University of Southern California Dissertations and Theses

Dopamine dependent: examining the link between learning and treatment-resistant depression

(USC Thesis Other)

Dopamine dependent: examining the link between learning and treatment-resistant depression

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content
DOPAMINE DEPENDENT:
EXAMINING THE LINK BETWEEN LEARNING AND TREATMENT-RESISTANT
DEPRESSION

by

Xiao Liu

A Thesis Presented to the
FACULTY OF THE USC DORNSIFE SCHOOL OF ARTS AND SCIENCES
UNIVERSITY OF SOURTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF ARTS
PSYCHOLOGY

August 2021

Copyright 2021 Xiao Liu
ii

TABLE OF CONTENTS
List of Tables ........................................................................................................................... iii
List of Figures ............................................................................................................................ v
ABSTRACT ..............................................................................................................................vi
INTRODUCTION ..................................................................................................................... 1
Background .........................................................................................................................................................1
Depression Heterogeneity and Treatment ...........................................................................................................2
Dopamine and Anhedonia ...................................................................................................................................3
Experiment 1 .............................................................................................................................. 8
METHODS ..........................................................................................................................................................9
RESULTS ..........................................................................................................................................................16
DISCUSSION ...................................................................................................................................................24
Experiment 2 ............................................................................................................................ 26
METHODS ........................................................................................................................................................30
RESULTS ..........................................................................................................................................................35
DISCUSSION ...................................................................................................................................................55
REFERENCES......................................................................................................................... 62
APPENDIX .............................................................................................................................. 77

iii

List of Tables
Table 1. Accuracy across blocks by bean valence…………………………………………….....16
Table 2. Response bias on novel beans……………………………………………………..........18
Table 3. Descriptive statistics by diagnostic group ……………………………………………..22
Table 4. 2-factor solution of the MEI……………………………………………........................35
Table 5. 2-factor solution of the TEPS……………………………………………......................37
Table 6. 3-factor solution of the MEI and TEPS items…………………………………………..40
Table 7. Missing data rates…………………………………………….………………………...43
Table 8. Self-report scales descriptive statistics by group……………………………………….43
Table 9. Personality subscales by diagnostic group……………………………………………...47
Table 10. Learning and response bias during test……………………………………………......49
Table 11. PHI coefficient by block……………………………………………............................49
Table 12. Learning bias by group……………………………………………..............................53
Table 13. Response bias by group…………………………………………….............................54
APPENDIX
Table 1A. Sample of DARS write-in answers……………………………………………...........77
Table 2A. Correlations between all measures……………………………………………............78
Table 3A. Experiment 1 Questionnaire…………………………………………….....................79
iv

Table 4A. Beck Depression Inventory (BDI) ……………………………………………...........81
Table 5A. Beck Anxiety Inventory (BAI) ……………………………………………................85
Table 6A. Inventory of Depression and Anxiety Symptoms-II (IDAS-II)....................................86
Table 7A. Experiment 2 Screening Questionnaire……………………………………………....90
Table 8A. Patient Health Questionnaire-2 (PHQ-2) …………………………………………….92
Table 9A. Motivation and Energy Inventory (MEI) …………………………………………….92
Table 10A. Dimensional Anhedonia Rating Scale (DARS) ………………………………….....94
Table 11A. Big Five Aspect Scales (BFAS) …………………………………………….............96

v

List of Figures
Figure 1. Bean variation and valence………………………………………………………….......7
Figure 2. BeanFest task interface…………………………………………………………….......12
Figure 3. Approach behavior across blocks for learning beans…………………………….........17
Figure 4. Depression scale distributions……………………………………………....................19
Figure 5. Scatterplot of the moderating influence of depression on strategy................................21
Figure 6. Approach across blocks by diagnostic group………………………………………….23
Figure 7. Parallel analysis scree plots for overall anhedonia construct.........................................38
Figure 8. Mean depression severity by diagnostic group………………………………………..42
Figure 9. Number correct by bean valence……………………………………………................50
Figure 10. Learning bean accuracy across blocks by group……………………………………..50
Figure 11. Approach across blocks……………………………………………............................51
Figure 12. Group differences in BeanFest…………………………………………….................59

vi

ABSTRACT
Treatment resistant depression (TRD) is a subtype of Major Depressive Disorder (MDD)
that is not responsive to typical antidepressants such as SSRIs. There is some evidence that
treatment responsiveness may be related to dopamine neuron dysfunction, specifically in the
mesolimbic region of the brain. To address the gap in understanding of how this translates to
behavioral and personality differences in TRD, we used a reinforcement learning task
administered online to objectively measure individual variation in learning from positive and
negative feedback. In Experiment 1, we found that participants on average were better at learning
the negative stimuli than the positive. Response bias during the test round was significantly
correlated with the IDAS-II for General Depression subscale, such that greater depression
severity was related to a more negative response bias (r = -.18, p = .03). In Experiment 2, we
empirically defined three underlying facets of anhedonia by factor analyzing two anhedonia
scales administered online to a sample of 137 participants. The major latent factors were found to
be motivation or energy, inventive salience, and consummatory liking. We found that TRD was
associated with significantly lower anticipation scores, whereas MDD overall was associated
with significantly lower consummatory pleasure. Additionally, the TRD group had significantly
lower extraversion scores than responders. Only antidepressant responders with MDD did not
demonstrate significantly higher learning of bad stimuli over good stimuli on our reinforcement
learning task. We postulated this was due to the effect of SSRIs; however, further research is
needed to determine a causal effect.

1

INTRODUCTION
Background
Depression may be defined as both a disorder and a symptom. As a disorder, Major
Depression (MDD) encompasses a heterogeneous criterion of mood, cognitive, and somatic
impairments. As a symptom, depression is colloquially defined as an inability to enjoy what was
once pleasurable. Due to the varied nature of symptoms grouped in the depression disorders
category, subtypes have emerged to preserve diagnostic validity. One pragmatic method of
subtyping depression is based on treatment response to traditional antidepressants. First-line
antidepressant medications (selective serotonin/norepinephrine reuptake inhibitors; SSRI/SNRIs)
have non-response rates of up to 60% (Fava, 2003), and their primary mechanism acts to directly
and immediately increase levels of serotonin/norepinephrine in the extracellular matrix.
Therefore, non-responsiveness to antidepressants may indicate a type of depression that results
from greater impairment of neuromodulator systems not directly affected, such as dopamine.
The symptom profile specific to treatment resistant depression (TRD) is not well
characterized, and most clinicians employ a trial-and-error strategy of testing various
antidepressants on patients to determine treatment responsiveness. However, due to the long (4
to 6 week) treatment response delay characteristic of most SSRI/SNRI medications, this can lead
to exponentially greater financial and emotional burden for non-responders. This study aims to
address these problems by helping to define behavioral and personality differences specific to
TRD. We will first review the existing literature on associations between depression, the
dopaminergic reward system, and treatment responsiveness. We postulate that reinforcement
2

learning mediated by the mesolimbic dopamine system represents one area for further research in
characterizing defining differences around this subtype of depression.
Depression Heterogeneity and Treatment
In the Diagnostic and Statistical Manual of Mental Disorders - 5
th
Edition (DSM-5;
American Psychiatric Association, 2013), there are 681 possible combinations of symptoms that
meet the criteria for a singular diagnosis of MDD (Akil et al., 2018). Uncovering and defining
the etiology of various types of depression is an ongoing subject of interest that has important
implications for treatment. One study conducted an exploratory factor analysis on the combined
items of three main psychometric scales for depression: Beck Depression Inventory (BDI; Beck
et al. 1961), Montgomery-Asberg Depression Rating Scale (MADRS ; Montgomery & Asberg,
1979), and Hamilton Depression Rating Scale (HAMD-17; Hamilton, 1959), producing a three
factor solution: cognitive (composed of guilt, suicide, pessimism, self-criticism), affective (mood
and anxiety symptoms), and neurovegetative (somatic symptoms) (Uher et al., 2008). A one-
factor solution of the combined 48 items accounted for 45% of the explained variance but
yielded poor model fit indices. This evidence supports the idea that depression is composed of
more than one latent construct. The serotonin system has been extensively studied in relation to
depression. The serotonin-1A receptor (5-HT1A) is an inhibitory G-protein coupled receptor
abundantly found in the raphe nucleus where it functions as an autoreceptor to downregulate the
release of serotonin (Lemonde et al., 2004). 5-HT1A autoreceptor over-expression has been
posed as a reason for the delay in treatment response present with SSRI treatment (Samuels et
al., 2011). However, solely focusing on one neuromodulator may not tell the full story, as
evidenced by high non-response rates.
3

To date, studies on treatment resistance have emphasized the greater symptom severity of
this subtype of depression, however there is a dearth of research on the qualitative profile of
treatment resistance. This could be partially due to inconsistencies in defining treatment-resistant
depression (TRD), and partially due to contention over accuracy of high non-response rates in
clinical antidepressant trials. TRD is generally recognized as failure to achieve remission from
adequate AD therapy. This has been defined in several ways, from treatment with at least one
AD medication of proven efficacy at adequate dosage and duration (Fava, 2003), to failure of at
least 2 ADs from different pharmacological classes for 6-8 weeks (Little, 2009; Wijeratne &
Sachdev, 2008). The largest (N = 3671) study of treatment resistant depression was the
Sequenced Treatment Alternatives to Relieve Depression (STAR*D) trial, which used a decision
tree algorithm to determine treatment procedure at 4 successive steps of non-responsiveness
(Gaynes et al., 2009). The STAR*D trial aimed to address concerns over high clinical trial non-
response rates by tailoring the dosage and treatment plan to each individual participant. After
level 2, remission rates experienced a sharp drop-off (after failure of 2 adequate dose-duration
treatments). Furthermore, likelihood of relapse after 1 year were strongly correlated with failed
trials. Therefore, treatment resistant depression is a more severe and persistent form of
depression without a well-defined predictive symptom profile. However, simply relying on
overall severity without examining differences in qualitative symptom expression may be an
oversimplification and does not contribute to greater understanding of underlying
neurobiological causes.
Dopamine and Anhedonia
One major qualitative symptom that characterizes depression is anhedonia, which is
experienced as a lack of enjoyment of once pleasurable activities. Anhedonia is a symptom of
4

interest in the current study due to its relationship with reward responding and thus, the
mesolimbic dopamine system. The measurement and neurobiology of anhedonia has been
extensively studied, but it remains a complex and heterogeneous concept. Anhedonia severity is
included as an item in most depression scales. One of the most commonly used psychometric
scales solely focused on measuring anhedonia is the Snaith–Hamilton Pleasure Scale (SHAPS;
Snaith et al., 1995), which treats it as a unitary construct. Anhedonia has also been measured
behaviorally through reward responsiveness in a reinforcement learning task (Rizvi, Pizzagalli,
Sproule, & Kennedy, 2016). The neurobiology of reward learning occurs in the ventral striatum,
and involves dopaminergic neurons projecting from the ventral tegmental area (VTA) of the
midbrain. Burst signaling triggers presynaptic release of DA in the striatum, which activates D2
receptor binding. These phasic bursts are proportional to the strength of reward prediction error
(RPE; Pessiglione, Seymour, Flandin, Dolan, & Frith, 2006). RPE is defined as the difference
between expected reward, based on value calculations of environmental stimuli in the Nucleus
Accumbens (NAc), and perceived reward received. This error drives learning and modulates
approach and avoidance behaviors toward environmental stimuli. Research has shown impaired
RPE signals in patients with MDD as well as lower performance on reward learning tasks (for a
review see: Chen, Takashi, Nakagawa, Inoue, & Kusumi, 2015).
There is some evidence that SSRI treatment can sometimes moderate changes in reward
responsiveness. In a study of participants who took an 8-week course of SSRIs, depressed
patients exhibited a loss of motivation (Pringle et al., 2013), whereas healthy controls tended to
show increased effort expenditure for reward (Meyniel et al., 2016). Other studies have found
that SSRIs contributed to increased reward sensitivity and motivation (Tang et al., 2009; Yuen et
al., 2014). One study found that greater anhedonia measured at baseline was predictive of
5

responsiveness to SSRIs, such that non-response rates were higher (Wang, Leri, & Rizvi, 2021).
Neural markers of reward sensitivity have also been found to be predictive of treatment
responsiveness, implicating the mesolimbic dopamine system (Wang, Leri, & Rizvi, 2021).
Compared to non-responders, patients who did improve with SSRI medication exhibited
increased striatal D2 receptor binding, the degree of which correlated with improvements in
HAM-D scores (Wijeratne & Sachdev, 2008). Pramipexole, a DA agonist targeting D2 receptors,
was found to be an effective addition to treatment for bipolar and unipolar depressives who were
drug-resistant to TCAs or SSRIs (Lattanzi et al., 2002). Anhedonia is therefore unsurprisingly
associated with treatment responsiveness of SSRIs (Malhi, Barker, Crawford, Wilhelm, &
Mitchell, 2005; Rizvi et al., 2015).
BeanFest
In this study, we will use a reinforcement learning task called BeanFest, developed by
Russ Fazio (Fazio, Eiser, and Shook, 2004), to operationalize measurement of mesolimbic
dopamine system impairment for people with depression. BeanFest is a game that approximates
the learning of novel stimuli through reward and loss. Participants must choose whether to
approach or avoid the stimulus in each trial, and there is evidence that approach and avoidance
behavior may be moderated by depression through the striatal dopamine system (for a review
see: Trew, 2011). This task departs from commonly used reinforcement learning paradigms such
as the Probabilistic Reward Task (PRT; Pizzagalli, Jahn, & O’Shea, 2005), which provide
feedback on every trial. In BeanFest, participants are not given information on the valence of the
stimuli if they do not choose to approach it. Therefore, learning is contingent on behavior. This
more closely mimics real-world situations where there is a risk-reward trade-off. Furthermore,
the BeanFest task contrasts learning from differently valenced feedback instead of different rates
6

of reward; studies have shown that loss activates brain circuits that are distinct from reward
(Elliot, Friston, & Dolan, 2000; Knutson, Westdorp, Kaiser, & Hommer, 2000; O’Doherty,
Critchley, Deichmann, & Dolan, 2003; Zalla, et al., 2000).
The authors have found reliable evidence for a bias in greater learning rates from
negative versus positive information (Fazio, Pietri, Rocklage, & Shook, 2015), which is
consistent with the literature that heightened amygdala activity during punishing situations may
lead to a stronger, more persistent memory (Coleman-Mesches, Salina, & McGaugh, 1996;
Murty, LaBar, & Adcock, 2012). Furthermore, BeanFest not only assesses information
acquisition rates but also information integration to predict outcomes in the form of
generalization bias. The authors have defined this as the weighting of learned positive and
negative information when applied to the judgement of novel stimuli. We used the originally
defined metric of generalization bias as a measure of response bias, as it was calculated from the
tendency to categorize more beans as either good or bad post-learning. The original authors also
found a tendency for participants to generalize more from negative attitudes than positive ones,
thus finding a negative generalization bias on average (Eiser, Fazio, & Shook, 2004).
BeanFest is a virtual game where participants are presented with a series of schematics of
beans. These stimuli vary along two dimensions: elongation and speckles, with 10 degrees of
variation along each dimension (Figure 1). Of the 100 total beans, 36 unique beans are presented
in random order, once each for 3 learning blocks. During each trial, participants have 5 seconds
to decide whether to approach, or eat, a bean. If eaten, the bean can either give the participant 10
energy points or cause the participant to lose 10 energy points. If the participant times out or
decides to avoid the bean, their point total will not be affected. However, the participant will lose
1 energy point every trial as a function of time, to simulate hunger over time, and to encourage
7

exploration behavior. Participants begin with 50 points at the start of the game, and they can win
the game by reaching 100 points or lose the game by reaching 0. Once either of these two
scenarios occur, the participant will see either a green message telling them they have won the
game, or a red message telling them they have lost, after which the game will reset to 50 points.
Winning or losing does not affect the number of trials or learning blocks, it simply sets the goal
to maximize points. Lastly, there is a test block during which participants will be presented with
all 100 beans, and they must categorize each bean as either good or bad. No feedback is given
during this block, and points are not shown.
Objectives
We hypothesize that greater learning from negative information relative to positive
information and greater generalization of negative features of novel stimuli during BeanFest
could indicate likelihood of AD non-responsiveness in MDD patients. Our aims are to (1)
replicate the findings of Fazio, Eiser, and Shook (2004) in a population of MDD participants, (2)
determine whether certain subtypes of depression, specifically TRD, contribute to individual
differences in learning rates and response bias, and (3) improve treatment efficiency by
providing objective information on whether a patient is likely to be treatment resistant.

