Journal Pre-proof Approach and avoidance patterns in reward learning across domains: An initial examination of the Social Iowa Gambling Task Julia A.C. Case, Thomas M. Olino PII:
S0005-7967(19)30233-5
DOI:
https://doi.org/10.1016/j.brat.2019.103547
Reference:
BRT 103547
To appear in:
Behaviour Research and Therapy
Received Date: 12 May 2019 Revised Date:
24 November 2019
Accepted Date: 29 December 2019
Please cite this article as: Case, J.A.C., Olino, T.M., Approach and avoidance patterns in reward learning across domains: An initial examination of the Social Iowa Gambling Task, Behaviour Research and Therapy (2020), doi: https://doi.org/10.1016/j.brat.2019.103547. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 Published by Elsevier Ltd.
Contributor Roles Taxonomy Author Statement Julia A. C. Case, M.A.: Conceptualization, Methodology, Software, Validation, Formal Analysis, Investigation, Resources, Data Curation, Writing – Original Draft, Visualization, Project Administration, Funding Acquisition Thomas M. Olino, Ph.D.: Conceptualization, Methodology, Validation, Formal Analysis, Writing – Review and Editing, Visualization, Supervision
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING
Approach and Avoidance Patterns in Reward Learning Across Domains: An Initial Examination of the Social Iowa Gambling Task
Julia A. C. Case, M.A., and Thomas M. Olino, Ph.D.
Temple University, Department of Psychology 1701 North 13th Street 6th Floor Weiss Hall Philadelphia, PA 191221
Corresponding Author: Julia A. C. Case, M.A.
[email protected], (267)585-4213
Manuscript Total Word Count: 7,206 Abstract Total Word Count: 180
1
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING Abstract The current study examines learning patterns in response to both monetary and social incentives through both approach and avoidance behaviors using modified versions of the Iowa Gambling Task. Specifically, we investigated learning in response to both positive and negative feedback in a sample of 191 undergraduate students. The social task was a novel paradigm, and social feedback were images of faces displaying positive and negative emotions. We examined internal validity of the tasks through modeling changes in approach and avoidance. We also explored associations between approach and avoidance learning and individual differences in anxiety and social anxiety, depression and well being, general anhedonia and social closeness, and fun-seeking, using multilevel models (MLMs). Results showed that both the monetary and social tasks demonstrated learning as shown by decreases in plays on disadvantageous decks across the task. Additionally, we found that overall task performance on the monetary task was associated with fun-seeking and overall task performance on the social task was associated with fun-seeking and depressive symptoms. Initial findings suggest promise for the novel task in the examination of social avoidance learning.
Keywords: social, Iowa Gambling Task, approach and avoidance, learning, emotions
2
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING Approach and Avoidance Patterns in Reward Learning Across Domains: An Initial Examination of the Social Iowa Gambling Task The Positive Valence System (PVS) of the Research Domain Criteria (RDoC) framework (Insel et al., 2010) describes responses to positive motivational situations or contexts. Specific domains within the PVS include reward responsiveness, or hedonic responses during consummation of rewards, reward valuation, or evaluation of the satiety of rewards, and reward learning, or the linking of information about stimuli and contexts with positive outcomes (National Institute of Mental Health, 2011). Notably, although rewards come from multiple domains including money, food, sex, and social interactions (Lehner, Balsters, Herger, Hare, & Wenderoth, 2017), much research on the PVS and on reward learning specifically has frequently considered monetary rewards. As such, this study extends the extant work on the PVS by examining reward learning in both monetary and social contexts. Additionally, this study examines both approach and avoidance reward learning behaviors, or movement towards appetitive or away from punishing outcomes, respectively. The Iowa Gambling Task (IGT; Bechara, Damasio, Damasio, & Anderson, 1994) is a reward learning task relying on monetary feedback. In the original task, participants draw cards from four separate decks that lead to winning or losing money. Two of the decks have smaller immediate rewards, but result in greater net gains (classified as advantageous, or “good” decks), and two decks are associated with larger immediate rewards, but result in greater net losses (classified as disadvantageous, or “bad” decks; Bechara et al., 1994). Results from the initial task showed that performance was driven entirely by participant exploratory behavior and could not distinguish between greater attraction to the good and/or greater aversion to the bad decks. Peters and Slovic (2000) modified the original task to present each deck to the participant equally often,
3
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING and provided participants with the opportunity to play or pass from the presented deck. This play or pass modification permitted for the independent quantification of approach (i.e., playing on the good decks over the course of the task) as well as of avoidance behaviors (i.e., not playing on the bad decks over the course of the task; Cauffman et al., 2010). Associations between performance on the original monetary IGT and multiple dimensions of individual differences in clinical constructs, including depressive symptoms, reward sensitivity, anxiety symptoms, and impulsivity, have previously been examined. Studies have found that depressive symptoms are associated with poorer learning on the task (Cella, Dymond, & Cooper, 2010; McGovern, Alexopoulos, Yuen, Morimoto, & Gunning, 2014; Moniz, Jesus, Gonçalves, Pacheco, & Viseu, 2016; Must et al., 2006; Smoski et al., 2008) and one study found that anhedonia, or the inability to experience pleasure, was specifically associated with slower reward learning on the IGT (Must, Horvath, Nemeth, & Janka, 2013). In contrast, anxiety symptoms have been shown to be associated with faster reward learning throughout the task (F. Zhang, Xiao, & Gu, 2017; L. Zhang, Wang, Zhu, Yu, & Chen, 2015). Finally, the performance on the IGT has also been found to be associated with impulsivity, where children and adolescents high in impulsivity make more disadvantageous choices across the task blocks (Bubier & Drabick, 2008; Burdick, Roy, & Raver, 2013). Similar to learning behaviors using monetary incentives, learning in response to social feedback also can be divided into approach and avoidance behaviors. In this case, social approach and avoidance behaviors have been shown to be directly related to the emotional valence of the social stimuli (Roelofs, Elzinga, & Rotteveel, 2005), as facial expressions are a means of communicating social information, including the emotional state of the poser and the intentions or action demands to the perceiver (Rolls, 2000; Horstmann, Villringer, & Neumann,
4
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING 2012). Specifically, appetitive or positive social stimuli typically elicit approach behaviors from the perceiver, and aversive or negative social stimuli typically elicit avoidance behaviors from the perceiver (Chen & Bargh, 1999). While a number of tasks exist to assess social reward responsiveness and reward valuation via neural response (e.g., Jarcho et al., 2013; Silk et al., 2012; Williams, Yeager, Cheung, & Choi, 2012), few studies have attempted to assess learning in response to affective stimuli using a behavioral task. One exception to this is a study conducted by Pittig et al., where researchers used a modified version of the IGT to examine avoidant decision-making processes in social anxiety disorder (Pittig, Pawlikowski, Craske, & Alpers, 2014). Similar to the original IGT, participants selected from four decks of cards, two of which were advantageous, and two of which were disadvantageous. However, dissimilar from the original task, upon selection of each deck, participants were shown a picture of a facial expression. Following the selection of advantageous decks, participants viewed an angry facial expression, and following the selection of disadvantageous decks, participants viewed a happy facial expression; this mismatch allowed for a decision conflict between the approach of reward and the avoidance of fear-relevant angry faces. Results indicated that socially avoidant individuals were more likely to show avoidance of these angry faces under uncertainty. Beyond the task described above, there are currently limited behavioral tasks assessing both approach and avoidance learning in response to social stimuli. In another study, Kringelbach and Rolls (2003) used a reversal task to examine individuals’ ability to change behavior in response to social stimuli, finding that individuals exhibit greater blood-oxygen level-dependent responses and faster reversal following angry faces than neutral faces. Additionally, Marsh, Ambady, and Kleck (2005) had participants either push or pull a lever in
5
