Dysfunctional expectations constitute a core feature of depression (Kube et al., 2018b; Rief & Joormann, 2019; Strunk & Adler, 2009). As such, they are predictive of the course of depressive symptoms, e.g., by predicting relapse (Gopinath et al., 2007), treatment outcome (Gordon et al., 2011) and symptom severity at post treatment (Ludman et al., 2003). Thus, modifying dysfunctional expectations in depression is of high clinical importance (Rief & Glombiewski, 2016; Wilhelm et al., 2022). However, this is often challenging, since people with elevated levels of depression tend to hold on to dysfunctional expectations even when they make positive, expectation-disconfirming experiences (Everaert et al., 2018; Kube et al., 2019b). In other words, people with depressive disorders (vs. non-clinical controls) show difficulties in changing expectations into positive direction after receiving novel positive information (Kube et al., 2019), while they do not differ from non-clinical controls in changing expectations into negative direction after receiving novel negative information (Kube et al., 2019). Thus, a better understanding of the mechanisms that contribute to the maintenance of dysfunctional expectations is necessary to be able to effectively modify them. To this end, the present study investigated the influence of state affect1 on expectation change in patients with depression. This question is of significance because people with depression often have negative affect and if such negative affect were to hamper the ability to process and integrate novel positive information, this would be a plausible explanation for why people with depression have difficulty with learning from new experiences.
As reviewed by Kube (2023a), previous studies in this field have put a special focus on the investigation of cognitive mechanisms, such as interpretation inflexibility (Everaert et al., 2018; Liknaitzky et al., 2017), lack of optimistic update bias (Hobbs et al., 2022; Korn et al., 2014), and the devaluation of disconfirmatory evidence, referred to as cognitive immunization (Kube et al., 2019; Kube, Rief, Kube et al., 2019a, b, c). However, research on potential affective mechanisms of the persistence of dysfunctional expectations in depression is still scarce. Elucidating possible affective mechanisms could be vital to the understanding of the failure to integrate new positive learning experiences in depression, because they may amplify maladaptive cognitive processes and result in a self-reinforcing negative feedback loop (Smith et al., 2018). The general idea that current affect influences the update of self-beliefs and expectations draws from different theoretical lines of reasoning.
First, research from the general psychology literature suggests that the question of how well an information can be processed depends on whether its valence is congruent with people’s current affect, referred to as mood-congruency approach (Ziegler, 2010, 2014). This means that if people are currently experiencing negative affect, they might be able to effectively process negatively valent information, but would have difficulty processing positively valent information – and vice versa. Thus, depending on people’s current affect, it will be more vs. less difficult for people to adjust their expectations in response to positive vs. negative information. Second, with reference to the impaired cognitive disengagement hypothesis in depression (Koster et al., 2011), it is plausible that the incongruence between negative affect and positive information is a particularly important problem in clinically depressed people. According to this hypothesis, the occurrence of negative affect is often accompanied by dysfunctional rumination, which further amplifies negative affect, and so forth. As a result, people with depression have difficulty disengaging from negative contents and, accordingly, have problems using new information to revise their beliefs. Third, in terms of interoceptive predictive processing (Allen, 2019; Barrett et al., 2016; Barrett & Simmons, 2015), people use their interoceptive state as an additional source of information to weight sensory input. For instance, if the externally received information is positive (e.g., “You did well”), but the information from the interoceptive state is negative because of negative affect (e.g., “I’m feeling down”), this discrepancy may bias the update of people’s beliefs towards the momentary affect.
Some of these predictions have been confirmed by previous research in the context of depression. Preliminary research has shown that negative affect, which is particularly prevalent in people with depression, can hinder the adjustment of negative performance expectations in response to positive performance feedback. Specifically, in an inpatient sample of patients with depression, the activation of negative affect before receiving positive performance feedback lowered the likelihood of integrating the feedback, relative to the activation of positive affect (Kube et al., 2023). In a non-clinical sample and a subclinical sample, there was no significant difference between the activation of negative and positive affect or a neutral control condition, respectively; however, there was evidence that with increasing levels of depression, expectations were less adjusted to positive feedback and this relationship was more pronounced when negative affect was induced (Kube, Glombiewski, Kube et al., 2019a, b, c, 2023). In another study using a non-clinical sample, the induction of negative affect hindered the update of beliefs about the self in response to positive social feedback, relative to the induction of positive affect (Kube & Korn, 2024). Additionally, a recent study in the field of predictive processing by Ramos-Grille et al. (2022) found that participants with depression (vs. non-clinical controls) showed greater prediction errors (i.e., less trial-by-trial flexibility in their expectations) and this was even more pronounced after inducing sadness.
However, from previous research, it remains unclear whether negative affect only limits integrating novel positive information in depression or any information regardless of its valence. The present research addressed this issue by not only manipulating affect (sadness vs. amusement) but also the valence of feedback (negative vs. positive). To build our hypotheses, we considered the following related bodies of literature.