Figure 1. Bean variation and valence
8

(Left) There are 100 unique bean schematics in BeanFest, varying along 10 levels of elongation and speckles.
(Right) The beans varied in terms of shape (X1 = circular to X10 = oblong) and speckles (Y1 = 1 speckle to Y10 =
10 speckles). The 36 game beans are denoted by their assigned valence of good (+) or bad (-). The other 64 beans
were only presented during the test round. (originally printed in Fazio, Pietri, Rocklage, & Shook, 2015)

Experiment 1
The comorbidity of depression and anxiety is highly prevalent and may affect learning
from feedback in BeanFest. Lifetime comorbidity of anxiety disorders has been found to range
from 50.6% in subjects diagnosed with MDD (Fava et al., 2003) up to 75% in subjects diagnosed
with any depressive disorder (Lamers et al., 2011). Both anxiety and depressive symptoms have
been studied in terms of dual-systems models of motivation (Trew, 2011). One such model is
Reinforcement Sensitivity Theory, which posits that the Behavioral Inhibition (BIS)/Behavioral
Approach Systems (BAS) govern inhibitory/approach behaviors respectively of an individual
navigating their environment (Gray, 1987). The theory proposed that abnormally high inhibitory
behavior was characteristic of anxiety (Gray, 1987). Both depression and anxiety disorders have
been associated with higher BIS and lower BAS (Gray, 1987; Johnson, Turner, & Iwata, 2002;
Kimbrel, Nelson-Gray, & Mitchell, 2012; Trew, 2011). Johnson, Turner, and Iwata (2002) found
evidence that BAS was significantly lower in lifetime depressed individuals; however, BIS was a
9

better predictor of current depressive episode. In depressed patients, BAS has been found to
relate to both symptom severity and poorer long-term clinical outcome (Kasch, Rottenberg,
Arnow, & Gotlib, 2002; McFarland, Shankman, Tenke, Bruder, & Klein, 2006). BAS
improvement in the first two weeks of antidepressant administration was associated with lower
symptom severity and anhedonia at the end of a full course of treatment (Allen, Lam, & Milev,
2019).
Clinical predictors of TRD significantly replicated in 2 studies (resulting in a total
participant sample of 1,618 depressed patients) were found to be: severity of symptoms, suicide
risk, number of lifetime depressive episodes, and anxiety comorbidity (Souery et al., 2007;
Kautzky et al., 2019). While abnormal BIS/BAS have been implicated in anxiety and depression
disorders, how these systems are differentially impacted in TRD is not fully understood. The
aims of this experiment were (1) to replicate the original findings of the BeanFest creators Fazio,
Eiser, and Shook (2004) in a sample of self-identified people with depression and diagnosis
naïve controls, (2) determine whether BeanFest outcomes during information acquisition
(learning) or information integration (generalization, or response) was associated with depression
or anxiety severity, and (3) test whether antidepressant responsiveness generated differences in
task outcomes. In Experiment 1, we included a measure of anxiety as well as a comprehensive
measure of depression, anxiety, and stress to assess whether comorbidity would contribute to an
even greater negative response bias and favoring of negative information. This study was
approved by the USC IRB review board and funded by the USC Department of Psychology.
METHODS
Participants
10

Participants (N = 141) aged 18-65 were recruited online from ResearchMatch.org, a
National Institutes of Health (NIH) funded website that matches volunteers with researchers.
Due to the difficulty of recruiting treatment resistant volunteers, we ended up with sample sizes
of TRD = 15, MDD without TRD = 77, No Diagnosis = 43, and unknown = 6. We used the
following exclusion criteria: all people with bipolar depression, psychosis, ADHD, and any
personality disorder, in order to minimize confounding variables due to different treatments for
these disorders. We also excluded people taking bupropion, stimulants, pramipexole, or L-dopa
medication due to potential confounding effects on the dopamine system. All participants were
compensated a flat rate of $15.00 in the form of an Amazon e-gift card for their time.
Procedure
Due to the exploratory nature of our hypotheses, we planned to conduct this study fully
online. Participants received an anonymous link to the consent information and study contents.
After completing the consent form, they first answered 4 self-report questionnaires: one
qualitative survey about their diagnosis and medications as well as 3 validated scales of
depression and anxiety symptoms (see Appendix). Next, they read a detailed explanation and
instructions for the BeanFest game, which consisted of 3 blocks of 36 learning trials, and 1 block
of 100 test trials. Participants were informed that the objective of BeanFest was to win as many
times as possible by reaching 100 energy points, and to avoid losing the game by reaching 0
energy points. Every subject would commence with 50 points each time they won or lost and
began a new game.
11

BeanFest task. The user interface consisted of a black background with a randomized
bean schematic displayed in the center of the screen (see Figure 2). A point meter was displayed
at the bottom of the screen as a green bar representing current points overlaid on top of a red bar
representing total possible points and included a numerical display tracking total points in white
text. Participants were instructed to press either the “k” key to approach or the “d” key to avoid a
bean, and these instructions were displayed along the top of the screen during all learning block
trials. After 5000ms, the trial timed out and would default to “avoid”. The subject’s decision
(approach or avoid the bean), effect of the bean (+10 or -10 points), and net loss or gain in points
were then displayed in the bottom right-hand corner of the screen in bold white text for 2000ms,
at which point the next trial would begin.
Subjects would first undergo 3 learning blocks (10-15 minutes each) during which 36
game beans (18 good, 18 bad) were presented, one per trial, in randomized order. Game beans
and their valence remained the same across all subjects. Positively and negatively valenced beans
were located in random groupings around the 10 by 10 feature grid (Figure 1). However, these
groupings were not linearly separable by valence as to minimize model-based learning, which
may be moderated by cognitive ability (Chen et al., 2015). Participants were asked to decide
whether to approach or avoid each bean by pressing the respective keys. If approached, the bean
could be good – leading to a gain of 10 points, or bad – leading to a loss of 10 points. If avoided,
the participant would not be given information on the valence of the bean. 1 point was deducted
every trial to simulate energy lost as a function of time. The 3 learning blocks were followed by
a final test block, during which no point meter or feedback were displayed. The test block
consisted of 100 beans (including the 36 game beans as well as 64 novel beans varying along the
same 2 dimensions), which the participant was instructed to categorize as either good or bad.
12

Once the game was complete, participants were shown a unique code on the screen which they
could email to the researcher to receive compensation.
We aimed to replicate the original BeanFest analyses of Fazio, Eiser and Shook (2004) in
their initial development of the task. Behavioral response bias was determined from the average
proportion of good/bad beans correctly approached/avoided during each block. For novel beans
in the test block, we computed a Euclidean distance score to the nearest good or bad bean. If the
bean was closer to a good bean, they were categorized as good-leaning (n = 29). If they were
closer to a bad bean, they were categorized as bad-leaning (n = 29), and if they were equidistant
they were categorized as neutral (n = 6). Learning Bias (LB; called learning asymmetry by the
original authors) scores were defined as the difference between the proportion of good learning
beans correctly classified and the proportion of bad learning beans correctly classified during the
test block. Response bias during test (RB; termed Generalization Asymmetry by the original
authors) was calculated from the sum of novel beans classified as good (coded as +1) and bad
(coded as -1) during the test block.

Figure 2. BeanFest task interface

13

(Left) One bean is presented per trial. Key actions are displayed at the top of the screen and a point meter is
displayed at the bottom. (Right) Participants have 5000ms to make a decision during each trial. If no decision is
made, the trial will time out and default to avoiding the bean, resulting in no feedback about the bean’s valence.

Self-Report Measures
Beck Depression Inventory (BDI; Beck, Ward, Mendelson, Mock, & Erbaugh, 1961) is
a widely used 4-point Likert scale for assessing depression severity based on 21 categories of
symptoms and attitudes constructed from common clinical complaints. It was originally
validated on 2 samples of 226 patients and 183 patients taken from the psychiatric outpatient
department of the Hospital of the University of Pennsylvania and the Philadelphia General
Hospital. 97 patients from the original sample were selected for calculating Spearman-Brown
split half reliability coefficient of .93. Discriminant validity between adjacent Depth of
Depression levels was significant at p < .0004.
Beck Anxiety Inventory (BAI; Beck, Epstein, Brown, & Steer, 1988) is a 21-item self-
report scale to measure anxiety symptom severity in psychiatric populations. It has a high
internal consistency (α = .92) and 1 week test–retest reliability (r = .75). It was moderately
correlated with the revised Hamilton Anxiety Rating Scale (r = .51), and mildly correlated with
the revised Hamilton Depression Rating Scale (r = .25). The BAI was further validated on a
sample of 40 outpatients with high internal consistency (α = .94) and moderate test-retest
reliability over 11 days (r = .67). In a sample of 71 outpatients the BAI outperformed the Trait
Anxiety measure of the State-Trait Anxiety Inventory (STAI; Spielberger, Gorsuch, Lushene,
Vagg, & Jacobs, 1983) in tests of convergent and discriminant validity (Fydrich, Dowdall, &
Chambless, 1992).
14

Inventory of Depression and Anxiety Symptoms Expanded Version (IDAS-II;
Watson et al., 2012) contains 99 items that represent a conglomeration of 18 factor analytically
derived nonoverlapping subscales plus General Depression, which samples items from these
other factors. In a sample of undergraduate psychology students, outpatient, and community
adults living in Eastern Iowa, internal consistency ranged from .80 to .90. The Dysphoria
subscale had the strongest association with MDD as diagnosed by DSM-IV criteria in a clinical
interview setting (r = .67) and Generalized Anxiety Disorder (GAD; r = .35).

Analysis
Statistical power. Using G*power (Faul, Erdfelder, Lang, & Buchner , 2007), we
determined that a sample of N = 158 would have a statistical power of 80.2% to detect at least a
medium effect size of 0.25.
Direct replication of BeanFest outcomes. We used this first experiment to replicate the
analysis in the Fazio, Eiser, and Shook (2004) study, Experiment 1. The two main dependent
variables of interest were accuracy and response bias. Accuracy was operationalized as a phi
coefficient for learning beans across all 108 learning trials and 36 test trials. Individual phi
coefficients were averaged per block to determine learning rate. We conducted one-way repeated
measures analysis of variance (ANOVA) to determine whether there were differences in (1)
overall approach behavior across blocks, and (2) overall accuracy across blocks. We also
conducted a 2 x 3 repeated-measures ANOVA to examine differences in learning by bean
valence and block as within-subjects factors. Response bias was calculated for each valence
15

category of test beans (good-leaning, bad-leaning, and neutral) and a one-way repeated-measures
ANOVA conducted to determine whether differences were present.
Mental health associations. We related the primary BeanFest outcomes of learning bias
and response bias to self-reported depression and anxiety scores using the significance test for
Pearson’s correlation coefficient. We also conducted a correlation analysis between points scored
(0-100) and behavior (coded as 1 for approach and 2 for avoid) to gauge how approach and
avoidance behavior may change as a function of performance. We then related this to mental
health scores, as we believed we would see differences in avoidance rates when points were low.
We hypothesized that healthy subjects would be more likely to employ a risk averse (avoidance
dominated) strategy when points were high and an explorative (approach dominated) strategy
when points were low, whereas those with depression and anxiety would be more inhibited when
points were low as well, due to an overactivation of BIS (Trew, 2011). First, we calculated the
point biserial correlation for overall association between points and approach for all learning
block trials. Then, we conducted Pearson’s r significance tests to determine whether these
correlations were significantly related to either BDI, BAI, or the IDAS-Depression subscale.
TRD analysis. We labeled participants as (1) No MDD Diagnosis: self-identified healthy
volunteers, (2) TRD Diagnosis: self-identified people with a diagnosis of treatment resistant
depression, or (3) MDD without TRD: self-identified people with a diagnosis of MDD and who
responded well to typical antidepressants. We conducted mixed-measures ANOVAs on learning
accuracy and approach behavior, with group as the between-subjects variable and bean valence
as the within-subjects variable.

16

RESULTS
Replication analysis
Did participants learn?
Average phi coefficients for each block of the learning phase were .05, .11, and .14. We
used a repeated measures one-way ANOVA and found these differences to be significant (F =
9.267, p < .001). During the test phase, the average phi coefficient was .15, which represents a
weak to moderate effect. For this analysis we excluded 3 participants who did not categorize any
beans as bad during the test block.
Did learning vary as a function of valence?
Learning of good beans did not display a linearly increasing or decreasing trend (see
Table 1) and was not significantly different from chance during the test block (M = .48, t = -1.23,
p = .220). However, learning of bad beans steadily increased over time and was significantly
higher than chance in the test block (M = .65, t = 6.76, p < .001). Overall there was a
significantly negative learning bias of -.17 (t = -5.02, p < .001).

Table 1. Accuracy across blocks by bean valence
Good Correct Bad Correct
LB 1 56.0% 48.7%
LB 2 56.5% 53.4%
LB 3 56.5% 55.5%
TEST 47.6% 64.6%

17

How did approach behavior change over time?
Overall approach behavior trended down over time but did not reach significance (F =
2.63, p = .074). However, overall accuracy for learning beans did improve significantly over
time (F = 5.10, p = .002). Next, we examined approach behavior across both block and bean
valence using a two-way repeated measures ANOVA. We found a significant interaction
between block and valence (F = 6.19, p < .001). Accuracy of good beans was significantly
different across blocks (F = 35.63, p < .001) as was accuracy of bad beans, which followed an
increasing trend (F = 27.10, p < .001).

Figure 3. Approach behavior across blocks for learning beans

Proportion of approach behavior by bean valence over time. The TEST factor reflects the proportion of learning
beans classified as “good”.

Did positive and negative attitudes generalize during test?
0%
10%
20%
30%
40%
50%
60%
LR 1 LR 2 LR 3 TEST
% of beans approached
Blockname
Good App
Bad App
18

Responses to the novel beans (scored as +1 and -1 for beans categorized as good vs. bad,
respectively) were marginally significantly affected by their valence categories (see Table 2; F =
2.927, p = .055). Furthermore, we found a trend similar to Fazio et al. (2004) such that novel
beans closer to a bad learning bean (M = -.20) were more likely to be classified as bad than were
equidistant beans (M = -.13) or beans closer to a good learning bean (M = -.17). However,
Bonferroni-corrected pairwise comparisons found no significant difference between good and
bad-leaning scores (p = .786).

Table 2. Response bias on novel beans
Novel Beans Sum Score
Bad-leaning (n = 29) -5.91 -.20
Good-leaning (n = 29) -4.86 -.17
Neutral (n = 6) -0.75 -.13
Mean response bias for the 64 novel beans during the test round. Sum is defined as the total response to each
category of novel beans coded as +1 for good and -1 for bad. Score is the Sum divided by the number of novel beans
in that category.

Is there a valence asymmetry in response behavior?
Novel beans that were equidistant between bad and good neighbors were significantly
more likely to be regarded as bad (M = -.13, t = -2.87, p = .005). The novel beans as a whole
were also more likely to be classified as bad than as good (M = -.18, t = -5.31, p < .001).
Greater learning of the bad beans relative to good beans in the test block was strongly
associated with a greater likelihood of skewing toward categorizing the novel beans as bad (r =
.90, p < .001).
19

Depression and anxiety results
Both depression measures were mildly right skewed (Figure 4; BDI score skewness =
.33, IDAS Depression skewness = .18). According to the rule of thumb proposed by Bulmer
(1979), these scores are close enough to 0 to consider them approximately symmetrical. The
mean test response bias was right skewed (skewness = .48), mainly due to some participants
responding to all beans during test as good. Therefore, we excluded those participants, resulting
in an adjusted skewness value of .26. The total sample size for completed BDI scores was n =
127, while the sample for IDAS and BAI variables was 135.

Figure 4. Depression scale distributions

The BDI (left) and IDAS Depression (right) scale score distributions for all participants. A score of 21 or above on
the BDI indicates some level of clinical depression.

Do depression and anxiety scores correlate with response bias?
20

The subset of items in IDAS-II for General Depression (IDAS-Dep) did reach the
threshold for significance in a Pearson’s r correlation test (r = -.18, p = .03), while BDI (p = .22)
scores did not. Anxiety as assessed by the BAI was on the cutoff value for significance (p = .06).
Therefore, greater depression and anxiety scores tended to correlate with a more negative RB.
No mental health scores correlated with learning bias during the test block.
Do depression scores predict response bias controlling for learning?
No. Learning bias accounted for almost all of the variance in response bias during test
when fitting a general linear model (B = .05, t = 22.87, p < .001). When added into the regression
model, the predictors for BDI, IDAS-Dep, and BAI did not reach significance.
Do mental health scores moderate the relationship between approach behavior and points?
First, we calculated a point biserial correlation for each participant on the relationship
between their response (coded as approach = 1 and avoid = 2) and point total over all learning
block trials. Then, we fit a general linear model with this correlation regressed on each of the
depression and anxiety scale scores. BDI score was found to significantly predict individual
differences in trial-by-trial correlation between point total and approach behavior (t = -3.20, p =
.002). Similarly, IDAS-Dep scores were also significantly related to participants’ correlation
between point total and approach (t = -2.57, p = .01). Anxiety as measured by the BAI was not
found to be a significant regressor (p = .06). Therefore, depression, but not anxiety was found to
moderate the relationship between points and behavior.
Could it be that people with more severe depression become more inhibited as their points near
0, while healthy people take on a more exploratory strategy?
21

To test this, we only looked at trials where points were less than 50 (low condition)
versus trials where points were greater than or equal to 50 (high condition). Contrary to our
hypothesis, we found that for the low condition, correlation between points and approach (corL;
see Figure 5) was significantly positive across all participants (t = 2.43, p = .02), but depression
did not moderate this relationship (p = .90). The high condition mean correlation (corH; Figure
5) did not significantly differ from zero (p = .53), depression played a significant moderating role
(t = -2.11, p = .04). Therefore, people who are depressed seem to increase their approach
behavior when they have over 50 points, whereas healthy controls will tend to hold back more as
their point total increases. Furthermore, when points are low, people across the board will engage
in more approach behavior as their points decrease.

