) = .15, p = 0.65). We also checked for an interaction between gender and

LDSB, which did not explain any additional variance in IGT performance (R2 change = 0.01, F change (1, 84) = .51, p = .36).

Given the counter intuitive direction of the relationship between LDSB and IGT performance (Figure 1), we considered that our results might be due to an effect specific to people with lower working memory performance. In order to examine specific vulnerabilities for individuals with low average to impaired working memory performance, we created a categorical LDSB variable with three groups: below average range (i.e., span of 3-4), average 37

range (i.e., span of 5), and above average range (i.e., span of 6 -8), based on WAIS-IV normative data for individuals 20 years of age. Below average performance made up the largest group (i.e., n = 43), while the average and above average group sizes were even (i.e., n’s = 23). This categorical variable was plotted with depression and IGT performance (Figure 2). Given the relatively large slope for individuals in the below average group (R2 linear = 0.24) compared to the average and above average groups (for both R2 linear = 0.004), we entered LDSB dichotomized at less below average into our model; group sizes were 43 (below average) and 46

(average or above). As before, the main effect of below average LDSB was not initially significant, but after adding an interaction between below average LDSB and depression to the model, both a main effect and a significant interaction were observed (Table 9). This interaction indicated that worse IGT performance was associated with depression, but only for individuals with below average working memory (Figure 3).

Regression models examining HRV. To explore whether or not autonomic flexibility

(HRV) predicts IGT performance, we utilized a baseline, tonic (i.e., resting) measure of HRV

(i.e., RSMMD log transformed). Six participants with irremovable artifacts in the HRV data and one extreme outlier (z = 4.53) needed to be removed for this analysis, leaving the final analytic sample size at 83. Using multiple regression, greater HRV at baseline did not predict better IGT performance after controlling for gender, depression, and LDSB (Table 10, model 1). In an exploratory analysis including working memory, a hierarchical regression model revealed that, while two-way interactions between HRV and LDSB or depression were not significant, a three- way interaction was. HRV interacted with LDSB and depression (dichotomized at mild depression), such that in depressed individuals with below average LDSB, lower HRV was associated with worse IGT performance (Table 10, model 2; Figure 5). HRV did not appear to 38

predict IGT performance among individuals with at least average LDSB score or individuals without depression symptomology (Figures 4 and 5).

Regression models examining GSR. To test the hypothesis that greater sympathetic reactivity would be associated with better performance on the IGT, we examined several GSR measures at baseline and across task performance. For this portion of the analysis, data for eight participants were excluded due to artifacts or other recording problems with the GSR equipment, resulting in an analytic sample size of 81.

We first examined measures of the average amplitude of SCRs in responses to gains and losses across all decision-making tasks (i.e., IGT, BART, and GDT). These scores were log transformed in order to correct for skew and kurtosis (see methods). After controlling for gender and depression, average SCR amplitude in response to losses, but not in response to gains, demonstrated a significant trend (p = .06) toward predicting IGT performance (Table 11).

Interestingly, when both measures were included in the model, they were both highly significant

(Table 9, adjusted model); however, a significant positive correlation (i.e., r = .76) between the average loss and gain amplitudes may have inflated the R2 value for this model. The average

SCR amplitude response to losses was also found to interact with gender, such that greater reactivity was a stronger predictor of IGT performance among women (Table 12; Figure 6).

Depression did not interact with average SCR amplitude response to losses.

We further examined the average frequency of SCRs of any type (i.e., both event related and non-event related) per minute across all decision making tasks, as well as during the three minute resting baseline period, in order to capture overall sympathetic reactivity, irrespective of task gains and losses events (Table 13). Average frequency of SCRs, either across tasks or across baseline, did not predict IGT performance, suggesting that the previously described relationship 39

(Table 12) between loss-specific SCR event amplitude and IGT performance is likely unique to loss events, and not simply an effect driven by overall sympathetic reactivity.

Regression models examining interoceptive sensitivity. In order to test our hypotheses that interoceptive sensitivity moderates the relationship between affect and sympathetic reactivity in the prediction of IGT performance, we examined percent accuracy on the HBPT.

Two participants were missing data for the HBPT, and therefore our initial sample size for this section of the analysis was 87.

To first examine the relationship between HBPT and IGT performance, we computed the main effect of HBPT accuracy, as well as potential interactions with gender and the affective variables, in the prediction of IGT performance. The main effect of HBPT accuracy was not significant (Table 14, model 1). Anxiety, but not depression or worry, interacted with HBPT accuracy, such that better accuracy was associated with worse performance for anxious participants, while for those without anxiety, the reverse was true – HBPT accuracy predicted better IGT performance (Table 14, model 2; Figure 7). HBPT accuracy was not correlated with anxiety (r = .05), worry (r = .07), or depression (r = .06).

We next added SCR loss amplitude back into the model, along with HBPT, gender, and anxiety predicting IGT performance. For this stage of the analysis, the sample size was restricted by the available GSR data. With the participants missing GSR and HBPT data excluded, the analytic sample size was 79. We first tested for potential main effects of interoceptive sensitivity and SCR loss amplitude. Consistent with the previous observed trend (Table 11), greater SCR loss amplitude predicted better IGT performance, while the relationship between HBPT and IGT performance was not significant (Table 13, model 1). We then entered the interaction between

HBPT and SCR loss amplitude into the model; this relationship was not significant using 40

continuous or median split HBPT measures for the interaction term (Table 15, model 2a). We further tested for triple interactions, alternating gender and anxiety as additional factors in the interaction term, given their previously observed interactions with HBPT performance (anxiety) and interoceptive sensitivity (gender). The triple interaction term including anxiety was significant, and this final model ultimately explained 30% of the variance in IGT performance

(Table 15, model 2b). The positive predictive relationship between loss associated SCR amplitude and IGT performance was strongest in non-anxious individuals when HBPT accuracy was greater, while in anxious individuals, it was strongest when HBPT was lower (Figure 8).

Given the negative association between depression and IGT performance, we also repeated this set of analyses alternating depression for anxiety. Although the main effect of depression persisted, depression did not interact with interoceptive sensitivity and/or sympathetic reactivity to predict IGT performance (results not shown).

IGT results summary. Gender and depression were both consistent predictors of IGT performance, with women and depressed participants doing more poorly on the task. In particular, participants who reported significant depression symptomology, and had either low tonic HRV or below average working memory scores, were the most likely to do poorly on the

IGT. Greater sympathetic reactivity in response to losses (across decision-making tasks), but not gains, was associated with better IGT performance, and this association was stronger among women. Interoceptive sensitivity was associated with worse IGT performance among individuals with anxiety, and better performance in individuals without anxiety. When sympathetic reactivity and interoceptive sensitivity were examined in the same model, it appeared that anxious participants performed best when sympathetic reactivity was higher, but interoceptive sensitivity was lower. 41

Analysis of GDT Performance

Outliers in our dependent variable, GDT performance (i.e., final balance) were identified as any values three or more standard deviations above the mean, and one case was excluded based on this criterion. Additionally, main effects models were screened for multivariate outliers using the Cook’s distance, leverage, and standardized residual diagnostics described previously.

Based on this process, two additional cases were excluded and the final analytic sample size was

90.

Demographic variables. We first conducted exploratory main effects analyses examining demographic variables (i.e., age, gender, and race) to determine which factors might account for significant variance in GDT performance. Gender and age were not significant predictors of GDT performance. Race did predict performance, such that individuals of Asian and Middle Eastern background performed significantly better that individuals identifying as white, African American, or Latino (R2 = .097, F (1, 89) = 9.5, p < .01). Gender and age were still not significant in the model after controlling for race.

Regression models examining race and affect. Independent regression models for depression, anxiety, and worry, indicated that none of these affective variables were significant predictors of GDT performance after controlling for Asian and Middle Eastern race (here on referred to as just race); however, significant trends were observed for both anxiety and worry, such that each was associated with better performance (Table 16, models 2 and 3). Interactions between the continuous affective variables were also conducted and were not significant (results not shown). Hierarchical models adding interactions between race and the affective variables were not significant; however, the main effect of worry became statistically significant after including a worry by race interaction in the model (Table 17). 42

Exploratory regression models examining working memory. WAIS digit span subtest variables, auditory attention (i.e., LDSF) and working memory (i.e., LDSB), were entered independently into linear regression models predicting GDT performance. Neither LDSB nor

LDSF were significant predictors of performance after adjusting for race (Table 18). To determine if working memory might moderate a relationship between our affective variables and

GDT performance, we utilized the categorical LDSB variable previously described in the IGT results section, and examined interactions with anxiety and worry using hierarchical regression.

LDSB interacted with anxiety, but not worry, such that among those with average or lower

LDSB scores, anxiety was associated with better performance (Table 19, Figure 9). This finding should be interpreted with caution given the unequal groups sizes for the above average LDSB (n

= 23) versus average or below LDSB (n = 67) groups.

Regression models examining HRV. To explore whether or not autonomic flexibility

(i.e., HRV) would predict GDT performance, we again utilized our baseline measure of HRV

(i.e., RSMMD log transformed), and six participants with data artifacts and one extreme outlier were again removed, leaving the final analytic sample size at 83. After adjusting for race and worry, a main effect of HRV was not significant (Table 20, model 1); however, hierarchical regression revealed a significant interaction between HRV and worry, such that worry was associated with better GDT performance, unless HRV was low (i.e., in the lower 3rd percentile), in which case the inverse relationship was observed (Table 20, model 2; Figure 10). HRV did not significantly interact with race, anxiety, or LDSB (results not show).

Regression models examining GSR. To test the hypothesis that greater sympathetic reactivity would be associated with better performance on the GDT, we examined several GSR measures at baseline and during across task performance. For this portion of the analysis, data 43

for 10 participants were excluded due to artifacts or other recording problems with the GSR equipment, leaving the total analytic sample size at 80.

We first examined our log-transformed measures of SCR amplitude in response to gains and losses averaged across all decision-making tasks. After controlling for race, neither loss- nor gain-associated SCR amplitude predicted GDT performance (Table 21). Including anxiety and worry in the models did not improve the predictive power of the model, and neither SCR amplitude measure was found to interact with race, gender, anxiety, or worry (results not show).

We further examined the average frequency of SCRs of any type (i.e., both event related and non-event related) per minute across all decision making tasks, as well as during the three minute resting baseline period, in order to capture overall sympathetic reactivity, irrespective of task gains and losses events. Average frequency of SCRs, either across tasks or across baseline, did not predict GDT performance (results not shown).

GDT results summary. GDT performance was found to be sensitive to race, with participants of Asian and Middle Eastern descent achieving higher gains in the task, while gender and age were not related to performance. Significant trends were observed for worry and anxiety as positive predictors of performance. Individuals reporting anxiety appeared to do better on the GDT when they had average or below average working memory performance, a finding that should be interpreted cautiously. Additionally, worry was found to be detrimental to performance among participants with low HRV. Measures of sympathetic reactivity did not predict significant variance in GDT performance.

Analysis of BART Performance

Due to technical problems, three participants were missing data for the BART and therefore could not be included in this section of the analysis. Our dependent variable for BART 44

performance, total money earned, was used in order to reflect both overall performance and the risk aversion indicated by low scores. As previously described, models were screened for multivariate outliers by examining Cook’s distance, leverage, and large standardized residuals.

Based on this process, two additional cases were identified as overly influential and excluded from this section of the analysis, creating a final analytic sample size of 88.

Demographic variables. We first conducted exploratory main effect analyses examining demographic variables (i.e., age, gender, and race). None of these variables were significant predictors of BART performance.

Regression models examining affect. Depression, anxiety, and worry were not significant predictors of BART performance (Table 22), and gender did not interact with any of these affective variables to predict BART performance. Interactions between the continuous affective variables were not significant (results not shown).

Exploratory regression models examining working memory. WAIS digit span subtest variables (i.e., LDSF and LDSB) were entered independently into regression models predicting

BART performance. No significant main effects were observed for either of these variables and neither interacted with gender (Table 23). To explore whether working memory could moderate a relationship between affective distress and BART performance, we again utilized the categorical LDSB variable to examine potential interactions with affective variables. Using hierarchical linear regression, we observed that below average LDSB scores predicted BART performance, and anxiety interacted with dichotomized LDSB to explain additional variance in

BART performance (Table 24), such that at least average LDSB scores were associated with worse BART performance (i.e., greater risk aversion) in individuals reporting significant anxiety symptomology (Figure 11). Furthermore, worry also interacted with above average LDSB scores 45

(Table 25), such that worry was associated with worse BART performance, but only among those with above average LDSB scores (Figure 12).

Regression models examining HRV. To explore whether or not autonomic flexibility

(HRV) would predict BART performance, we again used the baseline HRV variable (i.e.,

RSMMD, log transformed), and the smaller analytic sample size of 80. After accounting for gender, worry, and anxiety, HRV did not predict BART performance (Table 26). Furthermore,

HRV did not interact with gender, LDSB, depression, worry, or anxiety to predict significant variance in BART performance (results not shown).

Regression models examining GSR. To test the hypothesis that greater sympathetic reactivity would be associated with worse performance on the BART, we examined several GSR measures at baseline and during across task performance. For this portion of the analysis, data for 9 participants were excluded due to artifacts or other recording problems with the GSR equipment, leaving the total analytic sample size at 79.

We first examined our log-transformed measures of average SCR amplitude in response to gains and losses, averaging across tasks. Neither loss, nor gain, associated SCR amplitude predicted BART performance (Table 27). Including anxiety or worry in the models did not improve the predictive power of either SCR amplitude measure, and neither measure was found to interact with race, gender, anxiety, or worry.

We also examined the average number of SCR events (both event and non-specific) per minute, averaging across decision-making tasks, and during the three minute resting baseline period, as measures of overall sympathetic reactivity. Average SCR events across tasks or across baseline did not predict BART performance, and these measures also did not interact with any of the affective variables to predict BART performance (results not shown). 46

BART results summary. BART performance was not sensitive to gender, age, or race.

None of the affective variables alone accounted for significant variance in BART performance.

Anxiety and worry were found to interact with working memory to predict BART performance, such that anxiety and worry both appeared to be detrimental for performance among participants with better working memory ability, while the performance of those with worse working memory appeared less affected by anxiety and worry. Measures of HRV and sympathetic reactivity did not predict significant variance in BART performance.

Affective Group Analysis

Given the frequent co-occurrence of anxiety, depression, and worry, we conducted an additional set of analyses characterizing physiological differences between three groups based on self-report: 1) anxiety and/or worry (n = 30), 2) depression and anxiety and/or worry (n = 35), 3) no significant affective symptoms (n =28). These groups were defined using cutoffs of 10 on the

BAI, for mild anxiety, 14 on the BDI, for mild depression, and a median split for the PSWQ. A total of four participants where characterized as depressed without significant anxiety or worry symptoms were included in the depression and anxiety and/or worry group; results did not differ significantly as a result of this inclusion. To test our hypothesis that anxiety would be associated with greater physiological reactivity, and depression with less, ANOVA was used to assess group differences in HRV, average SCR amplitude across tasks for gains and losses, and average

SCR events per minute across tasks and baseline. After adjusting for gender, affective grouping did not predict any of the physiological measures (Table 28).