According to the aforementioned mood-congruency approach, the congruency between people’s mood and the valence of new information is crucial for effective information processing (Ziegler, 2010, 2014). Similarly, research on mood-biased predictive judgements (for a review, see Eldar et al., 2016) has indicated that positive affect leads to more optimistic and less pessimistic predictions about future events (Paul & Pourtois, 2017), whereas negative affect leads to less optimistic and more pessimistic predictions (Mayer et al., 1992; Wright & Bower, 1992). Applying this to our research question, sadness should only reduce the integration of positive feedback, with which it is incongruent, but not of negative feedback.
Another body of literature suggests that learning (Bakic et al., 2014) and cognitive flexibility (Dreisbach & Goschke, 2004; Goschke & Bolte, 2014) is enhanced under positive affect while transfer and learning (Brand et al., 2007) and the application of contextual information (Masuyama & Mochizuki, 2020) is hampered under negative effect. According to these lines of research, expectation change would be supposed to be impaired under negative affect, regardless of the valence of the feedback received.
Acknowledging the theoretical plausibility of both – a mood-congruency effect and a general impairment of expectation change under negative affect – we opted to ground our hypotheses in the former for the following reason. We believe that our rationale for investigating expectation change in response to negative vs. positive performance feedback in dependence of negative vs. positive state affect is more closely related to the methodology of research on the mood-congruency effect, which also puts a focus on how affect effects the content of thinking by giving it a mood-congruent color. On the other hand, the above-referenced line of research on impaired learning under negative affect might be more loosely related to our rationale as it focuses rather on the relation between affect and formal impairments in thinking.
Accordingly, we tested the following hypotheses. Based on previous research indicating reduced integration of novel positive information in depression (as reviewed by Kube, 2023a), we first predicted that receiving positive feedback will lead to less expectation change than receiving negative feedback. Second, in terms of a mood-congruency effect, we hypothesized that the influence of feedback on expectation change is greater if the valence of feedback (positive vs. negative) is congruent with participants’ affect (amusement vs. sadness).
Furthermore, it is important to consider that people do not passively experience affective states, but rather actively respond to them (Joormann & Stanton, 2016). This is of particular interest in depression, since depression is associated with altered affect regulation resulting in an increased experience of negative affect and reduced experience of positive affect (Joormann & Stanton, 2016). Relevant affect regulation strategies in this context include rumination in response to negative affect, suppression of negative as well as positive affect and dampening of positive affect (Bean et al., 2022; Beblo et al., 2012; Joormann & Stanton, 2016). Moreover, in recent years, these strategies have been widely associated with cognitive biases, such as interpretation inflexibility (Everaert et al., 2020), over-general autobiographical memories (Joormann & Stanton, 2016), or cognitive costs, such as impaired memory (Gross, 2002). Given the importance of such response styles in depression, we will also explore how rumination, suppression of positive and negative feelings, and dampening of positive feelings relate to expectation change in depression.
Rumination has been defined as repetitive thinking about one’s depressive symptoms and their causes, implications and meanings in response to negative affect (Nolen-Hoeksema et al., 2008). As rumination worsens and maintains negative affect (e.g., Vickers & Vogeltanz-Holm, 2003), we expected that the impact of our sadness-induction on expectation change will be boosted in individuals with high levels on rumination. Suppression is defined as an attempt to reduce affective reactions, including expressional, psychological and psychophysiological characteristics (Beblo et al., 2012). As suppression is cognitively demanding and as such impairs working memory (e.g., Richards & Gross, 1999), we expected suppression to hinder expectation change. Dampening has been defined as cognitive downregulation of positive affect by negatively appraising positive affective experiences (Bean et al., 2022). As dampening has been shown to be linked to negative interpretation bias (Everaert et al., 2020), we expected reduced expectation change in response to positive feedback in individuals with high levels in dampening.
Method
We report how we determined our sample size, all data exclusions, all manipulations and all measures in the study. All data, analysis code, and research materials are available at https://osf.io/7q2v5/?view_only=ad992506a8754d1d93b51121a1013586 (Rapo et al., 2023). Data were analyzed using R-Studio Version 4.2.2 (R Core Team, 2022) using the following packages: “tidyverse” (Wickham et al., 2019), “ggpubr” (Kassambara, 2022b), “reshape2” (Wickham, 2007), “psych” (Revelle, 2022), “utils” (R Core Team, 2022), “car” (Fox & Weisberg, 2019), “lme4” (Bates et al., 2015), “lmerTest” (Kuznetsova et al., 2017), “emmeans” (Lenth, 2023), “effectsize” (Ben-Shachar et al., 2020), “datarium” (Kassambara, 2019), “patchwork” (Pedersen, 2023), and “rstatix” (Kassambara, 2022a). The local ethics committee (reference number 2019 − 216) approved the study. Study design, hypotheses and analysis plan were preregistered at ‘AsPredicted’ (https://aspredicted.org/FDG_5F7). All participants gave written informed consent.