Figure 5. Scatterplot of the moderating influence of depression on strategy

Correlation scores were calculated for each participant between their behavior (coded as approach = 1, avoid = 2)
and point totals. BDI scores did not affect the relationship between approach behavior and points in the low
condition (points < 50; left), however in the high condition (points ≥ 50; right), higher BDI meant a negative (as
opposed to positive for low BDI) correlation between behavior and points.
22

Treatment responsiveness results
Were there differences in depression and anxiety severity between diagnostic groups?
There was a significant main effect of diagnostic group on both BDI (F = 13.40, p <
.001) and BAI (F = 13.66, p < .001). Bonferroni adjusted pairwise comparisons revealed a
significant difference in BDI scores between the no diagnosis group with (1) the TRD group (M
= -15.61, p < .001) and (2) the MDD without TRD group (M = -10.90, p < .001) and BAI scores
(1: M = -17.33, p = .002; 2: -15.21, p < .001). However, there was no significant difference
between the two depression groups in either BDI (p = .573) or BAI (p = 1.000) scores (see Table
3).
Were there differences in learning or response bias during test by diagnostic group?
No. Learning bias during test was not significantly different between diagnostic groups
(F = .06, p = .95), nor was response bias (F = .17, p = .85).

Table 3. Descriptive statistics by diagnostic group
Mean BDI (se) Mean BAI (se)
No MDD Diagnosis (n = 41) 12.24 (2.00) 11.17 (1.78)
TRD Diagnosis (n = 14) 27.86 (2.97) 28.50 (4.34)
MDD without TRD (n = 71) 23.14 (1.44) 26.38 (2.09)
Mean (std error) for each diagnostic group. No MDD diagnosis group have never diagnosed with clinical
depression. TRD diagnosis group self-identified as received a diagnosis of TRD or did not respond to > 1 course of
antidepressant medication.

23

Does diagnostic group moderate overall learning of good or bad beans?
No. Group did not have a significant effect on total good beans classified correctly (F =
1.77, p = .175) or bad beans classified correctly (F = 1.85, p = .162)
Does diagnostic group moderate overall approach behavior across learning blocks?
Classification behavior during test?
We conducted a 3 (diagnostic group) x 3 (learning block) mixed measures ANOVA on
approach behavior with group as the between-subjects factor and block as the within-subjects
factor. There were no significant main effects of block (F = 1.771, p = .174) or block x group (F
= 1.078, p = .368), but there was a significant main effect of group (F = 3.328, p = .039), such
that overall the TRD group had the highest approach behavior across blocks (see Figure 6).
Furthermore, group did not moderate classification behavior during test (F = .166, p = .847).

Figure 6. Approach across blocks by diagnostic group

24

95% confidence interval plot of the mean number of trials participants pressed the approach key across learning
blocks (LB) and by group (x-axis).

DISCUSSION
The findings in this study demonstrated that learning of the beans stimuli did indeed
occur in BeanFest, which replicated the results of the original study (Fazio, Eiser, & Shook,
2004). Furthermore, this was driven by learning of bad beans, which increased steadily over
time, while learning of good beans remained steady. We used the term learning bias to describe
the inequality in learning rate of good and bad beans during test. The bias was demonstrated by a
decrease in approach behavior toward bad beans across blocks. The authors of the original study
found average approach of good beans to remain steady across blocks, while we found
significant differences in approach behavior for good beans across blocks. This discrepancy can
solely be attributed to the strongly attenuated classification of beans as “good” during the test
block. Due to this trend, we witnessed a sharp decrease in accuracy of categorizing good beans
during the test block and an increase in accuracy of categorizing bad beans.
Depression and anxiety scores were found to correlate with response bias such that
people with more severe symptoms had a more negative bias. We defined response bias as the
tendency to classify novel beans during test as either good or bad. A negative number indicated
greater weighting of feature information from bad beans, while a positive bias would indicate the
reverse. However, mental health scores did not have any relationship to learning bias during test,
nor did depression severity significantly predict response bias when controlling for learning bias.
Overall behavior during the test round significantly trended towards categorizing beans as “bad”
regardless of whether they were learned or novel, and this trend did not differ by depression
severity or group. This could (1) reflect a general tendency to form a more negative impression
25

to varied positive and negative feedback, or (2) be due to feedback contingency, participants may
falsely believe there are more negatively valenced beans due to lack of feedback for positive
beans incorrectly avoided. However, this could also be due to the limitations of our online study.
We did find evidence to suggest that some participants were simply employing a strategy of
“button mashing” during the test round. This may have been facilitated by distractions in their
environment, since participation occurred completely online. Additionally, during the test block
point totals were no longer tracked and feedback was not given; therefore, the participant did not
receive external motivation to perform well during this portion of the game. Future studies
should seek to conduct in-person sessions when possible to minimize distractions and incentivize
subjects to perform well during the test round (i.e. providing a monetary bonus for correct
classification).
We also examined how approach and avoidance behavior may have been influenced by
the running point total during learning trials, and whether this was related to anxiety or
depression severity. Depression appeared to have a significant effect on whether behavior was
influenced by points. Anxiety also had an effect, although it did not reach significance. We found
that the correlation between points and avoid tended to be positive for people with low
depression scores, and negative for people with high depression scores. According to
exploration-exploitation theory (March, 1991), the normative strategy would be to increase
approach behavior as points decreased towards 0, and to be more risk averse (increase
avoidance) as points approach 100. Thus, from our data it seems that people with low depression
scores are abiding by the optimal strategy; they are decreasing their approach behavior as points
increase. However, for people with higher levels of depression, we found a positive correlation
such that as points increased, approach increased. This is the opposite of what we would expect.
26

We tested to see if this directional discrepancy between low-depression and high-depression
people would be moderated by a scarcity condition. That is, did people change their approach
and avoidance strategy when faced with few points (<50) versus ample points (>50)? In the low
points condition, all participants maintained the optimal strategy of approaching more beans as
points decreased. In the high points condition, people without depression maintained the
normative strategy, but people with higher levels of depression tended to approach more as their
points increased, as well as when their points decreased. This suggests a curvilinear relationship
between points and approach behavior for these participants, such that as points trended towards
the extremes (0 or 100), they would approach more. One explanation could be the temporal
nature of getting to extreme point values. Initially, each game begins at 50 points; therefore,
participants with depression may demonstrate a prolonged period of initial avoidance. As the
game progresses, they may start to modify their behavior and begin to approach beans, leading to
a positive correlation between approach and extreme point values.
In this experiment, participants were asked to self-identify their mental health diagnoses.
This was another significant limitation, as we could not verify the definition and truthfulness of
their diagnostic status. Furthermore, we did not incorporate a pre-screening procedure into our
recruiting methodology, and this resulted in a much lower sample size of self-identified TRD
patients. To improve diagnostic power, future studies may use official medical records or a
clinician interview to determine treatment-resistant status.
Experiment 2
The presence of anhedonia is not only central to MDD, but there is some evidence for
heightened anhedonia in antidepressant non-responders (Klein, 1974 as cited in Gorwood, 2008;
27

McMakin et al., 2012; Rizvi et al., 2015). However, there has been a dearth of research
investigating the contribution of dimensionality in anhedonia to treatment resistance, and it is
often treated as a unitary construct. Psychometric scales developed to measure anhedonia such as
the PID-5 Anhedonia and Depressivity subscales (Krueger, Derringer, Markon, Watson, &
Skodol, 2012) and the BDI anhedonia subscore include the same 3 measures: lassitude, loss of
interest, and loss of pleasure. We know that each of these behaviors are governed by distinct
neural networks. Lassitude, or excess fatigue, has been associated with both depression and
anxiety over insomnia, and also showed specificity for depression over anxiety (Koffel &
Watson, 2009). It has been identified as both a residual symptom after SSRI and SNRI treatment
and as a side-effect of aforementioned treatment (Targum & Fava, 2011). In the former case,
lassitude is generally responsive to administration of the dopamine receptor agonist bupropion
(Targum & Fava, 2011). Lassitude can also be treated with medications acting primarily on the
noradrenergic system and psychostimulants. There is some evidence for reduced reward
signaling and reward anticipation as risk factors for depression (Whitton, Treadway, &
Pizzagalli, 2015). Studies have found lower ventral striatum activity during reward anticipation
in high-risk youth for unipolar depression (Gotlib et al., 2010; Olino et al., 2014). The ventral
striatum is frequently activated during the anticipatory phases of reward and loss tasks, which is
also the site of phasic dopamine signaling which has been found to code for reward prediction
error (Oldham et al., 2018; Pagnoni, Zink, Montague, & Berns, 2002; Wilson et al., 2018).
However, we do not know if there are endemic differences in anticipatory anhedonia in
antidepressant responders versus non-responders. Berridge (2003) found that hedonic pleasure
may recruit the opioid system over the dopamine system, and is associated with an independently
operating region of the Nucleus Accumbens (NAcc). Current measures of anhedonia use
28

quantitative severity to distinguish antidepressant non-responsiveness, however, we are lacking
in research on the qualitative anhedonic nature of TRD.
Both anhedonia and depression are related to the personality traits of extraversion and
neuroticism (Kotov, Gamez, Schmidt, & Watson, 2010). DeYoung and Krueger (2018) argued
that most psychopathology symptoms could be characterized as maladaptive extremes of
variation in normative personality constructs. Both extraversion and reward-seeking behavior
have been linked to the neurotransmitter dopamine, specifically via the limbic pathways
governing reward and reinforcement learning (Cohen, Young, Baek, Kessler, & Ranganath,
2005). A study on depression and reward sensitivity using the PRT (Pizzagalli, Jahn, & O’Shea,
2005), in which participants had to choose between two differentially rewarded stimuli, found
that over time participants tended to develop a bias for selecting the stimulus that was more
frequently rewarded, and this change in response bias negatively correlated with depression
severity. In a recent study replicating the protocol from Pizzagalli et al. (2005), Blain and
colleagues (2020) found that Extraversion was a strong predictor of latent reward sensitivity as
well as latent depression, but depression did not explain additional variance in reward sensitivity
over individual differences in Extraversion and Neuroticism. Thus, they concluded that having
depressive symptoms may be an indicator of low Extraversion and high Neuroticism.
Neuroticism is generally defined as the tendency to experience negative affect and
engage in avoidance behaviors, and it is associated with many psychopathologies (Gray, 1991;
Widiger, 2011). Studies have found greater amygdala activity in response to negative versus
neutral stimuli in individuals with high neuroticism. This was linked to possession of a short
variant of the serotonin transporter-linked promoter region (5-HTTLPR) gene, which reduces
transcription and leads to lower levels of the serotonin transporter in the synaptic cleft (Ormel et
29

al., 2013). Furthermore, individuals with high neuroticism have been found to also have negative
biases in attention and interpretation of emotional information (Ormel et al., 2013). There is
some evidence to suggest that acute SSRI administration may moderate the relationship between
Neuroticism and activity in the subgenual Anterior Cingulate Cortex (ACC) in response to
fearful facial expressions, such that Neuroticism scores were negatively correlated with neural
activity in this region under citalopram, and positively correlated under acute tryptophan
depletion (L-tryptophan is the chemical pre-cursor to serotonin; Hornboll et al., 2018). Blood-
oxygen-level-dependent (BOLD) signal in the subgenual ACC has been found to correlate with
anhedonia severity, and this region has been linked to MDD and several other psychopathologies
(Rudebeck et al., 2014). Quilty, Meusel, & Bagby (2008) additionally found that patients treated
with SSRIs demonstrated a greater decrease in Neuroticism, which correlated with improvement
in depression severity. Therefore, depression that is treatment resistant to SSRIs may be an
indicator of relatively more stable or more extreme levels of Neuroticism. These lines of
evidence indicate that some of the heterogeneity in depression responsiveness to monoamine
antidepressants may be explained by personality variations in Extraversion and Neuroticism.
The aims of this study are: first, to explore how qualitative aspects of anhedonia are
related to antidepressant responsiveness and task outcomes. Specifically, we hypothesize that
anhedonia will factor analyze into 3 components: lassitude, wanting, and consummatory
pleasure. Furthermore, the components of anhedonia with the strongest neural associations to the
mesolimbic dopamine reward system (wanting and lassitude) should drive differences between
antidepressant responders and non-responders. We also hypothesize that these differences in
incentive salience and appetitive motivation may have an effect on BeanFest outcomes such that
the TRD group will demonstrate (1) poorer learning accuracy for good beans and (2) decreased
30

approach behavior driven by greater impairments to dopaminergic network connections. Second,
we aim to investigate whether there are distinct individual differences in Neuroticism and
Extraversion related specifically to TRD. We hypothesize that the TRD group will have higher
neuroticism than the regular MDD group, and that this will result in a more negative learning and
response bias due to a greater sensitivity to negative feedback. Furthermore, we hypothesize that
depression in general will be associated with lower Extraversion, and that this will also be related
to a more negative learning bias on BeanFest. To test these hypotheses, we factor analyzed the
items in three anhedonia and motivation scales as well as the Extraversion and Neuroticism
component of the Big Five Aspect Scale (DeYoung, Quilty, & Peterson, 2007) to confirm latent
structure, and then investigated whether each component of these scales had discriminant
validity for TRD compared to regular MDD.

METHODS
Participants
Participants were again recruited from ResearchMatch.org (N = 137). This study used a
prescreen survey to determine eligibility and control for group sizes. We recruited from three
populations of participants: those with a self-identified diagnosis of MDD who responded well to
a time appropriate course of antidepressants, those who did not respond to two different
antidepressants or were diagnosed with treatment-resistant depression, and healthy controls.
Exclusion criteria were identical to Experiment 1. Participants were told that they would be
compensated $10.00 for their time, as well as have an opportunity to earn an additional $5.00
contingent on their performance in the test block of BeanFest.
31

Procedure
Due to COVID-19 safety protocols and restrictions, we modified the original protocol of
this study to be able to conduct it fully online. After passing the eligibility survey, participants
were emailed an anonymous link to the online study hosted on the PsyToolkit server (Stoet,
2010; 2017). Participants gave informed consent and then completed a short survey on their
diagnosis and medication, including a two-item self-report assessment of their depression
symptoms. Then, they completed 3 validated scales relating to anhedonia and motivation. Lastly,
they played 3 learning blocks and 1 test block of the BeanFest game as in Experiment 1.
Self-Report Measures
Patient Health Questionnaire-2 (PHQ-2; Kroenke, Spitzer, & Williams, 2003):
Shortened version of the PHQ-9 consisting of 2 items on frequency of depressed mood and
anhedonia over the past 2 weeks. Construct validity was assessed using the General Health
Survey, reported sick days, clinic visits, and symptom-related difficulty. A score of 3 or more
has been validated against the mental health professional interview in a sample of 580 patients,
with a sensitivity of 83% and a specificity of 92% for major depression (Kroenke, Spitzer, &
Williams, 2003).
Temporal Experience of Pleasure Scale (TEPS; Gard, Gard, Kring, & John, 2006):
This is an 18 item 6-point Likert scale consisting of two subscales that measure consummatory
(TEPS-CON) and anticipatory (TEPS-ANT) experience of pleasure. Participants were recruited
from undergraduate psychology student population at University of California at Berkeley and
the University of Miami. Cronbach’s alpha = .79. From the original validation sample of 665
women and 369 men, 92 women and 61 men retook the scale after 5-7 weeks. The test-retest
32

reliability was high at r = .81 (p < .001). Factor analysis consistently revealed two factors, with
TEPS-ANT alpha = .72, and TEPS-CON alpha = .64, and correlation between the two scales was
r = .46.
Motivation and Energy Inventory (MEI; Fehnel, Bann, Hogue, Kwong, & Mahajan,
2004): A 27-item Likert scale that was validated on a total of 809 subjects (60.7% women) who
at the time of study were experiencing a recurrent major depressive episode for at least 8 weeks
but no more than 24 months, and scored a minimum of 20 on the HAM-D. Internal consistency
analyses of the subscales resulted in a Cronbach’s α range of .70 – .90. Confirmatory factor
analysis of the three subscales revealed high correlations between them (range: .84 - .93). The
MEI subscales had high correlation with the Quality of Life in Depression Scale (QLDS; Hunt &
McKenna, 1992), and moderate negative correlations with the Hamilton Depression Scale
(HAM-D; Hamilton, 1960). Furthermore, each MEI subscale was able to distinguish between
responder and non-responders in an 8-week antidepressant vs placebo trial (p < 0.001 for all
pairwise t-tests).
Dimensional Anhedonia Rating Scale (DARS; Rizvi et al., 2015): A 17-item 5-point
Likert scale that measures desire, motivation, effort and consummatory pleasure across 4 hedonic
categories. Following item selection procedures and reliability testing using data from
community participants (N=229) (Study 1), the 17-item scale was validated in an online study
with community participants (N=150) (Study 2). The DARS was also validated in unipolar or
bipolar depressed patients (n=52) and controls (n=50) (Study 3). Principal components analysis
of the 17-item DARS revealed a 4-component structure mapping onto the following categories:
hobbies, food/drink, social activities, and sensory experience. Reliability of the DARS subscales
was high across studies (Cronbach's α = .75–.92). The DARS also demonstrated good convergent
33