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING response to affective facial expressions and found that participants avoided viewing angry expressions. Other studies have utilized similar lever pull tasks to examine approach and avoidance of facial expressions in relation to mood induction (Vrijsen, van Oostrom, Speckens, Becker, & Rinck, 2013), comorbid symptomatology such as social anxiety (Heuer, Rinck, & Becker, 2007; Ly & Roelofs, 2009; Roelofs et al., 2010), and personality variables (Louise von Borries et al., 2012). Findings from these studies have converged to show approach behaviors towards positive valence emotions such as happiness, and avoidance behaviors of negative valence emotions such as anger. These previous studies, however, have not considered several important factors related to approach and avoidance behaviors. First, similar to the studies examining monetary reward learning, the studies assessing general approach and avoidance behavioral responses to social stimuli have examined these responses in relation to a narrow set of clinical constructs. Although prior studies have examined social approach and avoidance behaviors in relation to social anxiety symptoms (Heuer et al., 2007; Ly & Roelofs, 2009; Roelofs et al., 2010), few have examined relationships with related constructs, such as depressive symptoms, impulsivity, or social anhedonia, the lack of pleasure obtained from expressing feelings and interacting with people (Der-Avakian & Markou, 2012). Second, most studies examining responses to social stimuli have only examined approach and avoidance in relation to happiness and anger. Thus, other relevant negative valence emotions, such as sadness, disgust, and fear have been unexamined. Third, although the studies examining responses to social stimuli have examined general approach and avoidance tendencies, almost none of these studies have examined learning pattern over time. Although Pittig et al. (2014) examined learning patterns, this learning was specific approaching monetary rewards versus avoiding negative social stimuli. Because of the
6
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING concurrent use of monetary and social feedback in this task, the individual examination of social learning patterns based on this task design was not possible. Given these limitations of the previous literature, we propose a modified Iowa Gambling Task utilizing positive and negative social feedback to examine social learning patterns of approach and avoidance behaviors, or the Social Iowa Gambling Task (S-IGT). Current Study This study examines learning in response to both monetary and social feedback, demonstrated by approach toward positive and avoidance of negative feedback. First, we examined the internal validity of the IGT and S-IGT as behavioral tasks. Our a priori goal was to test whether there were significant block X deck type interaction for each of the behavioral tasks. We hypothesized that learning on the IGT and S-IGT would be demonstrated by increases in plays on good decks and decreases in plays on bad decks across the task. Second, we explored associations between IGT and S-IGT learning and individual differences in clinical constructs. We explored whether individual differences in approach- related clinical constructs (e.g., wellbeing, impulsivity, depressive symptoms, general anhedonia, and social anhedonia) would be associated with greater approach of positive feedback whereas avoidance-related clinical constructs (e.g., anxiety symptoms) would be associated with greater learning of avoidance of negative feedback. Finally, we explored whether social-specific clinical constructs (e.g., social closeness and social anxiety symptoms) would be associated with approach and avoidance learning for the S-IGT, respectively, but not the IGT. Methods Participants Participants were 191 undergraduate students from a metropolitan university in the
7
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING northeastern United States. This sample size was selected to ensure adequate power to examine changes in task performance between approach and avoidance decks. Based on a priori power analyses, we had substantial power (> .90) to detect moderate sized (d = .30) interactions between change and time on our behavioral measures. Participants ranged in age from 18 to 46 years (M=21.37; SD=3.78) and were predominantly female (n = 134; 71%). Over half of the sample identified as White or Caucasian (52%, n = 100), 23% (n = 43) of participants identified as Black or African American, 15% (n = 28) identified as Asian, 5% (n = 9) identified as bi- or multiracial, 4% (n = 7) identified as unlisted race, and 1% (n = 4) chose not to list their racial identity. Of these participants, 10% (n = 19) identified as Hispanic, 73% (n = 139) identified as non-Hispanic, and 17% (n = 33) chose not to list their ethnicity. Iowa Gambling Task The IGT was composed of three blocks, with each block consisting of 40 trials. Participants were presented with a card from one of four decks on the screen and were asked whether they wanted to “Play” or “Pass” on that card. If they selected to pass, they were presented with a card from another deck; if they selected to play, participants won money, lost money, or had no monetary gain or loss for that trial. Each of the decks varied on the frequency and rate of reward and punishment. Participants were not made aware of the contingencies during the task and needed to learn from feedback which decks were most profitable to draw. Social Iowa Gambling Task The S-IGT was a modified version of the IGT that used affective facial feedback rather than monetary feedback. Faces were drawn from the Facial expressions of emotion: Stimuli and Test (FEEST; Ekman & Friesen, 1971; Young, Perrett, Calder, Sprengelmeyer, & Ekman, 2002) that included Pictures of Facial Affect (POFA) differing in intensity. The intensity of each
8
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING emotion was used reflect small (percent-morph) versus large (percent-morph) social rewards and punishments, with proportions chosen to mirror the monetary gains and losses in the monetary IGT. Positive social feedback was represented by faces displaying happiness, and negative social feedback was represented by faces displaying one of four emotions: anger, sadness, fear, and disgust. Four separate tasks (a happiness-anger version, a happiness-sadness version, a happiness-fear version, and a happiness-disgust version) were then created utilizing each negative emotion separately as negative feedback. Figure 1 depicts faces displaying each emotion (positive and negative) from POFA that were utilized in the current study. Self-Report Measures Clinical constructs broadly assessing social closeness, pleasure, fun-seeking, well being, depressive symptoms, anxiety symptoms, and social anxiety symptoms were included in the current study. We relied on the composite approach to constructs detailed in Olino, McMakin, and Forbes (2018) in order to reduce the number of individual measures and analyses in the current study. Descriptives for measures assessing clinical constructs can be found in Table 1. Social closeness was measured using the Anticipatory and Consummatory Interpersonal Pleasure Scale (ACIPS; Gooding & Pflum, 2014), the Broad Autism Phenotype Questionnaire Social subscale (BAPQ-S; Hurley, Losh, Parlier, Reznick, & Piven, 2007), the social closeness subscale of the MPQ (MPQ-SC; Tellegen & Waller, 2008), and the Chapman Revised Social Anhedonia Scale (RSAS; Chapman, Chapman, & Raulin, 1976). The ACIPS is 17-item selfreport measure assessing one's ability to experience pleasure in the interpersonal domain. The ACIPS is scored on a 6-point Likert scale, ranging from “very false for me” to “very true for me.” Internal consistency for this scale in the current sample was good (α = .85). The BAPQ Social subscale is composed of 24 self-report items assessing individuals’ lack of interest in or
9
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING enjoyment of social interaction. Items are rated on a 6-point Likert scale rating how often these situations arise for respondents, ranging from “very rarely” to “very often.” Internal consistency for this scale in the current sample was good (α = .86). The MPQ-SC consists of 12 self-report items, with higher scores on these items suggesting that individuals describe themselves as sociable, liking to be with people, taking pleasure in and valuing close personal ties, and turning to others for comfort and help. Internal consistency in this sample for the MPQ-SC was good (α = .80). Finally, the RSAS is a 40 item self-report scale that assesses deficits in the ability to experience pleasure from other people, such as talking or exchanging expressions of feelings. Respondents may mark each item as “true” or “false” as much as it pertains to them and their opinions. Internal consistency for this scale in the current sample was good (α = .86). The average association between the ACIPS, BAPQ-S, the social closeness subscale of the MPQ, and RSAS was .61; given this moderate-strong association, these scales were z-scored and averaged together to create a composite social closeness score. Pleasure was measured using the Fawcett-Clark Pleasure Scale (FCPS; Fawcett, Clark, Scheftner, & Gibbons, 1983), the Chapman Revised Physical Anhedonia Scale (RPAS; Chapman, Chapman, & Raulin, 1976), the Snaith-Hamilton Pleasure Scale (SHAPS; Snaith et al., 1995), and the Temporal Experience of Pleasure Scale (TEPS; Gard, Gard, Kring, & John, 2006). The FCPS is a 36-item self-report questionnaire asking participants to rate imagined hedonic reactions to hypothetical pleasurable situations. Responses are made on a 5-point Likert scale ranging from “no pleasure at all” to “extreme and lasting pleasure.” Internal consistency for this scale in the current sample was excellent (α = .92). The RPAS is a 61-item self-report scale asking participants to respond true or false to statements about their typical feelings about normally pleasurable stimuli and activities. Internal consistency for this scale in the current
10