Task Comparison Analysis

To explore the relationships between our three affective decision-making tasks, we first conducted a Pearson’s correlation analysis with our primary task associated dependent variables 47

(Table 3). The IGT and GDT were positively associated, while the BART was not associated with either of these two tasks. In an exploratory analysis, we also examined these relationships within our task-specific regression models. After adjusting for gender and depression, BART performance (i.e., total money earned) significantly predicted IGT performance, such that higher scores on the BART were associated with worse IGT performance (Table 29). After controlling for gender, depression, and race, GDT performance significantly predicted IGT performance, such that better GDT performance was associated with better IGT performance; however, this effect was only observed when GDT performance was entered as the percentage of dice selections that were safe versus risky (Table 30). BART performance did not predict GDT performance or visa versa (Table 31).

DISCUSSION

The present study sought to expand on prior research suggesting that anxiety is associated with risk-averse decision-making, while depression is often associated with disadvantageous decision-making. Toward this end, we examined associations between self-reported anxiety and depression and performance on three different decision-making tasks (i.e., the IGT, GDT, and

BART), which had not previously been compared within the same study. Measures of working memory, autonomic flexibility and reactivity, and interoceptive sensitivity were also analyzed as potential moderating factors, to better understand mechanisms involved in the relationships between affective decision-making and anxiety and depression. Given the central role of decision-making in self-regulation and the development of maladaptive behaviors, this area of research has the potential to inform interventions (e.g., for anxiety, depression, eating disorders, and/or substance abuse), particularly those that combine approaches (e.g., mindfulness, 48

biofeedback) to address physiological, as well as cognitive and emotional aspects of these disorders.

IGT

Affect and working memory. Consistent with the literature (Bolla et al., 2004; van den

Bos, Harteveld, & Stoop, 2013; Cella et al., 2010; Must et al., 2006), gender and self-reported depressive symptoms were both predictors of IGT performance, with female gender and at least mild levels of depression symptoms associated with lower total scores. Surprisingly, neither anxiety nor worry predicted IGT performance, a finding that is inconsistent with several previous studies (Maner et al., 2007; Mueller et al., 2010).

Working memory was not directly associated with IGT performance, but the existing literature is equivocal on this relationship. It is possible that such associations are not as strong in normal college samples with relatively unimpaired working memory performance in comparison with clinical samples. Additionally, more cognitively demanding working memory tasks, such as the n-back or running span task, could potentially produce more robust findings in this type of cognitively unimpaired population. Direct associations between IGT performance and working memory have been reported in populations expected to have working memory impairment, such as individuals with substance abuse disorders (Bechara & Martin, 2004; Duarte, Woods, Rooney,

Atkinson, & Grant, 2012) and frontal lobe brain lesions (Bechara, Tranel, Damasio, & Damasio,

1996). Other findings have been mixed across different populations and different measures of working memory (e.g., Delayed nonmatch to sample, spatial span, and Paced Auditory Serial-

Addition Task). For example, Brevers and colleagues found that IGT performance was associated with performance on the Ospan working memory task, but not with the WAIS-IV

Digit Spain task in problematic gamblers (Brevers et al., 2012). 49

A more in depth analysis of potential associations between working memory and performance on specific blocks of the IGT could be the key to revealing a significant main effect of working memory. Previous investigations conducted by our team (Golub, Thompson, &

Kowalczyk, 2016) and others (Brand et al., 2007b) have distinguished between early (i.e., first three blocks) and late (i.e., last two block) stages of decision making on the IGT, with early stages representing decision making under conditions of ambiguity and later stages representing decision making under conditions of risk. One reason for this approach is to capture performance differences related to the shifting cognitive demands of the task, as some research has shown that other aspects of executive functioning (i.e., working memory and cognitive flexibility) are engaged more during the later trials of the task (Brand et al., 2007b; Brevers et al., 2012).

Reduced working memory capacity has been well documented in individuals with depression (Joormann & Gotlib, 2008; Rose & Ebmeier, 2006; Harvey et al., 2004); however, to our knowledge, working memory has not previously been examined in studies of IGT performance in individuals with depression. We found that after accounting for gender, working memory moderated the relationship between depression and IGT performance, such that depression was associated with worse IGT performance only among individuals with below average range working memory performance based on WAIS-IV normative data. This finding suggests that limited working memory may be an important mechanism involved in worse decision-making performance among individuals with depression. Consistent with this interpretation, research has suggested that depression and rumination are associated with prolonged activation of negative information in working memory as a result of failure to inhibit irrelevant (or previously relevant) negative stimuli (Joormann & Gotlib, 2008). A failure to efficiently update working memory with information about rewards and losses during the IGT 50

may lead to skewed, inflexible representations of the decks, resulting in delayed learning and worse performance.

Tonic HRV. Contrary to our hypothesis and the literature (Kemp et al., 2010), depression was not directly associated with reduced tonic HRV in our sample; however, when tonic HRV was added to our IGT predictive model containing gender, depression, and working memory, a three-way interaction emerged. Lower tonic HRV predicted worse IGT performance in depressed participants with below average working memory performance, but not in those with at least average working memory performance. Among participants reporting minimal depression symptomology, there was no relationship between working memory, tonic HRV, and IGT performance.

Although we hypothesized that tonic HRV would be associated with IGT performance, we did not find enough information in the literature to make a prediction regarding the direction of this relationship or its potential associations with depression or working memory. Our finding regarding tonic HRV should thus be interpreted as exploratory. Nonetheless, the directionality of the HRV interaction in our IGT model is consistent with our theoretical conceptualization of reduced autonomic flexibility and working memory as potential features of depression that affect self-regulation and emotional learning. Moreover, our finding suggests that examining the convergence of these features in depression can explain more variance in IGT performance than self-reported depression alone.

Sympathetic reactivity (GSR). The association between IGT performance and sympathetic reactivity, operationalized as SCR amplitude in response to rewards and loses across all three decision-making tasks, was also examined to address our hypothesis that physiological reactivity would positively predict IGT performance. Our findings indicated that, after 51

controlling for depression and gender, greater sympathetic reactivity in response to losses, but not gains, was associated with better IGT performance; however, this main effect association was only a trend (i.e., p = .06). Sympathetic reactivity to losses interacted with gender, such that greater reactivity was a stronger predictor of performance among women. After the interaction with gender was included in the IGT predictive model, a significant main effect of sympathetic reactivity to losses emerged.

This finding adds support to the existing evidence for the role of GSR in affective decision-making, and extends previous findings that anticipatory SCR amplitude predicts IGT performance in a number of populations, including healthy women (Carter & Pasqualini, 2004) and in patients with VMPFC and amygdala lesions (Bechara et al., 1996; Bechara et al., 1999).

Consistent with findings in a college sample reported by Suzuki and colleagues, our finding suggests that response SCRs, in addition to anticipatory SCRs, may predict outcomes in the IGT

(Suzuki, Hirota, Takasawa, & Shigemasu, 2003). Though we did not predict that an association between sympathetic reactivity and IGT performance would be specific to losses, larger SCR responses for losses have been previously documented, and are consistent with the evidence that weighing losses more than gains results in better IGT performance (Bechara et al., 1996; Suzuki et al., 2003; Weller, Levin, & Bechara, 2010).

Contrary to our hypothesis, sympathetic reactivity was not positively associated with anxiety or negatively associated with depression, and sympathetic reactivity to losses did not interact with anxiety or depression to predict IGT performance. Future, more in-depth analysis of different GSR parameters, including an examination of anticipatory SCRs, may reveal more associations in our data. The few previous investigations in this area have been limited to studies of anxiety, and findings have produced mixed results; while one study found that anxiety was 52

associated with larger SCRs (i.e., larger anticipatory amplitudes) during the IGT and predicted better performance (Werner et al., 2009), another found that anxiety was associated with larger

SCRs, but predicted worse performance (Miu et al., 2008). The lack of association between self- reported anxiety and IGT performance in our sample may have limited our ability to observe potential relationships between sympathetic reactivity and anxiety in the prediction of IGT performance.

Interoceptive sensitivity. Our findings revealed an association between anxiety and interoceptive sensitivity in the prediction of IGT performance. Interoceptive sensitivity, defined by HBPT accuracy, did not directly predict IGT performance, but was associated with worse IGT performance among individuals with anxiety, and better performance in individuals without anxiety. This finding is consistent with Wölk and colleague’s study showing that greater interoceptive sensitivity predicts worse IGT performance in individuals with panic disorder

(Wölk et al., 2014). A negative predictive relationship between interoceptive sensitivity and affective decision-making performance in anxiety may be explained by evidence that individuals with anxiety have greater interoceptive sensitivity, but worse interoceptive accuracy (Ehlers &

Breuer, 1995), two distinct and dissociable dimensions of interoception (Garfinkel, Seth, Barrett,

Suzuki, & Critchley, 2015). Interoceptive cues may be more detrimental than beneficial for performance in individuals with anxiety, if they tend to misinterpret bodily sensations.

Consistent with this interpretation, evidence suggests that interoceptive sensitivity is associated with intuitive decision-making ability and has the potential to improve or hinder performance, depending on whether interoceptive cues arise in anticipation of advantageous or disadvantageous choices (Dunn et al., 2010).

Although depression did not interact with interoceptive sensitivity to predict IGT performance in our sample, interoceptive sensitivity has been shown to be reduced in individuals with clinical depression, and preliminary research has shown that reduced interoceptive sensitivity is associated with self-reported decision-making difficulty in this population (Avery et al., 2013; Furman et al., 2013).

To investigate our hypothesis that interoceptive sensitivity would moderate the relationship between sympathetic reactivity and affective-decision making performance, we added sympathetic reactivity (i.e., mean SCR amplitude response to loss) back into the IGT predictive model. This analysis revealed an interaction showing that a combination of higher sympathetic reactivity and lower interoceptive sensitivity predicted better performance among participants with anxiety. In non-anxious participants, greater sympathetic reactivity appeared to be associated with better IGT performance regardless of interoceptive sensitivity. This finding may explain some of the variability in the literature regarding the association between anxiety and IGT performance, and lends further support to the idea that, while sympathetic reactivity generally aids IGT performance, this relationship may be modulated by the strength and accuracy of interoceptive awareness in certain individuals. Future research in a clinically diagnosed sample would help to clarify the extent to which poor affective decision-making ability may be associated with a combination of heightened physiological reactivity and interoceptive sensitivity, and whether this profile is more common in certain disorders (e.g., panic disorder).

GDT

Demographic factors. Gender did not predict GDT performance, but race did, such that participants of Asian and Middle Eastern descent performed better on the task. The reason for 54

this effect and its specificity to the GDT is unclear. We did not ask individuals to report on their approaches or reactions to the individual tasks, which may have provided some insight. Our sample is unique its degree of racial and cultural diversity, with over 30% of participants reporting that they were born outside of the U.S. It is therefore important to consider the possible impact of non-western cultural beliefs and attitudes about gambling on task performance. Asian cultures may have more positive attitudes toward gambling (Raylu & Oei, 2004), and particularly in Chinese culture, gambling has historically been integrated into tradition and lifestyle as an acceptable form of entertainment (Loo, Raylu, & Oei, 2008). In contrast, in some

Middle Eastern cultures gambling has been condemned historically, and there is general disapproval toward gambling. Although we cannot determine whether specific cultural factors contributed to the observed race effect on GDT performance in this study, one possible explanation could be that the explicit gain and loss contingencies of the GDT facilitated closer associations between the participant’s gambling attitudes and expectations and actual outcomes in the game. Future studies utilizing the GDT and other decision-making tasks in diverse samples may benefit from eliciting feedback about the tasks and incorporating questions about personal beliefs toward gambling to better understand the role of such factors.

Affect and working memory. Overall, the GDT differed substantially from the IGT in its associations with affective variables. Depression did not predict performance, while trends were observed for worry (i.e., p = .07) and anxiety (i.e., p = .08) as positive predictors of performance. This difference, if replicated, could have important implications for future investigations that strive to utilize a decision-making task that is more or less sensitive to depression. The key difference between the IGT and the GDT lies in the transparency of their gain and loss contingencies; therefore, the uncertainty in decision-making inherent to the IGT 55

may be why it poses a particular challenge to individuals with depression. If poor IGT performance in individuals with depression is a manifestation of impaired emotional learning, one might intuit that the GDT is less sensitive to depression because the gain-loss contingencies are apparent.

We found that working memory did not directly predict performance, but interacted with anxiety, such that anxiety was associated with better performance in individuals with average or lower working memory scores. This finding was unexpected and should be interpreted with caution, as the distinction between above average working memory performance versus average or below working memory performance is less clear in its meaning and clinical relevance (in contrast to the distinction between below average and at least average performance). The small size of the above average working memory group (n = 23), further suggests that caution should be taken in interpreting this finding. Additionally, Brand and colleagues have shown that working memory performance is not associated with GDT performance (Brand et al., 2005), which further suggests that our finding could be misleading, though potential interactions with anxiety have not previously been examined.

Physiological variables. Contrary to prediction, measures of tonic HRV and sympathetic reactivity did not predict significant variance in GDT performance. Worry, but not anxiety or depression, was found to interact with tonic HRV, such that worry was associated with better

GDT performance in individuals with greater HRV, and worse performance in individuals with lower HRV. It is unclear why this effect was seen for worry specifically, and not for anxiety.

Nonetheless, given that the GDT rewards a risk-averse approach to decision-making, it could be interpreted that worry may be beneficial for performance in individuals with stronger self- regulatory systems (i.e., autonomic flexibility). 56

The absence of further associations between physiological variables and GDT performance is a useful contribution to the literature, as the GDT has previously been used in relatively few studies examining the role of depression, anxiety, or physiological factors in decision-making. One possible explanation for the lack of associations found is that the explicit win and loss contingencies in the task may allow for a more calculated, rational approach that, in comparison with the IGT, leaves less of a role for physiological feedback and emotional factors.

Yet Brand and colleagues have argued that GDT performance involves a combination of emotional (i.e., feedback from previous trials) and cognitive (i.e., strategic) factors, and the importance of emotional feedback in the GDT has been demonstrated in a study of individuals with amygdala lesions (Brand, Grabenhorst, Starcke, Vandekerckhove, & Markowitsch, 2007).

These individuals were shown to perform poorly on both the IGT and the GDT, and they generated reduced anticipatory and response SCRs, which was attributed to the role of the amygdala in attaching affective attributes to stimuli. In conclusion, though not observed in our study, there is some preliminary evidence that physiological factors play a role in GDT performance, and future studies should further investigate these relationships.

BART

Affect and working memory. Demographic (i.e., age, gender, race) and affective variables did not account for significant variance in BART performance. We had predicted that anxiety would be associated with worse BART performance based on prior research showing that anxiety and negative state affect predict increased risk aversion on the BART (i.e., worse performance) in undergraduate samples (Maner et al., 2007; Heilman et al, 2010). However, other studies have failed to show an association between BART performance and measures of anxiety (BAI) and fear of anxiety symptoms (Anxiety Sensitivity Index) in undergraduates 57

(Lejuez et al, 2002; Hunt et al., 2005; Buelow & Barnhart, 2015). In contrast to findings for the

IGT, depression was not associated with BART performance, which was also consistent with at least one prior study (Lejuez et al., 2002).