and divergent validity. Hierarchical regression analysis revealed the DARS showed additional
utility over the SHAPS in predicting reward function and distinguishing MDD subgroups.
Big Five Aspect Scale (BFAS; DeYoung, Quilty, & Peterson, 2007). A 100-item 5-
point Likert scale assessing the two factor components of each of the Big Five personality
constructs. It is thought that Extraversion and Neuroticism are personality factors related to a
diathesis for psychopathology. The scale was validated on two populations: (1) 480
undergraduates in southern Ontario (299 women) and (2) 423 members of the Eugene-
Springfield community sample (ESCS) (281 women). A retest sample was taken approximately
one month after the initial scale administration, with a sample size of 90 undergraduates. The
scale demonstrated good reliability, with Cronbach’s alpha for the ESCS = .83, the initial
university sample = .81, and a retest university sample = .83. Test-retest correlation for the
community sample was 0.81. Correlations between the BFAS and the Big Five Inventory (BFI)
were high, ranging from .72 - .96. Correlations in the ESCS with NEO-PI-R scale was also high,
ranging from .80 - .92, demonstrating that the scale was indeed measuring the standard Big Five.
Only subscores for Extraversion and Neuroticism were used in the analysis.
Analyses
Anhedonia factor validity. First we conducted principal components analyses (PCA)
and confirmatory factor analyses (CFA) of the MEI, TEPS, and DARS self-report scales to
determine latent structure of each scale and assess conformity to the theoretical subscales with
which they were developed. Then, we ran an exploratory factor analysis on the combined items
of all 3 anhedonia scales (MEI, TEPS, and DARS) to uncover empirical latent factors of
anhedonia in our sample. These analyses were conducted using R version 3.5.1. PCA and CFA
analyses were conducted using the psych package (v 2.1.3; Revelle, 2021). PCA analyses were
34

conducted with a promax rotation, using parallel analysis to determine the number of
components to extract (Hayton, Allen, & Scarpello, 2004).
Diagnostic group effects. Similar to Experiment 1, we categorized participants into 3
groups: self-identified antidepressant non-responders (TRD; N = 23), self-identified
antidepressant responsive depressed (regular MDD; N = 73), and diagnosis naïve controls (HC;
N = 41). Nine one-way ANOVA tests for group differences were conducted on the social
motivation, physical energy, and mental energy subscales of the MEI, the anticipatory and
consummatory subscales of the TEPS, the four reward categories of the DARS. We also
explored effects on BeanFest end points. A one-way repeated-measures ANOVA was conducted
to determine if learning occurred on the average PHI coefficient per block. Next, we examined
95% CI graphs on mean accuracy by bean valence and approach by bean valence to determine
trends over time. To determine the effect of diagnosis on task outcomes, we conducted a one-
way ANOVA on learning bias with group as the between-subjects variable as well as an analysis
of covariance (ANCOVA) on response bias holding constant good and bad learning rates.
Analyses were conducted using IBM SPSS Statistics software (v26).
Personality and anhedonia symptomatology. 6 one-way ANOVAs were conducted on
E, N, and their subcomponents: E-Enthusiasm, E-Assertiveness, N-Withdrawn, and N-Volatility
to explore the effect of diagnostic groups.

35

RESULTS
Anhedonia factor analyses
MEI
Reliability analysis for this scale was good (Chronbach’s alpha = .92, 95% CI =
[.90,.94]). A principal components analysis (PCA) was performed, and examination of the scree
plot suggested 2 factors as the best solution. For theoretical purposes, a 3-factor solution was
also examined as the scale was developed to reflect 3 subscales: mental energy (ME), social
motivation (SM), and physical energy (PE). In the 3-factor solution, Q1, 2, 19-30 loaded onto
Factor 1 (which loosely mapped onto SM and PE), and Q2-18 loaded onto Factor 2 (which
loosely mapped onto ME). Only Q11 (social conversations), 29 (interest in talking with others),
and 30 (interest in social activities) had loadings greater than .3 on the third factor.

Table 4. 2-factor solution of the MEI
36

* starred items are reverse coded

The 2-factor solution explained 54.2% of the variance whereas the 3-factor solution
explained only 58.0% of the total variance. Therefore, we maintained the 2-factor solution.
Factor 1 loosely mapped onto interest and enthusiasm for recreational activities while Factor 2
loosely mapped onto energy and motivation for tasks and chores (see Table 4). The exceptions
were: Q2 (satisfied with accomplishments) originally loaded mostly onto mental energy and now
loaded onto Factor 1 (SM & PE). Q4 (energy during the day), 17 (physically tired), & 18
(exhausted) originally loaded onto physical energy but now loaded onto Factor 2 (ME), and Q13
(prefer to be alone) originally mapped onto SM but now loaded onto Factor 2 (ME). For
37

purposes of this study, we re-interpreted Factor 1 as Pursuit of Fun and Factor 2 as Motivation
and Drive.
TEPS
Reliability analysis indicated good reliability (Chronbach’s alpha = .84, 95% CI = [.81,
.88]). Examination of the Scree plot suggested 4 factors as the best solution. 2- and 4-factor
solutions were examined, because theoretically the scale was created on only 2 factors:
anticipation and consummation. The 4-factor solution yielded a cumulative variance of 41.8%
while the 2-factor solution yielded a cumulative variance of 36.3%. The 2-factor solution is
shown below as it is more parsimonious and theoretically based. The items group exactly onto
TEPS-con and TEPS-ant, except for item 9 (I love it when people play with my hair) which did
not have a loading above .30 on either factor. Due to human error Q18 (TEPS-ant) was left out of
data collection.

Table 5. 2-factor solution of the TEPS

38

* starred items are reverse scored

DARS
Reliability analysis indicated good reliability (Chronbach’s alpha = .93, 95% CI = [.91,
.95]). Parallel analysis suggested 4 factors as the best solution. The 4-factor solution explained a
total of 71.3% of the variance. Each DARS item loaded onto its respective reward category.
Factor 1 represents hobbies, Factor 2 was sensory, Factor 3 was social, and Factor 4 was food.
Total Anhedonia EFA
All items from TEPS, MEI, and DARS were entered into one dataset. Reliability analysis
indicated good reliability (Chronbach’s alpha = .94, 95% CI = [.93, .96]). Examination of the
Parallel analysis suggested 6 factors as the best solution (Figure 7a).

Figure 7. Parallel analysis scree plots for overall anhedonia construct
(a) (b)

39

Scree plots showing (a) the 6-factor solution for all scales combined, and (b) the 3-factor solution when the DARS
was excluded.

The 6-factor solution explained a total of 55.6% of the variance. Examination of the 6
factors fell as follows. Factor 1 (enthusiasm) consisted of the Pursuit of Fun empirical factor of
the MEI grouped with most of the TEPS Anticipatory factor (exceptions were all food-based:
TEPS 3 – favorite food anticipation, TEPS 7 – anticipation of eating at a restaurant, TEPS 14 –
anticipation of dessert). Factor 2 (motivation) consisted of the Motivation and Drive empirical
factor of the MEI grouped with TEPS 7. Factor 3 consisted of DARS hobbies and food/drink
items as well as TEPS 14 and TEPS 3 (food/hobbies). Factor 4 (hedonic experience) consisted
only of the TEPS-consummatory items. Factor 5 (sensations) consisted of only the 5 DARS
sensory experiences items. Factor 6 (social enjoyment) was composed of only the 4 DARS social
items and MEI 29 – speaking with others. Some of the factor interpretations contained
overlapping concepts. In the interest of obtaining a more parsimonious and interpretable solution,
we excluded DARS items and conducted another EFA on only the items from MEI and TEPS. A
scree plot analysis suggested a 3-factor solution, which explained 47.8% of the total variance
(Figure 7b).
Factor 1 contained the Pursuit of Fun component of the MEI and several TEPS-ant items.
Factor 2 included the Motivation and Drive component of the MEI and TEPS 7. Factor 3
contained mostly TEPS-con items. Exceptions included: TEPS 3 – anticipation of food loaded
onto Factor 3, TEPS 9 – play with hair and TEPS 14 – anticipation of dessert did not have
loadings > .3 on any factor. We used this 3-factor solution for the empirically determined latent
structure of anhedonia in the below SEM analysis, because the factors agreed with the
theoretically determined latent structure of anhedonia present in the literature. Factor 1 was
40

interpreted as overall enthusiasm (wanting), Factor 2 was interpreted as energy and drive
(lassitude), and Factor 3 was interpreted as consummatory “liking”. The DARS items, TEPS 7,
TEPS 9, and TEPS 14 were excluded from all future anhedonia structural analyses.

Table 6. 3-factor solution of the MEI and TEPS items
41

42

Diagnostic group effects
Depression severity and diagnostic group
A one-way ANOVA on depression score by group was significant (F = 12.78, p < .001),
and Bonferroni adjusted pairwise comparisons revealed significant differences between regular
MDD and TRD (Mdiff = -2.22, se = .44, p <.001) as well as HC and TRD (Mdiff = -1.88, se =
.48, p < .001), such that the TRD group had a significantly higher mean PHQ-2 score (Figure 8).
No significant differences were found between HC and regular MDD (p = 1.000).

Figure 8. Mean depression severity by diagnostic group

Participants diagnosed with TRD scored significantly higher on the PHQ-2 than the other 2 groups.

Anhedonia subscales and diagnostic group
Due to the length and online format of the self-report surveys (over 150 items across 5
questionnaires) we had extensive casewise missing data; especially in the BFAS scale, which
43

was ordered last. Multiple imputation was conducted using fully conditional specification
methods in IBM SPSS v26 (IBM Corporation, 2021). Multiple imputation has been shown
superior to other ways of dealing with missing data due to its method of using correlations to
produce estimates based on the underlying distribution of a variable (Manly & Wells, 2014).
SPSS generated 5 imputed datasets, which were pooled for analysis using Rubin’s rules (Rubin,
1987).

Table 7. Missing data rates
N Complete
Cases
% Complete
Cases
MEI 91 66.42%
TEPS 103 75.18%
DARS 129 94.16%
BFAS-E 46 33.58%
BFAS-N 74 54.01%
The first column denotes number of participants who fully completed each scale, out of 137 total subjects. The
PHQ-2 was excluded due to negligible missing cases.

Table 8. Self-report scales descriptive statistics by group
Original
data
Group Statistics TEPS MEI DARS
Diagnosis
Naive
N 30 27 36
Mean 79.53 56.89 60.00
Std. Error 2.02 4.36 1.25
Regular MDD
N 52 46 70
Mean 69.42 58.83 50.59
Std. Error 1.86 2.61 1.42
44

TRD
N 21 18 23
Mean 60.43 39.22 40.30
Std. Error 2.81 3.57 3.17
Total
N 103 91 129
Mean 70.53 54.37 51.38
Std. Error 1.40 2.12 1.16
Pooled
Imputed
Diagnosis
Naive
N 41 41 36
Mean 79.11 65.75 60.00
Std. Error 1.72 3.99 1.25
Regular MDD
N 73 73 70
Mean 70.59 62.68 50.59
Std. Error 1.62 2.24 1.42
TRD
N 23 23 23
Mean 59.29 44.91 40.30
Std. Error 2.68 3.65 3.17
Total
N 137 137 129
Mean 71.24 60.61 51.38
Std. Error 1.23 1.89 1.16
Descriptive statistics for anhedonia scales by group for original and pooled imputed data

TEPS
One-way ANOVAs revealed significant main effect of group on TEPS consummatory
(TEPSc; F = 13.58, p < .001, partial η
2
= .17; imputed partial η
2
range = [.132, .139]) and TEPS
anticipatory (TEPSa; F = 11.51, p < .001, partial η
2
= .18; imputed partial η
2
range = [.195,
.202]) subscales. Bonferroni corrected pairwise comparisons for TEPSc showed both depression
groups had significantly lower scores than the HC group (HC-MDD Mdiff = 6.11, se = 1.44, p <
45

.001; HC-TRD Mdiff = 9.03, se = 1.90, p < .001). However, only the TRD group scored
significantly lower on the TEPSa subscale (Mdiff = 10.74, se = 2.24, p < .001).
MEI
One-way ANOVAs revealed significant main effect of group on MEI mental energy
(MEIme; F = 4.25, p = .017, partial η
2
= .08; imputed partial η
2
range = [.045, .055]), social
motivation (MEIsm; F = 4.56, p = .013, partial η
2
= .08; imputed partial η
2
range = [.088,
.095]), and physical energy (MEIpe; F = 7.36, p = .001, partial η
2
= .11; imputed partial η
2

range = [.128, .132]) subscales. From Bonferroni corrected pairwise comparisons for MEIme,
only the TRD group had significantly lower scores from regular MDD (Mdiff = -5.76., se = 2.12,
p < .023). On social motivation, the TRD group scored significantly lower than both HC (Mdiff =
-7.03, se = 2.60, p = .024) and regular MDD (Mdiff = -6.53, se = 2.34, p = .018). On physical
energy, the TRD group again scored significantly lower than both HC (Mdiff = -7.23, se = 2.00,
p = .001) and regular MDD (Mdiff = -6.05, se = 1.79, p = .003), which were not significantly
different.
DARS
We did not impute DARS data as there was only 2 missing casewise from the HC group,
and they were left out of the analysis. 4 one-way ANOVAs were computed for each reward
category of DARS (hobbies, DARSh; food, DARSf; social, DARSsoc; sensory, DARSsen) and
Bonferroni corrected pairwise comparisons were conducted on significant test results. Group had
a significant effect on hobbies (F = 20.32, p < .001, partial η
2
= .235) such that HC had
significantly higher DARSh than regular MDD (Mdiff = 1.94, se = .68, p = .014), and regular
MDD had significantly higher DARSh than TRD (Mdiff = 3.76, se = .81, p < .001). Group had a
46

significant effect on food (F = 7.50, p < .001, partial η
2
= .103) such that HC had significantly
higher DARSf than TRD (Mdiff = 3.28, se = .85, p = .001). Group had a significant effect on
social activities (F = 13.93, p < .001, partial η
2
= .180) such that HC had significantly higher
DARSsoc than regular MDD (Mdiff = 2.95, se = .77, p = .001), and TRD (Mdiff = 5.12, se =
1.01, p < .001). TRD had a non-significantly lower DARSsoc score than regular MDD (Mdiff = -
2.17, se = .91, p = .057). Group had a significant effect on sensory experiences (F = 8.63, p <
.001, partial η
2
= .117) such that HC had significantly higher DARSsen than regular MDD
(Mdiff = 2.53, se = .97, p = .031), and TRD (Mdiff = 5.26, se = 1.28, p < .001). TRD had a non-
significantly lower DARSsen score than regular MDD (Mdiff = -2.74, se = 1.15, p = .058). The
only reward category where TRD group scored significantly lower than regular MDD was
DARSh.
Extraversion
Group had a significant effect on E (F = 8.61, p < .001, partial η
2
= .195; imputed partial
η
2
= .063), and Bonferroni corrected pairwise comparisons revealed the TRD group scored
significantly lower than HC (Mdiff = -14.08, se = 3.40, p < .001) and regular MDD (Mdiff = -
8.57, se = 3.18, p = .027), which were not significantly different. Furthermore, differences
between diagnosed TRD and non-TRD groups were driven mainly by the E-Enthusiasm (F =
9.223, p < .001, partial η
2
= .160; imputed partial η
2
range = [.113, .117]) and not the E-
Assertiveness (F = 7.28, p = .001, partial η
2
= .142; imputed p = [.313, .271]) component of this
trait.
Neuroticism
47

A one-way ANOVA using listwise deletion showed that group did not have a significant
effect on N (p = .73) due to extremely high rates of casewise missingness (cases with missing
items: 56% HC, 62% regular MDD, and 100% TRD). Therefore, the following results are
reported on the 5 imputed datasets, with statistical ranges shown. Group had a significant effect
on N in all imputations (F range = [8.08, 9.35], p < .001, partial η
2
range = [.105, .122]).
Bonferroni pairwise comparisons revealed HC had significantly lower N than regular MDD
(Mdiff = [-5.43, -6.02], se = [2.22, 2.24], p = [.049, .021]) and TRD (Mdiff = [-11.90, -12.67], se
= [2.93, 3.00], p < .001). Regular MDD also had significantly lower N than TRD for 2 out of 5
imputations (Mdiff = [-6.92, -7.05], se = [2.71, 2.73], p = [.036, .033]). Furthermore, these
differences were more driven by N-Withdrawn (imputed F = [12.81, 14.54], p < .001, partial η
2

= [.161, .178]) than N-Volatility (imputed F = [2.67, 3.72], p = [.073, .027])

Table 9. Personality subscales by diagnostic group
Group Statistics E N
Original
data
Diagnosis
Naïve
N 24 18
Mean 68.54 63.44
Std. Error 1.69 2.55
Regular MDD
N 37 28
Mean 63.03 64.64
Std. Error 1.65 2.24
TRD
N 13 0
Mean 54.46 N/A
Std. Error 3.32 N/A
Total N 74 46
48

Mean 63.31 64.17
Std. Error 1.26 1.67
Pooled
Imputed
Diagnosis
Naïve
N 41 41
Mean 67.73 64.72
Std. Error 1.96 1.79
Regular MDD
N 73 73
Mean 66.33 70.42
Std. Error 1.40 1.39
TRD
N 23 23
Mean 58.54 77.02
Std. Error 2.85 2.13
Total
N 137 137
Mean 65.44 69.82
Std. Error 1.09 1.03

Diagnostic group and BeanFest outcomes
Did learning occur?
We calculated the average PHI coefficient in the same way as Experiment 1 and
conducted a one-way repeated-measures ANOVA across blocks to confirm that learning
occurred in this sample (see Table 10 for descriptive statistics). We found a significant omnibus
difference (F = 22.22, p < .001), and pairwise Bonferroni adjusted comparisons showed
significant increases in PHI from block 1 to block 2 (Mdiff = -.08, se = .02), block 1 to block 3
(Mdiff = -.11, se = .02), and block 1 to block 4 (Mdiff = -.15, se = .02). This learning was driven
by improvements in learning of bad beans (Figure 10), which significantly differed across blocks
(F = 54.13, p < .001) and non-significant changes in accuracy for good beans across learning
49

blocks 1-3 (see figure below). These results held for all diagnostic groups and align with our
results in experiment 1. While all groups showed learning across blocks, accuracy improved
most for TRD participants (Figure 11).