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING sample was good (α = .87). The SHAPS is a 14-item self-report questionnaire instructing participants whether they agree or disagree with statements of hedonic response in pleasurable situations (e.g., “I would enjoy my favorite television or radio program”). Responses are rated on a 4-point Likert scale ranging from “strongly disagree” to “strongly agree.” Internal consistency for this scale in the current sample was acceptable (α = .77). The TEPS is a measure composed of 18 self-reported items assessing both anticipatory pleasure (10 items) and consummatory pleasure (8 items). Respondents are asked about responses to specific pleasurable experiences that may have happened to them over the past week and are asked to rate how true or false these statements are to their own experience on a 6-point Likert scale. Internal consistency for the total scale in the current sample was excellent (α = .91). The average correlation between the FCPS, RPAS, SHAPS, and TEPS in the current study was .42; given this moderate average association, these scales were z-scored and averaged together to create a composite score. Fun-seeking was measured using the behavioral activation subscale of the BIS/BAS scales (BIS/BAS; Carver & White, 1994), and a short-form of the Barratt Impulsiveness Scale (SF-BIS; Barratt, 1959). The BIS/BAS is a self-report scale containing 24 statements that a participant may agree or disagree with; participants rate whether this statement is “very true for [me],” “somewhat true for [me],” “somewhat false for [me],” or “very false for [me].” The BAS subscale consists of 17 items corresponding to appetitive motives, in which the goal is to move toward something desired. Internal consistency for this scale in the current sample was good (α = .82). The SF-BIS contains 18 self-report questions designed to assess the behavioral construct of impulsivity. Respondents rate how often they engage in specific behaviors on a 4-point Likert scale ranging from “rarely/never” to “almost always.” Internal consistency for this scale in the current sample was acceptable (α = .78). The correlation between the BAS fun-seeking scale and
11
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING the SF-BIS was .34; given this moderate correlation, these scales were z-scored and averaged together to create a composite fun-seeking score. Well being was measured using the positive affect subscale of the Positive and Negative Affect Schedule (PANAS; Watson, Clark, & Tellegen, 1988), and the well-being subscale of the Multidimensional Personality Questionnaire (MPQ-WB; Tellegen & Waller, 2008). The PANAS consists of two 10-item mood subscales evaluating experiences of both positive and negative affect on that day on a 5-point Likert scale ranging from “very slightly or not at all” to “extremely.” The internal consistency in this sample for the positive affect subscale was excellent (α = .91). The MPQ-WB consists of 14 self-report items assessing an individual’s disposition to experience positive emotions, with higher scores on these items indicating greater well-being, including having a cheerful or happy disposition, feeling good about the self, seeing a bright future ahead, and enjoying the things that they are doing. Internal consistency in this sample for the MPQ-WB was good (α = .86). The correlation between the positive affect subscale of the PANAS and the well-being subscale of the MPQ in the current study was .48; given this moderate correlation, these scales were z-scored and averaged together to create a composite well being score. Depressive symptoms were measured using the Center for Epidemiologic Studies: Depression (CES-D; Radloff, 1977), and Patient-Reported Outcomes Measurement Information System Depression (PROMIS-D) scales. The CES-D scale is a 20 item self-report scale with Likert responses reflecting the frequency with which symptoms were experienced in the previous two week period. Internal consistency in this sample was excellent (α = .90). The PatientReported Outcomes Measurement Information System Depression scale (PROMIS-D) is an 8 item self-report scale asking participants how often they have experienced symptoms of
12
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING depression over the past 7 days on a 5-point Likert-scale ranging from “never” to “always.” Internal consistency in this sample was also excellent (α = .95). The correlation between the CES-D and the PROMIS-D in the current study was .77; given this strong correlation, these scales were z-scored and averaged to create a composite depressive symptoms score. Anxiety symptoms were measured using the behavioral inhibition subscale of the BIS/BAS scales (BIS/BAS; Carver & White, 1994), and the state subscale of the State-Trait Anxiety Inventory (STAI; Spielberger, Gorsuch, Lushene, Vagg, & Jacobs, 1983). The BIS/BAS is a self-report scale containing 24 statements that a participant may agree or disagree with; participants rate whether this statement is “very true for [me],” “somewhat true for [me],” “somewhat false for [me],” or “very false for [me].” The BIS subscale consists of seven items corresponding to motivation to avoid aversive outcomes. The STAI is a 20-item including items rated on a 4-point Likert scale from “not at all” to “very much so,” with higher scores indicating greater anxiety symptoms. Internal consistency for this scale in the current sample was excellent (α = .92). Internal consistency for this scale in the current sample was acceptable (α = .71). The correlation between the STAI and the behavioral inhibition subscale of the BIS/BAS in the current study was also .48; given this moderate correlation, these scales were z-scored and averaged together to create a composite anxiety symptoms score. Social anxiety symptoms were measured using the Liebowitz Social Anxiety Scale (LSAS; Liebowitz, 1987), and the Social Phobia Scale (SPS; Mattick & Clarke, 1998). The LSAS is a widely-used self-report measure of social anxiety symptoms comprised of questions evaluating 24 social situations within performance or social domains, and each rated for level of anxiety and avoidance. Internal consistency for the total scale was excellent (α = .95). The SPS is a 20 item self-report scale that measures the level of anxiety symptoms associated with scrutiny
13
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING or observation by other people while performing a task or action (e.g., working, eating, drinking, writing, or using a public restroom). Internal consistency for this scale in the current sample was excellent (α = .95). The correlation between the LSAS and the SPS in the current study was .75; given this strong correlation, these scales were z-scored and averaged together to create a composite anxiety symptoms score. Procedure Data come from a larger study. The full parent study procedures are detailed here. All study procedures were completed after receipt of IRB approval. Participants were recruited for the study using SONA Systems for research management. Informed consent was obtained from all participants. Following consent, participants first completed two behavioral tasks, then a series of self-report questionnaires, then two additional behavioral tasks, and finally a second series of self-report questionnaires. All study procedures took place in-person, and participants were compensated for their time with course credit. Behavioral tasks were administered using EPrime Stimulus Presentation Software (Schneider, Eschman, & Zuccolotto, 2002) and self-report questionnaires were administered electronically. Administration of the IGT and S-IGT tasks were spaced so that participants completed one of these tasks first and the other task last. Additionally, the order of presentation for the IGT and S-IGT was counterbalanced across participants. Further, two separate versions of each task were created, differing on which two decks were advantageous or “good,” and which two decks were disadvantageous or “bad;” these deck assignments were counter-balanced between tasks within each participant to address practice effects. Finally, participants were also randomized into one of the four negative emotion deck (fear, anger, disgust, and sadness) conditions for the S-IGT, with 48 participants per cell. Task data for one participant in the sadness condition were missing resulting from an issue with
14
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING task administration due to human error; as such, this cell contained data for 47 participants. Data Analysis Task Performance Scoring. Probabilities of plays on good decks and plays on bad decks were calculated for each block of the task, as well as for the overall task. Because proportion scores are bound between 0 and 1, these values were then transformed using an arcsine transformation (McDonald, 2014). Task Functioning. To clearly describe task function, paired samples T-tests examined changes in plays for good and bad decks between the first and last blocks for the IGT and S-IGT. Additionally, exploratory paired samples T-tests examined differences in changes in plays for each of the individual four negative emotions used in the S-IGT separately. Multi-Level Modeling. Multilevel models (MLMs) estimated overall task functioning and the relationships between individual differences in clinical constructs and approach and avoidance learning. For all models, arcsine transformations of play proportions were level one outcomes. To examine task functioning, deck type (good vs. bad) was the level one predictor, block number (first vs. last) was the level two predictor. We also included the deck type X block interaction to examine conditional changes in play. We also examined models for our exploratory analyses that included individual differences in clinical constructs as a level three predictor. For these exploratory analyses, models included the main effect terms, as well as all two- and three-way interactions between variables, inclusive of variables to evaluate task functioning. After examining each clinical construct as a predictor individually, a comprehensive model estimated associations for all clinical constructs in the model simultaneously excluding social anxiety symptoms, to account for shared variance between anxiety and social anxiety clinical symptom constructs. Models
15
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING were estimated for the IGT and S-IGT separately. Analyses were conducted using Mplus Version 8.3 (Muthén & Muthén, 1998-2017), using maximum likelihood estimation with robust standard errors. When significant interactions were found, we followed principles from Aiken and West (1991) to interpret the results, focusing on principles of centering as a method of estimating simple slope effects at specific values of moderators. Results Table 2 presents the inter-correlations among the construct composite scores and measures of overall task performance for both the IGT and S-IGT. Table S1 within the online supplement also presents inter-correlations for the individual self-report measures used to create the composites scores, respectively. Task Functioning Table 2 shows rates of plays on good and bad decks for the IGT and S-IGT. For good decks on the IGT, the proportion of plays in block 3 did not significantly differ from the proportion of plays in block 1. However, for bad decks, the proportion of plays in block 3 was significantly less than the proportion of plays in block 1. Similarly, for the good decks on the SIGT, the proportion of plays in block 3 did not significantly differ from the proportion of plays in block 1. However, for bad decks, the proportion of plays in block 3 was significantly lower than the proportion of plays in block 1. Additionally, we examined associations between performance on the IGT and the S-IGT, with performance estimated as the arcsine transformed values of the change scores of plays between the first and last blocks of the task. We did find significant associations for performance between the tasks on bad decks, but not on good decks. Finally, when examining bad decks by emotion type, the proportion of plays in block 3 was significantly less than the probabilities of plays in block 1 for anger, sadness, fear, and disgust. Notably,
16
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING across both tasks, the reduction on plays on bad decks was similar. Thus, we focused the remainder of analyses on the combined negative affective emotions. Multilevel Models IGT. Overall task performance of the IGT was examined by estimating the main effects of and interactions between deck type and block number on plays. In this model, we found main effects of deck type (b = .15, SE = .012, t = 12.87, p < .001) and block number (b = -.07, SE = .013, t = -5.65, p < .001), but these main effects were qualified by the interaction between deck type and block number (b = .17, SE = .021, t = 8.21, p < .001). Additionally, post-hoc tests showed that the reductions in plays over time was significant for bad (b = -.20, SE = .027, t = 7.39, p < .001), but not for good (b = .03, SE = .028, t = 1.16, p = .24) decks. Further exploratory MLMs estimated relationships between individual differences in each clinical construct and proportion of plays over time for the IGT (Table 3). The overall task functioning was consistent regardless of the clinical constructs included in the models. Thus, for a simplified presentation, we focus on the terms including each clinical construct separately (though full model results are available in the online supplement). For individual models, the deck type X fun-seeking, block number X fun-seeking, and deck type X block number X funseeking interactions were all significant. No other clinical constructs were associated with deck type, block number, or the interaction between deck type and block number. As this three-way interaction was significant, we conducted post-hoc simple slopes to identify at what levels of fun-seeking the two-way interactions were significant. We found that the deck type X block number interactions were significant at low (b = .31, SE = .05, t = 6.15, p < .001), moderate (b = .23, SE = .034, t = 6.83, p < .001), and high (b = .16, SE = .05, t = 3.07, p = .002) levels of funseeking. These patterns are shown in Figure 2 and indicate that at as fun-seeking increased, the
17
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING difference in proportion of plays on good and bad decks diminished in the later block. Finally, we also found that the block number X well-being interaction was significant. Post-hoc simple slopes showed that at low (b = -.13, SE = .03, t = -4.18, p < .001) and moderate (b = -.08, SE = .02, t = -3.91, p < .001) levels of well-being, there were greater reductions in the amount of plays between blocks. However, this effect was not found for high (b = -.04, SE = .03, t = -1.38, p = .17) levels of well-being. Finally, in the comprehensive multivariate multilevel model including all clinical constructs (presented in the online supplement Table S3), we did not find any significant threeway interactions. However, we found significant a significant block number X well-being (b = .10, SE = .04, t = 2.43, p = .015) interaction. Results of post-hoc analyses were consistent with those from the univariate model. S-IGT. Overall task performance of the S-IGT was examined by estimating the main effects of and interactions between deck type and block number on plays. In this model, we found a main effect of block number (b = -.07, SE = .013, t = -5.65, p < .001), but not deck type (b = .15, SE = .012, t = 12.87, p < .001). However, the interaction between deck type and block number (b = .06, SE = .019, t = 3.26, p = .001) was significant. Post-hoc tests showed that the reductions in plays over time was significant for bad (b = -.13, SE = .032, t = -3.99, p < .001), but not for good (b = -.01, SE = .029, t = -.27, p = .79) decks. Further MLMs estimated relationships between individual differences in each clinical construct and proportion of plays over time for the S-IGT (Table 4) following the same modeling procedures as were used for the IGT. MLMs estimated relationships between each clinical construct and proportion of plays over time for the S-IGT. For individual models, the deck type X block number X fun-seeking and the deck type X block number X depressive symptoms
18
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING interactions were associated with proportion of plays. As the three-way interaction was significant, we conducted post-hoc simple slopes to identify at what levels of fun-seeking and depressive symptoms the two-way interactions were significant. The same patterns were found across both of the clinical constructs. We found that the deck type X block number interaction was significant at low (b = .20, SE = .046, t = 4.34, p < .001) and moderate (b = .20, SE = .033, t = 3.59, p < .001), but not high (b = .039, SE = .041, t = .93, p = .352) levels of fun-seeking. Similarly, the deck type X block number interaction was significant at low (b = .21, SE = .047, t = 4.42, p < .001) and moderate (b = .12, SE = .033, t = 3.57, p < .001), but not high (b = .028, SE = .039, t = .72, p = .473) levels of depressive symptoms. These patterns are shown in Figure 3 (panels a and b) and indicate that at as fun-seeking and depressive symptoms increased, the difference in proportion of plays on good and bad decks diminished in the later block. Finally, in the comprehensive multivariate multilevel model including all clinical constructs (available in the online supplement Table S4), the deck type X block number X depressive symptoms interaction and the deck type X block number X fun-seeking interaction continued to be associated with proportion of plays. The pattern of post-hoc simple slopes for the two-way interactions was consistent across both fun-seeking and depressive symptoms, such that as fun-seeking and depressive symptoms increased, the difference in proportion of plays on good and bad decks diminished in the later block. Additionally, the deck type X block number X social closeness interaction was now also significantly associated with the proportion of plays (b = -.134, SE = .048, t = -2.78, p = .005). Post-hoc simple slopes showed that the deck type X block number interaction was significant at low (b = .229, SE = .049, t = 4.67, p < .001) and moderate (b = .117, SE = .033, t = 3.58, p < .001), but not high (b = .005, SE = .055, t = .083, p = .93) levels of social closeness. Finally, the deck type X well-being interaction was also
19
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING associated with plays (b = .078, SE = .037, t = 2.11, p = .035). Post-hoc simple slopes showed that although the associations for good (b = .031, SE = .028, t = 1.10, p = .271) and bad (b = .047, SE = .025, t = -1.90, p = .057) decks were significantly different from each other, neither slope was significantly different from 0. Discussion In this study, we investigated learning patterns in response to both monetary and social rewards, indexed by approach toward and avoidance of positive and negative feedback, respectively. We used an adapted behavioral task, the S-IGT, to assess social learning patterns. We found evidence of individuals learning to avoid negative feedback for both the monetary and social tasks. However, we did not find evidence of individuals learning to increase positive feedback on either the monetary or social task. In addition to these overall learning patterns, we found evidence of associations for specific clinical constructs, including fun-seeking and both monetary and social learning, depressive symptoms and social learning, and social closeness and social learning. When examining task functioning for both the monetary and social task, we did not find increases in approach from the first to last blocks. The lack of increases in plays on good decks may be due to participants initially exploring the decks at high frequencies to obtain feedback. As a result, rates of playing were high at the beginning of each task and continued at a high rate throughout each task for good decks. One way to adjust the tasks to reduce this ceiling effect could be to present monetary losses or negative social feedback on early trials of the first block for good decks, leading to fewer plays on good decks that would increase over time. Notably, however, when examining the functioning of each task, we found avoidance of negative feedback on both the monetary and social versions of the task. This finding of a significant
20