Below average working memory performance was associated with worse BART performance, suggesting that working memory could play a role in optimizing the amount of risk taking needed to maximize gains in the task. This is a unique finding not previously reported in the literature. To our knowledge, only a few prior studies have examined the relationship between working memory and BART performance in clinical samples. One found no association between working memory and BART performance in pathological gamblers with and without substance use disorders (Ledgerwood, Alessi, Phoenix, and Petry, 2009). The other found that individuals in treatment for stimulant use who underwent working memory training did not show any pre- post-treatment differences in BART performance (Bickel, Yi, Landes, Hill, & Baxter,

2011). Considering that we found an effect of working memory on BART performance in a normal college sample, it is possible that in populations that engage in high rates of risk taking behavior, other factors (e.g., impulsivity) may be stronger determinants of BART performance than working memory.

Further analysis revealed an interaction between anxiety and working memory in the prediction of BART performance, such that at least average working memory was associated with worse BART performance only among individuals endorsing anxiety. Worry also appeared to interact with working memory to predict performance; however, this association was such that only above average working memory was associated with worse performance in individuals endorsing worry. As mentioned previously, the small of the group with above average LDSB scores, suggests that such findings should be interpreted with caution. 58

The directionality of the observed interactions between working memory and anxiety and worry are interesting in light of our prediction that individuals with anxiety and/or worry would perform worse on the BART due to greater risk aversion in decision-making. As previously discussed in the context of the IGT, the prolonged maintenance of negative information in working memory due to failed inhibition of negative or threatening stimuli, may be a key characteristic of cognitive dysfunction in mood and anxiety disorders (Joormann & Gotlib,

2008). Anxious individuals with greater working memory capacity may engage in more extensive perseverative thought processing, thereby amplifying the salience of negative experience (e.g., a balloon pop / loss of money) throughout the task, and resulting in a risk averse approach that leads to suboptimal performance. In sum, though our findings are preliminary, they suggest that the BART engages working memory, and that the effect working memory on performance may vary with other factors such as anxiety. Additional research is needed to clarify the potential impact of depression, which did not show associations with BART performance in our sample regardless of whether working memory was taken into account.

Physiological variables. Contrary to prediction, measures of tonic HRV and sympathetic reactivity did not predict significant variance in BART performance, and they did not interact with any of the affective variables. We had hypothesized that sympathetic reactivity would predict worse BART performance, based on the assumption that greater reactivity to losses would result in less risk taking; however, no such relationships were found even when potential interactions with anxiety and working memory were taken into account. In contrast to the IGT, there has been little investigation of potential physiological factors in BART performance. To our awareness, only two studies have been done using fairly rudimentary physiological measures

(i.e., baseline adjusted mean SCRs and heart rate during task performance), and both failed to 59

show significant associations between these variables and BART performance (Lejuez et al,

2002; Hunt et al., 2005). This is somewhat surprising, particularly for SCRs, given that the

BART involves stimuli that are startling (i.e., a loud pop), which have generally been shown to produce large SCRs.

There were a few methodological factors that could have potentially introduced noise in our BART data and limited our ability to find significant associations. The first was the degree of variability in the amount of time that participants spent completing the task – completion time ranged from less than a minute to 15 minutes. Though we excluded participants who spent less than one minute on the task, this finding nonetheless raised the question, to what extent might risk aversion and rushed performance due to low effort be conflated in our dataset? Information about average BART completion times could not be found in the literature, and there was no post-hoc way for us to address this question in our sample. Another, perhaps related, methodological factor was that we did not provide real monetary incentivizes for task performance. Though numerous studies have been published using the BART with and without financial incentives, a few studies have shown that participants approach the task differently when real money is at stake (Xu, Fang, & Rao, 2013; Xu et al., 2016).

Task Comparisons

In order to further clarify the relationships between our three decision-making tasks, we examined BART and GDT performance as predictors of IGT performance, and BART performance as a predictor of GDT performance. After controlling for gender and depression,

GDT performance positively predicted IGT performance, while BART performance negatively predicted IGT performance. BART performance did not predict GDT performance. These findings are consistent with the literature showing that the IGT and GDT reward a risk averse 60

approach to decision making, while the BART rewards risk taking, though it is unclear why we did not observe a relationship between BART and GDT performance. In sum, our findings indicate that individuals who do well on the IGT and GDT may be more likely to have low scores on the BART, and suggest that the BART provides a substantially different context in which to examine decision-making behavior.

The relationship we observed between the IGT and GDT is consistent with prior research showing a positive association between the two tasks (Brand, Grabenhorst, Starcke,

Vandekerckhove, & Markowitsch, (2007a); Bayard, Raffard, & Gely-Nargeot, 2011). The GDT appears to be most closely correlated with the last two blocks of the IGT, and is more closely associated with other aspects of executive functioning (e.g., cognitive flexibility) relative to the

IGT (Brand et al., 2007b). Studies comparing the IGT and GDT have also found the GDT to be more sensitive to personality traits, including perfectionism, behavioral avoidance and inhibition, and impulsivity (Brand & Altstotter-Gleich, 2008; Bayard et al., 2011).

The negative relationship we observed between BART and IGT performance is consistent with prior studies that have used these tasks together and found them to differ significantly in their associations with other study variables. Interestingly, though the IGT has been shown to predict a number of real-world risk behaviors, in a comparison study, Lejuez and colleagues found that the BART, but not the IGT, distinguished smoking and non-smoking students (Lejuez et al., 2003). Buelow and Barnhart (2015) found that anxiety, social concern/stress, and test anxiety predicted poor IGT performance, but not BART performance in a college sample. Balaguero and colleagues found no correlation between BART and IGT performance in participants with acquired brain injuries (Balaguero et al., 2014). In a recent factor analysis, Buelow and Blaine (2015) found that the IGT and BART held as separate factors 61

in analyses with the IGT, BART and Columbia Card Task (cold and hot conditions). These studies suggest that the BART and IGT measure different types of decision-making.

Implications

Our findings, as well as those mentioned in the previous section, serve to highlight the complexity of affective decision-making processes, and the degree of variability in research outcomes that can come from choosing to use one task over the other. Most importantly, they suggest that future investigations of affective decision-making can benefit from employing more than one of these tasks, toward drawing stronger, more generalizable conclusions.

The three tasks used in this study were found to differ substantially in their associations with participant demographics, anxiety, depression, and working memory, suggesting that these may be particularly important factors to consider in cross study comparisons. Anxiety did not directly predict performance on any of the tasks, but was ultimately found to factor in through interactions with other variables (i.e., working memory, sympathetic reactivity) for each task.

The IGT was the only task found to be sensitive to depression, perhaps due to its unique condition of uncertainty regarding the contingencies of each deck during early trials. Though further analysis is needed to determine if depression affected performance on specific stages of the IGT in our study, previous research suggest that depressed individuals have greater difficulty adjusting their appraisals of the decks and are slower to learn the IGT (Cella et al., 2010; Must et al., 2006). There has been relatively less investigation of BART and GDT performance in individuals with depression. Given the sensitivity of the GDT to executive dysfunction, one might predict that depression would affect performance. Future research in a sample of individuals with clinically diagnosed depression would be beneficial in clarifying potential associations between GDT and BART performance and depression. 62

All three tasks appeared to be sensitive to working memory, albeit not directly. In individuals with depression, below average working memory was detrimental to IGT performance, an effect that was compounded by low autonomic flexibility (i.e., tonic HRV). In individuals with anxiety, average or higher working memory was detrimental to BART performance. Though we also saw some detrimental effect of working memory on GDT performance, we were not confident in this finding, as the effect was driven by only a small group of individuals with above average working memory performance. Nonetheless, we can conclude that working memory was closely tied to affective factors in our sample, and appeared to be a potential mechanism through which anxiety and depression enacted their predicted effects on task performance. This is consistent with the literature showing that working memory is sensitive to anxiety and depression (Harvey et al., 2004; Joormann & Gotlib, 2008; Schmeichel et al., 2008; Shackman et al., 2006) and is an important cognitive substrate in decision-making

(Bechara & Martin, 2004; Hinson et al., 2003).

Our efforts toward characterizing physiological self-regulatory profiles in anxiety and depression and determining their association with affective decision-making produced limited results. Physiological variables did not distinguish individuals with pure anxiety or worry from those with comorbid anxiety and depression or no emotional distress as anticipated. We had aimed to extend the literature on physiological characteristics of emotional learning in decision- making based on studies using the IGT, but were ultimately unsuccessful in finding associations between sympathetic reactivity and performance on the GDT or BART. Our study was somewhat limited in the extent of physiological data analysis that could be completed; it is possible that a more in-depth analysis of different GSR parameters in our data could reveal more significant findings. This may be particularly true given that the GSR parameters of interest in 63

affective decision-making research have largely been defined by the IGT and the SMH; there could very well be different GSR features and data time points that are relevant to outcomes on the BART and the GDT. Additionally, challenges with artifacts introduced noise into our data and limited the size of the sample that could be used for the physiological section of the analysis.

Collecting a larger sample of data could improve the power of our statistical analyses and yield more interesting results.

Nonetheless, there was a novel finding related to sympathetic reactivity, interoceptive sensitivity, and IGT performance in our data. Greater sympathetic reactivity (i.e., SCR amplitude) to losses was associated with better performance on the IGT, particularly among women, and this finding was consistent with the SMH and literature showing a relationship between SCR amplitude and task performance in general and clinical populations (Bechara et al.,

1996; Suzuki et al., 2003; Carter & Pasqualini, 2004). Furthermore, a combination of higher sympathetic reactivity and lower interoceptive sensitivity was found to predict better performance among individuals with anxiety, while in non-anxious participants, interoceptive sensitivity did not appear to impact the positive association between sympathetic reactivity and performance. This finding supported our hypothesis that interoceptive sensitivity moderates the relationship between sympathetic reactivity and affective decision-making in some individuals, and suggests that it may be a productive target for interventions that seek to improve decision- making in anxiety.

Interoceptive exposure (IE) has been used as a method for reducing anxiety sensitivity, or fear of arousal-related sensations, which has been shown to be elevated in several types of anxiety disorders. Studies have shown that CBT treatments incorporating IE, which may involve exercises such as brief hyperventilation, spinning, or breathing through a narrow straw, are 64

effective in treating panic disorder (Barlow, Gorman, Shear, & Woods, 2000; Craske & Barlow,

2008; Gloster et al., 2011), and may also be an effective tool in the treatment of PTSD (Wald &

Taylor, 2005, 2007). Despite these promising findings, more research is needed to examine the specific efficacy of IE and potential mechanisms through which is may impact anxiety-related cognitions (Deacon et al., 2013).

There were also unique findings in our data related to tonic HRV and task performance.

Low HRV predicted worse performance on the GDT in participants who endorsed worry, and worse performance on the IGT in participants who endorsed depression and had below average working memory. The finding that a relationship between reduced tonic HRV and poor affective decision-making arises specifically in the context of depression or worry is consistent with the

PCH. Rumination and worry are pathological patterns of cognitive perseveration in depression and anxiety that have been related to physiological disruptions in self-regulation (e.g., reduced autonomic flexibility) in several studies (Thayer et al., 2006; Aldao et al., 2013; Nahshoni et al.,

2004; Agelink et al., 2001, 2002).

Future, more in depth, research in this vein has the potential to reveal the nuances of these moderating mechanisms, toward a better understanding of how they can be targeted in interventions. Recent studies have suggested that HRV biofeedback treatment may have the potential to reduce clinical depression (Karavidas et al., 2007; Siepmann, Aykac, Unterdorfer,

Petrowski, & Mueck-Weymann, 2008), improve executive function and emotion regulation in individuals with severe brain injury (Kim et al., 2012), and reduce drug and alcohol cravings in individuals with low HRV in inpatient substance use treatment (Eddie, Kim, Lehrer, Deneke, and

Bates, 2014). Furthermore, significantly increased HRV has been observed in individuals who

responded successfully to CBT (Carney et al., 2000), as well as the antidepressant venlafaxine

(Davidson et al., 2005).

Limitations

The aims of the present study were fairly broad and largely exploratory in nature, toward our goal of characterizing differences between the IGT, GDT, and BART and their associations with common variables of interest in the affective decision-making literature. As such, there are some key limitations and areas where a more refined and in depth analysis is warranted in order to draw more thorough conclusions. While many smaller areas of limitation have been address already in our discussion, the following points are considered to be the broader, study-wide limitations of this project.

First, our sample size was depleted by missing or invalid data for some tasks, as well as noise or other problems with physiological data recording. In the interest of maximizing statistical power, instead of excluding all cases that could not be used in all aspects of analysis, the analytic sample size was adjusted differently between each major phases of analysis (i.e., different cases were excluded depending on which task was the dependent variable, and additional cases were excluded for physiological data analysis). While these adjustments were small, typically involving 2-3 cases, it could be argued that this approach potentially weakens the validity of our findings and the integrity of our conclusions to a degree. Ultimately, more uniform exclusions and/or additional data collection will be needed prior to publishing some findings.

Second, our sample was particularly diverse in terms of racial and ethnic background, with nearly 40% of individuals identifying as Asian, and approximately 30% of participants reporting countries other than the U.S. as their birthplace. This is not so much a limitation, but 66

rather an important consideration for the generalizability of our findings, as the majority of college samples used in research in the U.S. are less diverse. Moreover, our finding that GDT performance was sensitive to race highlighted the importance of measuring factors that might mediate such effects, including attitudes toward risk taking and gambling, strategies employed during task performance, and reactions to the tasks. Unfortunately, we did not include any measures of this nature, making it difficulty to further characterize our sample and evaluate potential variations in task performance that could be influenced by such factors.

Third, though our study assessed participant’s anxiety and depression symptoms during the week prior to the experiment, it was limited by not assessing the state mood of participants.

The inclusion of a measure, such as the positive and negative affect schedule (PANAS), at the beginning of the study would have allowed us to control for potential variability in performance due to individual differences in baseline mood at the time of the assessment. This oversight is significant because the IGT and BART have been shown to be sensitive to stress and state mood

(Vries, Holland, & Witeman, 2008; Lighthall et al., 2009; Buelow & Suhr, 2013). It is possible that participant’s moods at the time of the experiment differed significantly from their ratings of depression and anxiety, and this could have influenced some aspects of performance. We would expect the potential impact of random variability in state mood to have less of an impact on our physiological data, as the dysfunction of self-regulatory systems that results in profiles of autonomic rigidity, and/or heightened interoceptive sensitivity and sympathetic reactivity, has generally been described as a phenomenon that occurs through chronic allostatic disruption (i.e.,

MDD, chronic stress).

Fourth, specific background questions pertaining to potential factors that could impact

HRV (e.g., heart conditions and medications, nicotine and caffeine use, or fitness level) we not 67

included, which prevented us from controlling for these potential sources of variability in our data. Previous research suggests that these are all factors important to consider in the measurement of HRV (for a recent review see Quintana & Heather, 2014), and they may have added unaccounted noise to our data that reduced the robustness of our findings.

Fifth, given the large number of regression analyses that were conducted in order to examine our three different decision-making tasks, there is an increased risk of familywise error that should be considered. A number of our analyses (e.g., examining working memory) were exploratory in nature, and findings unrelated to our main hypotheses should be replicated to reduce the possibility of familywise error prior to publication.