Table 10. Learning and response bias during test

Learning
Beans
Novel Beans
Group Statistics LB Bad-leaning Good-leaning Neutral
Healthy Controls

N 41 41 41 41
Mean -0.28 -0.41 -0.27 -0.19
Std. Error 0.05 0.06 0.06 0.08
Regular MDD

N 71 71 71 71
Mean -0.08 -0.15 -0.05 -0.12
Std. Error 0.04 0.05 0.05 0.06
TRD

N 22 22 22 22
Mean -0.35 -0.40 -0.28 -0.26
Std. Error 0.08 0.09 0.09 0.12
Total

N 134* 134 134 134
Mean -0.19 -0.27 -0.16 -0.16
Std. Error 0.03 0.04 0.04 0.04
*3 participants were excluded from the analysis due to 100% good-classification in test block.

Table 11. PHI coefficient by block
95% Confidence Interval
Blockname Mean Std Error Lower Bound Upper Bound
LB1 .015 .016 -.017 .047
50

LB2 .091 .018 .054 .127
LB3 .127 .020 .088 .166
Test .168 .020 .128 .207

Figure 9. Number correct by bean valence

95% CI plots of good beans correctly approached and bad beans correctly avoided by block (Corr4 displays number
of good and bad beans correctly classified during test). Top chart pertains to bad beans and bottom chart shows
metrics for good beans.

Figure 10. Learning bean accuracy across blocks by group
51

Accuracy steadily improved across blocks (Left error bar = LB1 to right error bar = Test) for all participants.

Were differences in approach behavior present over time?
Approach behavior showed a trending decline across the 3 blocks but did not
significantly decrease until the test block (Figure 12; F = 45.64, p < .001), during which
participants were instructed to classify beans instead of approach or avoid them.

Figure 11. Approach across blocks
52

App4 shows classification of beans as “good” during test

Does diagnostic group predict learning bias?
To assess learning differences by bean valence across diagnostic groups we conducted a
one-way ANOVA on LB, which yielded a significant omnibus test (F = 7.04, p = .001). Pairwise
posthoc Bonferroni adjusted comparisons revealed significant differences between HC and
regular MDD (Mdiff = -.21, se = .07, p = .013) such that the HC group had a more negative
learning bias, as well as regular MDD and TRD (Mdiff = .28, se = .09, p = .007) such that the
TRD group had a more negative learning bias. Also, no significant differences were found for
HC and TRD (see Table 12 for group means). We also ran an ANCOVA test to determine
whether group differences would hold controlling for depression severity as measured by the
PHQ-2. We found group to be a significant predictor of LB (F = 5.36, p = .006), holding
constant depression severity (F = 1.38, p = .243), which did not have a significant effect on LB.
Posthoc pairwise Bonferroni corrected comparisons showed significant differences in adjusted
means for HC and regular MDD (Mdiff = -.20, se = .07, p = .017) as well as TRD and regular
MDD (Mdiff = -.23, se = .10, p = .047) such that the regular MDD group had a significantly less
negative adjusted mean LB than either of the other two groups.
53

We hypothesized these effects were due to the efficacy of antidepressants on the regular
MDD group, as the regular MDD group and diagnosis naïve group did not greatly differ in
depression severity. To test this, we conducted an independent samples t-test on participants who
were diagnosed with MDD to compare differences in learning bias and response bias between
those taking traditional antidepressants and those not on medication at the time of study
participation. Learning bias differed non-significantly between participants taking
antidepressants versus no medication (p = .092), and response bias also differed non-significantly
(p = .140). However, learning bias did differ significantly according to self-reported
antidepressant efficacy (Mdiff = .22, t = 2.82, p = .006), such that the subjects whose symptoms
improved with medication had a less negative learning bias. We also noticed similar effects of a
less negative response bias in the regular MDD group and performed the same within group t-
test, finding that response bias differed almost significantly based on perceived medication
efficacy (Mdiff = .15, t = 1.82, p = .074).

Table 12. Learning bias by group
Group Mean N Std Dev
Diagnosis Naïve -.28 41 .34
Regular MDD -.08 71 .38
TRD -.35 22 .36
Total -.19 134 .38

Does diagnostic group predict response bias when controlling for learning?
54

We entered the proportion of good beans classified correctly during test and the
proportion of bad beans classified correctly during test as covariates in an ANCOVA model
regressing RB on diagnostic group. Levene’s test for differences in error variance across groups
was not significant (F = .96, p = .387). While there were significant main effects of both good
and bad beans classified correctly during the test block on novel bean bias, there was no
significant additional effect of diagnostic group (p = .436).

Table 13. Response bias by group
Group Mean N Std Dev
Diagnosis Naïve -.33 41 .39
Regular MDD -.10 71 .37
TRD -.34 22 .40
Total -.21 134 .40

Anhedonia as a mediator between personality and TRD
Pearson’s correlations for pooled imputed Extraversion and Neuroticism scores, anhedonia
scales, and BeanFest performance are shown in Table 2A of the Appendix. The personality
constructs were significantly correlated with each other as well as the anhedonia scale totals, and
the BeanFest bias measures and accuracy were all significantly correlated with each other.
However, the self-report scales and BeanFest behavioral measures were not significantly
correlated with each other. Point biserial correlations were also computed between binary coded
diagnostic variables (MDD = 1 vs not MDD = 0; TRD = 1 vs not TRD = 0), empirically-derived
anhedonia factors, E and N, and BeanFest outcomes in Table 2A of the Appendix. Again, all
55

anhedonia and personality metrics were strongly correlated with each other. Additionally,
Anhedonia F2: motivation/energy was significantly correlated with BeanFest learning bias,
response bias, and accuracy such that lower motivation/energy was associated with greater
accuracy (driven by accuracy of bad beans) and more negative biases. Learning bias was
significantly correlated with both diagnoses such that TRD vs non-TRD had a more positive
bias, and MDD vs non-MDD had a more negative bias, while response bias was only
significantly negatively correlated with a diagnosis of MDD. Anhedonia F1: fun-seeking was
significantly negatively correlated with both diagnoses. N was significantly negatively correlated
with both diagnoses, while E was only significantly positively correlated with TRD.
DISCUSSION
Fractionating anhedonia
The first aim of this study was to explore the statistically separable qualitative aspects of
the anhedonia endophenotype related to antidepressant non-responsiveness. We hypothesized
based on existing anhedonia scales that there are three such facets: lassitude, hedonic liking, and
incentive salience (“wanting”). Indeed, our empirical self-report measures of anhedonia factor
analyzed into these facets in a sample of self-identified diagnosed MDD and diagnostically naive
subjects. The TEPS items split neatly between anticipatory (“When something exciting is coming
up in my life, I really look forward to it”) and consummatory (“The sound of crackling wood in
the fireplace is very relaxing”) subscales. The MEI items grouped into a factor measuring
motivation or drive (“you have trouble finishing things you started because you lost interest in
them”) and a second factor measuring pursuit of interests (“did you engage in recreational
56

activities or hobbies?”). This second factor grouped with the TEPS-ant subscale to form the
overall wanting factor of anhedonia.
Group differences in anhedonia
The next objective aimed to test whether these qualitative factors were differentially
relevant to antidepressant responders versus non-responders. We found depression severity as
assessed by the PHQ-2 was significantly higher for non-responders compared to responders and
controls. We also observed greater anhedonia severity on all self-report scales for non-
responders compared to responders, and for responders compared to diagnosis naïve participants.
This adds to existing evidence that treatment non-responders have more severe symptoms and
higher relapse rates (Gaynes, Rush, & Trivedi, 2008; Nie et al., 2018). Next, we tested for group
effects on the theoretically defined self-report subscales, which reflected the qualitative aspects
of anhedonia. The TEPS-ant and TEPS-con subscales had well defined items mapping onto
anticipatory and consummatory pleasure. Any diagnosis of MDD was associated with a
significant decrease in consummatory pleasure, but what separated the TRD group was their
significantly lower TEPS-ant scores. This could (1) be due to medication closing the gap in
anticipation scores for the regular MDD group, (2) mean that decreased anticipation is a residual
symptom post-treatment, or (3) indicate that lower anticipatory compared to consummatory
pleasure is an indicator of treatment resistance. The latter two options are more aligned with
existing evidence, as treatment with monoamine antidepressants have been found to induce
apathy and affective flattening (Goodwin, Price, Bodinat, & Laredo, 2017; Price, Cole, &
Goodwin, 2009).
The MEI assessed how often people were able to engage in everyday activities in the past
4 weeks through 3 subscales: mental energy (i.e. feeling overwhelmed), physical energy (i.e. felt
57

exhausted), and social motivation (i.e. interest in talking with others). An MDD diagnosis
without TRD did not significantly differ from controls on any of the MEI subscales. In fact,
MDD responders actually trended higher than controls in mental energy, perhaps due to the
efficacy of their medication. However, the TRD group scored significantly lower than their
regular MDD counterparts in all three components. The aforementioned attenuation of
anticipatory pleasure is likely related to these motivation and energy results.
The DARS scale aimed to assess the willingness of effort expenditure, interest,
motivation, and enjoyment of rewards in 4 hedonic categories, individualized to the self-
identified favorite experiences of the participant (see Table 1A in Appendix for examples). The
hedonic categories were non-social hobbies (i.e. playing video games), food/drinks (i.e.
brownies), sensory experiences (i.e. a hot bath), and social activities (i.e. going to a sporting
event). On average, the control group scored highest in all categories, and the non-responders
group scored lowest in all categories. A depression diagnosis resulted in significantly lower
scores in all categories. However, non-responders only scored significantly lower than
responders in the non-social hobbies category. Hobbies was non-significantly lower than food
scores for the TRD group, but hobbies scores were significantly higher than food in the control
group and trended higher in the regular MDD group. While anhedonia in regular depression
seems to affect functioning in all hedonic experiences, those with a TRD diagnosis tend to fare
worse in more effortful activities requiring internal motivation.
Group differences in BeanFest
Neuroticism was significantly lower for both groups of MDD compared to diagnosis
naïve controls, but not significantly lower for antidepressant non-responders, contrary to our
hypothesis. Furthermore, we hypothesized that higher N in the TRD group would be associated
58

with a more negative learning and response bias during test. The learning bias was indeed more
negative for the TRD group than the regular MDD group, but the TRD group did not
significantly differ from HC. Therefore, we believe that the effect of medication acted on
increasing approach behavior. Indeed, approach of learning beans was higher during all rounds
of the game for the regular MDD group. To further examine whether the cause of the less
negative bias for the regular MDD group was due to more positive classification of good or bad
beans, we graphed the 95% confidence intervals around the mean accuracy of learning beans as
well as response bias to novel beans during test (Figure 13). We found that the regular MDD
group was better at classifying good beans, but significantly worse at classifying bad beans than
the other 2 groups. This means there was an overall classification bias during test; they had a
much lower rate of bad-bean classification than the other 2 groups. For novel beans the regular
MDD group tended to have a significantly less negative classification bias toward the good-
leaning and bad-leaning beans than the other 2 groups. SSRIs and other antidepressant
medication may be moderating these results as the other 2 groups would not experience their
mood enhancing effect. Our evidence does not support the hypothesis that N contributes to
differences in learning and classification biases between groups. However, due to the high rates
of missingness from the N subscale, these results should be interpreted with caution.
Extraversion was significantly higher for diagnosis naïve controls than both MDD
groups, and also significantly higher for antidepressant responders compared to non-responders.
Furthermore, the component of E that was most depressed in the TRD group was enthusiasm (i.e.
make friends easily, have a lot of fun). Since Extraversion is predictive of reward sensitivity
(Blain et al., 2020), we can interpret these results as a dampening of predicted reward value. In
BeanFest, accuracy of good beans did not significantly change across learning blocks, while
59

accuracy of bad beans significantly increased from Block 1 to Block 2. It’s possible that this
evidence points toward decreased reward sensitivity coupled with increased punishment
sensitivity in people with TRD. Extraversion is strongly negatively correlated with all
components of anhedonia (see Table 2A in Appendix) and has been identified as a risk factor for
MDD. Yet there is a paucity of research on low
Figure 12. Group differences in BeanFest
(a)

(b)
60

BeanFest performance by group during the test block. (a) 95% confidence interval error bars around the mean group
differences in learning bias by bean valence. Overall the MDD had significantly less negative LB than both the
diagnosis naïve and TRD groups. (b) Error bar differences in mean response bias for the novel beans during test.
Diagnosis Naïve subjects responded significantly more negatively to the bad-leaning beans than the other novel bean
types, but this difference was not seen in regular MDD and TRD groups. The antidepressant responders tended to
have a more positive response bias than the other two groups.

extraversion, and especially E-enthusiasm as a possible predictor of TRD (for a review, see
Murphy, Sarris, & Byrne, 2017). Furthermore, we cannot conclude from this data whether low
E-enthusiasm is a predictor of TRD or an effect of it, only that it is an indicator of antidepressant
non-responsiveness. Future studies may focus on determining the causal relationship of E-
enthusiasm with treatment resistance, by examining a sample of treatment naïve people with
depression. Furthermore, presence of antidepressant usage in our current sample was not
manipulated and may confound the BeanFest biases. SSRIs have not directly been associated
with reward sensitivity, but it would be interesting for future studies to confirm the effect of
SSRIs on creating a more positive learning and response bias in an MDD population.
61

General Limitations
There were several important limitations in this study. First of all, due to COVID-19
regulations, we made changes to our original protocol to partner with USC Psychiatry and
conduct in-person data collection with students who had a clinical diagnosis. We modified data
collection to take place completely online. However, because of this we were not able to regulate
the environment as participants undertook the task, which may have led to impairments in
attention and learning as compared to an in-lab sample. This was evident in the fact that we had
to remove 3 participants from our data due to only classifying beans as good in the test block.
Second, we relied on self-identification of treatment resistance, which may have created an
imprecise group definition. Third, we did not control for presence or absence of SSRI medication
while participating in the study, as we could not randomly assign participants into groups.
Fourth, depression symptomatology was high in the Diagnosis Naïve group, as they did not
significantly differ from the regular MDD group in their PHQ-2 assessment. Due to the
pandemic, there has been widespread economic hardship, unemployment, and increased rates of
affective disorders. Therefore, some of the effects may have been confounded due to
undiagnosed clinical depression present in the control group. Fifth, we conducted data collection
from February-March 2021 for Experiment 2; therefore, some of the scales assessing social
motivation and frequency of desired social contact may have been confounded by the current
social climate and distancing protocols in effect for COVID-19. Future studies should take these
limitations into account.