Participants
We determined the required sample size with an a-priori power simulation using the packages “lme4” (Bates et al., 2015) and “simr” (Green & MacLeod, 2016) in R-Studio (R Core Team, 2022). Thereby, we specified a linear mixed model as indicated in the main analysis below. Based on the literature on the modification of expectations in depression (Kube et al., 2019) and the influence of affect on expectation change in depression (Kube et al., 2023), we predefined the following regression weights: β = 0.27 for Feedback, β = 0.1 for Affect and β = 0.1 for Baseline expectations. For the main effect of Feedback, a power of 83.3% was found for a sample of N = 42. To be able to address our research question regarding the affective mechanisms, we pre-registered the aim to recruit a larger sample of N = 80 individuals, even though according to our a-priori power analysis this sample size appeared to be insufficient to reliably detect an interaction effect of feedback and affect. In fact, according to our a-priori power simulation a sample size of N = 305 would have been required to reliably detect an interaction effect of feedback and affect (1-β = 80.5). We decided to address our research questions using the sample size of N = 80 nevertheless, as a similar sample size proved to be sufficient to detect an interaction effect of affect and feedback in a previous study using the EXPEC (Kube et al., 2023). We recruited patients from an university outpatient clinic who met the criteria of an episodic major depressive disorder or Dysthymia (ICD-10: F32.0-F32.2, F33.0-F33.2, F34.1) as the primary diagnosis, according to a diagnostic interview (usually the German version of the Structured Clinical Interview for DSM SCID-5-CV; Beesdo-Baum et al., 2019) or the German „Diagnostisches Kurz-Interview bei psychischen Störungen“ (Mini-Dips; Margraf, 1994), as determined by trained clinical psychologists. The SCID-5-CV is a reliable and valid diagnostic instrument with good to excellent interrater-reliability (κ ≥ 0.7; Osório et al., 2019). Dysthymia could also be a secondary diagnosis if patients met the criteria of a major depression as the primary diagnosis (reflecting a “double depression”). The diagnoses were based on ICD-10 classification which represented the common classification system in clinical practice in Germany at the time the study was conducted. Further inclusion criteria were: minimum age of 18 years and fluency in German. Exclusion criterion was a bipolar affective disorder. Participants received financial compensation (8€) for their participation. We recruited N = 84 participants from September 2021 until June 2023. Four participants were excluded due to diagnostic failure, resulting in a final sample size of N = 80 (53 female, 25 male, 2 diverse). Their age ranged from 18 to 64 years (M = 35.01, SD = 13.95). For an overview of the sample characteristics, see Table 1. There were unintended significant differences in age between affect-conditions, with age being significantly lower in the amusement-condition (M = 31.95, SD = 13.02) than in the sadness-condition (M = 38.08, SD = 14.33; t(78) = -2.00, p = .049). There were no significant differences in baseline affect or severity of depressive symptoms.
Table 1
Sample characteristics
Variable | |
|---|---|
Age in years, M (SD) | 35.01 (13.95) |
Gender, N (%) | |
Male | 25 (31.25) |
Female | 53 (66.25) |
Diverse | 2 (2.5) |
Highest educational level, N (%) | |
School student | 3 (3.75) |
School diploma | 28 (35) |
Qualification for university entrance | 34 (42.5) |
University diploma | 10 (12.5) |
Other | 5 (6.25) |
Employment status, N (%) | |
In Training | 6 (7.5) |
Student | 16 (20) |
Employee | 37 (46.25) |
Self-employed | 8 (10) |
Unemployed | 8 (10) |
Other | 5 (6.25) |
Diagnoses (ICD-10), N (%) | |
Primary diagnosis | |
Major depressive episode Recurrent depressive disorder Dysthymia | 26 (32.5) 41 (63.75) 3 (3.75) |
Secondary diagnosis | |
None | 42 (52.5) |
Substance use disorder | 4 (5) |
Dysthymia | 10 (12.5) |
Phobic disorder | 10 (12.5) |
Generalized anxiety disorder | 5 (6.25) |
Obsessive compulsive disorder | 2 (2.5) |
Posttraumatic stress disorder | 3 (3.75) |
Eating disorder | 2 (2.5) |
Other | 2 (2.5) |
Tertiary diagnosis | |
None | 67 (83.75) |
Phobic disorder | 5 (6.25) |
Obsessive-compulsive disorder | 1 (1.25) |
Eating disorder | 1 (1.25) |
Insomnia | 1 (1.25) |
Personality disorder | 1 (1.25) |
Other | 4 (5) |
BDI-II, M (SD) | 24.15 (9.41) |
General Procedure
A graphical illustration of the experimental procedure is shown in Fig. 1. The study was conducted via www.soscisurcey.de. We collected data at a laboratory room of the RPTU Kaiserslautern-Landau. The procedure is an adapted version of the well-established “EXperimental Paradigm to investigate Expectation Change (EXPEC; Kube et al., 2018). Before starting the study, participants red a study information which contained the following cover story: “The aim of the study is to investigate the extent to which the presence of a sad or happy mood influences one’s own expected and actual performance. Previous studies have shown that people with depressive symptoms tend to underestimate their own performance. In our study, we are now interested in how the current mood influences this effect.”
Following this, either positive (“The tasks in this test are very easy”) or negative (“The tasks in this test are very difficult.”) baseline expectations were induced. Then baseline expectations and affect were assessed. Subsequently, participants received the first affect induction (amusement vs. sadness), which was followed by another assessment of affect.
Participants then took the first part of the performance test, which was followed by expectation-disconfirming feedback. That is, participants who previously received an induction of positive expectations now received negative feedback and vice versa. This was followed by another measurement of affect and expectations.