A final limitation of our study was that the order of the tasks was not randomized. There were several reasons for our decision to use a fixed order for the experiment’s components. First, we wanted to give the IGT priority in order, given that the basis of this study and the majority of our hypotheses were based on prior research with the IGT. The IGT has been shown to be sensitive to priming and learning, potentially due to its implicit learning component, whereas the

BART has not (Overman & Pierce, 2013; Hart, Schwabach, & Solomon, 2010; Schrag, Tremea,

Lagger, Ohana, & Mohr, 2016). A second factor was our desire to have the participants complete the tasks in one block of time that was undisrupted by the examiner. The IGT has historically been explained to participants through scripted instruction from the examiner to ensure participant comprehension, and this standardized administration has been shown to be important for performance (Balodis, MacDonald & Olmstead, 2006; Fernie & Tunney, 2006). The BART and GDT are relatively less complex tasks and have instructions provided on the computer screen. Ideally, the instructions for each task could have been read by the examiner, but this was not feasible due the time demands involved; therefore, we placed the IGT at the beginning. 68

Conclusions

Improving our understanding of the cognitive and physiological disruptions that occur in anxiety and mood disorders has the potential to reveal novel ways to target and improve treatments. Better treatment options are of particular importance given the prevalence of these disorders, their high rates of relapse and treatment resistance, and the frequency with which they co-occur with other psychiatric problems. Clarifying decision-making processes in depression and anxiety may specifically aid in addressing problematic behavioral patterns that occur in these disorders, including avoidance and substance use.

The present study drew upon several theoretical frameworks, including the SMH and the

PCH, to explore the role of working memory, autonomic physiology, and interoceptive awareness in affective decision-making performance in individuals with self-reported anxiety and depression. The literature on affective decision-making in anxiety and depression has largely depended on the use of the IGT, and therefore, another major aim or our study was to compare the IGT with two other affective decision-making measures, the BART and the GDT, to characterize their sensitivity to anxiety and depression, and to determine the extent to which findings can be generalized across these tasks.

Our findings are preliminary, and their utility lies primarily in the insight they provide into areas worthy of more targeted investigations with refined hypotheses experimental techniques. In particular, there were two novel findings that should be replicated and further characterized in future research. First, we showed that tonic HRV serves as a potential moderator of the negative associations between affective decision making performance and depression (in the IGT) or worry (in the GDT). Second, we showed that interoceptive sensitivity serves as a potential moderator of the association between physiological reactivity and IGT performance in 69

anxiety. These are substantial findings supported by previous literature that have the potential to inform recent treatment advances involving HRV biofeedback and interoceptive exposure methods. Finally, perhaps the most significant conclusion to be drawn from our findings is that the IGT, BART, and GDT differ significantly in their associations with each other, and with other common variables of interest (i.e., anxiety, depression, working memory, and physiological markers), suggesting that conclusions about affective decision making should not be based on the use of any one of these tasks in isolation.

Appendix A

Table 1

Sample demographic and affective characteristics Full sample Analytic Variable name = 100 sample = 93 Analytic %

Gender Female 64 58 62.4

Age ≤ 20 77 71 76.3

Race: Black 9 9 9.7 Latino 20 17 18.3 White 22 21 22.6 Asian 38 36 38.7 Other 11 10 10.8

Origin Born in U.S. 68 63 67.7

Income (family) < $40,000 annual 41 (5 missing) 37 (5 missing) 42

BDI Depression Minimal 62 58 62.4 Mild 13 13 14 Moderate 20 17 18.3 Severe 5 5 5.4

BAI Anxiety Minimal 54 49 52.7 Mild 28 27 29 Moderate 14 13 14 Severe 4 4 4.3

PSW Worry Screening cut off 73 68 73.1 Clinical cut off 27 25 26.9

Digit Span Backwards Below average 51 46 49.5 Average 24 23 24.7 Above average 25 24 25.8

Note. BDI = Beck Depression Inventory. BAI = Beck Anxiety Inventory, PSW = Penn State Worry Questionnaire. The analytic sample size reflects participants excluded due to missing data or invalid performances.

Table 2

Decision-making task variables, N = 93 Variable name Mean Median SD Min - Max IQR IGT Scores Total Earned -620.5 -720.0 1524.6 -5970 - 3070 -1667.5, 225.5 Good Deck – Bad Deck Selections 8.5 4.0 31.3 -92 - 98 -11, 33

BART Scores (n = 90*) Total Earned 27.2 27.3 12.2 3.0 – 61.2 18.1, 35.4 Total Pops 6.7 6.0 3.8 0 -17 3, 9

GDT Scores Total Earned -887.1 0 2466.8 -8800 - 2700 -2650, 1000

Digit Span Longest Digit Span Forward 6.9 7.0 1.2 4 - 9 6, 8 Longest Digit Span Backward 4.8 5.0 1.2 3 - 8 4, 6 Reliable Digit Span 10.1 10 1.9 7 -17 9, 11

BAI Anxiety Total 10.1 8 8.9 0 - 49 3, 15 BAI Anxiety Transformed 4.1 4.0 2.6 0-12.1 2, 6

PS Worry Total 51.6 53 13 19, 74 42.5, 62

BDI Depression Total 12.5 10 8.8 0, 37 6, 19 Note. IGT = Iowa Gambling Task, BART = Balloon Analogue Risk Task, GDT = Game of Dice Task. BAI Anxiety Transformed = Box-Cox transformation lambda x26. *Excluding 3 participants missing BART data due to technical problems.

Table 3 Spearman correlations between decision-making tasks, affect, and digit span variables, N = 93

1 2 3 4 5 6 7 8 9 10 11

1. BAI Anxiety ǂ --

2. BDI Depression 0.57** 3. PSW Worry 0.48** 0.62** 4. IGT Total -0.16 -0.19 -0.15 5. IGT Good - Bad Selections -0.05 -0.09 -0.08 0.86** 6. BART Total -0.05 -0.01 -0.06 -0.08 -0.11 7. BART Pops -0.04 -0.03 -0.11 -0.02 -0.01 0.58** 8. GDT Total 0.21* -0.06 0.09 0.22* 0.27** 0.03 0.05 9. GDT Total 1 outcome bets -0.12 -0.02 -0.04 -0.19 -0.23* -0.08 -0.02 -0.77** 10. LDSF -0.12 -0.18 -0.28* 0.22* 0.12 -0.03 0.13 0.02 0.07 11. LDSB -0.10 -0.06 0.04 0.03 -0.07 0.18 0.09 -0.08 0.02 0.24* 12. RDS -0.20 -0.23* -0.23* 0.14 0.08 -0.00 0.12 -0.06 0.04 0.75** 0.54** Note. * p ≤ .05. ** p ≤ .01. ǂ The correlation between the BAI transformed and non-transformed variables was 1.00, and both variables shared the same significant correlations with other variables of interest. LDSF = longest digit span forward. LDSB = longest digit span backward. RDS = reliable digit span. BDI = Beck Depression Inventory, BAI = Beck Anxiety Inventory, PSW = Penn State Worry Questionnaire.

Table 4

Physiological and interoceptive sensitivity variable descriptives

Variable name Mean Median SD Skewness Kurtosis Min - Max IQR

Tonic HRV 14.6 7.7 15.4 1.1 0.0 0.0 – 55.0 2.4, 25.7 (n = 86) PNN50 (raw)* RMSSD (raw) 34.1 30.1 16.1 .71 -.13 8.9 – 78.9 22.0, 43.5 RMSSD Log trans. 3.4 3.4 .49 -.26 -.42 2.2 -4.4 3.1, 3.8

Mean SCR amplitude Losses (raw) 76.9 60.7 60.8 1.4 2.5 4.0 – 321.3 30.3, 107.9 (n = 83) Gains (raw) 74.5 59.3 63.3 1.8 5.3 .33 – 375.0 1.2, 3.7 Log trans. losses 10.7 10.4 4.4 .22 -.29 1.9 – 22.7 7.3, 13.7 Log trans. gains 10.4 10.3 4.6 .24 .18 -.89 – 24.3 6.8, 13.7

Mean NS-SCR amplitude At baseline 537 489 366.1 .82 .51 0 – 1742 263, 760 (n = 83) During tasks 916.5 893 450.7 .73 .28 146 – 2224 568, 1162

Mean NS- SCRs/min At baseline 3.4 3.0 2.3 .96 1.3 0.0 – 12.0 1.7, 4.7 (n = 83) Log trans. baseline 1.3 1.4 .57 -.43 -.21 .00 – 2.56 .98, 1.7 During tasks 2.6 2.4 1.7 .26 .33 7.2 1.2, 3.7

HBPT percent accuracy 65.0 65.7 20.1 -.22 -.67 17.7 – 99.3 51.7, 80.7 ( n = 91)

Note. Tonic HRV = Resting heart rate variability during the 3-minute baseline. Mean SCR amplitude is calculated separately for loss and gain events across tasks. NS-SCR = Non-event specific skin conductance responses. HBPT = Heart Beat Perception Task. PNN50 (parasympathetic fraction of consecutive normal sinus intervals exceeding 50 milliseconds) and RMSSD (root mean square differences) are two time-domain measures of HRV. *The correlation between PNN50 and RMSSD was 0.94.

Table 5

Spearman correlations between physiological, affect, and working memory variables Mean Loss Mean Gain Mean Tonic Mean SCRs/min Amplitude Across Amplitude Across SCRs/min HRV Across Tasks Tasks Tasks Baseline 1. HBPT Accuracy 0.01 0.14 0.08 0.05 -0.08

2. BAI Anxiety 0.03 0.12 0.07 0.04 -0.07

3. PSW Worry -0.10 0.05 -0.03 0.08 -0.09

4. BDI Depression -0.11 0.07 -0.04 0.20 -0.03

5. LDSB 0.23* -0.17 -0.03 -0.12 -0.07 Note. Tonic HRV = Resting heart rate variability (RMSSD) during the 3-minute baseline. Mean loss and gain amplitude were correlated r = .078, p < .05. HBPT = Heart Beat Perception Task; accuracy was not correlated with the physiological variables. BDI = Beck Depression Inventory, BAI = Beck Anxiety Inventory, PSW = Penn State Worry Questionnaire total score.

Table 6

One-way ANOVA results for variables of interest predicting physiological outcomes Physiological Variables M (SD) Tonic Mean SCR Loss Mean SCR Gain Mean NS- Mean NS- HRV Amplitude Amplitude SCRs / Min SCRs / Min Factors Across Tasks Across Tasks Baseline Across Tasks Gender Male 3.6 (0.5) 10.1(3.5) 9.6 (4.2) 1.3 (0.6) 2.9 (1.7) Female 3.3 (0.5) 11.1(5.0) 10.9 (4.8) 1.3 (0.5) 2.5 (1.6) F 3.5ǂ 1.0 1.6 0.1 1.4

BDI Minimal 3.5 (0.5) 10.5 (3.8) 10.3(3.9) 1.3(0.6) 2.8 (1.7) Depression At least mild 3.4 (0.5) 10.9(5.3) 10.5(5.7) 1.4(0.5) 2.4(1.5) F 0.95 0.22 0.05 1.6 0.92

BAI Anxiety Minimal 3.4 (0.5) 10.5 (4.0) 10.3 (4.3) 1.3 (0.6) 2.8(1.8) At least mild 3.4 (0.5) 10.8 (4.9) 10.5 (4.9) 1.4 (0.6) 2.4(1.4) F 0.12 0.06 0.05 0.41 1.22

PSW Worry Lower Worry 3.4 (0.5) 10.4 (3.9) 10.5(4.5) 1.2 (0.70) 2.8 (1.8) (median split) Higher Worry 3.4 (0.5) 10.9 (4.9) 10.2 (4.8) 1.4 (0.4) 2.45 (1.4) F 0.2 0.35 0.07 2.6 1.0

LDSB Below average 3.3 (0.5) 11.3(4.5) 10.4 (4.5) 1.4 (0.6) 2.7 (1.5) Average 3.5 (0.6) 11.3 (4.8) 10.9 (5.2) 1.4 (0.4) 2.7 (1.8) Above average 3.5 (0.4) 9.0 (3.6) 9.7 (4.3) 1.3 (0.6) 2.5 (1.8) F 2.8ǂ 2.3 0.4 1.5 0.2

Interoceptive Lower sensitivity 3.35 10.1 (4.5) 10.0 (4.4) 1.3 (0.7) 2.9 (1.7) Sensitivity (0.5) (median split) Higher sensitivity 3.5 (0.5) 10.9 (4.3) 10.4 (4.7) 1.4 (0.5) 2.4 (1.6) F 1.8 0.67 0.12 0.21 1.8 Note. ǂ = p < 0.08 BDI = Beck Depression Inventory, BAI = Beck Anxiety Inventory, PSW = Penn State Worry Questionnaire. BDI and BAI were dichotomized based on published clinical cut off scores. LDSB (longest digit span backwards) was split into performance categories based on WAIS-IV normative data.

Table 7

Independent regression models of affect and gender predicting IGT performance, N=89

Model 1 (depression) Model 2 (anxiety) Model 3 (worry) Adjusted Model

Variable B SE B β B SE B β B SE B β B SE B β

.263. 292. Gender -791.0 -.30** -767.1 267.5 -.29** -706.8 -.27* -821.0 293.1 -.311** 4 2 BDI Depression -.33.7 14.6 -.23* -30.2 19.2 -.21 BAI Anxiety -96.2 50.3 -.19 -57.2 59.5 -.12 PSW Worry -12.7 11.4 -.12 6.5 14.2 .10

R2 .15 .14 .11 .16

F 7.68** 6.77** 5.42** 4.05** Note. * p ≤ .05. ** p ≤ .01. BDI = Beck Depression Inventory, BAI = Beck Anxiety Inventory, PSW = Penn State Worry Questionnaire.

Table 8

Independent regression for gender, depression, and LDSB predicting IGT performance, N=89

Model 1 Model 2 Adjusted model LDSB interaction model

Variable B SE B β B SE B β B SE B β B SE B β

279. 270. 257. Gender -793.7 -.30 -822.7 -.31** -728.3 275.6 -.28** -759.4 -.29** 3 3 0

111. LDSF 62.8 .06 69.1 117.4 .07 6

113. LDSB -84.9 -.07 -142.7 118.4 -.13 -164.7 56.7 -1.12** 2

BDI 172. Depression -34.6 14.9 .24* -436.2 -.39* 8 BDI Depression 27.1 11.5 .92* x LDSB R2 .10 .11 .17 .22

F 4.9** 5.1** 4.2** 5.7** Note. * p ≤ .05. ** p ≤ .01. LDSF = longest digit span forward. LDSB = longest digit span backward. BDI = Beck Depression Inventory

Figure 1

Figure 2

Table 9

Hierarchical regression for gender, depression, and LDSB predicting IGT performance, N = 89

Model 1 Model 2

Variable B SE B β B SE B β

Gender -790.8 264.7 -.30** -764.5 255.5 -.29**

Below Average LDSB -97.3 257.6 -.04 1090.8 442.1 -.42*

BDI Depression -34.1 14.7 -.23* 5.2 20.3 .04 Below Average LDSB x BDI Depression -77.1 28.4 -.55**

R2 .15 .22

F 5.1** 6.0**

2 Δ R .068

F for Δ R2 7.4* Note. * p ≤ .05. ** p ≤ .01. LDSB = longest digit span backward. LDSB was dichotomized at below average range performance based on WAIS-IV age normative data.