62

REFERENCES
Akil, H., Gordon, J., Hen, R., Javitch, J., Mayberg, H., McEwen, B., Meaney, M. J., & Nestler,
E. J. (2018). Treatment resistant depression: A multi-scale, systems biology approach.
Neuroscience & Biobehavioral Reviews, 84, 272–288.
https://doi.org/10.1016/j.neubiorev.2017.08.019
Allen, T.A., Lam, R.W., Milev, R., et al. (2019). Early change in reward and punishment
sensitivity as a predictor of response to antidepressant treatment for major depressive
disorder: a CAN-BIND-1 report, 49(10), 1629-1638. doi:10.1017/S0033291718002441
American Psychiatric Association. (2013). Diagnostic and Statistical Manual of Mental
Disorders, 5th Edition: DSM-5 (5th ed.). American Psychiatric Publishing.
Beck, A. T., Epstein, N., Brown, G., & Steer, R. A. (1988). An inventory for measuring clinical
anxiety: Psychometric properties. Journal of Consulting and Clinical Psychology, 56(6),
893–897. https://doi.org/10.1037/0022-006X.56.6.893
Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An Inventory for
Measuring Depression. Archives of General Psychiatry, 4(6), 561–571. doi:
10.1001/archpsyc.1961.01710120031004
Berridge, K. C. (2003). Pleasures of the brain. Brain and Cognition, 52(1), 106–128.
https://doi.org/10.1016/s0278-2626(03)00014-9
Blain, S. D., Sassenberg, T. A., Xi, M., Zhao, D., & DeYoung, C. G. (2020). Extraversion but
not depression predicts reward sensitivity: Revisiting the measurement of anhedonic
63

phenotypes. Journal of Personality and Social Psychology. Published.
https://doi.org/10.1037/pspp0000371
Chen, C., Takahashi, T., Nakagawa, S., Inoue, T., & Kusumi, I. (2015). Reinforcement learning
in depression: A review of computational research. Neuroscience & Biobehavioral
Reviews, 55, 247–267. https://doi.org/10.1016/j.neubiorev.2015.05.005
Cohen, M. X., Young, J., Baek, J. M., Kessler, C., & Ranganath, C. (2005). Individual
differences in extraversion and dopamine genetics predict neural reward responses.
Cognitive Brain Research, 25(3), 851–861.
https://doi.org/10.1016/j.cogbrainres.2005.09.018
Coleman-Mesches, K. & McGaugh, J.L. (1995) Differential involvement of the right and left
amygdalae in expression of memory for aversively motivated training. Brain Research,
670, 75-81.
DeYoung, C. G., & Krueger, R. F. (2018). A Cybernetic Theory of Psychopathology.
Psychological Inquiry, 29(3), 117–138. https://doi.org/10.1080/1047840x.2018.1513680

DeYoung, C.G., Quilty, L.C., Peterson, J. B. (2007). Between facets and domains: 10 aspects of
the Big Five. Journal of Personality and Social Psychology, 93(5), 880-896. doi:
10.1037/0022-3514.93.5.880. PMID: 17983306
Elliott, R., Friston, K. J., & Dolan, R. J. (2000). Dissociable Neural Responses in Human Reward
Systems. The Journal of Neuroscience, 20(16), 6159–6165.
https://doi.org/10.1523/jneurosci.20-16-06159.2000
64

Fava, M. (2003). Diagnosis and definition of treatment-resistant depression. Biological
Psychiatry, 53(8), 649–659. https://doi.org/10.1016/s0006-3223(03)00231-2
Faul, F., Erdfelder, E., Lang, A. G., & Buchner, A. (2007). G*Power 3: A flexible statistical
power analysis program for the social, behavioral, and biomedical sciences. Behavior
Research Methods, 39(2), 175–191. https://doi.org/10.3758/bf03193146
Fazio, R. H., Eiser, J. R., & Shook, N. J. (2004). Attitude formation through exploration:
Valence asymmetries. Journal of Personality and Social Psychology, 87(3), 293–311.
https://doi.org/10.1037/0022-3514.87.3.293
Fazio, R. H., Pietri, E. S., Rocklage, M. D., & Shook, N. J. (2015). Positive Versus Negative
Valence: Asymmetries in Attitude Formation and Generalization as Fundamental
Individual Differences. In J. M. Olson & M. P. Zanna (Eds.), Advances in Experimental
Social Psychology (1st ed., Vol. 51, pp. 98–142). Academic Press.
Fehnel, S. E., Bann, C. M., Hogue, S. L., Kwong, W. J., Mahajan, S. S. (2004). The development
and psychometric evaluation of the motivation and energy inventory (MEI). Quality of
Life Research, 13(7), 1321–1336. https://doi.org/10.1023/b:qure.0000037502.64077.4d
Fydrich, T., Dowdall, D., & Chambless, D. L.. (1992). Reliability and validity of the beck
anxiety inventory. Journal of Anxiety Disorders, 6(1), 55–61.
https://doi.org/10.1016/0887-6185(92)90026-4
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. doi:10.1016/j.jrp.2005.11.001
65

Gaynes, B. N., Rush, A. J., Trivedi, M. H., Wisniewski, S. R., Spencer, D., & Fava, M. (2008).
The STAR*D study: treating depression in the real world. Cleveland Clinic Journal of
Medicine, 75(1), 57–66. https://doi.org/10.3949/ccjm.75.1.57
Gaynes, B. N., Warden, D., Trivedi, M. H., Wisniewski, S. R., Fava, M., & John Rush, A.
(2009). What Did STAR ✫D Teach Us? Results From a Large-Scale, Practical, Clinical
Trial for Patients With Depression. FOCUS, 10(4), 510–517.
https://doi.org/10.1176/appi.focus.10.4.510
Goodwin, G., Price, J., De Bodinat, C., & Laredo, J. (2017). Emotional blunting with
antidepressant treatments: A survey among depressed patients. Journal of Affective
Disorders, 221, 31–35. https://doi.org/10.1016/j.jad.2017.05.048
Gorwood, P. (2008). Neurobiological mechanisms of anhedonia. Dialogues in Clinical
Neuroscience, 10(3), 291–299. https://doi.org/10.31887/dcns.2008.10.3/pgorwood
Gotlib, I. H., Hamilton, J. P., Cooney, R. E., Singh, M. K., Henry, M. L., & Joormann, J. (2010).
Neural Processing of Reward and Loss in Girls at Risk for Major Depression. Archives of
General Psychiatry, 67(4), 380. https://doi.org/10.1001/archgenpsychiatry.2010.13
Gray, J. A. (1987). The psychology of fear and stress (2nd ed.). New York, NY: Cambridge
University Press.
Hamilton, M. (1960). A rating scale for depression. Journal of Neurology, Neurosurgery and
Psychiatry, 23, 56–62.
66

Hayton, J. C., Allen, D. G., & Scarpello, V. (2004). Factor Retention Decisions in Exploratory
Factor Analysis: a Tutorial on Parallel Analysis. Organizational Research Methods, 7(2),
191–205. https://doi.org/10.1177/1094428104263675

Hornboll, B., Macoveanu, J., Nejad, A., Rowe, J., Elliott, R., Knudsen, G. M., Siebner, H. R., &
Paulson, O. B.. (2018). Neuroticism predicts the impact of serotonin challenges on fear
processing in subgenual anterior cingulate cortex. Scientific Reports, 8(1).
https://doi.org/10.1038/s41598-018-36350-y
Hunt, S.M., McKenna, S.P. (1992). The QLDS: A scale for measurement of quality of life in
depression. Health Policy, 22, 307–319.
IBM Corporation. (2021). Method (Multiple Imputation). Method (Multiple Imputation).
https://www.ibm.com/docs/en/spss-statistics/SaaS?topic=imputation-method-multiple
Johnson, S.L., Turner, R.J. & Iwata, N. (2003). BIS/BAS Levels and Psychiatric Disorder: An
Epidemiological Study. Journal of Psychopathology and Behavioral Assessment 25, 25–
36. https://doi.org/10.1023/A:1022247919288
Kasch, K. L., Rottenberg, J., Arnow, B. A., & Gotlib, I. H. (2002). Behavioral activation and
inhibition systems and the severity and course of depression. Journal of Abnormal
Psychology, 111(4), 589-597. http://dx.doi.org/10.1037/0021-843X.111.4.589
Kautzky, A., Dold, M., Bartova, L., Spies, M., Kranz, G. S., Souery, D., … Kasper, S. (2018).
Clinical factors predicting treatment resistant depression: affirmative results from the
67

European multicenter study. Acta Psychiatrica Scandinavica, 139(1), 78–88. doi:
10.1111/acps.12959
Kimbrel, N. A., Nelson-Gray, R. O., & Mitchell, J. T. (2012). BIS, BAS, and bias: The role of
personality and cognitive bias in social anxiety. Personality and Individual
Differences, 52(3), 395–400. doi: 10.1016/j.paid.2011.10.041
Klein, D. F. (1974). Endogenomorphic Depression. Archives of General Psychiatry, 31(4), 447.
https://doi.org/10.1001/archpsyc.1974.01760160005001
Knutson, B., Westdorp, A., Kaiser, E., & Hommer, D. (2000). FMRI Visualization of Brain
Activity during a Monetary Incentive Delay Task. NeuroImage, 12(1), 20–27.
https://doi.org/10.1006/nimg.2000.0593
Koffel, E., & Watson, D. (2009). The two-factor structure of sleep complaints and its relation to
depression and anxiety. Journal of Abnormal Psychology, 118(1), 183–194.
https://doi.org/10.1037/a0013945
Kotov, R., Gamez, W., Schmidt, F., & Watson, D. (2010). Linking “big” personality traits to
anxiety, depressive, and substance use disorders: A meta-analysis. Psychological
Bulletin, 136(5), 768–821. https://doi.org/10.1037/a0020327

Kroenke, K., Spitzer, R. L., & Williams, J. B. W. (2003). The Patient Health Questionnaire-2:
Validity of a Two-Item Depression Screener. Medical Care, 41(11), 1284–1292.
https://doi.org/10.1097/01.mlr.0000093487.78664.3c
68

Krueger, R. F., Derringer, J., Markon, K. E., Watson, D., & Skodol, A. E. (2012). Initial
construction of a maladaptive personality trait model and inventory for. DSM–5.
Psychological Medicine, 42, 1879–1890. http://dx.doi.org/10.1017/S0033291711002674
Lamers, F., Oppen, P. V., Comijs, H. C., Smit, J. H., Spinhoven, P., Anton J. L. M. Van Balkom,
… Penninx, B. W. J. H. (2011). Comorbidity Patterns of Anxiety and Depressive
Disorders in a Large Cohort Study. The Journal of Clinical Psychiatry, 72(03), 341–348.
doi: 10.4088/jcp.10m06176blu

Lemonde, S., Du, L., Bakish, D., Hrdina, P., & Albert, P. R. (2004). Association of the C(–
1019)G 5-HT1A functional promoter polymorphism with antidepressant response. The
International Journal of Neuropsychopharmacology, 7(4), 501–506.
https://doi.org/10.1017/s1461145704004699
Lattanzi, L., Dellosso, L., Cassano, P., Pini, S., Rucci, P., Houck, P. R., … Cassano, G. B.
(2002). Pramipexole in treatment-resistant depression: a 16-week naturalistic
study. Bipolar Disorders, 4(5), 307–314. doi: 10.1034/j.1399-5618.2002.01171.x
Little, A. (2009). Treatment-Resistant Depression. American Family Physician, 80(2), 167-172.
Malhi, G. S., Parker, G. B., Crawford, J., Wilhelm, K., & Mitchell, P. B. (2005). Treatment-
resistant depression: resistant to definition? Acta Psychiatrica Scandinavica, 112(4),
302–309. https://doi.org/10.1111/j.1600-0447.2005.00602.x
69

Manly, C. A., & Wells, R. S. (2014). Reporting the Use of Multiple Imputation for Missing Data
in Higher Education Research. Research in Higher Education, 56(4), 397–409.
https://doi.org/10.1007/s11162-014-9344-9
March, J. G. (1991). Exploration and Exploitation in Organizational Learning. Organization
Science, 2(1), 71–87. https://doi.org/10.1287/orsc.2.1.71
Mcfarland, B. R., Shankman, S. A., Tenke, C. E., Bruder, G. E., & Klein, D. N. (2006).
Behavioral activation system deficits predict the six-month course of depression. Journal
of Affective Disorders, 91(2-3), 229–234. doi: 10.1016/j.jad.2006.01.012
McMakin, D. L., Olino, T. M., Porta, G., Dietz, L. J., Emslie, G., Clarke, G., Wagner, K. D.,
Asarnow, J. R., Ryan, N. D., Birmaher, B., Shamseddeen, W., Mayes, T., Kennard, B.,
Spirito, A., Keller, M., Lynch, F. L., Dickerson, J. F., & Brent, D. A. (2012). Anhedonia
Predicts Poorer Recovery Among Youth With Selective Serotonin Reuptake Inhibitor
Treatment–Resistant Depression. Journal of the American Academy of Child &
Adolescent Psychiatry, 51(4), 404–411. https://doi.org/10.1016/j.jaac.2012.01.011
Meyniel, F., Goodwin, G. M., Deakin, J. W., Klinge, C., MacFadyen, C., Milligan, H., Mullings,
E., Pessiglione, M., & Gaillard, R. (2016). A specific role for serotonin in overcoming
effort cost. ELife, 5. https://doi.org/10.7554/elife.17282
Montgomery, S. A., & Åsberg, M. (1979). A New Depression Scale Designed to be Sensitive to
Change. British Journal of Psychiatry, 134(4), 382–389.
https://doi.org/10.1192/bjp.134.4.382
70

Murphy, J. A., Sarris, J., & Byrne, G. J. (2017). A Review of the Conceptualisation and Risk
Factors Associated with Treatment-Resistant Depression. Depression Research and
Treatment, 2017, 1–10. https://doi.org/10.1155/2017/4176825
Murty, V. P., LaBar, K. S., & Adcock, R. A. (2012). Threat of Punishment Motivates Memory
Encoding via Amygdala, Not Midbrain, Interactions with the Medial Temporal Lobe.
Journal of Neuroscience, 32(26), 8969–8976. https://doi.org/10.1523/jneurosci.0094-
12.2012
Nie, Z., Vairavan, S., Narayan, V. A., Ye, J., & Li, Q. S. (2018). Predictive modeling of
treatment resistant depression using data from STAR*D and an independent clinical
study. PLOS ONE, 13(6), e0197268. https://doi.org/10.1371/journal.pone.0197268

O’Doherty, J., Critchley, H., Deichmann, R., & Dolan, R. J. (2003). Dissociating Valence of
Outcome from Behavioral Control in Human Orbital and Ventral Prefrontal Cortices. The
Journal of Neuroscience, 23(21), 7931–7939. https://doi.org/10.1523/jneurosci.23-21-
07931.2003
Oldham, S., Murawski, C., Fornito, A., Youssef, G., Yücel, M., & Lorenzetti, V. (2018). The
anticipation and outcome phases of reward and loss processing: A neuroimaging meta-
analysis of the monetary incentive delay task. Human Brain Mapping, 39(8), 3398–3418.
https://doi.org/10.1002/hbm.24184
Olino, T. M., McMakin, D. L., Morgan, J. K., Silk, J. S., Birmaher, B., Axelson, D. A.,
Williamson, D. E., Dahl, R. E., Ryan, N. D., & Forbes, E. E. (2014). Reduced reward
71

anticipation in youth at high-risk for unipolar depression: A preliminary study.
Developmental Cognitive Neuroscience, 8, 55–64.
https://doi.org/10.1016/j.dcn.2013.11.005
Ormel, J., Bastiaansen, A., Riese, H., Bos, E. H., Servaas, M., Ellenbogen, M., Rosmalen, J. G.
M., & Aleman, A. (2013). The biological and psychological basis of neuroticism: Current
status and future directions. Neuroscience & Biobehavioral Reviews, 37(1), 59–72.
https://doi.org/10.1016/j.neubiorev.2012.09.004
Pagnoni, G., Zink, C. F., Montague, P. R., & Berns, G. S. (2002). Activity in human ventral
striatum locked to errors of reward prediction. Nature Neuroscience, 5(2), 97–98.
https://doi.org/10.1038/nn802
Paris, J. (2005). Neurobiological dimensional models of personality: a review of the models of
Cloninger, Depue, and Siever. Journal of Personality Disorders, 19, 156–170.
Pessiglione, M., Seymour, B., Flandin, G., Dolan, R. J., & Frith, C. D. (2006). Dopamine-
dependent prediction errors underpin reward-seeking behaviour in humans. Nature,
442(7106), 1042–1045. https://doi.org/10.1038/nature05051
Pietri, E. S., Vasey, M. W., Grover, M., & Fazio, R. H. (2015). Predicting Changes in Depressive
Symptoms From Valence Weighting During Attitude Generalization. Journal of Social
and Clinical Psychology, 34(10), 859–875. https://doi.org/10.1521/jscp.2015.34.10.859
Pizzagalli, D. A., Jahn, A. L., & O'Shea, J. P. (2005). Toward an objective characterization of an
anhedonic phenotype: a signal-detection approach. Biological psychiatry, 57(4), 319–
327. https://doi.org/10.1016/j.biopsych.2004.11.026
72

Price, J., Cole, V., & Goodwin, G. M. (2009). Emotional side-effects of selective serotonin
reuptake inhibitors: qualitative study. British Journal of Psychiatry, 195(3), 211–217.
https://doi.org/10.1192/bjp.bp.108.051110
Pringle, A., McCabe, C., Cowen, P., & Harmer, C. (2013). Antidepressant treatment and
emotional processing: can we dissociate the roles of serotonin and noradrenaline?
Journal of Psychopharmacology, 27(8), 719–731.
https://doi.org/10.1177/0269881112474523
Quilty, L. C., Meusel, L. A. C., & Bagby, R. M. (2008). Neuroticism as a mediator of treatment
response to SSRIs in major depressive disorder. Journal of Affective Disorders, 111(1),
67–73. https://doi.org/10.1016/j.jad.2008.02.006
Revelle W (2021). psych: Procedures for Psychological, Psychometric, and Personality
Research. Northwestern University, Evanston, Illinois. R package version
2.1.3, https://CRAN.R-project.org/package=psych.
Rizvi, S. J., Pizzagalli, D. A., Sproule, B. A., & Kennedy, S. H. (2016). Assessing anhedonia in
depression: Potentials and pitfalls. Neuroscience & Biobehavioral Reviews, 65, 21–35.
https://doi.org/10.1016/j.neubiorev.2016.03.004
Rizvi, S. J., Quilty, L. C., Sproule, B. A., Cyriac, A., Michael Bagby, R., Kennedy, S. H. (2015).
Development and validation of the Dimensional Anhedonia Rating Scale (DARS) in a
community sample and individuals with major depression. Psychiatry Research, 229(1–
2), 109–119. https://doi.org/10.1016/j.psychres.2015.07.062
Rosseel Y (2012). “lavaan: An R Package for Structural Equation Modeling.” Journal of
Statistical Software, 48(2), 1–36. https://www.jstatsoft.org/v48/i02/.
73