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING reduction in plays on bad decks but no difference in plays on good decks is consistent with the literature suggesting that responses are faster and stronger to proximate negative events than to positive events (Cacioppo & Gardner, 1999). In this case, the large losses contained in the bad decks may be particularly salient for participants, contributing to greater learning on those decks specifically. Importantly, these results suggest the validity of the S-IGT as a task used to measure social learning patterns, specifically represented by avoidance of negative social feedback. Our analyses also examined associations between individual differences in clinical constructs and performances on the IGT and S-IGT. For the IGT, we found significant three-way interactions between deck type, block number, and fun-seeking, such that as fun-seeking increased, the difference in proportion of plays on good and bad decks diminished over time. Specifically, individuals higher in fun-seeking played less from good decks and more from bad decks over time. These findings are consistent with the previous literature examining associations between impulsivity—contained within the fun-seeking construct in the current study—and performance on the IGT. Specifically, individuals rated higher on impulsivity choose more disadvantageously across IGT blocks, continuing to play on the bad decks over time and valuing the high rewards contained within those decks, while attending less to the high losses contained within them. This pattern has been shown across populations, including in children and adolescents (Aklin, Lejuez, Zvolensky, Kahler, & Gwadz, 2005), as well as in non-clinical (Sweitzer, Allen, & Kaut, 2008) and clinical (Bechara, 2003; Goudriaan, Oosterlaan, de Beurs, & van den Brink, 2006; Lawrence et al., 2006; Malloy-Diniz, Fuentes, Leite, Correa, & Bechara, 2007; Stout, Rock, Campbell, Busemeyer, & Finn, 2005) adult samples. Unlike previous research finding significant effects for depression on the monetary IGT, we did not find associations between depressive symptoms and IGT performance. We speculate
21
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING that our results diverged from previous research due to the difference in the design of our task, which led us to use a distinctive analytic approach. Specifically, whereas previous studies have typically calculated their measure of learning on the task by subtracting the number of disadvantageous deck choices from the number of advantageous deck choices, the play or pass modification utilized in the current study allows for the independent quantification of approach and avoidance behaviors. Thus, we expect that in our examination of approach and avoidance independently from each other, we no longer found significant effects for depression. For the S-IGT, we found significant three-way interactions between deck type, block number, and fun-seeking similar to the IGT, such that as fun-seeking increased, the difference in proportion of plays on good and bad decks diminished in the later block. Again, these individuals higher in fun-seeking played less from good decks and more from bad decks over time. This finding is in line with the aforementioned literature suggesting that individuals high in impulsivity tend to value and favor decks with larger monetary rewards when completing the IGT, and learn to avoid the decks that contain smaller monetary rewards and losses over time. Based on our findings, this effect seems to generalize across reward domains for fun-seeking or impulsivity. Additionally, although we did not find significant associations for depressive symptoms on the monetary IGT, we did find significant three-way interactions between deck type, block number, and depressive symptoms for the S-IGT, such that as depressive symptoms increased, the difference in proportion of plays on good and bad decks diminished in the later block. Notably, this finding seemed to be driven by performance on good decks specifically, where individuals higher in depressive symptoms played less from good decks over time. A similar pattern of behavior was previously demonstrated in depressed adolescents on the IGT (Han et al., 2012), where individuals high in depressive symptoms selected fewer cards from the
22
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING good decks over time. These findings suggest that depressive symptoms may impair reinforcement processing, leading to difficulty integrating feedback in guiding future behavior. Indeed, Must et al. (2013) found similar results with depressed patients focusing on the immediate outcome of the task and preferred the decks that contained higher short-term rewards than the decks that contained long-term rewards or gains. Additionally, in the comprehensive multivariate multilevel model examining learning on the S-IGT and including all clinical constructs, a significant three-way interaction emerged for deck type, block number, and social closeness. Specifically, at lower levels of social closeness, there was a greater difference in plays on good and bad decks during the later blocks, or individuals low in social closeness learned to select fewer cards from the bad decks over time. In support of this finding, previous research has shown that individuals who are low in social closeness are hyper-reactive to negative faces, and that high social anhedonia is associated with exaggerated reactivity to negative social stimuli (Günther et al., 2017). This suggests that individuals low in social closeness would exhibit similar heightened responsiveness to negative social stimuli, as found for results of the current study. However, as this association emerged only in the presence of covariates, there are outstanding questions about the replicability and meaning of this finding. Notably, findings from the current study assessing approach and avoidance learning behaviors suggest that individuals higher in fun-seeking are less sensitive to negative feedback, as they continued to play from disadvantageous decks over time at higher rates than individuals lower in fun-seeking. Thus, interventions with individuals displaying high levels of impulsivity can target these individuals bolstering their focus on consequences or negative outcomes, in order for them to successfully adjust future behavior. Additionally, individuals higher on
23
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING depressive symptoms displayed a behavioral tendency to engage less in the task over time, regardless of whether the outcome was good or bad, as they played less from both advantageous and disadvantageous decks over time than individuals lower in depressive symptoms. However, this finding was specific to the social domain, rather than generalizing across both domains. This finding is in line with other work portraying unique associations between depression and salience for social feedback, even when controlling for responses to monetary feedback (Ait Oumeziane, Jones, & Foti, 2019), and underscores the importance of assessing approach and avoidance learning across multiple domains, particularly when considering these behaviors in relation to depression. Finally, these results implicate the importance of engagement with behavioral activation in the treatment of individuals displaying high levels of depressive symptoms, so that these individuals may learn to maintain engagement with activities while attending to positive outcomes. This may be particularly effective for behavioral activities of a social nature. There were several notable limitations to this work. First, the images that were used in the S-IGT task were black and white. It is possible that using alternative stimuli in color would elicit greater rates of learning on the task. Additionally, some research has shown that tasks utilizing dynamic feedback as opposed to static feedback, particularly when examining social behaviors, are more salient for participants (Rymarczyk, Żurawski, Jankowiak-Siuda, & Szatkowska, 2016). Finally, future tasks examining social approach and avoidance learning behaviors could employ feedback linked to real world social outcomes (e.g., Blake et al., 2015). Second, our sample included undergraduate students, which may limit the generalizability of our study across other stages of development. Third, multiple tests were conducted in the paper. Although we limited the number of tests we conducted by relying on clinical construct composites and aggregate negative affect stimuli in the S-IGT, we did not perform any
24
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING correction for multiple comparisons, given that many of these tests were exploratory and considered hypothesis-generating for future uses of the task. Thus, we are tentative in our results concerning associations between task performance and individual differences in clinical constructs; future work should attempt to replicate these associations. Fourth, our analyses did not include a direct comparison across the monetary and social tasks. In the current analytic framework, we were unable to fully integrate these associations within our models. Models were fit such that task parameters (i.e., deck type and block) were estimated as fixed effects. As such, we could not specify models with both tasks modeled simultaneously to estimate associations between task performance. Future examination of direct comparisons across the IGT and S-IGT are warranted. Finally, the current analyses compared differences in performance across blocks. However, it is possible that learning could be better assessed through the examination of changes in plays and passes on a trial level rather than a block level. Future examination of learning on a trial level may be warranted. Despite these limitations, our study also had several notable strengths. Specifically, we examined learning for approach and avoidance behaviors across both social and monetary domains, which has not previously been done within the literature. Additionally, our social task included multiple emotions for negative feedback (anger, fear, sadness, and disgust), which has not commonly been done in previous studies examining responses to social or affective stimuli. However, we did not find that learning to avoid negative feedback differed across types of negative affect. Our initial findings suggest promise for the S-IGT as a measure of social avoidance learning, in contrast to the initial design motivation. Future directions include adaptations to the task stimuli and task design, as well as examination of learning using other modeling designs, such as examining changes in performance on a trial-by-trial basis.
25
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING References Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Thousand Oaks, CA: Sage Publications, Inc. Ait Oumeziane, B., Jones, O., & Foti, D. (2019). Neural Sensitivity to Social and Monetary Reward in Depression: Clarifying General and Domain-Specific Deficits. Frontiers in Behavioral Neuroscience, 13. https://doi.org/10.3389/fnbeh.2019.00199 Aklin, W. M., Lejuez, C. W., Zvolensky, M. J., Kahler, C. W., & Gwadz, M. (2005). Evaluation of behavioral measures of risk taking propensity with inner city adolescents. Behaviour Research and Therapy, 43(2), 215–228. https://doi.org/10.1016/j.brat.2003.12.007 Barratt, E. S. (1959). Anxiety and impulsiveness related to psychomotor efficiency. Perceptual and Motor Skills, 9, 191–198. Bechara, A. (2003). Risky business: Emotion, decision-making, and addiction. Journal of Gambling Studies, 19(1), 23–51. Bechara, A., & Damasio, H. (2002). Decision-making and addiction (part I): Impaired activation of somatic states in substance dependent individuals when pondering decisions with negative future consequences. Neuropsychologia, 40(10), 1675–1689. Bechara, A., Damasio, H., Damasio, A., & Anderson, S. (1994). Insensitivity to future consequences following damage to human prefrontal cortex. Cognition, 50(1–3), 7–15. Blake, P. R., McAuliffe, K., Corbit, J., Callaghan, T. C., Barry, O., Bowie, A., … Warneken, F. (2015). The ontogeny of fairness in seven societies. Nature, 528(7581), 258–261. https://doi.org/10.1038/nature15703 Bubier, J. L., & Drabick, D. A. G. (2008). Affective decision-making and externalizing behaviors: The role of autonomic activity. Journal of Abnormal Child Psychology, 36(6), 941–953. https://doi.org/10.1007/s10802-008-9225-9 Burdick, J. D., Roy, A. L., & Raver, C. C. (2013). Evaluating the Iowa Gambling Task as a Direct Assessment of Impulsivity with Low-Income Children. Personality and Individual Differences, 55(7), 771–776. https://doi.org/10.1016/j.paid.2013.06.009 Cacioppo, J. T., & Gardner, W. L. (1999). Emotion. Annual Review of Psychology, 50, 191–214. https://doi.org/10.1146/annurev.psych.50.1.191 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. https://doi.org/10.1037/00223514.67.2.319
26
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING Cauffman, E., Shulman, E. P., Steinberg, L., Claus, E., Banich, M. T., Graham, S., & Woolard, J. (2010). Age differences in affective decision making as indexed by performance on the Iowa Gambling Task. Developmental Psychology, 46(1), 193–207. https://doi.org/10.1037/a0016128 Cella, M., Dymond, S., & Cooper, A. (2010). Impaired flexible decision-making in major depressive disorder. Journal of Affective Disorders, 124(1–2), 207–210. https://doi.org/10.1016/j.jad.2009.11.013 Chapman, L. J., Chapman, J. P., & Raulin, M. L. (1976). Scales for physical and social anhedonia. Journal of Abnormal Psychology, 85(4), 374–382. https://doi.org/10.1037/0021-843X.85.4.374 Chen, M., & Bargh, J. A. (1999). Consequences of automatic evaluation: Immediate behavioral predispositions to approach or avoid the stimulus. Personality and Social Psychology Bulletin, 25(2), 215–224. Der-Avakian, A., & Markou, A. (2012). The neurobiology of anhedonia and other reward-related deficits. Trends in Neurosciences, 35(1), 68–77. https://doi.org/10.1016/j.tins.2011.11.005 Ekman, P., & Friesen, W. V. (1971). Constants across cultures in the face and emotion. Journal of Personality and Social Psychology, 17(2), 124–129. https://doi.org/10.1037/h0030377 Fawcett, J., Clark, D. C., Scheftner, W. A., & Gibbons, R. D. (1983). Assessing anhedonia in psychiatric patients. Archives of General Psychiatry, 40(1), 79–84. Gard, D. E., Gard, M. G., Kring, A. M., & John, O. P. (2006). Anticipatory and consummatory components of the experience of pleasure: A scale development study. Journal of Research in Personality, 40(6), 1086–1102. https://doi.org/10.1016/j.jrp.2005.11.001 Gooding, D. C., & Pflum, M. J. (2014). The assessment of interpersonal pleasure: Introduction of the Anticipatory and Consummatory Interpersonal Pleasure Scale (ACIPS) and preliminary findings. Psychiatry Research, 215(1), 237–243. https://doi.org/10.1016/j.psychres.2013.10.012 Goudriaan, A. E., Oosterlaan, J., de Beurs, E., & van den Brink, W. (2006). Psychophysiological determinants and concomitants of deficient decision making in pathological gamblers. Drug and Alcohol Dependence, 84(3), 231–239. https://doi.org/10.1016/j.drugalcdep.2006.02.007 Günther, V., Zimmer, J., Kersting, A., Hoffmann, K.-T., Lobsien, D., & Suslow, T. (2017). Automatic processing of emotional facial expressions as a function of social anhedonia. Psychiatry Research. Neuroimaging, 270, 46–53. https://doi.org/10.1016/j.pscychresns.2017.10.002
27
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING 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. https://doi.org/10.1016/j.adolescence.2011.06.009 Heuer, K., Rinck, M., & Becker, E. S. (2007). Avoidance of emotional facial expressions in social anxiety: The Approach-Avoidance Task. Behaviour Research and Therapy, 45(12), 2990–3001. https://doi.org/10.1016/j.brat.2007.08.010 Horstmann, A., Villringer, A., & Neumann, J. (2012). Iowa Gambling Task: There is More to Consider than Long-Term Outcome. Using a Linear Equation Model to Disentangle the Impact of Outcome and Frequency of Gains and Losses. Frontiers in Neuroscience, 6. https://doi.org/10.3389/fnins.2012.00061 Hurley, R. S. E., Losh, M., Parlier, M., Reznick, J. S., & Piven, J. (2007). The broad autism phenotype questionnaire. Journal of Autism and Developmental Disorders, 37(9), 1679– 1690. https://doi.org/10.1007/s10803-006-0299-3 Insel, T., Cuthbert, B., Garvey, M., Heinssen, R., Pine, D. S., Quinn, K., … Wang, P. (2010). Research Domain Criteria (RDoC): Toward a New Classification Framework for Research on Mental Disorders. American Journal of Psychiatry, 167(7), 748–751. https://doi.org/10.1176/appi.ajp.2010.09091379 Jarcho, J. M., Leibenluft, E., Walker, O. L., Fox, N. A., Pine, D. S., & Nelson, E. E. (2013). Neuroimaging studies of pediatric social anxiety: Paradigms, pitfalls and a new direction for investigating the neural mechanisms. Biology of Mood & Anxiety Disorders, 3, 14. https://doi.org/10.1186/2045-5380-3-14 Kringelbach, M. L., & Rolls, E. T. (2003). Neural correlates of rapid reversal learning in a simple model of human social interaction. NeuroImage, 20(2), 1371–1383. https://doi.org/10.1016/S1053-8119(03)00393-8 Lawrence, N. S., Wooderson, S., Mataix-Cols, D., David, R., Speckens, A., & Phillips, M. L. (2006). Decision making and set shifting impairments are associated with distinct symptom dimensions in obsessive-compulsive disorder. Neuropsychology, 20(4), 409– 419. https://doi.org/10.1037/0894-4105.20.4.409 Lehner, R., Balsters, J. H., Herger, A., Hare, T. A., & Wenderoth, N. (2017). Monetary, Food, and Social Rewards Induce Similar Pavlovian-to-Instrumental Transfer Effects. Frontiers in Behavioral Neuroscience, 10. https://doi.org/10.3389/fnbeh.2016.00247 Liebowitz, M. R. (1987). Social phobia. Modern Problems of Pharmacopsychiatry, 22, 141–173. Louise von Borries, A. K., Volman, I., de Bruijn, E. R. A., Bulten, B. H., Verkes, R. J., & Roelofs, K. (2012). Psychopaths lack the automatic avoidance of social threat: Relation
28
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING to instrumental aggression. Psychiatry Research, 200(2–3), 761–766. https://doi.org/10.1016/j.psychres.2012.06.026 Ly, V., & Roelofs, K. (2009). Social anxiety and cognitive expectancy of aversive outcome in avoidance conditioning. Behaviour Research and Therapy, 47(10), 840–847. https://doi.org/10.1016/j.brat.2009.06.015 Malloy-Diniz, L., Fuentes, D., Leite, W. B., Correa, H., & Bechara, A. (2007). Impulsive behavior in adults with attention deficit/ hyperactivity disorder: Characterization of attentional, motor and cognitive impulsiveness. Journal of the International Neuropsychological Society: JINS, 13(4), 693–698. https://doi.org/10.1017/S1355617707070889 Marsh, A. A., Ambady, N., & Kleck, R. E. (2005). The effects of fear and anger facial expressions on approach- and avoidance-related behaviors. Emotion (Washington, D.C.), 5(1), 119–124. https://doi.org/10.1037/1528-3542.5.1.119 Mattick, R. P., & Clarke, J. C. (1998). Development and validation of measures of social phobia scrutiny fear and social interaction anxiety. Behaviour Research and Therapy, 36(4), 455–470. McDonald, J. H. (2014). Handbook of Biological Statistics (3rd ed.). Baltimore, MD: Sparky House Publishing. McGovern, A. R., Alexopoulos, G. S., Yuen, G. S., Morimoto, S. S., & Gunning, F. M. (2014). Reward-Related Decision Making in Older Adults: Relationship to Clinical Presentation of Depression. International Journal of Geriatric Psychiatry, 29(11), 1125–1131. https://doi.org/10.1002/gps.4200 Moniz, M., Jesus, S., Gonçalves, E., Pacheco, A., & Viseu, J. (2016). Decision-making in adult unipolar depressed patients and healthy subjects: Significant differences in Net Score and in non-traditional alternative measures. Neuropsychological Trends, (19), 7–15. https://doi.org/10.7358/neur-2016-019-moni Must, A., Horvath, S., Nemeth, V. L., & Janka, Z. (2013). The Iowa Gambling Task in depression – what have we learned about sub-optimal decision-making strategies? Frontiers in Psychology, 4. https://doi.org/10.3389/fpsyg.2013.00732 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(2–3), 209–215. https://doi.org/10.1016/j.jad.2005.12.005 Muthén, L. K., & Muthén, B. O. (2000). MPlus User’s Guide (Version Sixth Edition). Los Angeles, CA.
29
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING National Institute of Mental Health. (2011, June). Positive Valence Systems: Workshop Proceedings. Presented at the Research Domain Criteria (RDoC), Rockville, Maryland. Retrieved from https://www.nimh.nih.gov/research/research-funded-bynimh/rdoc/positive-valence-systems-workshop-proceedings.shtml Olino, T. M., McMakin, D. L., & Forbes, E. E. (2018). Toward an Empirical Multidimensional Structure of Anhedonia, Reward Sensitivity, and Positive Emotionality: An Exploratory Factor Analytic Study. Assessment, 25(6), 679–690. https://doi.org/10.1177/1073191116680291 Peters, E., & Slovic, P. (2000). The Springs of Action: Affective and Analytical Information Processing in Choice. Personality and Social Psychology Bulletin, 26(12), 1465–1475. https://doi.org/10.1177/01461672002612002 Pittig, A., Pawlikowski, M., Craske, M. G., & Alpers, G. W. (2014). Avoidant decision making in social anxiety: The interaction of angry faces and emotional responses. Frontiers in Psychology, 5. https://doi.org/10.3389/fpsyg.2014.01050 Radloff, L. S. (1977). The CES-D Scale A Self-Report Depression Scale for Research in the General Population. Applied Psychological Measurement, 1(3), 385–401. https://doi.org/10.1177/014662167700100306 Roelofs, K., Elzinga, B. M., & Rotteveel, M. (2005). The effects of stress-induced cortisol responses on approach-avoidance behavior. Psychoneuroendocrinology, 30(7), 665–677. https://doi.org/10.1016/j.psyneuen.2005.02.008 Roelofs, K., Putman, P., Schouten, S., Lange, W.-G., Volman, I., & Rinck, M. (2010). Gaze direction differentially affects avoidance tendencies to happy and angry faces in socially anxious individuals. Behaviour Research and Therapy, 48(4), 290–294. https://doi.org/10.1016/j.brat.2009.11.008 Rolls, E. T. (2000). Précis of The brain and emotion. The Behavioral and Brain Sciences, 23(2), 177–191; discussion 192-233. Rymarczyk, K., Żurawski, Ł., Jankowiak-Siuda, K., & Szatkowska, I. (2016). Do Dynamic Compared to Static Facial Expressions of Happiness and Anger Reveal Enhanced Facial Mimicry? PLOS ONE, 11(7), e0158534. https://doi.org/10.1371/journal.pone.0158534 Schneider, W., Eschman, A., & Zuccolotto, A. (2002). E-prime software (Version 2.0). Pittsburgh, PA. Silk, J. S., Stroud, L. R., Siegle, G. J., Dahl, R. E., Lee, K. H., & Nelson, E. E. (2012). Peer acceptance and rejection through the eyes of youth: Pupillary, eyetracking and ecological data from the Chatroom Interact task. Social Cognitive and Affective Neuroscience, 7(1), 93–105. https://doi.org/10.1093/scan/nsr044
30
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING Smoski, M. J., Lynch, T. R., Rosenthal, M. Z., Cheavens, J. S., Chapman, A. L., & Krishnan, R. R. (2008). Decision-making and risk aversion among depressive adults. Journal of Behavior Therapy and Experimental Psychiatry, 39(4), 567–576. https://doi.org/10.1016/j.jbtep.2008.01.004 Snaith, R. P., Hamilton, M., Morley, S., Humayan, A., Hargreaves, D., & Trigwell, P. (1995). A scale for the assessment of hedonic tone the Snaith-Hamilton Pleasure Scale. The British Journal of Psychiatry: The Journal of Mental Science, 167(1), 99–103. Spielberger, C. D., Gorsuch, R. L., Lushene, R., Vagg, P. R., & Jacobs, G. A. (1983). Manual for the State-Trait Anxiety Inventory. Palo Alto, CA: Consulting Psychologists Press. Stout, J. C., Busemeyer, J. R., Lin, A., Grant, S. J., & Bonson, K. R. (2004). Cognitive modeling analysis of decision-making processes in cocaine abusers. Psychonomic Bulletin & Review, 11(4), 742–747. Stout, J. C., Rock, S. L., Campbell, M. C., Busemeyer, J. R., & Finn, P. R. (2005). Psychological processes underlying risky decisions in drug abusers. Psychology of Addictive Behaviors: Journal of the Society of Psychologists in Addictive Behaviors, 19(2), 148–157. https://doi.org/10.1037/0893-164X.19.2.148 Sweitzer, M. M., Allen, P. A., & Kaut, K. P. (2008). Relation of individual differences in impulsivity to nonclinical emotional decision making. Journal of the International Neuropsychological Society: JINS, 14(5), 878–882. https://doi.org/10.1017/S1355617708080934 Tellegen, A., & Waller, N. G. (2008). Exploring Personality Through Test Construction: Development of the Multidimensional Personality Questionnaire. In The SAGE Handbook of Personality Theory and Assessment: Volume 2—Personality Measurement and Testing (pp. 261–292). https://doi.org/10.4135/9781849200479 Vrijsen, J. N., van Oostrom, I., Speckens, A., Becker, E. S., & Rinck, M. (2013). Approach and Avoidance of Emotional Faces in Happy and Sad Mood. Cognitive Therapy and Research, 37(1), 1–6. https://doi.org/10.1007/s10608-012-9436-9 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. Williams, K. S., Yeager, D. S., Cheung, C. K. T., & Choi, W. (2012). Cyberball. Retrieved from https://cyberball.wikispaces.com Young, A. W., Perrett, D. I., Calder, A. J., Sprengelmeyer, R., & Ekman, P. (2002). Facial expressions of emotion: Stimuli and tests (FEEST). Retrieved from https://www.researchgate.net/publication/252068424_Facial_expressions_of_emotion_St imuli_and_tests_FEEST
31
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING Zhang, F., Xiao, L., & Gu, R. (2017). Does Gender Matter in the Relationship between Anxiety and Decision-Making? Frontiers in Psychology, 8, 2231. https://doi.org/10.3389/fpsyg.2017.02231 Zhang, L., Wang, K., Zhu, C., Yu, F., & Chen, X. (2015). Trait Anxiety Has Effect on Decision Making under Ambiguity but Not Decision Making under Risk. PloS One, 10(5), e0127189. https://doi.org/10.1371/journal.pone.0127189
32
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING Table 1. Means, standard deviations, and ranges of individual self-report measures used to create composite scales. Measure ACIPS BAPQ-S MPQ-SC RSAS Pleasure FCPS RPAS SHAPS TEPS Fun-Seeking BAS SF-BIS Well Being MPQ-WB PANAS-P Depressive Symptoms CESD PRMD Anxiety Symptoms STAI BIS Social Anxiety Symptoms LSAS SPS Social Closeness
Mean 60.72 67.02 7.10 11.69 138.50 14.36 0.90 81.05 35.52 38.34 9.21 27.64 17.39 14.18 38.27 16.07 101.66 41.69
SD 6.06 15.08 3.13 6.71 16.83 8.88 1.73 14.67 4.82 7.18 3.79 8.94 10.87 7.35 11.84 2.79 24.52 16.33
Range 32.0-68.0 29.0-105.0 0.0-12.0 0.0-30.0 81.0-180.0 1.0-42.7.0 0.0-10.0 35.0-108.0 22.0-44.0 21.0-63.0 0.0-14.0 10.0-50.0 0.0-49.0 8.0-40.0 20.0-74.0 9.0-20.0 50.0-158.0 20.0-89.0
ACIPS = Anticipatory and Consummatory Interpersonal Pleasure Scale; BAPQ-S = Broad Autism Phenotype Questionnaire Social subscale; BAS, BIS = behavioral activation/inhibition subscales of the BIS/BAS scales, respectively; CESD = Center for Epidemiologic Studies: Depression Scale; FCPS = Fawcett-Clark Pleasure Scale; SF-BIS = Barratt Impulsiveness Scale; LSAS = Liebowitz Social Anxiety Scale; MPQ-WB = Multidimensional Personality Questionnaire well-being subscale; MPQ-SC = Multidimensional Personality Questionnaire social closeness subscale; PANAS = Positive and Negative Affect Schedule positive affect subscale; PRMD = Patient-Reported Outcomes Measurement Information System Depression Scale; RPAS = Chapman Revised Physical Anhedonia Scale; RSAS = Chapman Revised Social Anhedonia Scale; SHPS = Snaith-Hamilton Pleasure Scale; SPS = Social Phobia Scale; STAI = state subscale of the State-Trait Anxiety Inventory; and TEPS = Temporal Experience of Pleasure Scale
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING Table 2. Inter-correlations amongst the construct composite scores and measures of task performance. For both the IGT and S-IGT, task performance is displayed in arcsine transformed values of the change scores of plays between the first and last blocks of the task.
IGT Bad S-IGT Good S-IGT Bad Social Closeness Pleasure Fun-Seeking Well Being Depressive Symptoms Anxiety Symptoms Social Anxiety Symptoms
Mean (SD) ---.0023 (.84) -.0015 (.75) 0 (.82) 0 (.86) .0041 (.94) .0049 (.81) -.0006 (.88)
IGT Good .22** -.01 -.08 -.05 -.01 .10 .06 .09 .06 .09
IGT Bad
S-IGT Good
S-IGT Bad
Social Closeness
.08 .16* .06 .01 .31** .20** .05 .09 .05
.39** -.01 -.10 -.17* -.01 -.13 -.07 -.04
.09 -.14 .04 -.01 .08 .00 .06
.56** .04 .56** -.36** -.30** -.41**
Note: *Significant at the p < .05 level. **Significant at the p < .01 level.
Pleasure
-.04 .47** -.24** -.28** -.27**
FunSeeking
Well Being
Depressive Symptoms
Anxiety Symptoms
.19** .27** .24** .12
-.47** -.44** -.39**
.64** .57**
.62**
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING Table 3. Measures of task performance for both the IGT and S-IGT, separated by deck type. Additionally, for the S-IGT, performance has been shown for each emotion type separately. Proportion of Plays Block 1 M (SD)
Arcsine Transformed M (SD)
Proportion of Plays Block 3 M (SD)
Arcsine Transformed M (SD)
T Statistic Block 1 vs. Block 3 t (df)
Effect Size (d)
IGT 2.48 (.54) -1.10 (188) -.08 .83 (.17) 2.44 (.54) .85 (.17) Good Decks 1.82 (.61) 8.37 (188)*** .61 .76 (.14) 2.18 (.38) .61 (.24) Bad Decks S-IGT 2.31 (.58) 1.88 (189) .14 .82 (.14) 2.33 (.44) .78 (.19) Good Decks 2.22 (.60) 5.57 (189)*** .40 .84 (.13) 2.43 (.43) .75 (.21) Bad Decks 2.12 (.61) 2.98 (50)** .42 Anger .83 (.14) 2.39 (.45) .72 (.21) 2.37 (.61) 2.32 (44)* .35 Sadness .88 (.11) 2.57 (.43) .81 (.22) 2.27 (.61) 2.34 (47)* .34 Fear .83 (.14) 2.37 (.43) .76 (.20) 2.19 (.56) 3.49 (45)*** .51 Disgust .84 (.12) 2.40 (.39) .73 (.19) Note: *Performance between blocks differed significantly at the p < .05 level. **Performance between blocks differed significantly at the p < .01 level. ***Performance between blocks differed significantly at the p < .001 level.
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING Table 4. Multilevel model results for clinical construct composite scores predicting task performance on the IGT.
Social Closeness Pleasure Fun-Seeking Well Being Depressive Symptoms Anxiety Symptoms Social Anxiety Symptoms
Intercept b (SE) -.01 (.02) -.01 (.02) -.02 (.02) -.02 (.02) .01 (.02) -.01 (.02) -.01 (.02)
Deck Type b (SE) .00 (.03) -.01 (.03) -.08 (.03)** -.04 (.03) -.04 (.02) -.03 (.03) .00 (.02)
Block Number b (SE) -.01 (.03) .01 (.03) .09 (.03)*** .05 (.03)* .03 (.02) .04 (.03) .03 (.03)
Deck x Block b (SE) -.01 (.02) .00 (.05) -.09 (.05)* -.04 (.04) -.00 (.04) -.04 (.05) .01 (.05)
Note: *Significant at the p < .05 level. **Significant at the p < .01 level. ***Significant at the p < .001 level. For the intercept column, the coefficients indicate the association between the intercept and the construct identified in the row header. For other table entries, coefficients in table refer to the interaction between the construct identified in the row header and the column header.
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING Table 5. Multilevel model results for clinical construct composite scores predicting task performance on the S-IGT.
Social Closeness Pleasure Fun-Seeking Well Being Depressive Symptoms Anxiety Symptoms Social Anxiety Symptoms
Intercept b (SE) -.02 (.02) -.01 (.02) -.01 (.02) .04 (.02) -.01 (.02) -.02 (.02) -.01 (.02)
Deck Type b (SE) .01 (.02) -.02 (.02) -.03 (.03) .02 (.03) -.01 (.02) .01 (.02) .03 (.02)
Block Number b (SE) -.02 (.03) -.04 (.04) -.03 (.04) .00 (.03) -.01 (.03) -.02 (.03) .00 (.03)
Deck x Block b (SE) -.04 (.04) .01 (.05) -.10 (.04)** .01 (.04) -.10 (.03)*** -.04 (.04) -.05 (.04)
Note: *Significant at the p < .05 level. **Significant at the p < .01 level. ***Significant at the p < .001 level. For the intercept column, the coefficients indicate the association between the intercept and the construct identified in the row header. For other table entries, coefficients in table refer to the interaction between the construct identified in the row header and the column header.
Figure 1. Pictures of facial affect used in task. The following emotions are depicted: a) happiness; b) anger; c) sadness; d) fear; and e) disgust.
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING Figure 2. MLMs examining three-way interactions for the IGT between deck type, block number, and fun-seeking.
APPROACH AND AVOIDANCE PATTERNS IN REWARD LEARNING Figure 3. MLMs examining three-way interactions for the S-IGT between deck type, block number, and a) fun-seeking, b) depressive symptoms, and c) social closeness (from the comprehensive multivariate model).
a)
b)
c)
Highlights The Social Iowa Gambling Task assesses avoidance learning of social feedback. Individual differences in depression were associated with learning of social feedback. Differences in fun-seeking were associated with learning of monetary and social feedback.