Figure 3

Table 10

Hierarchical regression model with tonic HRV interactions predicting IGT performance, N = 83

Model 1 Model 2

Variable B SE B β B SE B β

Gender -808.2 283 -.30** -775.5 273.7 -.29**

Below Average LDSB 150.2 275.3 .06 676.5 336.0 .26*

BDI Depression dichotomized -631.5 273.1 -.24* -41.9 350.1 -.02

Tonic HRV 257.7 282.4 .10 314.1 273.7 .12 HRV x BDI Depression x Below Average LDSB -408.1 159.3 -.39*

R2 .194 .257

F 4.68** 5.33**

Δ R2 .063

F for Δ R2 6.57* Note. * p ≤ .05. ** p ≤ .01. LDSB = longest digit span backward. LDSB was dichotomized at below average range performance based on WAIS-IV age normative data. BDI depression scores were dichotomized at mild depression based on established measure cut offs. Tonic HRV = resting heart rate variability during the 3 minute baseline (BL).

Figures 4 and 5

Table 11

Independent regression models with mean SCR amplitude for gain and loss events (across all tasks) predicting IGT performance, N = 81

Model 1 (Loss Amplitude) Model 2 (Gain Amplitude) Adjusted Model

Variable B SE B β B SE B β B SE B β

Gender -947.9 263.3 -.36** -880.5 270.3 -.34** -881.0 252.5 -.34**

BDI Depression -39.3 15.3 -.26* -40.5 15.8 -.26* -47.0 14.9 -.31**

Mean Loss Amplitude 59.6 31.4 .19ǂ 162.0 46.3 .52**

Mean Gain Amplitude -10.9 31.2 -.04 -130.5 44.9 -.44** R2 .232 .198 .309

F 7.76** 6.32** 8.49** Note. ǂ = p = .06 * p ≤ .05. ** p ≤ .01. Mean loss and gain amplitude is defined as the mean SCR amplitude following each gain or loss of money across all three decision-making tasks. BDI = Beck Depression Inventory total score.

Table 12

Independent regression with SCR amplitude interactions predicting IGT performance, N = 81

Model 1 Model 2

Variable B SE B β B SE B β

Gender 197.5 579.1 .08 -916.6 264.6 -.35**

BDI Depression -41.7 15.0 -.27** 1.4 40.5 .01

Mean Loss Amplitude 117.6 40.4 .38** 117.7 62.2 .38ǂ

Gender x Mean Loss Amplitude -111.0 50.3 -.54*

Depression x Mean Loss Amplitude -4.2 3.9 -.34

R2 .278 .244

F 7.33** 6.13** Note. * p ≤ .05. ** p ≤ .01. Mean loss and gain amplitude is defined as the mean SCR amplitude following each gain or loss of money across all three decision-making tasks. BDI = Beck Depression Inventory total score. 82

Figure 6

Table 13

Independent regression models with mean frequency of non-specific SCRs per minute (during baseline and across all tasks) predicting IGT performance, N = 81

Model 1 (Tasks) Model 2 (BL)

Variable B SE B β B SE B β

Gender -811.6 274.5 -.31 -844.7 268.6 -.32**

BDI Depression -35.4 15.8 -.23 -36.7 15.7 -.24*

Mean SCRs/min. across tasks 79.0 81.9 .10

Mean SCRs/min. across baseline 23.2 231.4 .01

R2 .189 .179

F 6.0** 5.7** Note. * p ≤ .05. ** p ≤ .01. SCR = Skin conductance response. BDI = Beck Depression Inventory total score.

Table 14

Hierarchical regression model with interoceptive sensitivity (HBPT accuracy) and anxiety interacting to predict IGT performance N = 87

Model 1 Model 2

Variable B SE B β B SE B β

Gender -772.9 274.2 -.29** -700.5 265.6 -.26** BAI Anxiety -96.1 51.2 -.19ǂ 336.3 167.2 .68* HBPT Accuracy -3.3 6.7 -.05 26.6 12.8 .04* Anxiety x HBPT Accuracy -6.9 2.6 -1.03** R2 .135 .206 F 4.31** 5.3** Δ R2 .071 F for Δ R2 7.3** Note. ǂ = .06 * p ≤ .05. ** p ≤ .01. HBPT Accuracy = percent accuracy on the first block of counting trials of the heart beat perception task. BAI Anxiety = Beck Anxiety Inventory total score transformed.

Figure 7

Table 15

Independent regression models with interoceptive sensitivity (HBPT accuracy), mean SCR amplitude, and anxiety interacting to predict IGT performance N = 79

Model 1 Model 2a Model 2b

Variable B SE B β B SE B β B SE B β

Gender -952.6 280.3 -.36** -878.1 285.3 -.33** -827.7 268.4 -.31**

BAI Anxiety -100.6 52.9 -.20ǂ -113.1 53.6 -.23* -36.7 54.2 -.07

HBPT Accuracy -3.2 6.9 -.05 7.7 11.0 .12 -36.7 8.7 .22

Mean Loss Amplitude 67.3 32.3 .22* 89.5 36.6 .29* 14.5 32.5 .33** Median split HBPT x SCR Loss Amplitude -47.8 37.8 -.23 Median split HBPT x SCR Loss Amplitude x BAI Anxiety -16.9 5.5 -.46** R2 .208 .225 .300 F 4.85** 4.23** 6.25**

Δ R2 (from model 1) .017 .092

F for Δ R2 1.60 9.59** Note. * p ≤ .05. ** p ≤ .01. HBPT Accuracy = percent accuracy on the first block of heart beat perception task counting trials. HBPT accuracy was dichotomized at the median. BAI Anxiety = Beck Anxiety Inventory total score transformed.

Figure 8

Table 16

Independent regression models of with race and affective variables in the prediction of GDT performance, N = 90

Model 1 (depression) Model 2 (anxiety) Model 3 (worry) Adjusted Model

Variable B SE B β B SE B β B SE B β B SE B β

Race 1351.7 457.8 .30** 1324.6 450.9 .30** 1295.2 451.7 .29** 1293.4 454.6 .29** BDI Depression 23.6 26.4 .09 -18.8 34.2 -.07 BAI Anxiety 154.3 87.3 .18ǂ 113.5 109.6 .13

PSW Worry 32.3 17.5 .19ǂǂ 27.5 23.3 .16 R2 .105 .128 .131 .143

F 5.12** 6.41** 6.56** 3.53** Note. ǂ = p = .081. ǂǂ = p = .068. * p ≤ .05. ** p ≤ .01. Race = dichotomized as individuals identifying as Asian or Middle Eastern versus all other participants. BAI Anxiety = Beck Anxiety Inventory total score transformed. BDI = Beck Depression Inventory total score. PSW = Penn State Worry Questionnaire total score.

Table 17

Hierarchical regression model for race, worry, and their interaction in the prediction of GDT performance, N = 90

Model 1 Model 2

Variable B SE B β B SE B β

Race 1295.2 451.7 .29** 4526.8 1879.4 1.0*

PSW Worry 31.0 17.5 .19ǂǂ 57.0 22.2 .33*

Race x PSW Worry -62.7 35.4 -.77ǂ

R2 .131 .162

F 6.56** 5.53**

Δ R2 .031

F for Δ R2 3.1ǂ Note. ǂ = p = .08. ǂǂ = p = .068. * p ≤ .05. ** p ≤ .01. Race = dichotomized as individuals identifying as Asian or Middle Eastern versus all other participants. PSW = Penn State Worry Questionnaire total score.

Table 18

Independent regression models for race, LDSB, and LDSF predicting GDT performance, N = 90

Model 1 Model 2 Adjusted model

Variable B SE B β B SE B β B SE B Β

Race 1448.4 457.8 .32** 1385.6 455.1 .31** 1452.6 456.2 .32**

LDSF 175.3 184.8 .10 252.3 194.0 .14

LDSB -183.3 204.9 -.09 -271.3 215.0 -.13

R2 .106 .105 .123

F 5.18** 5.12** 4.0** Note. * p ≤ .05. ** p ≤ .01. Race = dichotomized as individuals identifying as Asian or Middle Eastern versus all other participants. LDSF = longest digit span forward. LDSB = longest digit span backward.

Table 19

Hierarchical models for anxiety and LDSB predicting GDT performance, N = 90 Model 1 Model 2

B SE B β B SE B β

Race 1312.2** 454.1 .29 1201.9** 447.8 .27**

BAI Anxiety 151 88.1 .17 223.3* 92.6 .26*

Above average LDSB -209.6 520.9 -.04 1918.3 1114.5 .37ǂ

BAI Anxiety x Above average LDSB -560.5* 261.0 -.47*

R2 .130 .175

F 4.3** 4.5**

Δ R2 .045

F for Δ R2 4.6* Note. ǂ = p = .09. * p ≤ .05. ** p ≤ .01. Race = dichotomized as individuals identifying as Asian or Middle Eastern versus all other participants. LDSB = longest digit span backward. LDSB was dichotomized at above average range performance based on WAIS-IV age normative data. BAI Anxiety = Beck Anxiety Inventory total score transformed. 87

Figure 9

Table 20

Hierarchical regression with worry and HRV interacting to predict GDT Performance, N = 90

Model 1 Model 2

B SE B β B SE B β

Race 1418.3 474.8 .31** 1377.4** 458.6 .30**

PSW Worry 25.1 18.8 .14 -350.6* 145.1 -2.0*

Tonic HRV -511.9 494.0 -.11 -6146.8** 2211.4 -1.3**

Tonic HRV x PSW 107.2** 41.1 2.4** Worry

R2 .141 .21

F 4.33** 5.2**

Δ R2 .069

F for Δ R2 6.81**

Note. * p ≤ .05. ** p ≤ .01. Race = dichotomized as individuals identifying as Asian or Middle Eastern versus all other participants. PSW = Penn State Worry Questionnaire. Tonic HRV = resting heart rate variability during the 3 minute baseline (BL).

Figure 10

Table 21

Independent regression models with average SCR amplitude for gain and loss events (across all tasks) predicting GDT performance, N= 80

Model 1 (Loss Amplitude) Model 2 (Gain Amplitude) Adjusted Model

Variable B SE B β B SE B β B SE B β

Race 1591.7 489.5 .35 1576.6 492.1 .34 1584.0 492.2 .35

Mean Loss Amplitude 57.0 58.6 .10 90.0 90.4 .16

Mean Gain Amplitude 24.1 57.0 .05 -42.1 87.6 -.08 R2 .127 .118 .129

F 5.59** 5.15** 3.77* Note. * p ≤ .05. ** p ≤ .01. Race = dichotomized as individuals identifying as Asian or Middle Eastern versus all other participants. Mean loss and gain amplitude is defined as the mean SCR amplitude following each gain or loss of money across all three decision-making tasks.

Table 22

Independent regression models of affective variables controlling for gender, predicting BART performance, N= 88

Model 1 (depression) Model 2 (anxiety) Model 3 (worry) Adjusted Model

Variable B SE B β B SE B β B SE B β B SE B β

Gender -3.2 2.6 -.13 -3.0 2.6 -.12 -3.2 2.8 -.13 -3.6 2.8 -.15 BDI Depression -.05 .14 -.04 .01 .19 .01

BAI Anxiety -.63 .52 -.13 -.85 .66 -.17

PSW Worry -.01 .10 -.01 -.08 .13 -.08 R2 .020 .035 .018 .040

F .85 1.53 .79 .86 Note. PSW = Penn State Worry Questionnaire total score. BAI Anxiety = Beck Anxiety Inventory total score transformed. BDI = Beck Depression Inventory total score.

Table 23

Independent regression models for LDSF and LDSB predicting BART performance, N = 88

Model 1 Model 2 Adjusted model

Variable B SE B β B SE B β B SE B β

Gender -3.4 2.7 -.14 -3.3 2.6 -.14 -3.7 2.7 -.15

LDSF -.15 1.1 -.02 -.72 1.1 -.08

LDSB 1.7 1.1 .16 1.9 1.2 .18

R2 .018 .042 .047

F .80 1.88 1.39 Note. LDSF = longest digit span forward. LDSB = longest digit span backward.

Table 24

Hierarchical regression model for anxiety, below average LDSB, and their interaction predicting BART performance, N = 88

Model 1 Model 2

B SE B β B SE B β

BAI Anxiety -.76 .51 -.16 -1.95** .70 -.40**

Below average LDSB -5.4 2.5 -.23* -14.9** 4.7 -.63**

BAI Anxiety x Below average LDSB 2.4* 1.0 .51*

R2 .071 .129

F 3.23* 4.16**

Δ R2 .059

F for Δ R2 5.66*

Note. * p ≤ .05. ** p ≤ .01. BAI Anxiety = Beck Anxiety Inventory total score transformed. LDSB = longest digit span backward. LDSB was dichotomized at below average range performance based on WAIS-IV age normative data.

Figure 11

Table 25

Hierarchical regression models for worry, above average LDSB, and their interaction predicting BART performance, N = 88

Model 1 Model 2

B SE B β B SE B β

PSW Worry -.04 .10 -.05 .11 .10 .12

Above average LDSB 2.2 2.9 .08 40.1** 11.9 1.5**

PSW Worry x Above average LDSB -.75 .23 -1.5**

R2 .009 .12

F .39 3.8*

Δ R2 .11

F for Δ R2 10.6**

Note. * p ≤ .05. ** p ≤ .01. LDSB = longest digit span backward. LDSB was dichotomized at above average range performance based on WAIS-IV age normative data. PSW = Penn State Worry Questionnaire total score.

Figure 12

Table 26

Independent regression models for gender, worry, anxiety, and HRV predicting BART performance, N = 80

Model 1 (Loss Amplitude) Model 2 (Gain Amplitude) Adjusted Model

Variable B SE B β B SE B β B SE B β

Gender -3.1 3.2 -0.1 -3.2 2.9 -0.1 -3.3 3.1 -0.1

PSW Worry -0.1 0.1 -0.1 -0.9 0.6 -0.2

BAI Anxiety -0.9 0.6 -0.2 -0.0 0.1 0.0

Tonic HRV -2.9 2.8 -0.1 -2.2 2.8 -0.1 -2.2 2.8 -0.1 R2 0.04 0.06 0.06

F 0.97 1.70 1.26 Note. PSW = Penn State Worry Questionnaire total score. BAI Anxiety = Beck Anxiety Inventory total score transformed. Tonic HRV = resting heart rate variability during the 3 minute baseline (BL).

Table 27

Independent regression models with mean SCR amplitude for gain and loss events (across all tasks) predicting BART performance N= 79

Model 1 (Loss Amplitude) Model 2 (Gain Amplitude) Adjusted Model

Variable B SE B β B SE B β B SE B β

Gender -2.3 2.7 -.10 -2.3 2.7 -.10 -2.3 2.7 -.10

Mean Loss Amplitude -.21 .30 -.08 -.18 .49 -.07

Mean Gain Amplitude -.17 .29 -.07 -.03 .47 -.01 R2 .018 .016 .018

F 0.68 0.61 0.45 Note. Mean loss and gain amplitude is defined as the mean SCR amplitude following each gain or loss of money across all three decision-making tasks.

