Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 May 1.
Published in final edited form as: J Psychiatr Res. 2023 Mar 20;161:324–332. doi: 10.1016/j.jpsychires.2023.03.026

History of major depressive disorder is associated with differences in implicit learning of emotional faces

Antonija Kolobaric 1,+, Akiko Mizuno 2,+, Xiao Yang 3, Charles J George 4, Andrew Seidman 2, Howard J Aizenstein 2,5, Maria Kovacs 2,*, Helmet T Karim 2,5,*
PMCID: PMC10202097  NIHMSID: NIHMS1887348  PMID: 36996725

Abstract

Major depressive disorder is often associated with worsened reward learning, with blunted reward response persisting after remission. In this study, we developed a probabilistic learning task with social rewards as a learning signal. We examined the impacts of depression on social rewards (facial affect displays) as an implicit learning signal. Fifty-seven participants without a history of depression and sixty-two participants with a history of depression (current or remitted) completed a structured clinical interview and an implicit learning task with social reward. Participants underwent an open-ended interview to evaluate whether they knew the rule consciously. Linear mixed effects models revealed that participants without a history of depression learned faster and showed a stronger preference towards the positive than the negative stimulus when compared to the participants with a history of depression. In contrast, those with a history depression learned slower on average and displayed greater variability in stimulus preference. We did not detect any differences in learning between those with current and remitted depression. The results indicate that on a probabilistic social reward task, people with a history of depression exhibit slower reward learning and greater variability in their learning behavior. Improving our understanding of alterations in social reward learning and their associations with depression and anhedonia may help to develop translatable psychotherapeutic approaches for modification of maladaptive emotion regulation.

Keywords: Depression, implicit learning, automatic behavior regulation, prediction error

General Scientific Summary.

Depression is often associated with social and emotional impairments, such as feeling disconnected from others and social withdrawal. In this study, we examine if history of depression is associated with the ability to learn from socially rewarding stimuli – a phenomena that may, in part, explain the underlying causes of social and emotional impairments in depression.

Major Depressive Disorder (MDD) is the leading cause of disability worldwide, affecting 4–6% of the global population (Mulders et al., 2015; Tadayonnejad & Ajilore, 2014). MDD is a complex disorder, traversing multiple symptom domains, such as cognitive impairment, dysphoric apathy, and vegetative symptoms (Kim et al., 2002; Parker et al., 2003; Vanheule et al., 2008). Such heterogenous symptoms are associated with a multitude of functional impairments, including lower reward learning, which play an important role in hedonic and motivational deficits present in depression (Admon & Pizzagalli, 2015; Alloy et al., 2016; Barch et al., 2015; Gong et al., 2017). Notably, increasing evidence suggests that blunted reward response persists in remitted MDD (Guath et al., 2022; Pechtel et al., 2013; Whitton et al., 2016). Understanding the mechanisms underlying alterations in reward learning is an essential first step towards developing novel treatments to efficiently treat reward learning deficits.

The Probabilistic Reward Task (PRT) has commonly been used to characterize the association between reward signal processing and depressive symptoms (Pizzagalli et al., 2008; Pizzagalli et al., 2005). During the task, participants choose between two response choices which, unknown to the participant, have different monetary reward probabilities. The goal is to integrate information from multiple trials over time to choose the more advantageous stimulus (i.e., a response bias), thus maximizing reward (Pizzagalli et al., 2008; Pizzagalli et al., 2005). The PRT capitalizes on basic principles of reinforcement learning (RL), whereby rewarding stimuli leads to strengthening in behavior (Rmus et al., 2023). In traditional RL, participants generally perform a behavior without awareness. Over time, the participants learn to take the actions that will maximize the cumulative reward (Rmus et al., 2023). Studies using the PRT have shown that individuals with depression develop weaker response bias and accumulate evidence slower when compared to non-depressed individuals, failing to maximize monetary reward (Lawlor et al., 2020; Liu et al., 2016; Pizzagalli et al., 2008; Pizzagalli et al., 2005; Vrieze et al., 2013). Similar tasks used other types of rewards, such as points or food, to study reward-processing in those with depression, including the Instrumental Learning Task (Dezfouli et al., 2019), Iowa Gambling Task (Hegedűs et al., 2018), and Go/No-Go Task (Moutoussis et al., 2018), and each have demonstrated presence of reward-processing alterations in those with depression (Halahakoon et al., 2020).

The PRT employs a probabilistic reward framework, with one stimulus receiving greater reinforcement (e.g., higher monetary reward) to promote response bias learning. As such, the PRT employs a type of learning based on self-accumulating evidence without conscious awareness called implicit learning (as opposed to learning that requires top-down conscious awareness, called explicit learning). Overall, the PRT measures an individual’s ability to modulate behavior in response to a history of monetary reinforcement (Reilly et al., 2020). Participants are aware of the monetary reward contingent with a selection of the correct response, although the delivery schedule of reinforcement is unknown to participants which produces a response bias, or implicit learning. While this task model captures the impact of depression on probabilistic monetary reward learning, it is unclear if these findings generalize to other types of cues, such as social rewards.

We highlight the importance of such generalization for three main reasons. First, in addition to developing weaker response bias in implicit learning tasks, those with depression tend to display additional biases when presented with ecological social stimuli, such as images depicting valenced facial expressions (Glaser & Glaser, 1989), whereby the effects of these biases tend to be twofold. Those with depression display attentional biases through which they prioritize negative-valence information over positive and neutral information (Duque & Vázquez, 2015; Koster et al., 2011; Lazarov et al., 2018). Additionally, the formation of attention bias resulting in an increased attention towards negative stimuli is commonly accompanied with a lack of protective bias resulting in decreased or absent attentional preferences for positive information (Duque & Vázquez, 2015; Koster et al., 2011; McCabe & Gotlib, 1995). Given these biases which are not present when evaluating monetary rewards, it is unknown if social stimuli will act as an implicit learning signal in the same way as monetary stimulus. Hence, incorporating social stimuli into an implicit learning task creates a novel framework that allows us to evaluate reward sensitivity and reward learning as it relates to processing of emotional information in depression.