Rubin, D.B. (1987) Multiple Imputation for Nonresponse in Surveys. John Wiley & Sons Inc.,
New York. http://dx.doi.org/10.1002/9780470316696
Rudebeck, P. H., Putnam, P. T., Daniels, T. E., Yang, T., Mitz, A. R., Rhodes, S. E. V., &
Murray, E. A. (2014). A role for primate subgenual cingulate cortex in sustaining
autonomic arousal. Proceedings of the National Academy of Sciences, 111(14), 5391–
5396. https://doi.org/10.1073/pnas.1317695111
Rush, A. J., Trivedi, M. H., Wisniewski, S. R., Nierenberg, A. A., Stewart, J. W., Warden, D.,
Niederehe, G., Thase, M. E., Lavori, P. W., Lebowitz, B. D., McGrath, P. J., Rosenbaum,
J. F., Sackeim, H. A., Kupfer, D. J., Luther, J., & Fava, M. (2008). Acute and Longer-
Term Outcomes in Depressed Outpatients Requiring One or Several Treatment Steps: A
STAR*D Report. FOCUS, 6(1), 128–142. https://doi.org/10.1176/foc.6.1.foc128
Salinas, J.A. & McGaugh, J.L. (1996) The amygdala modulates memory for changes in reward
magnitude: involvement of the amygdaloid GABAergic system. Behavioral Brain
Research, 80, 87±98.
Samuels, B. A., Leonardo, E. D., Gadient, R., Williams, A., Zhou, J., David, D. J., Gardier, A.
M., Wong, E. H., & Hen, R. (2011). Modeling treatment-resistant
depression. Neuropharmacology, 61(3), 408–413.
https://doi.org/10.1016/j.neuropharm.2011.02.017
Satel, S.L., Nelson, J.C. (1989). Stimulants in the treatment of depression: a critical overview.
The Journal of Clinical Psychiatry, 50(7), 241-249.
74

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. British
Journal of Psychiatry, 167(1), 99–103. https://doi.org/10.1192/bjp.167.1.99
Souery, D., Oswald, P., Massat, I., Bailer, U., Bollen, J., Demyttenaere, K., Kasper, S.,
Lecrubier, Y., Montgomery, S., Serretti, A., Zohar, J., & Mendlewicz, J. (2007). Clinical
Factors Associated With Treatment Resistance in Major Depressive Disorder. The
Journal of Clinical Psychiatry, 68(07), 1062–1070. https://doi.org/10.4088/jcp.v68n0713
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.
Stoet, G. (2010). PsyToolkit - A software package for programming psychological experiments
using Linux. Behavior Research Methods, 42(4), 1096-1104.
Stoet, G. (2017). PsyToolkit: A novel web-based method for running online questionnaires and
reaction-time experiments. Teaching of Psychology, 44(1), 24-31.
Stoy M., Schlagenhauf, F., Sterzer, P., Bermpohl, F., Ha ̈gele. C., Suchotzki, K., Schmack, K.,
Wrase, J., Ricken, R., Knutson, B., Adli, M., Bauer, M., Heinz, A., Stro ̈hle, A. (2012).
Hyporeactivity of ventral striatum towards incentive stimuli in unmedicated depressed
patients normalizes after treatment with escitalopram. Journal of Psychopharmacology,
26, 677–688.
Tang, T.Z., DeRubeis, R.J., Hollon, S.D., Amsterdam, J., Shelton, R., Schalet, B. (2009). A
placebo-controlled test of the effects of paroxetine and cognitive therapy on personality
risk factors in depression. Archives of General Psychiatry, 66, 1322–1330.
75

Targum, S. D., & Fava, M. (2011). Fatigue as a residual symptom of depression. Innovations in
clinical neuroscience, 8(10), 40–43.
Trew, J. L. (2011). Exploring the roles of approach and avoidance in depression: An integrative
model. Clinical Psychology Review, 31(7), 1156–1168. doi: 10.1016/j.cpr.2011.07.007
Uher, R., Farmer, A., Maier, W., Rietschel, M., Hauser, J., Marusic, A., Mors, O., Elkin, A.,
Williamson, R., Schmael, C., Henigsberg, N., Perez, J., Mendlewicz, J., Janzing, J.,
Zobel, A., Skibinska, M., Kozel, D., Stamp, A., Bajs, M., . . . Aitchison, K. (2007).
Measuring depression: comparison and integration of three scales in the GENDEP study.
Psychological Medicine, 38(2), 289–300. https://doi.org/10.1017/s0033291707001730
Wang, S., Leri, F., & Rizvi, S. J. (2021). Anhedonia as a central factor in depression: Neural
mechanisms revealed from preclinical to clinical evidence. Progress in Neuro-
Psychopharmacology and Biological Psychiatry, 110, 110289.
https://doi.org/10.1016/j.pnpbp.2021.110289
Watson, D., O’Hara, M. W., Naragon-Gainey, K., Koffel, E., Chmielewski, M., Kotov, R.,
Stasik, S. M., & Ruggero, C. J.. (2012). Development and Validation of New Anxiety
and Bipolar Symptom Scales for an Expanded Version of the IDAS (the IDAS-
II). Assessment, 19(4), 399–420. https://doi.org/10.1177/1073191112449857
Whitton, A. E., Treadway, M. T., & Pizzagalli, D. A. (2015). Reward processing dysfunction in
major depression, bipolar disorder and schizophrenia. Current Opinion in Psychiatry,
28(1), 7–12. https://doi.org/10.1097/yco.0000000000000122
Widiger, T. A. (2011). Personality and psychopathology. World Psychiatry, 10(2), 103–106.
https://doi.org/10.1002/j.2051-5545.2011.tb00024.x
76

Wijeratne, C., & Sachdev, P. (2008). Treatment-Resistant Depression: Critique of Current
Approaches. Australian & New Zealand Journal of Psychiatry, 42(9), 751–762. doi:
10.1080/00048670802277206
Wilson, R. P., Colizzi, M., Bossong, M. G., Allen, P., Kempton, M., & Bhattacharyya, S. (2018).
The Neural Substrate of Reward Anticipation in Health: A Meta-Analysis of fMRI
Findings in the Monetary Incentive Delay Task. Neuropsychology Review, 28(4), 496–
506. https://doi.org/10.1007/s11065-018-9385-5
Zalla, T., Koechlin, E., Pietrini, P., Basso, G., Aquino, P., Sirigu, A., & Grafman, J. (2000).
Differential amygdala responses to winning and losing: a functional magnetic resonance
imaging study in humans. European Journal of Neuroscience, 12(5), 1764–1770.
https://doi.org/10.1046/j.1460-9568.2000.00064.x

77

APPENDIX
Table 1A. Sample of DARS write-in answers*

*Edited for brevity.

78

Table 2A. Correlations between all measures

E
N
WANTING
(Anh F1)
ENERGY
(Anh F2)
LIKING
(Anh F3)
BF
Accuracy
BF
Learning
Bias
BF
Response
Bias
MDD(1=y)
TRD(1=y)
E
1
-.432**
.539**
.454**
.276**
0.033
0.116
0.046
0.11
.223**
N

1
-.342**
-.290**
-.319**
0.144
-0.006
0.005
-.286**
-.249**
WANTIN
G

1
.232**
.430**
0.079
0.088
-0.094
.318**
.360**
ENERGY

1
0.098
-.188*
.215*
.193*
-0.105
0.136
LIKING

1
0.136
-0.117
-0.11
.357**
.292**
Accuracy

1
-.379**
-.339**
0.093
-0.137
Learning
Bias

1
.881**
-.172*
.198*
Response
Bias

1
-.203*
0.145
MDD(1=y
)

1
.294**
79

TRD(1=y)

1
* denotes significance at the .05 level.
** denotes significance at the .01 level.
*** denotes significance at the .001 level.
Self-Report Measures
Table 3A. Experiment 1 Questionnaire
Please answer the following questions as accurately as possible:
1. Are you currently taking any prescription medication for treatment of Major Depression
(MDD)?
a. Yes
b. No
2. If so, please enter the name, dosage, and number of months continuously used of all
current medication you are taking for the treatment of MDD in the table below. You
may skip forward to access a list of commonly prescribed medications for depression.
NAME DOSAGE (mg/day) TIME TAKEN (months)

3. On a scale of 1 - 5, how well do you feel your current medication is working, compared
to before you started taking it?
1 – No difference or worse
2 – Slightly better
3 – Moderately well
4 – Mostly recovered
5 – Fully recovered
4. When were you diagnosed with MDD?
Enter month and year (MM/YY):
5. When did you start taking prescription medication for MDD?
Enter month and year (MM/YY):
6. Of the following, select all of the medications that you have taken for depression.
80

SSRI
 Fluoxetine (Prozac)
 Sertraline (Zoloft)
 Paroxetine (Paxil)
 Escitalopram (Lexapro)
 Fluvoxamine (Luvox)
 Citalopram (Celexa)
SNRI
 atomoxetine (Strattera)
 desvenlafaxine (Pristiq, Khedezla)
 duloxetine (Cymbalta, Irenka)
 levomilnacipran (Fetzima)
 milnacipran (Savella)
 tramadol (Ultram)
 venlafaxine (Effexor XR)
MAOI
 rasagiline (Azilect)
 selegiline (Eldepryl, Zelapar)
 isocarboxazid (Marplan)
 phenelzine (Nardil)
 tranylcypromine (Parnate)
TRICYCLIC
 amitriptyline (Elavil)
 desipramine (Norpramin)
 doxepine (Sinequan)
 Imipramine (Tofranil)
 nortriptyline (Pamelor)
 amoxapine (Asendin)
 clomipramine (Anafranil)
 maprotiline (Ludiomil)
 trimipramine (Surmontil)
 protriptyline (Vivactil)
ATYPICAL
 bupropion (Wellbutrin)
 mirtazapine (Remeron)
 nefazodone (Serzone)
 trazodone (Desyrel, Oleptro)
 vilazodone (Viibryd)
 vortioxetine (Brintellix)
81

 pramipexole (Mirapex)
OTHER
 lithium (Eskalith, Lithobid)
 dextroamphetamine (Adderall)
 lisdexamfetamine (Vyvanse)
 methylphenidate (Ritalin, Concerta)
 modafinil (Provigil)
7. Please select the medication(s) that you feel decreased your depressive symptoms. If
none did, leave blank.
[repeat checklist]
8. Has your doctor ever told you that you have treatment resistant depression?
a. Yes
b. No
9. If so, when did your doctor diagnose you with treatment resistant depression?
Enter month and year (MM/YY):
10. Have you been undergoing alternative treatment for treatment resistant depression?
a. Yes
b. No
11. If so, please list the alternative treatments that you have tried as well as your rating for
how effective you feel they were on a scale of 1-5 (1 = No difference or worse, 5 = fully
recovered)
NAME OF TREATMENT EFFECTIVENESS RATING

Table 4A. Beck Depression Inventory (BDI)
This questionnaire consists of 21 groups of statements. Please read each group of statements
carefully. And then select the one statement in each group that best describes the way you have
been feeling during the past two weeks, including today. If several statements in the group seem
to apply equally well, circle the highest number for that group.
1.
I do not feel sad.
I feel sad
I am sad all the time and I can't snap out of it.
82

I am so sad and unhappy that I can't stand it.

2.
I am not particularly discouraged about the future.
I feel discouraged about the future.
I feel I have nothing to look forward to.
I feel the future is hopeless and that things cannot improve.

3.
I do not feel like a failure.
I feel I have failed more than the average person.
As I look back on my life, all I can see is a lot of failures.
I feel I am a complete failure as a person.

4.
I get as much satisfaction out of things as I used to.
I don't enjoy things the way I used to.
I don't get real satisfaction out of anything anymore.
I am dissatisfied or bored with everything.

5.
I don't feel particularly guilty
I feel guilty a good part of the time.
I feel quite guilty most of the time.
I feel guilty all of the time.

6.
I don't feel I am being punished.
I feel I may be punished.
I expect to be punished.
I feel I am being punished.

7.
83

I don't feel disappointed in myself.
I am disappointed in myself.
I am disgusted with myself.
I hate myself.

8.
I don't feel I am any worse than anybody else.
I am critical of myself for my weaknesses or mistakes.
I blame myself all the time for my faults.
I blame myself for everything bad that happens.

9.
I don't have any thoughts of killing myself.
I have thoughts of killing myself, but I would not carry them out.
I would like to kill myself.
I would kill myself if I had the chance.

10.
I don't cry any more than usual.
I cry more now than I used to.
I cry all the time now.
I used to be able to cry, but now I can't cry even though I want to.

11.
I am no more irritated by things than I ever was.
I am slightly more irritated now than usual.
I am quite annoyed or irritated a good deal of the time.
I feel irritated all the time.

12.
I have not lost interest in other people.
I am less interested in other people than I used to be.
I have lost most of my interest in other people.
I have lost all of my interest in other people.
84

13.
I make decisions about as well as I ever could.
I put off making decisions more than I used to.
I have greater difficulty in making decisions more than I used to.
I can't make decisions at all anymore.

14.
I don’t feel that I look any worse than I used to.
I am worried that I am looking old or unattractive.
I feel there are permanent changes in my appearance that make me look unattractive.
I believe that I look ugly.

15.
I can work about as well as before.
It takes an extra effort to get started at doing something.
I have to push myself very hard to do anything.
I can't do any work at all.

16.
I can sleep as well as usual.
I don't sleep as well as I used to.
I wake up 1-2 hours earlier than usual and find it hard to get back to sleep.
I wake up several hours earlier than I used to and cannot get back to sleep.

17.
I don't get more tired than usual.
I get tired more easily than I used to.
I get tired from doing almost anything.
I am too tired to do anything.

18.
85

My appetite is no worse than usual.
My appetite is not as good as it used to be.
My appetite is much worse now.
I have no appetite at all anymore.

19.
I haven't lost much weight, if any, lately.
I have lost more than five pounds.
I have lost more than ten pounds.
I have lost more than fifteen pounds.

20.
I am no more worried about my health than usual.
I am worried about physical problems like aches, pains, upset stomach, or
constipation.
I am very worried about physical problems and it's hard to think of much else.
I am so worried about my physical problems that I can

21.
have not noticed any recent change in my interest in sex.
I am less interested in sex than I used to be.
I have almost no interest in sex.
I have lost interest in sex completely.

Scored on a scale ranging from 0 (first option) to 3 (last option) points per item.

Table 5A. Beck Anxiety Inventory (BAI)
Below is a list of common symptoms of anxiety. Please carefully read each item in the list.
Indicate how much you have been bothered by that symptom during the past month, including
today, by selecting the corresponding button next to each symptom.
Not at all
86

Mildly, but it didn’t bother me much
Moderately – it wasn’t pleasant at times
Severely – it bothered me a lot
- Numbness or tingling
- Feeling hot
- Wobbliness in legs
- Unable to relax
- Fear of worst happening
- Dizzy or lightheaded
- Heart pounding / racing
- Unsteady
- Terrified or afraid
- Nervous
- Feeling of choking
- Hands trembling
- Shaky / unsteady
- Fear of losing control
- Difficulty in breathing
- Fear of dying
- Scared
- Indigestion
- Faint / lightheaded
- Face flushed
- Hot / cold sweats
Scored on a scale ranging from 0 (Not at all) to 3 (Severely – it bothered me a lot) points per item.

Table 6A. Inventory of Depression and Anxiety Symptoms-II (IDAS-II)
Below is a list of feelings, sensations, problems, and experiences that people sometimes have.
Read each item to determine how well it describes your recent feelings and experiences. Then,
87

circle the choice that best describes how much you have felt or experienced things this way
during the past two weeks, including today.
Not at all
A little bit
Moderately
Quite a bit
Extremely
- I did not have much of an appetite
- I had little interest in my usual hobbies and activities
- I felt optimistic
- I slept less than usual
- I felt fidgety, restless
- I felt exhausted
- I felt a pain in my chest
- I felt depressed
- I had trouble making up my mind
- I was proud of myself
- I had trouble falling asleep
- I was furious
- I had thoughts of suicide
- I had disturbing thoughts of something bad that happened to me
- I felt self-conscious knowing that others were watching me
- I felt dizzy or lightheaded
- I woke up early and could not get back to sleep
- I was worried about embarrassing myself socially
- I thought a lot about food
- I became anxious in a crowded public setting
- I blamed myself for things
- I cut or burned myself on purpose
- I felt that I had accomplished a lot
88

- I ate when I wasn't hungry
- I woke up much earlier than usual
- I felt like eating less than usual
- I looked forward to things with enjoyment
- I had nightmares that reminded me of something bad that happened
- I slept more than usual
- It took a lot of effort for me to get going
- I felt inadequate
- I was trembling or shaking
- I thought that the world would be better off without me
- I had memories of something scary that happened
- I felt like breaking things
- I woke up frequently during the night
- I felt enraged
- I hurt myself purposely
- I felt faint
- I felt discouraged about things
- I found it difficult to make eye contact with people
- I got upset thinking about something bad that happened
- I had trouble waking up in the morning
- I lost my temper and yelled at people
- My heart was racing or pounding
- I thought about my own death
- I found it difficult to talk with people I did not know well
- I found myself worrying all the time
- I had a very dry mouth
- I felt hopeful about the future
- I slept very poorly
- I thought about hurting myself
89

- I felt that I had a lot to look forward to
- I felt much worse in the morning than later in the day
- I felt drowsy, sleepy
- I was short of breath
- I talked more slowly than usual
- I felt like I was choking
- I felt like I had a lot of interesting things to do
- I did not feel much like eating
- I had trouble concentrating
- Little things made me mad
- I ate more than usual
- I felt like I had a lot of energy
- I rearranged things so that they were in a certain order
- I washed my hands excessively
- I kept racing from one activity to the next
- I checked things over and over again
- I felt the urge to rearrange things so that they were “just right”
- I worried a lot about germs
- I spoke so rapidly that others could not understand me
- I felt elated for no special reason
- I tried not to think about bad things from my past
- I avoided small spaces
- I found myself checking things, even though I knew it wasn’t necessary
- I avoided handling dirty things
- It felt like my mind was moving “a mile a minute”
- I felt like I was “on top of the world”
- I avoided situations that bring up bad memories
- I was afraid of getting trapped in a crowd
- I felt the urge to check to make sure I had done something
90

- I followed the same, fixed order in performing everyday tasks
- My thoughts jumped rapidly from one idea to another
- I felt anxious in small spaces
- I felt compelled to follow certain rituals
- I had difficulty touching something that was dirty
- My thoughts were moving so quickly it was hard to keep up
- I had so much energy it was hard for me to sit still
- I tried to ignore upsetting memories
- I was afraid of tunnels
- I had to clean myself because I felt contaminated
- I felt that I could do things that other people couldn’t
- I avoided talking about bad experiences from my past
- I avoided tight, enclosed spaces
- I had little rituals or habits that took up a lot of my time
- I avoided using public restrooms
- I had much more energy than usual
- I used an object (such as a towel) so I could avoid touching something directly
- I was anxious about talking in public
Scored on a scale ranging from 0 (Not at all) to 4 (Extremely) points per item.