Then, affect was induced a second time. Participants who previously received an induction of amusement again received an induction of amusement and participants who previously received an induction of sadness again received an induction of sadness. This was followed by another measurement of affect.
Participants took the second part of the performance test and received feedback that disconfirmed the feedback from the first part. That is, participants who received negative feedback after the first part now received positive feedback and vice versa. Subsequently, expectations and affect were measured again.
At the end, emotion regulation and sociodemographic variables were assessed and participants were debriefed.
Fig. 1
Study procedure
×
![]()
Manipulation of Baseline Expectations
We manipulated participants’ baseline expectations to be able to systematically disconfirm them subsequently. Randomly, we induced either positive (“The tasks in the test were designed by the developers to be very simple and to be correctly solved by most people.”) or negative (“The tasks in the test were designed by the developers to be very difficult and to be correctly solved by only few people.”) baseline expectations at the beginning of the study. Both the induction of positive and negative baseline expectations were successfully used in previous studies on expectation change using the EXPEC task (Kube & Glombiewski, 2021; Kube et al., 2023; Kube, Rief, Kube et al., 2018a, b).
Performance Test
As in previous studies using the EXPEC task (Kube & Glombiewski, 2021; Kube et al., 2023; Kube, Rief, Kube et al., 2018a, b), participants completed the “Test of Emotional Intelligence” (TEMINT; Schmidt-Atzert & Bühner, 2002). The main reason for the choice of the test is that estimating their actual performance on the test is difficult for participants, which is an important prerequisite for the credibility of the standardized feedback. Another advantage of the TEMINT is that depression is unrelated to actual performance deficits in this test (Kube et al., 2022). For the current study, the TEMINT was split into two parts, each comprising six brief descriptions of situations that were actually experienced by a real person (e.g., “Computer scientist, age 30: ‘My cat was sick, I had to take her to the vet. I believed that I had poisoned her with the insect spray.’”). For each situation, participants had to empathize with the respective person and rate how much this person had experienced certain emotions (e.g., fear, sadness or guilt) on a scale from 1 (“non-existent or very weak”) to 3 (“strong or very strong”). Completing each part of the TEMINT was followed by standardized performance feedback as described below.
Performance Feedback
After each of the two parts of the TEMINT, participants received standardized performance feedback, which was independent of their actual performance. The feedback was supposed to convey a disconfirmation of participants’ expectations: Participants who underwent the induction of positive baseline expectations received negative feedback after the first part of the TEMINT (“You correctly solved 18 of 42 tasks. Thus, you belong to the bottom 42% of all participants.”); participants who underwent the induction of negative expectations received positive feedback (“You correctly solved 32 of 43 tasks. Thus, you belong to the best 26% of all participants.”). After completing the second part of the TEMINT, participants received feedback again. Those who received positive feedback for the first part received negative feedback for the second part and vice versa.
We chose the specific percentages fed back to participants based on the following reasoning. We assumed that for the feedback to be perceived as negative, people should be informed that they are in the bottom half of the distribution. On the other hand, the percentage should not be too low since the TEMINT tasks are in fact not very difficult and being told that one solved just a few tasks correctly could be considered implausible. Accordingly, we reasoned that being among the bottom 42% would be perceived as negative by most people, but not implausibly negative. For the positive feedback, we assumed that it should be clearly in the upper half of the distribution. Also, we reasoned that it must not be too positive to prevent doubts about its credibility. Indeed, research has shown that extremely positive feedback raises doubts about its credibility and entails little expectation change, as opposed to moderately positive feedback (Kube, 2023b; Kube et al., 2022b). Accordingly, we opted for the upper 26%.
Induction of State Affect
Participants were randomly assigned to one of two affect-conditions: sadness and amusement. We induced affect by showing a short video-clip before each of the two parts of the TEMINT. We decided to show a movie clip before both parts of the TEMINT because we wanted to make sure that the affect induction was as equally effective as possible for both parts of the study procedure. The order of the video-clips was randomized. Participants in the amusement-condition watched a sequence of “The curse of Mr. Bean” (2:58 min) and of “Harry & Sally” (2:46 min). Participants in the sadness-condition watched a sequence of “My Girl” (2:17 min) and “The Champ” (2:44 min). Previous studies demonstrated the effectiveness of the video-clips in inducing amusement or sadness, respectively (Gabert-Quillen et al., 2015; Gilman et al., 2017; Gross & Levenson, 1995; Jurásová, Kinga & Špajdel, Marián, 2013). To avoid non-respondents, we supplemented the videoclips with the following instruction: “Try to imagine as best as you can how you would feel if you were present in this situation and observing the scene.”, following the suggestion of Rottenberg et al. (2018). We decided to induce sadness and amusement due to the following reason: As this study aimed to replicate and supplement the results of Kube et al. (2023), who used the film-clips of “The Champ” and “When Harry met Sally”, we decided to use the same film-clips in order to retain as much of their study protocol as possible.