Second, previous research has demonstrated that monetary and social rewards may undergo different processing, and thus be differentially impacted by rewards deficits in depression (Ait Oumeziane et al., 2019; Goerlich et al., 2017; Gu et al., 2019; Zhang et al., 2022). Social stimuli, such as helping others, receiving interpersonal feedback, or seeing pictures of smiling faces, are highly rewarding to people and they will actively seek out opportunities to receive social stimuli (Tamir & Hughes, 2018). Notably, social rewards, such as receiving praise or seeing a person smile, function as unconscious reinforcers in natural settings, whereas in the case of monetary reward in a lab experiment, participants are explicitly instructed about the contingency between their response and the reward. Finally, even though depression is commonly associated with marked social impairments, there is a lack of studies examining social reward as an implicit learning signal (Hirschfeld et al., 2000; Kupferberg et al., 2016; Segrin, 2000; Weightman et al., 2014). Past studies have detected neural regions associated with social reward processing and established the importance of social reward in everyday experiences (Flores et al., 2018; Martins et al., 2021; Schwartz et al., 2019). While previous studies have advanced our understanding of neural underpinnings of social reward in health and disease, they most commonly did so in an explicit or non-probabilistic manner (Flores et al., 2018; Martins et al., 2021; Schwartz et al., 2019). Thus, despite marked social and emotional impairments in depression, the impact of depression on social reward as an implicit learning signal in the context of a probabilistic reward task remains poorly understood (Ait Oumeziane et al., 2019; Safra et al., 2019).

In this study, we evaluate the impact of history of depression on social reward as an implicit learning signal. To examine if individuals with a history of depression (current or remitted) can learn the hidden pattern of the advantageous stimuli from implicit social rewards, we developed a probabilistic reward task that emulates natural reward processing -- both as a social reward and without awareness of the reward delivery. We developed a task that used social reward (happy faces or sad faces) to induce implicit learning. We hypothesized that all individuals would develop a response bias, and that this learning would be driven by social reward. Further, we hypothesized that participants with depression history would be slower to develop a response bias compared to participants without depression history.

Materials and Methods

Participants and Study Design

We analyzed data from 126 participants between the ages of 18 to 59 with a history of depression or no history of depression. These participants represent a subsample of participants recruited for a family study investigating mood trajectories from young adulthood into midlife. The parent study examined adults without a history of psychiatric disorders and those with childhood-onset mood disorder, defined as major depressive or dysthymic disorder episode before the age of 14, or bipolar disorder episode before the age of 17 (Kovacs et al., 2009; Miller et al., 2002). Exclusion criteria included major medical disorder or sub-average IQ (Kovacs et al., 2009; Miller et al., 2002). The parent study (Kovacs & George, 2020) established diagnoses of current and/or past depression using the semi-structured clinical interview for DSM-V (First, 2014). For more details, see (Kovacs & George, 2020). The current sample included participants enrolled during 2018 and 2019.

To maximize the chances of assessing implicit emotional reward-learning, participants were told they will complete a computerized task testing their motor control and perception. All participants were debriefed about the true nature of the task after completing the study. Following debriefing, participants were reminded of their right to withdraw their consent and data they provided at any point without penalty or loss of compensation; no participants withdrew their informed consent. This study was approved by the University of Pittsburgh IRB and all participants gave written informed consent before starting study procedures.

Assessments

Standardized data sheets completed by clinicians were used to collect demographic information including age, sex, race, and education. Handedness was assessed using the Edinburgh Handedness Inventory (Oldfield, 1971). Attention was assessed with the Attentional Control Scale (ACS) (Derryberry & Reed, 2002). Current depressive symptoms were assessed using the Beck Depression Inventory-II, which yields total scores from 0 to 63 (Beck et al., 1996).

Task

The Implicit Bias Emotion Learning Task (IBELT) was designed to assess sequential learning patterns by implementing asymmetrically delivered social reward in the form of a happy face (Tamir & Hughes, 2018) or alternatively a sad face. The task was programmed in MATLAB 2018b (Mathworks) and presented using PsychToolbox-3 (http://psychtoolbox.org/). In the task, which included 120 trials, participants were presented with a face with a positively or a negatively valenced expression (i.e., happy or sad). Faces for this task were obtained from the Karolinska Directed Emotional Faces (Lundqvist et al., 1998). Every face was paired with two randomly chosen sex-matched names, one on the left side and one on the right side of the face (Figure 1). ). All faces were standardized as 10 × 20 cm (72 dots per inch and typically scaled to 70% of the original size) images and were presented in the center of a 61-cm diagonal wide computer screen. Participants were asked to match the name that best “fits” each face as quickly as possible. Participants responded by pressing the left arrow key with their left index finger to choose the name on the left side of the screen, or by pressing the right arrow key with their right index finger to select the name on the right side of the screen. Participants had 1.5 seconds to respond to each stimulus. This was followed by a 0.5 second inter-stimulus interval, during which a blank black screen was shown prior to the next cue.

Figure 1. Diagram of the task.

Figure 1.