Table 28

ANOVA and descriptives for physiological variables by gender and affect category, N = 84 Variable name Affect Category N Mean SD Tonic HRV (n = 86) none 25 3.5 0.5 anxiety +/or worry* 28 3.4 0.5 depression 33 3.3 0.5 Mean SCR Amplitude For Losses (n = 84) none 22 10.9 3.4 anxiety +/or worry 29 10.1 4.1 depression 33 10.9 5.2 Mean SCR Amplitude For Gains (n = 84) none 22 11.1 4.1 anxiety +/or worry 29 9.6 3.6 depression 33 10.4 5.6 Mean NS-SCRs/min Baseline (n = 84) none 22 3.3 2.5 anxiety +/or worry 29 3.2 2.7 depression 33 3.7 1.9 Mean NS-SCRs/min Tasks (n = 84) none 22 3.1 2.0 anxiety +/or worry 29 2.5 1.6 depression 33 2.4 1.5 Note: Affect category is defined as a) none = no endorsement of anxiety or depression symptoms, b/ anxiety +/or worry = endorsed anxiety or worry symptoms, but no depression symptoms, b) depression = endorsed only depression symptoms. Tonic HRV = resting heart rate variability during the 3 minute baseline (BL). Mean loss and gain amplitude is defined as the mean SCR amplitude following each gain or loss of money across all three decision-making tasks. Mean NS-SCRs/min = mean frequency of non-specific SCRs per minute (show separately for during baseline and across all tasks). 94

Table 29

Independent regression model for BART performance predicting IGT performance, N = 85

Model

Variable B SE B β

Gender -936.1 263.8 -.35**

BDI Depression -38.0 14.5 -.26

BART Performance -24.4 11.0 -.22*

R2 .22

F 7.6** Note. * p ≤ .05. ** p ≤ .01. BDI Depression = Beck Depression Inventory total score. BART Performance = total money earned in the BART.

Table 30

Independent regression models for GDT performance s predicting IGT performance, N = 86

Model 1 Model 2

Variable B SE B β B SE B β

Gender -837.6 266.8 -.32** -799.1 263.8 -.31**

BDI Depression -29.5 15.1 -.20 -30.6 14.9 -.21*

Race -275.8 271.6 -.11 -308.9 265.4 -.12

GDT Total Score .09 .06 .16

GDT % safe dice selections 32.3 15.2 .22*

R2 .177 .200

F 4.36** 5.07** Note. * p ≤ .05. ** p ≤ .01. Race = dichotomized as individuals identifying as Asian or Middle Eastern versus all other participants. BDI Depression = Beck Depression Inventory total score.

Table 31

Independent regression model for BART performance predicting GDT performance, N = 85

Model 1

Variable B SE B β

Race 1574.5 483.3 0.3**

BART Performance -1.5 21.1 -0.0 R2 .12

F 5.3** Note. * p ≤ .05. ** p ≤ .01. Race = dichotomized as individuals identifying as Asian or Middle Eastern versus all other participants. BART Performance = total money earned in the BART.

REFERENCES

Agelink, M. W., Boz, C., Ullrich, H., & Andrich, J. (2002). Relationship between major depression and heart rate variability: Clinical consequences and implications for antidepressive treatment. Psychiatry research, 113(1), 139-149. Agelink, M. W., Majewski, T., Wurthmann, C., Postert, T., Linka, T., Rotterdam, S., & Klieser, E. (2001). Autonomic neurocardiac function in patients with major depression and effects of antidepressive treatment with nefazodone. Journal of Affective Disorders, 62(3), 187- 198. Aldao, A., Mennin, D. S., & McLaughlin, K. A. (2013). Differentiating worry and rumination: evidence from heart rate variability during spontaneous regulation. Cognitive Therapy and Research, 37(3), 613-619. Allen, J. J., Chambers, A. S., & Towers, D. N. (2007). The many metrics of cardiac chronotropy: A pragmatic primer and a brief comparison of metrics. Biological Psychology, 74(2), 243-262. Apkarian, A. V., Sosa, Y., Krauss, B. R., Thomas, P. S., Fredrickson, B. E., Levy, R. E., ... & Chialvo, D.R. (2004). Chronic pain patients are impaired on an emotional decision- making task. Pain, 108(1), 129-136. Avery, J. A., Drevets, W. C., Moseman, S. E., Bodurka, J., Barcalow, J. C., & Simmons, W. K. (2013). Major depressive disorder is associated with abnormal interoceptive activity and functional connectivity in the insula. Biological psychiatry 76(3), 258-266. Balodis, I. M., MacDonald, T. K., & Olmstead, M. C. (2006). Instructional cues modify performance on the Iowa Gambling Task. Brain and Cognition, 60(2), 109-117. Barlow, D. H., Gorman, J. M., Shear, M. K., & Woods, S. W. (2000). Cognitive-behavioral therapy, imipramine, or their combination for panic disorder: A randomized controlled trial. Jama, 283(19), 2529-2536. Basso, M. R., Lowery, N., Ghormley, C., Combs, D., Purdie, R., Neel, J., ... & Bornstein, R. (2007). Comorbid anxiety corresponds with neuropsychological dysfunction in unipolar depression. Cognitive Neuropsychiatry, 12(5), 437-456. Batson, C. D., Engel, C. L., & Fridell, S. R. (1999). Value judgments: Testing the somatic- marker hypothesis using false physiological feedback. Personality and Social Psychology Bulletin, 25(8), 1021-1032. Bayard, S., Raffard, S., & Gely-Nargeot, M. C. (2011). Do facets of self-reported impulsivity predict decision-making under ambiguity and risk? Evidence from a community sample. Psychiatry Research, 190(2), 322-326. Beauchaine, T. P., Gatzke-Kopp, L., & Mead, H. K. (2007). Polyvagal theory and developmental psychopathology: Emotion dysregulation and conduct problems from preschool to adolescence. Biological Psychology, 74(2), 174-184. Bechara, A., Damasio, A. R., Damasio, H., & Anderson, S. W. (1994). Insensitivity to future consequences following damage to human prefrontal cortex. Cognition, 50(1), 7-15. Bechara, A., Damasio, H., & Damasio, A. R. (2000). Emotion, decision making and the orbitofrontal cortex. Cerebral Cortex, 10(3), 295-307.

Bechara, A., Damasio, H., Damasio, A. R., & Lee, G. P. (1999). Different contributions of the human amygdala and ventromedial prefrontal cortex to decision-making. The Journal of Neuroscience, 19(13), 5473-5481. Bechara, A., Damasio, H., Tranel, D., & Anderson, S. W. (1998). Dissociation of working memory from decision making within the human prefrontal cortex. Journal of Neuroscience, 18(1), 428-437. Bechara, A., & Martin, E. M. (2004). Impaired decision making related to working memory deficits in individuals with substance addictions. Neuropsychology, 18(1), 152. Bechara, A., Tranel, D., Damasio, H., & Damasio, A. R. (1996). Failure to respond autonomically to anticipated future outcomes following damage to prefrontal cortex. Cerebral cortex, 6(2), 215-225. Bechara, A., Tranel, D., & Damasio, H. (2000). Characterization of the decision-making deficit of patients with ventromedial prefrontal cortex lesions. Brain, 123(11), 2189-2202. Beck, A. T., Guth, D., Steer, R. A., & Ball, R. (1997). Screening for major depression disorders in medical inpatients with the Beck Depression Inventory for Primary Care. Behaviour Research and Therapy, 35(8), 785-791. Beck, A. T., & Steer, R. A. (1993). Beck anxiety inventory manual. San Antonio, Texas: The Psychological Corporation. Beck, A. T., Steer, R. A., & Brown, G. K. (1996). Beck depression inventory-II. San Antonio, Texas: The Psychological Corporation. Belzer, K., & Schneier, F. R. (2004). Comorbidity of anxiety and depressive disorders: issues in conceptualization, assessment, and treatment. Journal of Psychiatric Practice, 10(5), 296-306. Bickel, W. K., Yi, R., Landes, R. D., Hill, P. F., & Baxter, C. (2011). Remember the future: working memory training decreases delay discounting among stimulant addicts. Biological Psychiatry, 69(3), 260-265. Bolla, K. I., Eldreth, D. A., London, E. D., Kiehl, K. A., Mouratidis, M., Contoreggi, C., ... & Ernst, M. (2003). Orbitofrontal cortex dysfunction in abstinent cocaine abusers performing a decision-making task. Neuroimage, 19(3), 1085-1094. Bolla, K. I., Eldreth, D. A., Matochik, J. A., & Cadet, J. L. (2004). Sex-related differences in a gambling task and its neurological correlates. Cerebral Cortex, 14(11), 1226-1232. Braithwaite, J. J., Watson, D. G., Jones, R., & Rowe, M. (2013). A guide for analysing electrodermal activity (EDA) & skin conductance responses (SCRs) for psychological experiments. Psychophysiology, 49, 1017-1034. Brand, M., & Altstötter-Gleich, C. (2008). Personality and decision-making in laboratory gambling tasks–Evidence for a relationship between deciding advantageously under risk conditions and perfectionism. Personality and Individual Differences, 45(3), 226-231. Brand, M., Fujiwara, E., Borsutzky, S., Kalbe, E., Kessler, J., & Markowitsch, H. J. (2005). Decision-making deficits of korsakoff patients in a new gambling task with explicit rules: associations with executive functions. Neuropsychology, 19(3), 267-277. Brand, M., Grabenhorst, F., Starcke, K., Vandekerckhove, M. M., & Markowitsch, H. J. (2007a). Role of the amygdala in decisions under ambiguity and decisions under risk: evidence from patients with Urbach-Wiethe disease. Neuropsychologia, 45(6), 1305-1317.

Brand, M., Recknor, E. C., Grabenhorst, F., & Bechara, A. (2007b). Decisions under ambiguity and decisions under risk: correlations with executive functions and comparisons of two different gambling tasks with implicit and explicit rules. Journal of Clinical and Experimental Neuropsychology, 29(1), 86-99. Brevers, D., Cleeremans, A., Goudriaan, A. E., Bechara, A., Kornreich, C., Verbanck, P., & Noël, X. (2012). Decision making under ambiguity but not under risk is related to problem gambling severity. Psychiatry research, 200(2), 568-574. Brosschot, J. F., Gerin, W., & Thayer, J. F. (2006). The perseverative cognition hypothesis: A review of worry, prolonged stress-related physiological activation, and health. Journal of Psychosomatic Research, 60(2), 113-124. Brown, T. A., Campbell, L. A., Lehman, C. L., Grisham, J. R., & Mancill, R. B. (2001). Current and lifetime comorbidity of the DSM-IV anxiety and mood disorders in a large clinical sample. Journal of Abnormal Psychology, 110(4), 585-599. Brown, L. K., Tolou-Shams, M., Lescano, C., Houck, C., Zeidman, J., Pugatch, D., ... & Project SHIELD Study Group. (2006). Depressive symptoms as a predictor of sexual risk among African American adolescents and young adults. Journal of Adolescent Health, 39(3), 444-e1. Buelow, M. T., & Barnhart, W. R. (2015). The Influence of Math Anxiety, Math Performance, Worry, and Test Anxiety on the Iowa Gambling Task and Balloon Analogue Risk Task. Assessment, 24(1), 127-137. Buelow, M. T., & Blaine, A. L. (2015). The assessment of risky decision making: A factor analysis of performance on the Iowa Gambling Task, Balloon Analogue Risk Task, and Columbia Card Task. Psychological Assessment, 27(3), 777-785. Buelow, M. T., & Suhr, J. A. (2009). Construct validity of the Iowa gambling task. Neuropsychology Review, 19(1), 102-114. Buelow, M. T., & Suhr, J. A. (2013). Personality characteristics and state mood influence individual deck selections on the Iowa Gambling Task. Personality and Individual Differences, 54(5), 593-597

Canu, E., Kostic, M., Agosta, F., Munjiza, A., Ferraro, P. M., Pesic, D., ... & Filippi, M. (2015). Brain structural abnormalities in patients with major depression with or without generalized anxiety disorder comorbidity. Journal of Neurology, 262(5), 1255. Carney, R. M., Freedland, K. E., Stein, P. K., Skala, J. A., Hoffman, P., & Jaffe, A. S. (2000). Change in heart rate and heart rate variability during treatment for depression in patients with coronary heart disease. Psychosomatic Medicine, 62(5), 639-647. Carney, R. M., Saunders, R. D., Freedland, K. E., Stein, P., Rich, M. W., & Jaffe, A. S. (1995). Association of depression with reduced heart rate variability in coronary artery disease. The American Journal of Cardiology, 76(8), 562-564. Carter, S., & Smith Pasqualini, M. (2004). Stronger autonomic response accompanies better learning: A test of Damasio's somatic marker hypothesis. Cognition and Emotion, 18(7), 901-911. Carver, C. S., & White, T. L. (1994). Behavioral inhibition, behavioral activation, and affective responses to impending reward and punishment: The BIS/BAS Scales. Journal of Personality and Social Psychology, 67(2), 319-333.

Cella, M., Dymond, S., & Cooper, A. (2010). Impaired flexible decision-making in major depressive disorder. Journal of Affective Disorders, 124(1), 207-210. Conway, A. R., Kane, M. J., Bunting, M. F., Hambrick, D. Z., Wilhelm, O., & Engle, R. W. (2005). Working memory span tasks: A methodological review and user’s guide. Psychonomic Bulletin & Review, 12(5), 769-786. Craig, A. D. (2003). Interoception: the sense of the physiological condition of the body. Current Opinion in Neurobiology, 13(4), 500-505. Craske, M. G., & Barlow, D.H. (2008). Panic Disorder and Agoraphobia. New York, NY: Guilford Press. Creamer, M., Foran, J., & Bell, R. (1995). The Beck Anxiety Inventory in a non-clinical sample. Behaviour Research and Therapy, 33(4), 477-485. Critchley, H. D., Elliott, R., Mathias, C. J., & Dolan, R. J. (2000). Neural activity relating to generation and representation of galvanic skin conductance responses: a functional magnetic resonance imaging study. The Journal of Neuroscience, 20(8), 3033-3040. Damasio, A. R. (1994). Descartes’ Error: Emotion, Reason, and the Human Brain. New York: Putnam. Damasio, A. R., Everitt, B. J., & Bishop, D. (1996). The somatic marker hypothesis and the possible functions of the prefrontal cortex [and discussion]. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, 351(1346), 1413-1420. Davidson, J., Watkins, L., Owens, M., Krulewicz, S., Connor, K., Carpenter, D., ... & Nemeroff, C. (2005). Effects of paroxetine and venlafaxine XR on heart rate variability in depression. Journal of clinical psychopharmacology, 25(5), 480-484. Deacon, B., Kemp, J. J., Dixon, L. J., Sy, J. T., Farrell, N. R., & Zhang, A. R. (2013). Maximizing the efficacy of interoceptive exposure by optimizing inhibitory learning: A randomized controlled trial. Behaviour Research and Therapy, 51(9), 588-596. Dozois, D. J., Dobson, K. S., & Ahnberg, J. L. (1998). A psychometric evaluation of the Beck Depression Inventory–II. Psychological Assessment, 10(2), 83-89. Drost, J., Spinhoven, P., Kruijt, A. W., & Van der Does, W. (2014). The influence of worry and avoidance on the Iowa Gambling Task. Journal of Behavior Therapy and Experimental Psychiatry, 45(1), 74-80. Duarte, N. A., Woods, S. P., Rooney, A., Atkinson, J. H., Grant, I., & Translational Methamphetamine AIDS Research Center (TMARC) Group. (2012). Working memory deficits affect risky decision-making in methamphetamine users with attention- deficit/hyperactivity disorder. Journal of Psychiatric Research, 46(4), 492-499. Dunn, B. D., Dalgleish, T., & Lawrence, A. D. (2006). The somatic marker hypothesis: A critical evaluation. Neuroscience & Biobehavioral Reviews, 30(2), 239-271. Dunn, B. D., Dalgleish, T., Ogilvie, A. D., & Lawrence, A. D. (2007). Heartbeat perception in depression. Behaviour Research and Therapy, 45(8), 1921-1930. Dunn, B. D., Billotti, D., Murphy, V., & Dalgleish, T. (2009). The consequences of effortful emotion regulation when processing distressing material: A comparison of suppression and acceptance. Behaviour Research and Therapy, 47(9), 761-773.

100

Dunn, B. D., Galton, H. C., Morgan, R., Evans, D., Oliver, C., Meyer, M., Cusack, R., Lawrence, A.D., & Dalgleish, T. (2010a). Listening to your heart how interoception shapes emotion experience and intuitive decision making. Psychological Science, 21, 1835-1844. Dunn, B. D., Stefanovitch, I., Evans, D., Oliver, C., Hawkins, A., & Dalgleish, T. (2010b). Can you feel the beat? Interoceptive awareness is an interactive function of anxiety-and depression-specific symptom dimensions. Behaviour Research and Therapy, 48(11), 1133-1138. Eddie, D., Kim, C., Lehrer, P., Deneke, E., & Bates, M. E. (2014). A pilot study of brief heart rate variability biofeedback to reduce craving in young adult men receiving inpatient treatment for substance use disorders. Applied Psychophysiology and Biofeedback, 39(3- 4), 181-192. Ehlers, A., & Breuer, P. (1992). Increased cardiac awareness in panic disorder. Journal of Abnormal Psychology, 101(3), 371-382. Ehlers, A., & Breuer, P. (1995). Selective attention to physical threat in subjects with panic attacks and specific phobias. Journal of Anxiety Disorders, 9(1), 11-31. Eisenberg, D., Gollust, S. E., Golberstein, E., & Hefner, J. L. (2007). Prevalence and correlates of depression, anxiety, and suicidality among university students. American Journal of Orthopsychiatry, 77(4), 534-542. Ellis, R., Sollers III, J., Edelstein, E., & Thayer, J. (2008). Data transforms for spectral analyses of heart rate variability. Biomedical Sciences Instrumentation Symposium, 44, 392-397. Engelmann, J. B., Meyer, F., Fehr, E., & Ruff, C. C. (2015). Anticipatory anxiety disrupts neural valuation during risky choice. The Journal of Neuroscience, 35(7), 3085-3099. Epel, E., Lapidus, R., McEwen, B., & Brownell, K. (2001). Stress may add bite to appetite in women: a laboratory study of stress-induced cortisol and eating behavior. Psychoneuroendocrinology, 26(1), 37-49. Fellows, L. K., & Farah, M. J. (2005). Different underlying impairments in decision-making following ventromedial and dorsolateral frontal lobe damage in humans. Cerebral Cortex, 15(1), 58-63. Fernie, G., & Tunney, R. J. (2006). Some decks are better than others: the effect of reinforcer type and task instructions on learning in the Iowa Gambling Task. Brain and Cognition, 60(1), 94-102. Figner, B., & Murphy, R. O. (2011). Using skin conductance in judgment and decision making research. In M. Schulte-Mecklenbeck, A. Kuehberger, & R. Ranyard (Eds.), A handbook of process tracing methods for decision research. New York, NY: Psychology Press. Friedman, B. H. (2007). An autonomic flexibility–neurovisceral integration model of anxiety and cardiac vagal tone. Biological Psychology, 74(2), 185-199. Friedman, B. H., & Thayer, J. F. (1998). Autonomic balance revisited: panic anxiety and heart rate variability. Journal of psychosomatic research, 44(1), 133-151. Friedman, B. H., Thayer, J. F., & Borkovec, T. D. (2000). Explicit memory bias for threat words in generalized anxiety disorder. Behavior Therapy, 31(4), 745-756. Furman, D. J., Waugh, C. E., Bhattacharjee, K., Thompson, R. J., & Gotlib, I. H. (2013). Interoceptive awareness, positive affect, and decision making in Major Depressive Disorder. Journal of Affective Disorders, 151(2), 780-785.

101

Garfinkel, S. N., Seth, A. K., Barrett, A. B., Suzuki, K., & Critchley, H. D. (2015). Knowing your own heart: distinguishing interoceptive accuracy from interoceptive awareness. Biological Psychology, 104, 65-74. Gloster, A. T., Wittchen, H. U., Einsle, F., Lang, T., Helbig-Lang, S., Fydrich, T., ... & Gerlach, A. L. (2011). Psychological treatment for panic disorder with agoraphobia: a randomized controlled trial to examine the role of therapist-guided exposure in situ in CBT. Journal of Consulting and Clinical Psychology, 79(3), 406-420. Golub, S. A., Starks, T. J., Kowalczyk, W. J., Thompson, L. I., & Parsons, J. T. (2012). Profiles of executive functioning: Associations with substance dependence and risky sexual behavior. Psychology of Addictive Behaviors, 26(4), 895-905. Golub, S. A., Thompson, L. I., & Kowalczyk, W. J. (2016). Affective differences in Iowa Gambling Task performance associated with sexual risk taking and substance use among HIV-positive and HIV-negative men who have sex with men. Journal of Clinical and Experimental Neuropsychology, 38(2), 141-157. Gorman, J. M., & Sloan, R. P. (2000). Heart rate variability in depressive and anxiety disorders. American Heart Journal, 140(4), S77-S83. Gray, J. R. (2004). Integration of emotion and cognitive control. Current Directions in Psychological Science, 13(2), 46-48. Gross, J. J., & John, O. P. (2003). Individual differences in two emotion regulation processes: implications for affect, relationships, and well-being. Journal of Personality and Social Psychology, 85(2), 348-362. Han, G., Klimes-Dougan, B., Jepsen, S., Ballard, K., Nelson, M., Houri, A., ... & Cullen, K. (2012). Selective neurocognitive impairments in adolescents with major depressive disorder. Journal of Adolescence, 35(1), 11-20. Hart, J., Schwabach, J. A., & Solomon, S. (2010). Going for broke: Mortality salience increases risky decision making on the Iowa gambling task. British Journal of Social Psychology, 49(2), 425-432. Harvey, P. O., Le Bastard, G., Pochon, J. B., Levy, R., Allilaire, J. F., Dubois, B. E. E. A., & Fossati, P. (2004). Executive functions and updating of the contents of working memory in unipolar depression. Journal of Psychiatric Research, 38(6), 567-576. Heilman, R. M., Crişan, L. G., Houser, D., Miclea, M., & Miu, A. C. (2010). Emotion regulation and decision making under risk and uncertainty. Emotion, 10(2), 257-265. Heims, H. C., Critchley, H. D., Dolan, R., Mathias, C. J., & Cipolotti, L. (2004). Social and motivational functioning is not critically dependent on feedback of autonomic responses: neuropsychological evidence from patients with pure autonomic failure. Neuropsychologia, 42(14), 1979-1988. Hettema, J. M., Prescott, C. A., & Kendler, K. S. (2003). The effects of anxiety, substance use and conduct disorders on risk of major depressive disorder. Psychological Medicine, 33(08), 1423-1432. Hibbert, A. S., Weinberg, A., & Klonsky, E. D. (2012). Field Validity of Heart Rate Variability Metrics Produced by QRSTool and CMetX. Psychological Assessment, 24(3), 777-782. Hinson, J. M., Jameson, T. L., & Whitney, P. (2003). Impulsive decision making and working memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 29(2), 298-306.

102

Hooper, C. J., Luciana, M., Conklin, H. M., & Yarger, R. S. (2004). Adolescents' performance on the Iowa Gambling Task: implications for the development of decision making and ventromedial prefrontal cortex. Developmental Psychology, 40(6), 1148-1158. Hovland, A., Pallesen, S., Hammar, Å., Hansen, A. L., Thayer, J., Tarvainen, M., & Nordhus, I. (2012). The relationships among heart rate variability, executive functions, and clinical variables in patients with panic disorder. International Journal of Psychophysiology, 86(3), 269-275. Hughes, J. W., & Stoney, C. M. (2000). Depressed mood is related to high-frequency heart rate variability during stressors. Psychosomatic Medicine, 62(6), 796-803. Hunt, M. K., Hopko, D. R., Bare, R., Lejuez, C. W., & Robinson, E. V. (2005). Construct validity of the Balloon Analogue Risk Task (BART) associations with psychopathy and impulsivity. Assessment, 12(4), 416-428. James, W. (1884). What is an emotion? Mind, 9, 188 -205. Joormann, J., & Gotlib, I. H. (2008). Updating the contents of working memory in depression: interference from irrelevant negative material. Journal of Abnormal Psychology, 117(1), 182-192. Kaplan, J. S., Erickson, K., Luckenbaugh, D. A., Weiland-Fiedler, P., Geraci, M., Sahakian, B. J., ... & Neumeister, A. (2006). Differential performance on tasks of affective processing and decision-making in patients with panic disorder and panic disorder with comorbid major depressive disorder. Journal of Affective Disorders, 95(1), 165-171. Karavidas, M. K., Lehrer, P. M., Vaschillo, E., Vaschillo, B., Marin, H., Buyske, S., ... & Hassett, A. (2007). Preliminary results of an open label study of heart rate variability biofeedback for the treatment of major depression. Applied Psychophysiology and Biofeedback, 32(1), 19-30. Katkin, E. S., Cestaro, V. L., & Weitkunat, R. (1991). Individual differences in cortical evoked potentials as a function of heartbeat detection ability. International journal of neuroscience, 61(3-4), 269-276. Keller, J., Nitschke, J. B., Bhargava, T., Deldin, P. J., Gergen, J. A., Miller, G. A., & Heller, W. (2000). Neuropsychological differentiation of depression and anxiety. Journal of Abnormal Psychology, 109(1), 3-10. Kemp, A. H., Quintana, D. S., Gray, M. A., Felmingham, K. L., Brown, K., & Gatt, J. M. (2010). Impact of depression and antidepressant treatment on heart rate variability: a review and meta-analysis. Biological Psychiatry, 67(11), 1067-1074. Kessler, R. C., Berglund, P., Demler, O., Jin, R., Merikangas, K. R., & Walters, E. E. (2005). Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the National Comorbidity Survey Replication. Archives of General Psychiatry, 62(6), 593-602. Kessler, R. C., Nelson, C. B., McGonagle, K. A., & Liu, J. (1996). Comorbidity of DSM-III—R major depressive disorder in the general population: Results from the US National Comorbidity Survey. The British Journal of Psychiatry, 30(6) 3017-3030. . Key, B. L., Campbell, T. S., Bacon, S. L., & Gerin, W. (2008). The influence of trait and state rumination on cardiovascular recovery from a negative emotional stressor. Journal of Behavioral Medicine, 31(3), 237-248. Kim, T., Pollock, S., Lee, D., O'Brien, R., & Keall, P. (2012). Audiovisual biofeedback improves diaphragm motion reproducibility in MRI. Medical Physics, 39(11), 6921-6928.

103

Knutson, B., Bhanji, J. P., Cooney, R. E., Atlas, L. Y., & Gotlib, I. H. (2008). Neural responses to monetary incentives in major depression. Biological Psychiatry, 63(7), 686-692. Kyte, Z. A., Goodyer, I. M., & Sahakian, B. J. (2005). Selected executive skills in adolescents with recent first episode major depression. Journal of Child Psychology and Psychiatry, 46(9), 995-1005. Lawrence, N. S., Jollant, F., O'Daly, O., Zelaya, F., & Phillips, M. L. (2009). Distinct roles of prefrontal cortical subregions in the Iowa Gambling Task. Cerebral Cortex, 19(5), 1134- 1143. Lecrubier, Y., Sheehan, D. V., Weiller, E., Amorim, P., Bonora, I., Sheehan, K. H., ... & Dunbar, G. C. (1997). The Mini International Neuropsychiatric Interview (MINI). A short diagnostic structured interview: reliability and validity according to the CIDI. European Psychiatry, 12(5), 224-231. Ledgerwood, D. M., Alessi, S. M., Phoenix, N., & Petry, N. M. (2009). Behavioral assessment of impulsivity in pathological gamblers with and without substance use disorder histories versus healthy controls. Drug and Alcohol Dependence, 105(1), 89-96. Lejuez, C. W., Aklin, W. M., Jones, H. A., Richards, J. B., Strong, D. R., Kahler, C. W., & Read, J. P. (2003). The balloon analogue risk task (BART) differentiates smokers and nonsmokers. Experimental and Clinical Psychopharmacology, 11(1), 26. Lejuez, C. W., Read, J. P., Kahler, C. W., Richards, J. B., Ramsey, S. E., Stuart, G. L., ... & Brown, R. A. (2002). Evaluation of a behavioral measure of risk taking: the Balloon Analogue Risk Task (BART). Journal of Experimental Psychology: Applied, 8(2), 75-84. Lighthall, N. R., Mather, M., & Gorlick, M. A. (2009). Acute stress increases sex differences in risk seeking in the balloon analogue risk task. PLoS One, 4(7), 1-6. Linnet, J., Møller, A., Peterson, E., Gjedde, A., & Doudet, D. (2011). Dopamine release in ventral striatum during Iowa Gambling Task performance is associated with increased excitement levels in pathological gambling. Addiction, 106(2), 383-390. Loo, J. M., Raylu, N., & Oei, T. P. S. (2008). Gambling among the Chinese: A comprehensive review. Clinical Psychology Review, 28(7), 1152-1166. Luhmann, C. C., Ishida, K., & Hajcak, G. (2011). Intolerance of uncertainty and decisions about delayed, probabilistic rewards. Behavior Therapy, 42(3), 378-386. Malik, M., Bigger, J., Camm, J., Kleiger, R., Malliani, A., Moss, J., Schwartz, P., & the Task Force of the European Society of Cardiology (1996). Heart rate variability standards of measurement, physiological interpretation, and clinical use. European Heart Journal, 17, 354-381. Maner, J. K., Richey, J. A., Cromer, K., Mallott, M., Lejuez, C. W., Joiner, T. E., & Schmidt, N. B. (2007). Dispositional anxiety and risk-avoidant decision-making. Personality and Individual Differences, 42(4), 665-675. Manes, F., Sahakian, B., Clark, L., Rogers, R., Antoun, N., Aitken, M., & Robbins, T. (2002). Decision-making processes following damage to the prefrontal cortex. Brain, 125(3), 624-639. Martin, C. O., Denburg, N. L., Tranel, D., Granner, M. A., & Bechara, A. (2004). The effects of vagus nerve stimulation on decision-making. Cortex, 40(4), 605-612. Mathews, A., & MacLeod, C. (1994). Cognitive approaches to emotion and emotional disorders. Annual Review of Psychology, 45(1), 25-50.

104

Mathews, A., & MacLeod, C. (1986). Discrimination of threat cues without awareness in anxiety states. Journal of Abnormal Psychology, 95(2), 131-138. McFarland, B. R., & Klein, D. N. (2009). Emotional reactivity in depression: diminished responsiveness to anticipated reward but not to anticipated punishment or to nonreward or avoidance. Depression and Anxiety, 26(2), 117-122. Meyer, T. J., Miller, M. L., Metzger, R. L., & Borkovec, T. D. (1990). Development and validation of the penn state worry questionnaire. Behaviour Research and Therapy, 28(6), 487-495. Mezzacappa, E., Tremblay, R. E., Kindlon, D., Saul, J. P., Arseneault, L., Seguin, J., ... & Earls, F. (1997). Anxiety, antisocial behavior, and heart rate regulation in adolescent males. Journal of Child Psychology and Psychiatry, 38(4), 457-469. Mineka, S., Watson, D., & Clark, L. A. (1998). Comorbidity of anxiety and unipolar mood disorders. Annual review of psychology, 49(1), 377-412. Miu, A. C., Heilman, R. M., & Houser, D. (2008). Anxiety impairs decision-making: psychophysiological evidence from an Iowa Gambling Task. Biological Psychology, 77(3), 353-358. Moffitt, T. E., Harrington, H., Caspi, A., Kim-Cohen, J., Goldberg, D., Gregory, A. M., & Poulton, R. (2007). Depression and generalized anxiety disorder: cumulative and sequential comorbidity in a birth cohort followed prospectively to age 32 years. Archives of General Psychiatry, 64(6), 651-660. Molina, S., & Borkovec, T. D. (1994). Worrying: Perspectives on theory, assessment and treatment (pp. 265-283). Oxford, England: John Wiley & Sons. Mueller, E. M., Nguyen, J., Ray, W. J., & Borkovec, T. D. (2010). Future-oriented decision- making in Generalized Anxiety Disorder is evident across different versions of the Iowa Gambling Task. Journal of Behavior Therapy and Experimental Psychiatry, 41(2), 165- 171. Must, A., Szabó, Z., Bódi, N., Szász, A., Janka, Z., & Kéri, S. (2006). Sensitivity to reward and punishment and the prefrontal cortex in major depression. Journal of Affective Disorders, 90(2), 209-215. Nahshoni, E., Aravot, D., Aizenberg, D., Sigler, M., Zalsman, G., Strasberg, B., ... & Weizman, A. (2004). Heart rate variability in patients with major depression. Psychosomatics, 45(2), 129-134. Naqvi, N. H., & Bechara, A. (2006). Skin conductance: A psychophysiological approach to the study of decision making. In Methods in Mind. C. Senior, T. Russell & M.S. Gazzaniga, Eds.: 103-122. MIT Press. Cambridge, MA. Nelson, A. L., Purdon, C., Quigley, L., Carriere, J., & Smilek, D. (2015). Distinguishing the roles of trait and state anxiety on the nature of anxiety-related attentional biases to threat using a free viewing eye movement paradigm. Cognition and Emotion, 29(3), 504-526. Newman, M. G., Zuellig, A. R., Kachin, K. E., Constantino, M. J., Przeworski, A., Erickson, T., & Cashman-McGrath, L. (2002). Preliminary reliability and validity of the Generalized Anxiety Disorder Questionnaire-IV: A revised self-report diagnostic measure of generalized anxiety disorder. Behavior Therapy, 33(2), 215-233.

105

Noël, X., Bechara, A., Dan, B., Hanak, C., & Verbanck, P. (2007). Response inhibition deficit is involved in poor decision-making under risk in nonamnesic individuals with alcoholism. Neuropsychology, 21(6), 778-786. North, N. T., & O'Carroll, R. E. (2001). Decision making in patients with spinal cord damage: afferent feedback and the somatic marker hypothesis. Neuropsychologia, 39(5), 521-524. Nunan, D., Sandercock, G., & Brodie, D. (2010). A quantitative systematic review of normal values for short‐term heart rate variability in healthy adults. Pacing and Clinical Electrophysiology, 33(11), 1407-1417. Overman, W. H., & Pierce, A. (2013). Iowa Gambling Task with non-clinical participants: effects of using real+ virtual cards and additional trials. Frontiers in Psychology, 4, article 935. Parker, G., Wilhelm, K., Mitchell, P., Austin, M. P., Roussos, J., & Gladstone, G. (1999). The influence of anxiety as a risk to early onset major depression. Journal of Affective Disorders, 52(1), 11-17. Patton, J. H., & Stanford, M. S. (1995). Factor structure of the Barratt impulsiveness scale. Journal of Clinical Psychology, 51(6), 768-774. Paulus, M. P. (2007). Decision-making dysfunctions in psychiatry—altered homeostatic processing? Science, 318(5850), 602-606. Paulus, M. P., & Stein, M. B. (2010). Interoception in anxiety and depression. Brain structure and Function, 214(5-6), 451-463. Paulus, M. P., & Yu, A.J. (2012). Emotion and decision-making: affect-driven belief systems in anxiety and depression. Trends in Cognitive Sciences, 16(9), 476-483. Piazza, P. V., & Le Moal, M. (1997). Glucocorticoids as a biological substrate of reward: physiological and pathophysiological implications. Brain Research Reviews, 25(3), 359- 372. Pollatos, O., Gramann, K., & Schandry, R. (2007). Neural systems connecting interoceptive awareness and feelings. Human Brain Mapping, 28(1), 9-18. Pollatos, O., Kirsch, W., & Schandry, R. (2005). On the relationship between interoceptive awareness, emotional experience, and brain processes. Cognitive Brain Research, 25(3), 948-962 Pollatos, O., Traut-Mattausch, E., & Schandry, R. (2009). Differential effects of anxiety and depression on interoceptive accuracy. Depression and Anxiety, 26(2), 167-173. Porges, S. W. (1992). Vagal tone: a physiologic marker of stress vulnerability. Pediatrics, 90(3), 498-504. Porges, S. W. (2007). The polyvagal perspective. Biological Psychology, 74(2), 116-143. Preston, S. D., Buchanan, T. W., Stansfield, R. B., & Bechara, A. (2007). Effects of anticipatory stress on decision making in a gambling task. Behavioral Neuroscience, 121(2), 257-263. Quintana, D. S., & Heathers, J. A. (2014). Considerations in the assessment of heart rate variability in biobehavioral research. Frontiers in Psychology, 5 (805). Raylu, N., & Oei, T. P. (2004). Role of culture in gambling and problem gambling. Clinical Psychology Review, 23(8), 1087-1114. Rogers, R. D., Everitt, B. J., Baldacchino, A., Blackshaw, A. J., Swainson, R., Wynne, K., ... & London, M. (1999). Dissociable deficits in the decision-making cognition of chronic amphetamine abusers, opiate abusers, patients with focal damage to prefrontal cortex, and

106

tryptophan-depleted normal volunteers: evidence for monoaminergic mechanisms. Neuropsychopharmacology, 20(4), 322-339. Rose, E. J., & Ebmeier, K. P. (2006). Pattern of impaired working memory during major depression. Journal of Affective Disorders, 90(2), 149-161. Ruiz-Padial, E., Vila, J., & Thayer, J. F. (2011). The effect of conscious and non-conscious presentation of biologically relevant emotion pictures on emotion modulated startle and phasic heart rate. International Journal of Psychophysiology, 79(3), 341-346. Schandry, R. (1981). Heart beat perception and emotional experience. Psychophysiology, 18(4), 483-488. Schmeichel, B. J., Volokhov, R. N., & Demaree, H. A. (2008). Working memory capacity and the self-regulation of emotional expression and experience. Journal of Personality and Social Psychology, 95(6), 1526-1540. Schrag, Y., Tremea, A., Lagger, C., Ohana, N., & Mohr, C. (2016). Pro Free Will Priming Enhances “Risk-Taking” Behavior in the Iowa Gambling Task, but Not in the Balloon Analogue Risk Task: Two Independent Priming Studies. PloS One, 11(3), e0152297. Schroeder, R. W., Twumasi-Ankrah, P., Baade, L. E., & Marshall, P. S. (2012). Reliable Digit Span: A systematic review and cross-validation study. Assessment, 19(1), 21-30. Scott, S. G., & Bruce, R. A. (1995). Decision-making style: The development and assessment of a new measure. Educational and Psychological Measurement, 55(5), 818-831. Shackman, A. J., Sarinopoulos, I., Maxwell, J. S., Pizzagalli, D. A., Lavric, A., & Davidson, R. J. (2006). Anxiety selectively disrupts visuospatial working memory. Emotion, 6(1), 40- 61. Shiffman, S., & Waters, A. J. (2004). Negative affect and smoking lapses: a prospective analysis. Journal of Consulting and Clinical Psychology, 72(2), 192-201. Siepmann, M., Aykac, V., Unterdörfer, J., Petrowski, K., & Mueck-Weymann, M. (2008). A pilot study on the effects of heart rate variability biofeedback in patients with depression and in healthy subjects. Applied Psychophysiology and Biofeedback, 33(4), 195-201. Spielberger, C.D. (1983). Manual for the State-Trait Anxiety Inventory STAI (Form Y). Palo Alto, CA: Consulting Psychologists Press. Stein, P. K., Carney, R. M., Freedland, K. E., Skala, J. A., Jaffe, A. S., Kleiger, R. E., & Rottman, J. N. (2000). Severe depression is associated with markedly reduced heart rate variability in patients with stable coronary heart disease. Journal of Psychosomatic Research, 48(4), 493-500. Stevens, S., Gerlach, A. L., Cludius, B., Silkens, A., Craske, M. G., & Hermann, C. (2011). Heartbeat perception in social anxiety before and during speech anticipation. Behaviour Research and Therapy, 49(2), 138-143. Suzuki, A., Hirota, A., Takasawa, N., & Shigemasu, K. (2003). Application of the somatic marker hypothesis to individual differences in decision-making. Biological Psychology, 65(1), 81-88. Tchanturia, K., Liao, P. C., Uher, R., Lawrence, N., Treasure, J., & Campbell, I. C. (2007). An investigation of decision making in anorexia nervosa using the Iowa Gambling Task and skin conductance measurements. Journal of the International Neuropsychological Society, 13(04), 635-641.

107

Thayer, J. F., Friedman, B. H., & Borkovec, T. D. (1996). Autonomic characteristics of generalized anxiety disorder and worry. Biological Psychiatry, 39(4), 255-266. Thayer, J. F., Friedman, B. H., Borkovec, T. D., Johnsen, B. H., & Molina, S. (2000). Phasic heart period reactions to cued threat and nonthreat stimuli in generalized anxiety disorder. Psychophysiology, 37(03), 361-368. Thayer, J. F., & Lane, R. D. (2000). A model of neurovisceral integration in emotion regulation and dysregulation. Journal of Affective Disorders, 61(3), 201-216. Tice, D. M., Bratslavsky, E., & Baumeister, R. F. (2001). Emotional distress regulation takes precedence over impulse control: if you feel bad, do it! Journal of personality and social psychology, 80(1), 53. Tomb, I., Hauser, M., Deldin, P., & Caramazza, A. (2002). Do somatic markers mediate decisions on the gambling task? Nature Neuroscience, 5(11), 1103-1104. Towbes, L. C., & Cohen, L. H. (1996). Chronic stress in the lives of college students: Scale development and prospective prediction of distress. Journal of Youth and Adolescence, 25(2), 199-217. Udupa, K., Sathyaprabha, T. N., Thirthalli, J., Kishore, K. R., Lavekar, G. S., Raju, T. R., & Gangadhar, B. N. (2007). Alteration of cardiac autonomic functions in patients with major depression: a study using heart rate variability measures. Journal of Affective Disorders, 100(1), 137-141. van den Bos, R., Harteveld, M., & Stoop, H. (2009). Stress and decision-making in humans: performance is related to cortisol reactivity, albeit differently in men and women. Psychoneuroendocrinology, 34(10), 1449-1458. van den Bos, R., Homberg, J., & de Visser, L. (2013). A critical review of sex differences in decision- making tasks: Focus on the Iowa Gambling Task. Behavioural Brain Research, 238, 95-108. Verdejo-García, A., López-Torrecillas, F., Calandre, E. P., Delgado-Rodríguez, A., & Bechara, A. (2009). Executive function and decision-making in women with fibromyalgia. Archives of Clinical Neuropsychology, 24(1), 113-122. de Vries, M., Holland, R. W., & Witteman, C. L. (2008a). In the winning mood: Affect in the Iowa gambling task. Judgment and Decision Making, 3(1), 42-50. de Vries, M., Holland, R. W., & Witteman, C. L. (2008b). Fitting decisions: Mood and intuitive versus deliberative decision strategies. Cognition and Emotion, 22(5), 931-943. de Visser, L., Van Der Knaap, L. J., van de Loo, A. J., van der Weerd, C. M., Ohl, F., & Van Den Bos, R. (2010). Trait anxiety affects decision-making differently in healthy men and women: towards gender-specific endophenotypes of anxiety. Neuropsychologia, 48(6), 1598-1606. Wald, J., & Taylor, S. (2005). Interoceptive exposure therapy combined with trauma‐related exposure therapy for post‐traumatic stress disorder: A case report. Cognitive Behaviour Therapy, 34(1), 34-40. Wald, J., & Taylor, S. (2007). Efficacy of interoceptive exposure therapy combined with trauma- related exposure therapy for posttraumatic stress disorder: A pilot study. Journal of Anxiety Disorders, 21(8), 1050-1060.

108

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(6), 1063-1070. Weber, E. U., Blais, A. R., & Betz, N. E. (2002). A domain‐specific risk‐attitude scale: Measuring risk perceptions and risk behaviors. Journal of Behavioral Decision Making, 15(4), 263-290. Wechsler, D. (2008). Wechsler adult intelligence scale–Fourth Edition (WAIS–IV). San Antonio, TX: NCS Pearson. Weller, J. A., Levin, I. P., & Bechara, A. (2010). Do individual differences in Iowa Gambling Task performance predict adaptive decision making for risky gains and losses?. Journal of Clinical and Experimental Neuropsychology, 32(2), 141-150. Werner, N. S., Duschek, S., & Schandry, R. (2009). Relationships between affective states and decision-making. International Journal of Psychophysiology, 74(3), 259-265. Werner, N. S., Schweitzer, N., Meindl, T., Duschek, S., Kambeitz, J., & Schandry, R. (2013). Interoceptive awareness moderates neural activity during decision-making. Biological Psychology, 94(3), 498- 506. Wiens, S., Mezzacappa, E. S., & Katkin, E. S. (2000). Heartbeat detection and the experience of emotions. Cognition & Emotion, 14(3), 417-427. Williams, D., Cash, C., Rankin, C., Bernardi, A., Koenig, J., & Thayer, J. (2015). Resting heart rate variability predicts self-reported difficulties in emotion regulation: a focus on different facets of emotion regulation. Frontiers in Psychology, 6, article 261. Wölk, J., Sütterlin, S., Koch, S., Vögele, C., & Schulz, S. M. (2014). Enhanced cardiac perception predicts impaired performance in the Iowa Gambling Task in patients with panic disorder. Brain and Behavior, 4(2), 238-246. Xu S., Fang Z. & Rao H. (2013). Real or Hypothetical Monetary Rewards Modulates Risk Taking Behavior. Acta Psychol Sinica 45, 874–886. Xu, S., Pan, Y., Wang, Y., Spaeth, A. M., Qu, Z., & Rao, H. (2016). Real and hypothetical monetary rewards modulate risk taking in the brain. Nature Publishing Group. Scientific Reports, 6, 29520. Yates, A. J., Jones, K. E., Marie, G. V., & Hogben, J. H. (1985). Detection of the heartbeat and events in the cardiac cycle. Psychophysiology, 22(5), 561-567. Young, E. A., Abelson, J. L., & Cameron, O. G. (2004). Effect of comorbid anxiety disorders on the hypothalamic-pituitary-adrenal axis response to a social stressor in major depression. Biological Psychiatry, 56(2), 113-120. Zhang, W. N., Chang, S. H., Guo, L. Y., Zhang, K. L., & Wang, J. (2013). The neural correlates of reward-related processing in major depressive disorder: a meta-analysis of functional magnetic resonance imaging studies. Journal of Affective Disorders, 151(2), 531-539.

109