Measures
Performance Expectations
We assessed expectations using the Performance Expectations Scale (Kube et al., 2018). The scale consists of two items to assess task-specific performance expectations (e.g., “Solving the tasks of the test will be easy for me.”) and two items to assess generalized performance expectations (e.g., “Solving unknown tasks in general will be difficult for me.”; reversely scored) with a seven-point Likert-scale from 1 (“totally disagree”) to 7 (“totally agree”). High scores on the scale represent positive performance expectations. In our sample, internal consistency was good to excellent for task-specific expectations (α = 0.88–0.91) and excellent for generalized expectations (α = 0.90–0.95). Performance expectations were measured three times: After the induction of baseline expectations, after receiving the first feedback and after receiving the second feedback.
Affect
We assessed affect, using the subscales Sadness and Joviality2 of the German version of the Positive And Negative Affect Scale - expanded form (PANAS-X; Röcke and Grühn (2003). The subscale Sadness contains five and the subscale Joviality eight items, each measuring the intensity with which one experiences a facet of the respective affect (e.g., “alone” in the Subscale Sadness or “enthusiastic” in the subscale Joviality) using a five-point Likert scale from 1 (“not at all”) to 5 (“extremely”). We decided to use the subscales instead of the full PANAS, since our manipulation of affect aimed to induce specific affective states (amusement vs. sadness) as opposed to general positive vs. negative affect. The PANAS-X is a reliable and valid measure for the assessment of self-reported affect (Watson & Clark, 1994). High scores on the scales represent a high intensity of self-reported joviality or sadness, respectively. In our sample, internal consistency was good to excellent for Sadness (α = 0.86–0.93) and excellent for Joviality (α = 0.90–0.97). Affect was measured five times: Right before and after each of the two affect manipulations and at the end of the study after receiving the last feedback.
Cognitive Immunization
As previous studies using the EXPEC task put a special focus on cognitive immunization, we also included cognitive immunization in this study. As it did not constitute a central construct in our study and there were no significant results regarding cognitive immunization, results regarding cognitive immunization will be presented in the supplement only (Online Resource 1). We assessed cognitive immunization using the Cognitive Immunization after Performance Feedback scale (CIPF; Kube et al., 2019). The scale consists of six items: two assessing the perceived credibility of the feedback received, two assessing the perceived relevance of the feedback and two assessing the extent to which people consider the feedback an exception to their normal performance. High scores on the scale represent high levels of cognitive immunization. In our sample, internal consistency was acceptable to good (α = 0.77–0.81).
Depressive Symptoms
We assessed depressive symptoms using the German translation of the Beck-Depression-Inventory Revised Version (BDI-II; Hautzinger et al., 2009). The BDI-II consists of 21 items, each measuring the manifestation of typical symptoms of depression on a scale from 0 to 3. High sum scores on the scale represent high severity of depressive symptoms. In our sample, internal consistency was good (α = 0.89).
Affect Regulation and Other Measures
As described above, we also included self-reported rumination, suppression and dampening of positive feelings. We decided to include these constructs because they are linked to depressive symptoms (Bean et al., 2022; Beblo et al., 2012; Joormann & Stanton, 2016), cognitive biases (Everaert et al., 2020; Joormann & Stanton, 2016) and cognitive costs (Gross, 2002) and could thus offer an additional exploratory value for altered expectation change in depression. We measured rumination using the shortened German version of the Response Style Questionnaire (RSQ-D; Bürger & Kühner, 2007), Suppression of positive and negative feelings using the subscales Suppression/Dampening of the Fragebogen zur Erfassung der Akzeptanz von unangenehmen und angenehmen Gefühlen (FrAGe; Beblo et al., 2011) and dampening of positive feelings using the subscale Dampening of the German version of the Responses to Positive Affect Questionnaire (RPA; Voss et al., 2019). The RSQ-D and the RPA measure the frequency with which people use the respective affect regulation strategy. The RSQ-D comprises 21 items, and the RPA comprises eight items with a four-point Likert-scale ranging from 1 (“almost never”) to 4 (“almost always”). The subscale Suppression/Dampening of the FrAGe measures the tendency to suppress feelings using eight items each for positive and negative feelings on a six-point Likert-Scale ranging from 1 (“Not at all”) to 6 (“Extremely”). The measures show good psychometric properties (Beblo et al., 2011; Bürger & Kühner, 2007; Voss et al., 2019). In our sample, the internal consistencies of all measures were good (α = 0.81–0.89). Further measures who were integrated in the study, but were not included in the analysis for this paper were the German version of the Trait-Meta-Mood-Scale (Otto et al., 2001) and Die Skala Angst vor negativer Bewertung 5 (SANB-5, Kemper et al., 2012).
Sociodemographic Variables
We assessed gender, age, level of formal education, employment status and level of German language fluency.
Statistical Analysis
We screened the data and tested the assumptions of analysis of variance and linear mixed models by analyzing residual plots and Q-Q-plots, conducting Levene-tests and determining Cook’s distance for each dependent variable. We checked for potential baseline differences between affect-conditions by applying t-tests for continuous variables and Chi2-tests for categorial variables. We verified that no demand effects were present by reading participants answers in the follow-up-interview.
Manipulation Checks
To test whether the manipulation of baseline expectations was successful, we conducted two separate Welch’s t-tests with baseline expectations (positive vs. negative) as independent variable and task-specific expectations at t1 and generalized expectations at t1 as dependent variables, respectively.
To test whether our manipulation of affect (amusement vs. sadness) was successful, we specified two separate linear mixed models allowing the fixed effects Time (t1 vs. t2 vs. t3 vs. t4 vs. t5) and Affect-condition to interact in explaining the dependent variables amusement and sadness, respectively.
Main Analysis
To test whether positive or negative feedback has varying effects on expectation change depending on affect (amusement vs. sadness), we specified linear mixed models allowing for an interaction between the fixed effects Feedback and Affect in explaining task-specific and generalized performance expectations, respectively. For the factor Feedback we defined an a-priori sum contrast to compare the deviation in expectation change for positive and negative feedback from average level of expectation change.
Exploratory Analysis
We conducted model comparisons as a data-driven approach to identify potential interaction effects between Baseline Expectations, Feedback and Affect. Therefore, we compared a model with exclusively additive effects of the fixed effects to five models each allowing for different combinations of interactions between these three effects. The model allowing an interaction between Baseline expectations and Feedback showed the best performance in explaining task-specific expectations and significant effects were followed up by post-hoc t-tests.
Furthermore, we conducted model comparisons with Rumination, Suppression and Dampening, respectively, as additional fixed effects. For all linear mixed models, a random intercept for each participant was estimated.
Results
Manipulation Check
Baseline Expectations
We tested whether our manipulation of baseline expectations (positive vs. negative) was successful by comparing task-specific and generalized performance expectations at t1. As hypothesized, participants had significantly lower task-specific expectations (M = 4.25, SD = 1.4) when they believed the test would be difficult (negative expectations) than when they believed the test would be easy (positive expectations; M = 5.1, SD = 1.39; t(77.15) = -2.7, p = .008, d = 0.62(95%CI[-1.07;-0.16])), but did not differ significantly in generalized performance expectations (M = 4.21, SD = 1.42; M = 4.5, SD = 1.58; t(78) = -0.86, p = .39).
Amusement
We tested whether amusement decreased through watching the sad video-clips and increased through watching the amusing video-clips.
There was a significant main effect of Time (F(4,312) = 11.11, p < .001, η2 = 0.12(95%CI[0.07;1.00])) and a significant interaction between Time and Affect-condition (F(4,312) = 20.33, p < .001, η2 = 0.21(95%CI[0.14;1.00])). Post-hoc contrasts revealed that at t2 (t(187) = -7.14, p < .001, d = -1.02(95%CI[-1.33;-0.72])) and t4 (t(187) = -6.43, p < .001, d = -0.94(95%CI[-1.24;-0.64])), amusement was significantly lower in the sadness vs. amusement condition. Amusement significantly increased from t1 to t2 (t(312) = 2.47, p = .014, d = 0.14(95%CI[0.03;0.25])) and from t3 to t4 (t(312) = 4.05, p < .001, d = 0.23(95%CI[0.12;0.34])) in the amusement-condition and significantly decreased from t1 to t2 (t(312) = -7.33, p < .001, d = -0.41(95%CI[-0.53;-0.30])) and from t3 to t4 (t(312) = -3.34, p < .001, d = -0.19(95%CI[-0.30;-0.08])) in the sadness-condition. Thus, right before vs. after the manipulation of affect, amusement successfully increased in the amusement-condition and successfully decreased in the sadness-condition. However, this effect was not stable over time as amusement re-decreased from t2 to t3 (t(312) = -6.09, p < .001, d = -0.34(95%CI[-0.46;-0.23])) and from t4 to t5 (t(312) = -2.53, p = .012, d = -0.14(95%CI[-0.25;-0.03])) in the amusement-condition and re-increased in the sadness condition (t2 to t3: t(312) = 2.32, p = .021, d = 0.13(95%CI[0.02;0.24])) and t4 to t5: t(312) = 3.99, p < .001, d = 0.23(95%CI[0.11;0.34])). A graphical illustration of the course of amusement and a table with descriptive values can be found in the supplement (Online Resource 1).
Sadness
We tested whether sadness decreased through watching the amusing video-clips and increased through watching the sad video-clips.
There was a significant main effect of Time (F(4,312) = 5.67, p < .001, η2 = 0.07(95%CI[0.02;1.00])) and a significant interaction between Time and Affect-condition (F(4,312) = 20.9, p < .001, η2 = 0.21(95%CI[0.14;1.00])). Post-hoc contrasts revealed that at t2 (t(124) = 3.05, p = .003, d = 0.55(95%CI[0.19;0.91])) and t4 (t(124) = 4.83, p < .001, d = 0.87(95%CI[0.50;1.23])) sadness was significantly higher in the sadness vs. amusement condition. Sadness significantly increased from t1 to t2 (t(312) = 4.65, p < .001, d = 0.26(95%CI[0.15;0.38])) and from t3 to t4 (t(312) = 6.11, p < .001, b1 = 0.35(95%CI[0.23;0.46])) in the sadness-condition and significantly decreased from t1 to t2 (t(312) = -3.7, p < .001, d = -0.21(95%CI[-0.32;-0.10])) and from t3 to t4 (t(312) = -2.37, p = .018, d = -0.13(95%CI[-0.25;-0.02])) in the amusement-condition. As for amusement, sadness was successfully manipulated, but this effect was not stable over time as sadness re-decreased in the sadness condition (t2 to t3: t(312) = -3.28, p = .001, d = -0.19(95%CI[-0.30;-0.07])) and t4 to t5: t(312) = -5.88, p < .001, d = -0.33(95%CI[-0.45;-0.22])). A graphical illustration of the course of sadness and a table with descriptive values can be found in the supplement (Online Resource).
Main Analyses
Valence of Feedback Predicts Expectation Change
We tested, whether valence of feedback, affect and an interaction between feedback and affect significantly predicted expectation change. There was a significant main effect of Feedback on task-specific expectations (F(2,156) = 11.41, p < .001, η2 = 0.13(95%CI[0.05;1.00])) and generalized expectations (F(2,156) = 8.55, p < .001, η2 = 0.1(95%CI[0.03;1.00])), indicating that participants lowered their expectations in response to negative feedback, whereas there was no significant expectation change in response to positive feedback. There was neither a significant main effect of Affect on task-specific expectations F(1,136.94) = 0.66, p = .42) or generalized expectations (F(1,104.34) = 1.27, p = .26) nor a significant interaction effect (task-specific expectations F(2,156) = 1.35, p = .26, generalized expectations: F(2,156) = 2.07, p = .13; see Fig. 2). To control for the unintended differences in age between affect-conditions, we conducted additional analysis with age as control variable. As age did not systematically influence the results of our main analysis, we did not control for age in any further analysis.
Fig. 2
Performance expectations at baseline, after negative feedback and after positive feedback. Note. Error bars represent standard deviations
×
![]()
Exploratory Analysis
Baseline Expectations Interact with Feedback Based on State Affect
Given the surprisingly non-significant results for the effects of affect on expectation change, we explored whether this effect could have been masked by the manipulation of baseline expectations. To this end, we computed an exploratory model comparison (see Online Resource 1). In the case of task-specific expectations, the model showing the highest performance included an interaction between Baseline expectations and Feedback and an additive effect of Affect. A further graphical exploration of these effects (see Fig. 3), supplemented by post-hoc t-tests, indicated that there was a non-significant trend suggesting that believing that the test was difficult (vs. easy) hampered expectation change in response to positive feedback (p = .073, Cohen’s d = -0.41(95%CI[-0.86;0.04])). With regard to the additive effect of Affect, there was another non-significant trend suggesting that sadness (as opposed to amusement) did promote expectation change in response to negative feedback, but only in participants who believed that the test was going to be easy (p = .068, Cohen’s d = 0.61(95%CI[-0.04;1.26])).
Fig. 3
Task-specific expectations in participants who were manipulated to believe that (a) the test is going to be easy (positive baseline expectations) and (b) the test was going to be difficult (negative baseline expectations) at baseline and after receiving positive and negative feedback. Participants of group (a) first received negative and then positive feedback and participants of group (b) vice versa
×
![]()
Does Affect Regulation Predict Expectation Change?
We conducted some exploratory model comparisons considering self-reported affect regulation strategies (see Online Resource 1). Except for Suppression of negative feelings, all assessed affect-regulation strategies had an additional exploratory value for expectations. To keep the presentation of results succinct, we report only the results regarding rumination in the main text, while the results for the other strategies are presented in the supplementary material.
There was a significant main effect of Rumination, with higher levels of rumination predicting overall lower performance expectations (task-specific expectations t(80) = -3.84, p < .001, d = -0.43(95%CI[-0.66;-0.20]), generalized expectations: t(80) = -3.1, p = .003, d = -0.35(95%CI[-0.57;-0.12])). There were no significant interaction effects with Feedback, indicating that rumination did not significantly predict expectation change, or Affect. However, on a graphical level (see Fig. 4), there were hints that in the case of sadness, rumination promoted change in expectations in response to negative feedback.
Fig. 4
Regression line between task-specific expectations (at baseline versus after receiving negative and positive feedback) and rumination in participants who either received an induction of amusement (left) or an induction of sadness (right panel)
×
![]()
Discussion
The present work aimed to replicate and extend previous research showing that sadness can hinder the revision of negative expectations based on novel positive information in depression. In line with previous studies (Kube et al., 2019; Kube et al., 2019), we found that patients with depression showed more pronounced expectation change in response to negative as compared to positive performance feedback. However, unlike previous research (Kube et al., 2023), the induction of sadness vs. amusement before receiving feedback did not influence the extent to which participants integrated feedback. Thus, the non-significant effects of affect on expectation change are neither consistent with the hypothesized mood-congruency assumption, nor with a general deficit in updating expectations under sadness. The following explanations might account for the non-significant effects of affect on expectation change.
First of all, statistical power of our study was limited. In fact, as indicated by a post-hoc-power analysis, our model including the interaction term of feedback and affect showed a power of only 30.6% for task-specific expectations and a power of 41.9% for generalized performance expectations. Thus, our choice to aim for a sample size of N = 80 might have been too optimistic, as the effects of affect on updating in depression seem to be smaller than previously assumed (Kube et al., 2023). If we also look at the experimental affect manipulation, the data suggests that even though participants’ affect was successfully modulated right after the experimental manipulation, these effects were neutralized by the next affect measurement, which took place right after receiving the performance feedback. Thus, it is likely that the induction of amusement or sadness was no longer effective when receiving the feedback.
Another explanation for the non-significant effects of affect on expectation change could be the complexity of our study design containing manipulations of baseline expectations, feedback and affect, which might have led to masking effects or complex interactions. Therefore, we carried out an exploratory model comparison to uncover putative masking and interaction effects using a data-driven approach. We found that baseline expectations (previously preregistered as a control variable) moderated the effect of feedback on expectation change. Specifically, negative baseline expectations reduced the impact of positive feedback on expectation change with a small effect size. In other words, participants showed a stronger resistance to positive feedback when they believed that the test was difficult (vs. easy). This is also in line with the results of Kube et al. (2022Kube (2023b), showing that expectation change decreases if the discrepancy between expectations and information is too large. Additionally, among those participants with positive baseline expectations, there was a non-significant trend with a medium effect size indicating that sadness was related to more pronounced expectation change in response to negative feedback. This is in line with a recent study showing that sadness reduces the optimistic update bias, meaning that under acute sadness people are more prone to integrate negative information than they usually are; thus, acute sadness is thought to reduce one’s “psychological immune system” (Karnick et al., 2023).
Regarding affect regulation, model comparisons showed that rumination, dampening of positive feelings and suppression of positive feelings had an additional exploratory value for performance expectations. Higher scores on the respective scales were associated with lower expectations. These results are consistent with findings about associations between altered affect regulation and biased cognition in depression (Everaert et al., 2020; Gross, 2002; Joormann & Stanton, 2016). On a descriptive level, there were hints that higher levels in rumination, dampening and suppression were also related to altered expectation change among sad but not among amused participants. This finding indicates that especially in sadness maladaptive affect regulation may lead to biased expectation change. However, our results do not allow us to draw any firm conclusions about the role of affect regulation on the change in expectations.
Implications for Future Research
Our findings have important implications for future research on expectation change in depression. We think that our results point out to the need to further explore the impact of baseline expectations as (a) baseline expectations may alter the openness for positive vs. negative feedback and (b) this may additionally depend on state affect. Thus, the investigation of cognitive and affective responses that are triggered through respective inductions of baseline expectations and their influence on expectation change could be worthwhile. Furthermore, our results suggest that emotion-eliciting film clips only induced amusement or sadness for a limited period of time, such that the induced affect had already disappeared by the time participants received the feedback. Therefore, it will be necessary for future studies to identify longer-lasting emotion-induction methods. Possibly, researchers may consider to instruct people to recall happy vs. sad memories as it is conceivable with reference to the impaired disengagement in depression (Koster et al., 2011) that once patients with depression start thinking about what went wrong in their lives, they have difficulty disengaging from it, such that the effects of such an instruction could last longer than those of a film clip.
Additionally, our results suggest the potential of the further exploration of affect regulation on expectation change. This is especially relevant since expectation violations can be considered as affective experiences and thus may trigger the use of affect regulation strategies. Building on a recent theoretical framework of Kube (2023a), maladaptive affect regulation such as dampening of positive feelings may lead to increased negative affect and a reduced ability to integrate positive experiences. This may amplify the persistence of negative expectations and in turn lead to sustained negative affect resulting in a vicious cycle (Smith et al., 2018). Thus, future research may consider the examination of the intercorrelation between reduced expectation change and sadness in depression and how this may be mediated by affect regulation.
Strengths and Limitations
To our knowledge, our study was the first to systematically investigate the effects of amusement vs. sadness on the integration of positive and negative feedback in depression. Further, we used a well-defined clinical sample, an established experimental task for the investigation of expectation change in depression, and preregistered both hypotheses and analysis plan.
Despite these strengths, an important limitation of the study is the relatively low sample size and the limited effect of the affect-manipulation, as described above. Furthermore, it should be noted that the complexity of the design in terms of manipulating baseline expectations, feedback, and affect may have led to some methodological issues, such as masking and sequence effects. For instance, it is conceivable that for some participants the initial information on whether the test is easy vs. difficult was more important for their updated expectations than the feedback received. Additionally, because the study was designed to gain a better understanding of the affective and cognitive mechanisms of feedback processing in depression, we did not include a non-depressed control group. Therefore, the present study does not allow any conclusions about how people with depression differ from non-depressed people.
Concluding Remarks
This study was designed to shed light on affective mechanism that promote vs. hinder expectation change in depression. As in previous studies, our results supported the hypothesis that people with depression are more prone to update their expectations in response to negative as compared to positive information. Regarding the influence of state affect, there was no support for our hypothesis of a mood-congruency effect. However, there were hints in our data that baseline expectations and affect regulation may influence expectation change in sad participants with depression. Future studies using larger sample sizes and more enduring affect inductions are needed to draw robust conclusions about how sadness influences expectation change in depression.
Declarations
Ethics Approval
All procedures performed in this study were in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The study was approved by the local ethics committee of the University of Landau (reference number 2019 − 216).
Consent to Participate
Informed consent was obtained from all individual participants included in the study.
Consent to Publish
Participants gave informed consent regarding publishing their data.
Competing Interests
The authors declare that they have no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