Open in a new tab

Note. Participants are shown a single face with two sex-matched choices for names. They were instructed to choose the name that best ‘fits’ the face presented – the name chosen was irrelevant. However, if they chose the name on the right side, the next face that would appear would be happy 80% of the time, whereas if they chose the left side the next face that would appear would be sad 80% of the time. Participants were shown 80 cues and then the valence of the sides would flip for another 40 trials. The expectation was that participants would implicitly choose one side over another through automated processes without conscious awareness. The flip would demonstrate that participants are learning the affect and not the side.

Unbeknownst to participants, the name they chose was not the primary outcome; it was the side of the screen where the name was located that determined the valence of the next affective face. At the beginning of the task, one side of the screen was chosen as positively biased (i.e., associated with happy face), and the initial positively biased side was determined randomly for each participant. Choosing the name on the positively biased side of the screen would result with the next face having a happy expression over 80% of the trials, and sad expression for 20% of the trials. The other side of the screen had a bias for a sad expression for 80% of the trials and happy expression for 20% of the trials. The initial positively biased side of the screen was supposed to be determined randomly for each participant. However, due to a programming error, the positively biased side of the screen was initially set to “left” for all participants. After the first 80 trials, the positive bias switched to the right side of the screen. Thus, all participants completed 80 trials with the left side of the screen biased as positive, followed by 40 trials with the right side of the screen biased as positive. It is important to highlight that, despite the programming error, the task successfully induced learning as intended (please see details in results section).

Following the task, all participants completed a post-task semi-structured interview to assess awareness of the nature of the task and its goal. This interview had a hierarchical structure to assess participants’ level of awareness of the reward delivery and is included in the supplementary material (see Appendix 1): it yielded a rating from 1 (not aware of a rule at all) to 4 (aware of the implicit rule). The post-task interview was delivered by the experimenter and independently rated by two trained researchers blind to the task, which yielded 100% interrater agreement on the scores.

Statistical Analysis

Prior to statistical analysis, we excluded five subjects due to software failures during the IBELT computer task, and two additional subjects due to experimenter error. Thus, the final analysis sample consisted of 119 participants.

Participants successfully completed 97.3% of trials, with an average of 117 trials (SD of 3.4) out of 120 total trials. Trials without a detected response were discarded from further analyses. We created a binary variable “choice” to reflect if participants chose the negatively biased side on any given trial (choosing the name on the side that is 80% likely to result in a sad face on the next trial). Additionally, we created a binary variable “catch trial” which reflects whether a trial belonged to the 20% of trials which are an exception to the biased side rule. Subscales of Attention Focusing (ACD focusing) and Attention Shifting (ACS shifting) from the ACS were used to assess and control for participants’ abilities to focus and shift attention.

We fit the data to logistic regression models with random effects (glmer function from lme4 package in R (R Core Team, 2018)) to build a series of mixed-effects models with choice as the outcome variable. Fixed effects included: valence (was the observed face happy or sad), catch trial, participant’s race, participant’s age, participant’s sex, trial number, history of depression, Beck Depression Inventory (current depressive symptoms), handedness, ACD focusing, ACS shifting, and reaction time. Anonymized participant ID was used as a random variable. All models had random intercept per participant and unstructured variance-covariance matrix, whereas slope (fixed versus random) and model order (linear versus quadratic versus cubic) was varied across models.

The final model was chosen by comparing three models with random slopes using the anova function in lme4. The best fitting model was chosen by comparing the Akaike Information Criterion (AIC) to find the simplest model (the model built with the lowest number of independent variables) that explains the greatest amount of variance, where lower AIC suggests greater amount of variance explained. We first compared all models with random intercept and random slope and chose the best-fitting model. Next, we compared that model with the same order model with a random intercept and fixed slope.

Results

Table 1 summarizes the demographic and clinical characteristics of the sample used in the analysis (Pollard et al., 2018). This cohort had 57 participants without a history of depression and 62 participants with a history of depression. Most participants (69%) in the group with a history of depression were currently in remission (Table 1). As expected, the group with a history of depression had higher depressive symptoms. The group with a history of depression also had more white individuals compared to the group without a history of depression. Participants without a history of depression self-reported lower ability to focus compared to those with a history of depression. On the post-task questionnaire, all participants reported being explicitly unaware of the side of the screen bias or any such rules, with mean rating of 1.00 (range = 1, SD = 0). This is critical, as it suggests implicit, rather than explicit, emotional reward-learning.

Table 1.

Sample demographics.

Participants with no History of Depression
 Participants with a History of Depression
Variable Mean (SD)/N (%) Mean (SD)/N (%) t / χ2 (df) P-Value
N 57 62
Age (years) 35.4 (11.8) 37.4 (9.5) −0.99 (107) 0.325
Current Depression Diagnosis No 57 (100.0) 43 (69.4) 20.79 (1) <0.001
Yes 0 (0) 19 (30.6)
Current Depression Symptoms 6.7 (7.0) 16.2 (12.8) −4.97(117) <0.001
Years of Education 14.2 (2.4) 14.3 (2.9) −0.09 (117) 0.924
Average Response Time (s) 0.9 (0.2) 1.0 (0.2) −0.59 (117) 0.553
ACD Focusing 26.0 (4.4) 23.6 (4.6) 2.87 (117) 0.005
ACD Shifting 32.9 (4.8) 31.3 (5.6) 1.65 (117) 0.099
Handedness [Reference Left] 0.9 (0.2) 0.9 (0.3) 0.06 (1) 0.808
Sex F 38 (66.7) 40 (64.5) 0.06 (1) 0.957
M 19 (33.3) 22 (35.5)
Race Not White 25 (43.9) 15 (24.2) 4.30 (1) 0.038
White 32 (56.1) 47 (75.8)
Open in a new tab

Note. Current Depression Symptoms are measured using Beck Depression Inventory II.

The variance explained in the overall mixed-effects models with choice as the outcome variable was tested with a series of ANOVA tests, which showed that a quadratic model with a random intercept and random slope performed better than a linear model with a random intercept and random slope [χ2(5, 119) = 73.81, p < 0.0001, linear AIC = 18856, quadratic AIC = 18789]. Similarly, the cubic model with random intercept and fixed slope failed to perform better than the quadratic model [χ2(6,119) = 1.96, p > 0.05, quadratic AIC = 18789, cubic AIC = 18798]. Finally, the quadratic model with a random intercept and random slope performed better than the quadratic model [χ2(5,119) = 112.95, p < 0.0001, random slope quadratic AIC = 18789, fixed slope quadratic AIC = 18895].

The results of the quadratic mixed effects model with random intercept and random slope are shown in Figure 2 and Table 2. As expected, the slope for “valence” was significant. Participants were more likely to choose the side that corresponded to the happy face if they were shown a happy face, supporting that the happy face functioned as a reinforcer. The significant slopes for “trial” and “trial2” indicate that participants were more likely to choose the happy biased side over multiple subsequent trials in linear and quadratic time. Figure 2 depicts this result as continuous average downward curves in both groups, indicating that participants successfully continued learning (i.e., choosing the happy biased side) even after switching the bias side at the 80th trial. On average, reaction times were slower for choosing the negatively biased side compared to the positively biased side, and older adults were more likely to choose sad faces compared to younger adults.

Figure 2. Participant performance across task trials by group.

Figure 2.

Open in a new tab

Note. Participants with no history of depression (left) learned to choose positive stimuli faster than those with a history of depression (right). The learning continues even after switching the biased side (at 80 trials). Higher choice probability indicates greater likelihood of choosing the sad face side, while lower values indicate higher probability of choosing the happy face side. Values around 0.5 indicate random choices. Each individual learning curve for each participant is shown in colors for each plot in each group – the black line shows the average time course in each group. The choice probabilities were estimated from the inverse link based on the best fitting model adjusting for covariates (Table 2).

Table 2.

The best fitting mixed effects model with choice as the outcome variable.

Variable Name B (SD) Z (p-value)
Intercept 0.31 (0.17) 1.7
Valence [Reference Group: Sad] −0.68 (0.04) −19.14 ***
Catch trial 0.09 (0.09) 0.97
Participant Race [Reference Group: White] −0.02 (0.04) −0.42
Age 0.00 (0.00) 2.45 *
Participant Sex [Reference Group: Male] 0.05 (0.04) 1.32
Trial −22.16 (5.60) −3.95 ***
Trial2 −15.29 (4.76) −3.21 ***
History of Depression [Reference Group: No] −0.07 (0.04) −1.76
Beck Depression Inventory 0.00 (0.00) 0.90
Handedness [Reference Group: Left] 0.03 (0.07) 0.40
ACD Focusing 0.00 (0.01) 0.40
ACD Shifting −0.00 (0.00) −0.23
Reaction time −0.15 (0.07) −2.27 *
Trial × Depression History Interaction 16.10 (7.86) 2.05 *
Trial2 × Depression History Interaction 8.21 (6.70) 1.25
Open in a new tab

Note. Β refers to unstandardized beta coefficients. Coefficients are expressed in log-odds of choice. Significance levels for p values:

***

0.001

**

0.01

*

0.05.

Bold variable names indicate statistical significance. Trial indicates the trial number (1 to 120).

We found a significant trial by depression interaction (see Figure 2). Participants without a history of depression learned to choose the happy biased side faster (linear rate) compared to participants with a history of depression. When we evaluated subject-level responses, we found greater variability in learning in the group with a history of depression compared to the group without a history of depression. Furthermore, in our additional analysis of learning pattern differences in remitted depression, current depression, and no history of depression, we did not find any differences (Appendix 2) between participants who had current versus remitted depression. Further, we did not detect a trial by current depression severity interaction (Appendix 3), nor a trial by anhedonia symptom interaction, (Appendix 4).

Discussion

In this study, we investigated social reward as an implicit learning signal for participants with and without a history of depression. We created a probabilistic social reward task wherein positive social reward was operationalized as a happy face to promote response bias learning and negative social reward was operationalized as a sad face. In support of our first hypothesis, we found that participants with a depression history were slower to develop a response bias towards happy faces than were control participants without a depression history. In support of our second hypothesis, we found that after a period of exploration, most participants developed a response bias -- a preference for choosing the side of the screen associated with the positive social reward. Importantly, these preferences were evident even after switching the side of the screen that displayed positive social reward, indicating that learning has taken place. In other words, participants actively responded to the social reward of looking at happy faces instead of just picking one side of the screen consistently throughout the task. The post-task questionnaires confirmed that participants were not aware of the hidden cues. Additional analysis confirmed that the results could not be explained by current depressive or anhedonic symptoms. Our results suggest that a history of depression, specifically with childhood or early adolescence onset, may have a long-term impact on social reward learning, regardless of current depressive and anhedonic symptoms.

Our findings demonstrate that, on average, participants with a history of depression have slower reward learning processes compared to participants with no history of depression. Importantly, these findings are likely not driven by differences in participants’ ability to focus, since those participants without a history of depression reported lower ability to focus compared to participants with a history of depression. As such, our findings are consistent with the view that depression history significantly impacts reward learning (Admon & Pizzagalli, 2015; Huys et al., 2013; Lawlor et al., 2020; Vrieze et al., 2013). Prior studies examining reward learning alterations in depression focused on conscious learning often using monetary rewards. We extend these findings to social reward and implicit learning. The finding of slower reward learning in depression has several potential explanations. First, it may be driven by an underlying difference in evidence accumulation needed for decision-making. Higher evidence accumulation threshold could be driven by reduced reward sensitivity, implying reduced internal worth of a reward in depression (Alloy et al., 2016; Takamura et al., 2017). It may also reflect that people with depression histories need more evidence to make certain decisions, suggesting a more cautious response style. These are not mutually exclusive and likely work in tandem.

In addition, our findings indicate that participants with no history of depression had a greater bias towards positive affective cues compared to participants with a history of depression. These findings are consistent with previous reports of a normative positivity bias, or a protective bias; namely non-depressed individuals generally focus more on happy faces and avoid sad faces (Joormann & Gotlib, 2007; Lazarov et al., 2018; Leyman et al., 2007). In contrast, when focusing on within-group patterns of those with a history of depression, previous studies reported amplified attention to sad faces (attention bias) and diminished attention to happy faces among individuals with current and remitted depression (Duque & Vázquez, 2015; Joormann & Gotlib, 2007; Lazarov et al., 2018; Leyman et al., 2007). Contrary to previous studies (Duque & Vázquez, 2015; Joormann & Gotlib, 2007; Lazarov et al., 2018; Leyman et al., 2007), we found that on average during a sequential learning task, individuals with a history of depression developed a greater bias towards happy faces (on average). There is a high degree of variability though across participants especially in the depressed group. These findings possibly suggest that while depression-prone people may attend more to sad faces, the sad faces do not uniformly represent a strong social reward, highlighting an important distinction between attention and reward signals. In other words, the co-existence of attention bias and response bias may suggest that participants with depression attend more to sad faces but do not necessarily finding them rewarding, and thus do not learn from them. The higher variability among participants with history of depression may reflect the individual differences in which the valence of faces functions as reward signals, demonstrating that negative faces may act as a reinforcer in some individuals but not others. Future studies should combine already existing measures to assess attention bias (Duque & Vázquez, 2015; Lazarov et al., 2018), protective bias, and response bias, to fully understand how these biases interact and influence implicit social reward learning. While such a thorough bias assessment may allow to disentangle the impacts of attention, protective, and response biases in depression, it might be challenging to keep the social reward aspect of the task hidden from participants, thus risking the implicit nature of the task.

Our results show that individuals with a history of depression exhibited more variable modulation of their behavior in response to reinforcement in the form of social reward, resulting in a lower group-level learning rate. While those with no history of depression rapidly learned to choose a side that corresponded with a positive affective social cue, this learning was slower in participants with a history of depression. Significantly slower reward learning in depression in response to social reward may suggest underlying differences in reinforcement sensitivity. Reinforcement sensitivity is based on a set of biobehavioral processes that mediate the relationship between reward or punishment processing and individuals’ personality, including their emotion, motivation, cognition, and even psychopathology (Katz et al., 2020; Smillie, 2008). Reinforcement sensitivity likely also impacts individuals’ emotion regulation – their ability to adapt to any situation at hand by adjusting their own expression and experience of emotion (Altan-Atalay, 2019). In fact, many studies point to a strong relationship between depression, reinforcement sensitivity, and emotion regulation, with some authors suggesting that reinforcement sensitivity may inherently impact the development of emotion regulation (Altan-Atalay, 2019; Depue & Iacono, 1989; Katz & Yovel, 2022; Tull et al., 2010). In our case, significantly slower reward learning in depression in response to social reward may be a combination of alterations in reinforcement sensitivity and emotion regulation deficits – a common symptom in both current and remitted depression (Visted et al., 2018). Therapeutic approaches for depression, such as Cognitive Behavioral Therapy (CBT), are thought to work, in part, by modulating neural correlates of emotion regulation (Rubin-Falcone et al., 2020), but it is unclear if they impact reward sensitivity. Future studies should examine the influence of CBT on emotion regulation as well as reinforcement sensitivity to determine the associations between alterations in both processes and treatment outcomes. Furthering our understanding of the associations between depression, reinforcement sensitivity, and emotion regulation may aid us in development of more translatable psychotherapeutic approaches that ameliorate depressive symptoms by targeting a multitude of processes at once.

Our study has several limitations. There was a significant race difference between our groups, the implications of which are unclear regarding social reward learning. Specifically, the group with a history of depression had a significantly higher proportion of White participants compared to the group with no history of depression. These sample differences may be driven by racial and ethnic disparities in the diagnosis of depression (Bailey et al., 2019). Nevertheless, we controlled for race in all our models, and we did not find it to be a significant predictor of social reward learning. We would also like to highlight a limitation in the stimuli used for this study, which included White actors only. The failure to detect learning differences between those with current and remitted depression (Appendix 2), depressive symptom severity (Appendix 3), and anhedonic symptom severity (Appendix 4) could have been driven by the small sample size of participants with current depression. While we found that those with depression history learn slower due to lower reward sensitivity, the current task was not designed to pinpoint the exact driving factor for this effect. Further, even though the experiment was designed to randomize which side was initially associated with a happy face, this randomization failed, and the positive reinforcer was always associated with the left side for the first 80 trials, and then switched to the right side for the last 40 trials. Although our results suggest that learning was driven by a happy face rather than a side of button (nor an effect of handedness or attention), future studies should replicate these findings by correctly implementing the random assignment of the side of reinforcer across participants. In addition, our current manuscript and analysis does not focus on reversal learning, where studies have shown that individuals with depression are slower to adapt to reversals(Mukherjee et al., 2020). Our current task would likely need some adaptation to evaluate this further and in addition a more traditional reinforcement learning modeling approach is needed to better model such reversals.

In conclusion, by employing a probabilistic implicit learning task with a social reward presented without conscious awareness, we found that participants with a history of depression have altered learning patterns, thus replicating, and extending previous findings. Our results suggest the need to further understand the long-lasting effect of depression on reinforcement sensitivity and emotional regulation as this may aid the development of cognitive behavioral interventions targeting maladaptive emotion regulation.

Supplementary Material

MMC1

Acknowledgements.

The authors would like to thank participants in this study.

Funding.

This work was funded by NIH Grant number MH 113214.

Footnotes

Conflicts of Interest. The authors declare no conflicts of interest.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Admon R, & Pizzagalli DA (2015). Dysfunctional reward processing in depression. Current opinion in psychology, 4, 114–118. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Ait Oumeziane B, Jones O, & Foti D. (2019). Neural sensitivity to social and monetary reward in depression: clarifying general and domain-specific deficits. Frontiers in behavioral neuroscience, 13, 199. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Alloy LB, Olino T, Freed RD, & Nusslock R. (2016). Role of reward sensitivity and processing in major depressive and bipolar spectrum disorders. Behavior therapy, 47(5), 600–621. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Altan-Atalay A. (2019). Interpersonal emotion regulation: Associations with attachment and reinforcement sensitivity. Personality and Individual Differences, 139, 290–294. [Google Scholar]
  5. Bailey RK, Mokonogho J, & Kumar A. (2019). Racial and ethnic differences in depression: current perspectives. Neuropsychiatric disease and treatment, 603–609. [DOI] [PMC free article] [PubMed]
  6. Barch DM, Pagliaccio D, & Luking K. (2015). Mechanisms underlying motivational deficits in psychopathology: similarities and differences in depression and schizophrenia. Behavioral neuroscience of motivation, 411–449. [DOI] [PubMed]
  7. Beck AT, Steer RA, & Brown G. (1996). Beck depression inventory–II. Psychological assessment.
  8. Depue RA, & Iacono WG (1989). Neurobehavioral aspects of affective disorders. Annual review of psychology, 40(1), 457–492. [DOI] [PubMed] [Google Scholar]
  9. Derryberry D, & Reed MA (2002). Anxiety-related attentional biases and their regulation by attentional control. Journal of abnormal psychology, 111(2), 225. [DOI] [PubMed] [Google Scholar]
  10. Dezfouli A, Griffiths K, Ramos F, Dayan P, & Balleine BW (2019). Models that learn how humans learn: the case of decision-making and its disorders. PLoS computational biology, 15(6), e1006903. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Duque A, & Vázquez C. (2015). Double attention bias for positive and negative emotional faces in clinical depression: Evidence from an eye-tracking study. Journal of behavior therapy and experimental psychiatry, 46, 107–114. [DOI] [PubMed] [Google Scholar]
  12. First MB (2014). Structured clinical interview for the DSM (SCID). The encyclopedia of clinical psychology, 1–6.
  13. Flores LE, Eckstrand KL, Silk JS, Allen NB, Ambrosia M, Healey KL, & Forbes EE (2018). Adolescents’ neural response to social reward and real-world emotional closeness and positive affect. Cognitive, Affective, & Behavioral Neuroscience, 18, 705–717. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Glaser WR, & Glaser MO (1989). Context effects in stroop-like word and picture processing. Journal of Experimental Psychology: General, 118(1), 13. [DOI] [PubMed] [Google Scholar]
  15. Goerlich KS, Votinov M, Lammertz SE, Winkler L, Spreckelmeyer KN, Habel U, Gründer G, & Gossen A. (2017). Effects of alexithymia and empathy on the neural processing of social and monetary rewards. Brain Structure and Function, 222(5), 2235–2250. [DOI] [PubMed] [Google Scholar]
  16. Gong L, Yin Y, He C, Ye Q, Bai F, Yuan Y, Zhang H, Lv L, Zhang H, & Xie C. (2017). Disrupted reward circuits is associated with cognitive deficits and depression severity in major depressive disorder. Journal of psychiatric research, 84, 9–17. [DOI] [PubMed] [Google Scholar]
  17. Gu R, Huang W, Camilleri J, Xu P, Wei P, Eickhoff SB, & Feng C. (2019). Love is analogous to money in human brain: Coordinate-based and functional connectivity meta-analyses of social and monetary reward anticipation. Neuroscience & Biobehavioral Reviews, 100, 108–128. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Guath M, Willfors C, Avdic HB, Nordgren A, & Kleberg JL (2022). Pupillary response in reward processing in adults with major depressive disorder in remission. Journal of the International Neuropsychological Society, 1–10. [DOI] [PubMed]
  19. Halahakoon DC, Kieslich K, O’Driscoll C, Nair A, Lewis G, & Roiser JP (2020). Reward-processing behavior in depressed participants relative to healthy volunteers: A systematic review and meta-analysis. JAMA psychiatry, 77(12), 1286–1295. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Hegedűs KM, Szkaliczki A, Gál BI, Andó B, Janka Z, & Álmos PZ (2018). Decision-making performance of depressed patients within 72 h following a suicide attempt. Journal of affective disorders, 235, 583–588. [DOI] [PubMed] [Google Scholar]
  21. Hirschfeld R, Montgomery SA, Keller MB, Kasper S, Schatzberg AF, Hans-Jurgen M, Healy D, Baldwin D, Humble M, & Versiani M. (2000). Social functioning in depression: a review. Journal of Clinical Psychiatry, 61(4), 268–275. [DOI] [PubMed] [Google Scholar]
  22. Huys QJ, Pizzagalli DA, Bogdan R, & Dayan P. (2013). Mapping anhedonia onto reinforcement learning: a behavioural meta-analysis. Biology of mood & anxiety disorders, 3(1), 1–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Joormann J, & Gotlib IH (2007). Selective attention to emotional faces following recovery from depression. Journal of abnormal psychology, 116(1), 80. [DOI] [PubMed] [Google Scholar]
  24. Katz BA, Matanky K, Aviram G, & Yovel I. (2020). Reinforcement sensitivity, depression and anxiety: A meta-analysis and meta-analytic structural equation model. Clinical psychology review, 77, 101842. [DOI] [PubMed] [Google Scholar]
  25. Katz BA, & Yovel I. (2022). Reinforcement sensitivity predicts affective psychopathology via emotion regulation: Cross-sectional, longitudinal and quasi-experimental evidence. Journal of affective disorders, 301, 117–129. [DOI] [PubMed] [Google Scholar]
  26. Kim Y, Pilkonis PA, Frank E, Thase ME, & Reynolds CF (2002). Differential functioning of the Beck depression inventory in late-life patients: use of item response theory. Psychology and aging, 17(3), 379. [DOI] [PubMed] [Google Scholar]
  27. Koster EH, De Lissnyder E, Derakshan N, & De Raedt R. (2011). Understanding depressive rumination from a cognitive science perspective: The impaired disengagement hypothesis. Clinical psychology review, 31(1), 138–145. [DOI] [PubMed] [Google Scholar]
  28. Kovacs M, & George CJ (2020). Maladaptive mood repair predicts suicidal behaviors among young adults with depression histories. Journal of affective disorders, 265, 558–566. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Kovacs M, Rottenberg J, & George C. (2009). Maladaptive mood repair responses distinguish young adults with early-onset depressive disorders and predict future depression outcomes. Psychological medicine, 39(11), 1841–1854. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Kupferberg A, Bicks L, & Hasler G. (2016). Social functioning in major depressive disorder. Neuroscience & Biobehavioral Reviews, 69, 313–332. [DOI] [PubMed] [Google Scholar]
  31. Lawlor VM, Webb CA, Wiecki TV, Frank MJ, Trivedi M, Pizzagalli DA, & Dillon DG (2020). Dissecting the impact of depression on decision-making. Psychological medicine, 50(10), 1613–1622. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Lazarov A, Ben-Zion Z, Shamai D, Pine DS, & Bar-Haim Y. (2018). Free viewing of sad and happy faces in depression: A potential target for attention bias modification. Journal of affective disorders, 238, 94–100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Leyman L, De Raedt R, Schacht R, & Koster EH (2007). Attentional biases for angry faces in unipolar depression. Psychological medicine, 37(3), 393–402. [DOI] [PubMed] [Google Scholar]
  34. Liu W. h., Roiser JP, Wang L. z., Zhu Y. h., Huang J, Neumann DL, Shum DH, Cheung EF, & Chan RC (2016). Anhedonia is associated with blunted reward sensitivity in first-degree relatives of patients with major depression. Journal of affective disorders, 190, 640–648. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Lundqvist D, Flykt A, & Öhman A. (1998). Karolinska directed emotional faces. Cognition and Emotion.
  36. Martins D, Rademacher L, Gabay A, Taylor R, Richey J, Smith D, Goerlich K, Nawijn L, Cremers H, & Wilson R. (2021). Mapping social reward and punishment processing in the human brain: A voxel-based meta-analysis of neuroimaging findings using the social incentive delay task. Neuroscience & Biobehavioral Reviews, 122, 1–17. [DOI] [PubMed] [Google Scholar]
  37. McCabe SB, & Gotlib IH (1995). Selective attention and clinical depression: performance on a deployment-of-attention task. Journal of abnormal psychology, 104(1), 241. [DOI] [PubMed] [Google Scholar]
  38. Miller A, Fox NA, Cohn JF, Forbes EE, Sherrill JT, & Kovacs M. (2002). Regional patterns of brain activity in adults with a history of childhood-onset depression: Gender differences and clinical variability. American Journal of Psychiatry, 159(6), 934–940. [DOI] [PubMed] [Google Scholar]
  39. Moutoussis M, Rutledge RB, Prabhu G, Hrynkiewicz L, Lam J, Ousdal O-T, Guitart-Masip M, Fonagy P, & Dolan RJ (2018). Neural activity and fundamental learning, motivated by monetary loss and reward, are intact in mild to moderate major depressive disorder. PLoS One, 13(8), e0201451. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Mukherjee D, Filipowicz ALS, Vo K, Satterthwaite TD, & Kable JW (2020, Nov). Reward and punishment reversal-learning in major depressive disorder. J Abnorm Psychol, 129(8), 810–823. 10.1037/abn0000641 [DOI] [PubMed] [Google Scholar]
  41. Mulders PC, van Eijndhoven PF, Schene AH, Beckmann CF, & Tendolkar I. (2015). Resting-state functional connectivity in major depressive disorder: a review. Neuroscience & Biobehavioral Reviews, 56, 330–344. [DOI] [PubMed] [Google Scholar]
  42. Oldfield RC (1971). The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia, 9(1), 97–113. [DOI] [PubMed] [Google Scholar]
  43. Parker R, Flint EP, Bosworth HB, Pieper CF, & Steffens DC (2003). A three‐factor analytic model of the MADRS in geriatric depression. International journal of geriatric psychiatry, 18(1), 73–77. [DOI] [PubMed] [Google Scholar]
  44. Pechtel P, Dutra SJ, Goetz EL, & Pizzagalli DA (2013). Blunted reward responsiveness in remitted depression. Journal of psychiatric research, 47(12), 1864–1869. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Pizzagalli DA, Iosifescu D, Hallett LA, Ratner KG, & Fava M. (2008). Reduced hedonic capacity in major depressive disorder: evidence from a probabilistic reward task. Journal of psychiatric research, 43(1), 76–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Pizzagalli DA, Jahn AL, & O’Shea JP (2005). Toward an objective characterization of an anhedonic phenotype: a signal-detection approach. Biological psychiatry, 57(4), 319–327. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Pollard TJ, Johnson AE, Raffa JD, & Mark RG (2018). tableone: An open source Python package for producing summary statistics for research papers. JAMIA open, 1(1), 26–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Reilly EE, Whitton AE, Pizzagalli DA, Rutherford AV, Stein MB, Paulus MP, & Taylor CT (2020). Diagnostic and dimensional evaluation of implicit reward learning in social anxiety disorder and major depression. Depression and anxiety, 37(12), 1221–1230. [DOI] [PubMed] [Google Scholar]
  49. Rmus M, Zou A, & Collins AG (2023). Choice Type Impacts Human Reinforcement Learning. Journal of Cognitive Neuroscience, 35(2), 314–330. [DOI] [PubMed] [Google Scholar]
  50. Rubin-Falcone H, Weber J, Kishon R, Ochsner K, Delaparte L, Doré B, Raman S, Denny BT, Oquendo MA, & Mann JJ (2020). Neural predictors and effects of cognitive behavioral therapy for depression: the role of emotional reactivity and regulation. Psychological medicine, 50(1), 146–160. [DOI] [PubMed] [Google Scholar]
  51. Safra L, Chevallier C, & Palminteri S. (2019). Depressive symptoms are associated with blunted reward learning in social contexts. PLoS computational biology, 15(7), e1007224. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Schaub A-C, Kirschner M, Schweinfurth N, Mählmann L, Kettelhack C, Engeli EE, Doll JP, Borgwardt S, Lang UE, & Kaiser S. (2021). Neural mapping of anhedonia across psychiatric diagnoses: A transdiagnostic neuroimaging analysis. NeuroImage: Clinical, 32, 102825. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Schwartz KT, Kryza-Lacombe M, Liuzzi MT, Weersing VR, & Wiggins JL (2019). Social and non-social reward: A preliminary examination of clinical improvement and neural reactivity in adolescents treated with behavioral therapy for anxiety and depression. Frontiers in behavioral neuroscience, 13, 177. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Segrin C. (2000). Social skills deficits associated with depression. Clinical psychology review, 20(3), 379–403. [DOI] [PubMed] [Google Scholar]
  55. Smillie LD (2008). What is reinforcement sensitivity? Neuroscience paradigms for approach-avoidance process theories of personality. European Journal of Personality, 22(5), 359–384. [Google Scholar]
  56. Tadayonnejad R, & Ajilore O. (2014). Brain network dysfunction in late-life depression: a literature review. Journal of geriatric psychiatry and neurology, 27(1), 5–12. [DOI] [PubMed] [Google Scholar]
  57. Takamura M, Okamoto Y, Okada G, Toki S, Yamamoto T, Ichikawa N, Mori A, Minagawa H, Takaishi Y, & Fujii Y. (2017). Patients with major depressive disorder exhibit reduced reward size coding in the striatum. Progress in Neuro-Psychopharmacology and Biological Psychiatry, 79, 317–323. [DOI] [PubMed] [Google Scholar]
  58. Tamir DI, & Hughes BL (2018). Social rewards: from basic social building blocks to complex social behavior. Perspectives on Psychological Science, 13(6), 700–717. [DOI] [PubMed] [Google Scholar]
  59. Tull MT, Gratz KL, Latzman RD, Kimbrel NA, & Lejuez C. (2010). Reinforcement sensitivity theory and emotion regulation difficulties: A multimodal investigation. Personality and Individual Differences, 49(8), 989–994. [Google Scholar]
  60. Vanheule S, Desmet M, Groenvynck H, Rosseel Y, & Fontaine J. (2008). The factor structure of the Beck Depression Inventory–II: An evaluation. Assessment, 15(2), 177–187. [DOI] [PubMed] [Google Scholar]
  61. Visted E, Vøllestad J, Nielsen MB, & Schanche E. (2018). Emotion regulation in current and remitted depression: a systematic review and meta-analysis. Frontiers in psychology, 9, 756. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Vrieze E, Pizzagalli DA, Demyttenaere K, Hompes T, Sienaert P, de Boer P, Schmidt M, & Claes S. (2013). Reduced reward learning predicts outcome in major depressive disorder. Biological psychiatry, 73(7), 639–645. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Weightman MJ, Air TM, & Baune BT (2014). A review of the role of social cognition in major depressive disorder. Frontiers in psychiatry, 5, 179. [DOI] [PMC free article] [PubMed] [Google Scholar]
  64. Whitton AE, Kakani P, Foti D, Van’t Veer A, Haile A, Crowley DJ, & Pizzagalli DA (2016). Blunted neural responses to reward in remitted major depression: A high-density eventrelated potential study. Biological Psychiatry: Cognitive Neuroscience and Neuroimaging, 1(1), 87–95. [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Zhang D, Shen J, Bi R, Zhang Y, Zhou F, Feng C, & Gu R. (2022). Differentiating the abnormalities of social and monetary reward processing associated with depressive symptoms. Psychological medicine, 52(11), 2080–2094. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

MMC1
NIHMS1887348-supplement-MMC1.docx (61.1KB, docx)

RESOURCES