Table 7A. Experiment 2 Screening Questionnaire
Please answer the following questions as accurately as possible:
1. Are you currently taking any of the following medication:
• Esketamine (Spravato)
• Dextroamphetamine (Dexedrine, Zenzedi)
• Amphetamine (Adderall, Mydayis)
• Dexmethylphenidate (Focalin)
• Methlyphenidate (Aptensio, Concerta, Daytrana, Metadate, Methylin, Ritalin,
Quilichew)
• Lisdexamfetamine (Vyvanse)
• Bupropion (Wellbutrin)
91

• Pramipexole (Mirapex)
• Levodopa (Duopa, Parcopa, Rytary, Sinemet, Stalevo)
• Any antipsychotic medication (Chorpromazine, Haloperidol, Perphenazine,
Fluphenazine, Clozapine, Risperdone, Olanzapine, Quetiapine, Ziprasidone,
Aripiprazole, Paliperidone, Lurasidone)
a. Yes
b. No
2. Have you ever been diagnosed with depression, or taken prescription medications to
treat depression?
a. Yes
b. No
3. Have you currently been taking a selective serotonin reuptake inhibitor (SSRI/SNRI) for
longer than 6 weeks? The following are examples of SSRI/SNRI medications:
SSRI
• Fluoxetine (Prozac)
• Sertraline (Zoloft)
• Paroxetine (Paxil, Pexeva)
• Citalopram (Celexa, Cipramil)
• Escitalopram (Lexapro, Cipralex)
• Fluvoxamine (Luvox)
SNRI
• atomoxetine (Strattera)
• desvenlafaxine (Pristiq, Khedezla)
• duloxetine (Cymbalta, Irenka)
• levomilnacipran (Fetzima)
• milnacipran (Savella)
• tramadol (Ultram)
• venlafaxine (Effexor)
a. Yes [for regular recruitment end, eligible. For TRD, continue]
b. No [end, not eligible]
4. Did your depression symptoms noticeably improve from taking an SSRI/SNRI for longer
than 6 weeks?
a. Yes [end – not eligible]
b. No
5. Have you tried more than 1 SSRI/SNRI medication for longer than 6 weeks and your
symptoms did not improve OR have you been diagnosed with Treatment Resistant
Depression?
a. Yes
b. No

92

Table 8A. Patient Health Questionnaire-2 (PHQ-2)
Over the last 2 weeks, how often have you been bothered by either of the following problems?
Not at all
Several days
More than half the days
Nearly every day
- Little interest or pleasure in doing things
- Feeling down, depressed, or hopeless
Scored on a scale ranging from 0 (Not at all) to 4 (Nearly every day) points per item.

Table 9A. Motivation and Energy Inventory (MEI)
For each question below, please check one box to indicate the way you have been feeling during
the past 4 weeks. When answering, please try to consider every day of the week (including
weekends), as well as every setting that applies to you such as work, home, school, etc.
{score=0} Never
{score=1} Less than one day a week
{score=2} 1 or 2 days a week
{score=3} 3 or 4 days a week
{score=4} 5 or 6 days a week
{score=5} Every day or nearly every day
- During the past 4 weeks, how often did you feel enthusiastic when you began your day?
- During the past 4 weeks, how often did you feel satisfied with what you accomplished during
the day?
- {reverse} During the past 4 weeks, how often did you have trouble getting out of bed in the
morning because you didn’t want to face the day?
- {reverse} During the past 4 weeks, how often did you run out of energy before the end of the
day?
- {reverse} During the past 4 weeks, how often did you have trouble finishing things you started
because you lost interest in them?
- {reverse} During the past 4 weeks, how often did you feel overwhelmed even by small tasks?
93

- {reverse} During the past 4 weeks, how often did you procrastinate or put things off until
another day?
- {reverse} During the past 4 weeks, how often did you have trouble remembering information
(such as people’s names, where you put things, or what you needed from the grocery store)?
- {reverse} During the past 4 weeks, how often did you have problems concentrating?
- {reverse} During the past 4 weeks, how often did you have trouble making minor decisions?
- {reverse} During the past 4 weeks, how often did you avoid social conversations with others?
- During the past 4 weeks, how often did you take advantage of opportunities to get to know
other people better?
- {reverse} During the past 4 weeks, how much of the time did you prefer to be alone?
- {reverse} During the past 4 weeks, how much of the time did you have trouble focusing your
attention on your work or other activities?
- {reverse} During the past 4 weeks, how much of the time did you have trouble keeping things
organized?
- During the past 4 weeks, how much of the time were you able to keep up with chores around
the house such as laundry, cleaning, and doing the dishes?
- {reverse} During the past 4 weeks, how much of the time did you feel physically tired during
the day?
- {reverse} During the past 4 weeks, how much of the time did you feel exhausted?
- During the past 4 weeks, how much of the time did you feel energetic?
- During the past 4 weeks, how much of the time did you feel motivated?

{score=0} Never
{score=1} Less than once a week
{score=2} 1 or 2 times a week
{score=3} 3 or 4 times a week
{score=4} 5 or 6 times a week
{score=5} At least 7 times a week
- During the past 4 weeks, how often did you call, e-mail, or write letters to friends or family
members?
- During the past 4 weeks, how often did you get together with friends or family members who
don’t live with you?
- During the past 4 weeks, how often did you engage in recreational activities or hobbies?
94

- During the past 4 weeks, how often did you exercise (for example by walking, swimming, or
practicing yoga)?

{score=0} Not at all interested
{score=1} A little interested
{score=2} Somewhat interested
{score=3} Quite interested
{score=4} Extremely interested
- During the past 4 weeks, to what extent were you interested in sexual activity?
- During the past 4 weeks, to what extent were you interested in taking on additional tasks or
projects?
- During the past 4 weeks, to what extent were you interested in learning or trying new things?
- During the past 4 weeks, to what extent were you interested in meeting new people?
- During the past 4 weeks, to what extent were you interested in talking with others?
- During the past 4 weeks, to what extent were you interested in social activities like visiting
friends, going out to dinner, or parties?

Table 10A. Temporal Experience of Pleasure Scale (TEPS)
Below you will find a list of statements that may or may not be true for you. Please read each
statement carefully and decide how true that statement is for you in general.
Very false for me
Mostly false for me
Somewhat false for me
Somewhat true for me
Mostly true for me
Very true for me
- When something exciting is coming up in my life, I really look forward to it
- The sound of crackling wood in the fireplace is very relaxing
- When I think about eating my favorite food, I can almost taste how good it is
- I love the sound of rain on the windows when I’m lying in my warm bed
95

- The smell of freshly cut grass is enjoyable to me
- I enjoy taking a deep breath of fresh air when I walk outside
- {reverse} I don’t look forward to things like eating out at restaurants
- A hot cup of coffee or tea on a cold morning is very satisfying to me
- I love it when people play with my hair
- I really enjoy the feeling of a good yawn
- When I’m on my way to an amusement park, I can hardly wait to ride the roller coasters
- I get so excited the night before a major holiday I can hardly sleep
- I appreciate the beauty of a fresh snowfall
- When I think of something tasty, like a chocolate chip cookie, I have to have one
- Looking forward to a pleasurable experience is in itself pleasurable
- I look forward to a lot of things in my life
- When ordering something off the menu, I imagine how good it will taste
Items are scored on a scale of 1 (Very false for me) to 6 (Very true for me)

Table 10A. Dimensional Anhedonia Rating Scale (DARS)
For the next set of questions, you will be asked to list 2 of your favorite things in each of 4
categories: pastimes/hobbies (eg: taking a walk, theater), food & drink (eg: pasta, red wine),
social activities (eg: going to concerts, bowling), and sensory experiences (eg: smell of bread,
hot bath). For each of your favorite things in these categories, you will be asked a series of
questions related to how much you would enjoy them right now. Please continue when you are
ready.
Not at all
Slightly
Moderately
Mostly
Very much
Please list at least 2 of your favorite pastimes/hobbies that are NOT primarily social
- 1st pastime/hobby:
- 2nd pastime/hobby:
- I would enjoy these activities
96

- I would spend time doing these activities
- I want to do these activities
- These activities would interest me
Please list at least 2 of your favorite foods or drinks
- 1st pastime/hobby:
- 2nd pastime/hobby:
- I would make an effort to get/make these foods/drinks
- I would enjoy these foods/drinks
- I want to have these foods/drinks
- I would eat as much of these foods as I could
Please list at least 2 of your favorite social activities
- 1st social activity:
- 2nd social activity:
- Spending time doing these things would make me happy
- I would be interested in doing things that involve other people
- I would be the one to plan these activities
- I would actively participate in these social activities
Please list at least 2 of your favorite sensory experiences
- 1st sensory experience:
- 2nd sensory experience:
- I would actively seek out these experiences
- I get excited thinking about these experiences
- If I were to have these experiences I would savor every moment
- I want to have these experiences
- I would make an effort to spend time having these experiences
Items are scored on a scale of 0 (Not at all) to 4 (Very much)

Table 11A. Big Five Aspect Scales (BFAS)
97

Here are a number of characteristics that may or may not describe you. For example, do you
agree that you seldom feel blue? Please fill in the number that best indicates the extent to which
you agree or disagree with each statement listed below. Be as honest as possible, but rely on your
initial feeling and do not think too much about each item.
1 Strongly Disagree
2
3 Neither Agree Nor Disagree
4
5 Strongly Agree
- {reverse} Seldom feel blue.
- {reverse} Am not interested in other people's problems.
- Carry out my plans.
- Make friends easily.
- Am quick to understand things.
- Get angry easily.
- Respect authority.
- {reverse} Leave my belongings around.
- Take charge.
- Enjoy the beauty of nature.
- Am filled with doubts about things.
- Feel others' emotions.
- {reverse} Waste my time.
- {reverse} Am hard to get to know.
- {reverse} Have difficulty understanding abstract ideas.
- {reverse} Rarely get irritated.
- {reverse} Believe that I am better than others.
- Like order.
- Have a strong personality.
- Believe in the importance of art.
- {reverse} Feel comfortable with myself.
- Inquire about others' well-being.
98

- {reverse} Find it difficult to get down to work.
- {reverse} Keep others at a distance.
- Can handle a lot of information.
- Get upset easily.
- Hate to seem pushy.
- Keep things tidy.
- {reverse} Lack the talent for influencing people.
- Love to reflect on things.
- Feel threatened easily.
- {reverse} Can't be bothered with other's needs.
- {reverse} Mess things up.
- {reverse} Reveal little about myself.
- Like to solve complex problems.
- {reverse} Keep my emotions under control.
- {reverse} Take advantage of others.
- Follow a schedule.
- Know how to captivate people.
- Get deeply immersed in music.
- {reverse} Rarely feel depressed.
- Sympathize with others' feelings.
- Finish what I start.
- Warm up quickly to others.
- {reverse} Avoid philosophical discussions.
- Change my mood a lot.
- Avoid imposing my will on others.
- {reverse} Am not bothered by messy people.
- {reverse} Wait for others to lead the way.
- {reverse} Do not like poetry.
- Worry about things.
99

- {reverse} Am indifferent to the feelings of others.
- {reverse} Don't put my mind on the task at hand.
- {reverse} Rarely get caught up in the excitement.
- {reverse} Avoid difficult reading material.
- {reverse} Rarely lose my composure.
- Rarely put people under pressure.
- Want everything to be “just right.”
- See myself as a good leader.
- {reverse} Seldom notice the emotional aspects of paintings and pictures.
- Am easily discouraged.
- {reverse} Take no time for others.
- Get things done quickly.
- {reverse} Am not a very enthusiastic person.
- Have a rich vocabulary.
- Am a person whose moods go up and down easily.
- {reverse} Insult people.
- {reverse} Am not bothered by disorder.
- Can talk others into doing things.
- Need a creative outlet.
- {reverse} Am not embarrassed easily.
- Take an interest in other people's lives.
- Always know what I am doing.
- Show my feelings when I'm happy.
- Think quickly.
- {reverse} Am not easily annoyed.
- {reverse} Seek conflict.
- {reverse} Dislike routine.
- {reverse} Hold back my opinions.
- {reverse} Seldom get lost in thought.
100

- Become overwhelmed by events.
- {reverse} Don't have a soft side.
- {reverse} Postpone decisions.
- Have a lot of fun.
- {reverse} Learn things slowly.
- Get easily agitated.
- {reverse} Love a good fight.
- See that rules are observed.
- Am the first to act.
- {reverse} Seldom daydream.
- Am afraid of many things.
- Like to do things for others.
- {reverse} Am easily distracted.
- Laugh a lot.
- Formulate ideas clearly.
- Can be stirred up easily.
- {reverse} Am out for my own personal gain.
- Want every detail taken care of.
- {reverse} Do not have an assertive personality.
- See beauty in things that others might not notice.

Abstract (if available)

Abstract Treatment resistant depression (TRD) is a subtype of Major Depressive Disorder (MDD) that is not responsive to typical antidepressants such as SSRIs. There is some evidence that treatment responsiveness may be related to dopamine neuron dysfunction, specifically in the mesolimbic region of the brain. To address the gap in understanding of how this translates to behavioral and personality differences in TRD, we used a reinforcement learning task administered online to objectively measure individual variation in learning from positive and negative feedback. In Experiment 1, we found that participants on average were better at learning the negative stimuli than the positive. Response bias during the test round was significantly correlated with the IDAS-II for General Depression subscale, such that greater depression severity was related to a more negative response bias (r = −.18, p = .03). In Experiment 2, we empirically defined three underlying facets of anhedonia by factor analyzing two anhedonia scales administered online to a sample of 137 participants. The major latent factors were found to be motivation or energy, inventive salience, and consummatory liking. We found that TRD was associated with significantly lower anticipation scores, whereas MDD overall was associated with significantly lower consummatory pleasure. Additionally, the TRD group had significantly lower extraversion scores than responders. Only antidepressant responders with MDD did not demonstrate significantly higher learning of bad stimuli over good stimuli on our reinforcement learning task. We postulated this was due to the effect of SSRIs; however, further research is needed to determine a causal effect.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Examining Pavlovian-to-instrumental transfer effect in human mate selection behavior

PDF

Investigating the effects of Pavlovian cues on social behavior

PDF

Generalization of social rejection in social networks

PDF

Facebook habits: rewards, cues, and automaticity

PDF

Reinforcement learning based design of chemotherapy schedules for avoiding chemo-resistance

PDF

Using Pavlovian cues to boost self-control

PDF

Latent physiological state dynamics underlying everyday affect, affect regulation and cognition

PDF

The impact of mesoscale and submesoscale physical processes on phytoplankton biomass, community composition, and carbon dynamics in the oligotrophic ocean

PDF

How misinformation exploits moral values and framing: insights from social media platforms and behavioral experiments

PDF

Neural and behavioral correlates of fear processing in first-time fathers

PDF

Bridging possible identities and intervention journeys: two process-oriented accounts that advance theory and application of psychological science

PDF

The cultural expression of depression in Asian Americans and European Americans

PDF

Bayesian hierarchical and joint modeling of the reversal learning task

PDF

Neural correlates of learning and reversal of approach versus withdraw responses to in- and out-group individuals

PDF

Social exclusion decreases risk-taking

PDF

Investigating discrimination and depression within a couple context

PDF

How parental psychological control is related to adolescent anger and aggression

PDF

Beyond active and passive social media use: habit mechanisms are behind frequent posting and scrolling on Twitter/X

PDF

Dynamic analyses of the interrelationship between mothers and daughters on a measure of depressive symptoms

PDF

Learning repetition rules and non-adjacent dependencies from human actions in nine-month-old infants

Asset Metadata

Creator Liu, Xiao (author)

Core Title Dopamine dependent: examining the link between learning and treatment-resistant depression

School College of Letters, Arts and Sciences

Degree Master of Arts

Degree Program Psychology

Degree Conferral Date 2021-08

Publication Date 08/02/2021

Defense Date 07/29/2021

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag anhedonia,Depression,Dopamine,OAI-PMH Harvest,refractory depression,reinforcement learning,reward learning,treatment-resistant depression

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Read, Stephen (committee chair), Hackel, Leor (committee member), Monterosso, John (committee member)

Creator Email xiaoliuub@gmail.com,xliu5899@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC15673810

Unique identifier UC15673810

Legacy Identifier etd-LiuXiao-9966

Document Type Thesis

Format application/pdf (imt)

Rights Liu, Xiao

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu