Introduction

Adolescence is characterised by significant cognitive, emotional, hormonal, physical, and social changes1. During this phase, adolescents face various challenges, which include assuming more mature social roles that entail increased responsibilities and the need for independence, navigating peer and romantic relationships, and transitioning relationships with parents as they seek greater autonomy. These challenges place substantial demands on emotion regulation2, making adolescence a critical phase in the emergence of various mental disorders, specifically internalising disorders such as anxiety and mood disorders3. One self-regulatory process that is considered crucial in the development of effective emotion regulation is affective control4.

Affective control describes the application of cognitive control in affective contexts, that is, the ability to flexibly orient one’s attention and allocate cognitive resources to goal-relevant affective information while disengaging from irrelevant affective information4. For example, affective control enables us to maintain focus on studying in the library despite our friends making jokes in the background. Over the course of development, affective control appears to mature in a non-linear fashion, as research suggests that affective control is reduced during adolescence compared to childhood and adulthood, possibly creating a developmental window with increased vulnerability towards negative affective stimuli (for a comprehensive review, see4). This notion is in line with the findings of a large-scale meta-analysis of behavioural and neuroimaging data that identified deficits in affective control as a trans-diagnostic factor in mental disorders5. Specifically, impairments in affective control have been associated with internalising mental disorders6,7.

At the same time, these disorders have been linked to aberrant emotion regulation [ER;8,9,10]. ER can be broadly defined as an individual’s internal and external processes to modulate their emotional reactions in order to accomplish their goals11. For instance, higher levels of anxiety symptoms in adolescents were associated poorer emotional clarity (i.e., the ability to make sense of emotional responses), higher non-acceptance of emotions (i.e., unwillingness to accept certain emotional responses) as well as a more negative self-evaluation of one’s own ability to manage emotions12. Furthermore, higher levels of social anxiety and depression symptoms in adolescents were associated with decreased emotional awareness (i.e., the ability to recognize and understand one’s own emotions) and limited access to ER strategies13. Limited access to ER strategies and overall difficulties in trait ER also predicted depressive symptoms in adolescents, both cross-sectionally and longitudinally over a timespan of two years14. In addition, the choice of ER strategies in adolescents has been linked to psychopathology. A meta-analysis integrating 35 studies showed that adolescents with anxiety and depressive symptoms engaged in less cognitive reappraisal (i.e., the ability to reinterpret a situation to alter its subjective meaning), acceptance (i.e., allowing oneself to experience an emotion without attempting to change it), and problem-solving15. The utility and adaptiveness of distraction as an ER strategy is highly context dependent, as it can be applied in terms of positive refocusing or experiential avoidance16. Apart from the specific ER strategies applied, other factors tied to ER have been proposed to influence ER success, namely perceived ER success as a facet of ER-specific self-efficacy (i.e., one’s own beliefs about being able to achieve or having achieved ER success;12,17) as well as ER flexibility (i.e., the ability to evaluate contextual demands, have a range of strategies to choose from and modify the applied strategy based on its effectiveness as defined in the framework by Bonanno and Burton;18,19).

Although both affective control and ER have been linked to psychopathology and specifically internalising mental disorders, their relationship is poorly understood. The cognitive neuroscience perspectives of ER states that cognitive control processes that include affective control are vital for successful regulation of negative emotion20. What affective control and ER have in common is that they are both relying on someone’s goals and motivation21 and that abilities in one area are predictive of abilities in the other22. However, what components of affective control and ER are related to each other and how their interaction impacts internalising psychopathology comprises a gap in the research literature.

This gap in understanding their relationship could, in part, be attributed to previous approaches in assessing affective control. Affective control is commonly studied using the Emotional Stroop task (EST,23,24, for a meta-analysis see25), an adaptation of the classic Stroop task26. During the EST, participants view images of facial expressions depicting different emotions with an emotional word (e.g., “happy” or “fear”) printed across the face. Participants are instructed to identify the emotion expressed through the face whilst ignoring the emotion word. Depending on whether the emotion of the face and the word match, trials are marked as congruent or incongruent. Since incongruent trials contain an image and a word of different emotions causing semantic conflict, participants must exert affective control to inhibit their automatic tendency to read the word. As affective and cognitive information compete for processing resources27, the presence of incongruent affective information hampers cognitive processing, which manifests as longer reaction times (RT) and increased errors in incongruent trials relative to congruent trials. This phenomenon is called the congruency effect or Stroop interference effect28. The congruency effect in the EST represents an individual’s attentional bias towards emotional stimuli29 that manifests in failures in emotional control and has been found to be comparable across clinical and healthy participants30. A recent meta-analysis of the EST in depression and anxiety disorder found greater interference by diagnosis-related and negative stimuli, but not by positive stimuli31. When it comes to incongruent trials, individuals with high anxiety responded slower than individuals with low trait anxiety during incongruent trials showing a neutral face and a fearful word compared to a fearful face and a neutral word, suggesting a bias to fearful words in anxious subjects32.

Although the EST offers a robust way to assess affective control, and its individual task performance has been linked to ER, it remains unclear which aspects of ER difficulties and strategies are specifically linked to the EST. Two of these aspects are response inhibition and initiation to emotional contents, which develop linearly with ER in childhood and adolescence33. While cognitive and affective control abilities increase with age with a diminished cognitive performance in late adolescence34, the ability to regulate emotions also increases. Adolescents rely more on cognition-related ER strategies such as cognitive reappraisal and acceptance, while children rely more on behaviour-related ER strategies such as distraction (i.e., behavioural preoccupation with something else to not think about the emotion-eliciting stimulus) and are less able to switch between strategies, representing lower ER strategy repertoire and potentially lower ER flexibility35,36. To the best of our knowledge, this is the first study to look at ER as a multilevel process and to link ER difficulties, ER strategies, ER success and ER strategy repertoire with affective control in an experimental setting in young adults and adolescents with anxiety and depression.

In the present study, we investigated the relationship between affective control, operationalized as the congruency effect and the number of errors on the EST, and ER in adolescents and young adults (for simplicity, they will both be referred to as youth) with internalising disorders. First, we hypothesised that (i) youth with internalising mental disorders (an anxiety disorder and/or depression) would exhibit lower affective control compared to healthy youth. Second, we explored whether ER strategies and difficulties predict affective control following a stress induction task and hypothesised that across groups (ii) more cognitive reappraisal, distraction and a higher subjective success in implementing those strategies would positively predict affective control. Third, we explored which aspects of ER would predict affective control and hypothesized that (iii) higher levels of trait ER difficulties and the specific ER difficulties impulse control difficulties, higher lack of emotional awareness and poorer access to emotion regulation strategies would inversely predict affective control. Lastly, we explored the developmental trajectory of affective control and ER and hypothesised that (iv) age would moderate the relationship between the two variables.

Methods

This cross-sectional study was part of a larger study about emotion regulation and sleep (Herber et al., in preparation, see supplementary materials for a detailed overview of the study procedures) approved by the ethics committee of the University of Freiburg. All study procedures were carried out in accordance with the Declaration of Helsinki. The study was preregistered on As predicted (study number 106298), modifications of the preregistered design are reported in the supplementary materials.

Sample size calculation

Appropriate sample size for a MANOVA (Global effects) with two groups and three outcome variables was determined based on an a priori power analysis using G*Power (Version 3.1.9.4;37). We used an estimated effect of f2 = 0.16 based on the large effect size found in a meta-analysis on the Stroop effect in youth with depression38, an alpha criterion of 0.05 and a power of 0.80. The results indicated a minimum sample size of N = 74.

Participants

We recruited participants through the city’s population register, youth groups and the outpatient treatment centre of the Child and Adolescent Psychiatric Clinic. Inclusion criteria were German language proficiency (native language or fluent), aged 14–21 and having a current depressive or anxiety disorder (for the CLIN group) or no current or past disorder (for the HC group). Exclusion criteria included the presence of any mental disorder other than depressive or anxiety disorder, a prior psychotic episode, a history of drug or alcohol abuse, acute suicidal ideation, and psychoactive medication intake. The study included a total of 91 participants (22% male). The clinical group (CLIN) with internalising disorder comprised 31 participants (34.07%), half of which met the criteria for a depressive disorder (N = 16) and in part a comorbid anxiety disorder, while the other half met the criteria for only an anxiety disorder (N = 15). The decision to include a comorbid group was made given the high comorbidity between the two disorder groups (39; see supplementary materials for flow chart of study inclusion, group allocation, and proportion of disorder types and comorbidities). The healthy control (HC) group comprised 60 participants who did not have any current or lifetime mental disorder. Group allocation was based on DSM-5 research diagnoses established using the Diagnostic Interview for Mental Disorders (DIPS), either in the short version for older adolescents or in the child version for those in middle adolescence. The Kinder-DIPS is a standardized, open-access interview of mental disorders for German speaking youth40 of high interrater reliability41 which can be carried out either with the youth alone or including their parents. In this study, the youth were interviewed alone. To analyse age effects, we split the sample into two groups. Approximately half of the participants (N = 41, age M = 16.00, SD = 1.46) were in adolescence, aged between 14 and 18 and went to secondary school, while the other half (N = 50, age M = 20.08, SD = 0.8) was in early adulthood, aged between 19 and 21 and attended University.

Measures

Emotional Stroop task

We programmed the EST in E-Prime (Version 2.0). Stimuli were happy and fearful faces from the Karolinska Directed Emotional Faces database42. The task contained 96 trials separated into two equal blocks with 65% congruent and 35% incongruent trials each. Trial order was randomised. Each trial began with the presentation of a fixation cross for 400 ms followed by the picture-word stimulus for 2000 ms. Missing responses and responses after 2000 ms counted as errors. Affective control was operationalised as the congruency effect and the error rate. The congruency effect was calculated by subtracting the mean reaction time (RT) on all incongruent trials with correct responses from the mean RT of all congruent trials with correct responses, with higher scores indicating poorer affective control. Error rate was calculated as the proportion of trials with incorrect responses relative to the total trial number (Fig. 1).

Fig. 1
figure 1

Length and order of stimuli presentation during the EST.

Full size image

Stress induction

Psychosocial stress was induced using the Trier Social Stress Test (TSST) in a digital version modified for youth43,44,45. The TSST included two tasks followed by an ER phase (see Procedures). First, participants were given the beginning of a story and were asked to continue telling the story for five minutes (speech task). Then, participants were asked to perform mental arithmetic while being evaluated by a jury for another five minutes (math task). The difficulty of the math task was adjusted to the participant’s age. The TSST was carried out digitally via the video conferencing platform Zoom (Version 5.15.3) with the participant being in one laboratory room and the experimenter in another. Participants were informed that two additional jury members (one male, one female) would join the video conference. In fact, two pre-recorded videos of the jury members were inserted as virtual backgrounds in two separate Zoom accounts that were set to mute, as it is usually done for people in a video conference not actively speaking. After the TSST, participants underwent a deception check to ensure they perceived the digital TSST as natural and did not suspect deception regarding the videos of the two not actually present jury members. Participants were asked to indicate on a 5-point Likert scale (1 = “Not at all” to 5 = “Very much”) how interested and concentrated they thought the two jury members were and if they had had doubts about them participating in the video conference in real-time. None of the participants included in this analysis reported having doubted the presence of the two jurors. In addition, a manipulation check was performed to ensure successful psychosocial stress induction (see supplementary materials).

ER strategies, success and strategy repertoire

After the ER phase following the TSST, participants were asked which ER strategies they had used during the ER phase and could either give a free response or select a predefined strategy from a list. The predefined ER strategies were cognitive reappraisal, acceptance, distraction, social support, problem solving, mindfulness and “Other” (view supplementary materials for description of predefined strategies). Participants could chose as many strategies as they had used from a list with these descriptions. Free responses were later assigned to one of the predefined strategies or “Other”, if they did not match any of the other strategies for data analysis. In addition, participants were asked to indicate their perceived ER success, that is, how successful they believed they were at regulating their stress by rating their current stress level relative to the stress level at the end of the stress test on a 10-point Likert scale (1 = no difference to stress level since end of TSST, 10 = achieved complete relaxation). ER strategy repertoire was dummy coded to measure whether participants used one (0) or multiple (1) strategies.

Psychometrics

Before the TSST, participants filled in a series of questionnaires. Habitual ER ability was quantified using the Difficulties in Emotion Regulation Scale (DERS), a 36-item self-report questionnaire assessing the subdimensions Lack of emotional awareness, Lack of emotional clarity, Non-acceptance of emotions, Limited access to regulation strategies, Impulse control difficulties and Difficulties engaging in goal-directed behaviour when distressed (46; German Version47]). Depressive symptom severity was quantified using the Beck Depression Inventory II (BDI-II;48), a questionnaire with 21 items assessing the occurrence of different depressive symptoms within the last two weeks, with higher scores indicating greater levels of depression. The BDI-II can be used for youth aged 13 and above and was found to be of high validity49. Anxiety was quantified using the State Trait Anxiety Inventory (STAI), a questionnaire with 20 items assessing state anxiety and 20 items assessing trait anxiety, with higher scores indicating greater anxiety that has also been validated in an adolescent sample50. However, there has been criticism pointing out that the STAI trait scale is particularly elevated in depressed individuals and should therefore be seen as a nonspecific measure of negative affectivity51.

Procedures

The present study comprised two sessions. Prior to participation in the lab study, participants were screened for eligibility over the phone. In the first session, participants and their guardians gave their informed consent and participants took part in the diagnostic interview. In the second session, participants filled in the questionnaires before undergoing the TSST. During the subsequent ER phase, participants were instructed to do whatever helped them reduce their stress and relax for the next ten minutes in the laboratory. Afterwards, data on ER strategies and ER success was collected. Finally, participants completed the EST before being debriefed.

Statistical analyses

Prior to data analysis, statistical assumptions were validated. For the EST, we calculated mean RTs across correct trials, the congruency effect (i.e., mean RT differences between incongruent and congruent trials) and error rates (i.e., the proportion of incorrect trials relative to all 96 trials). To check for the presence of the congruency effect and potential effects of facial stimulus type (= manipulation check for the EST), two 2 × 2 mixed model ANOVAs were calculated using the within-subject factors congruency (congruent vs. incongruent) and facial stimulus (happy vs. fearful) and the between-subject factor group (CLIN vs. HC). The first ANOVA used the RTs as a dependent variable; the second ANOVA used the error rate. To test the first hypothesis, a MANOVA was computed to compare affective control (quantified as the magnitude of the mean RT, congruency effect and error rates in the EST) between the CLIN and HC group. The factor age (adolescents/young adults) was added as a fixed factor to explore potential age effects. A chi-square goodness of fit test showed that participant sex was unevenly distributed between the two groups (χ2[1[= 4.15, p = .042, φ = 0.42), thus sex was added as a covariate. Since detecting a group difference would have required a bigger sample size to be adequately powered (which was not possible due to the complexity of recruitment and financial restraints of the study), we decided to include regression analyses of dimensional measures of internalising symptoms as secondary analyses. Then, we computed two multiple linear regression analyses using the entry method with the two measures of affective control (congruency effect, error rates in the EST) as criteria to test our second and third hypothesis. Added predictors were the ER strategies Distraction and Cognitive reappraisal, ER success and ER strategy repertoire and for the second hypothesis. These strategies were chosen over the other strategies based on previous studies that tested them in the EST52. For the third hypothesis, the predictors were the ER difficulties Trait ER difficulties, Impulse control difficulties, Limited access to ER strategies and Lack of emotional clarity, as their conceptual framework seemed to overlap with the one of affective control.

Results

Affective control between internalising disorders and healthy participants

A one-way MANOVA was carried out to test if participants with internalising mental disorders would exhibit lower affective control compared to healthy youth. We detected no statistically significant differences between the CLIN and the HC group on the mean RT (see Fig. 2), congruency effect and error rates, F(3, 85) = 0.634, p = .60, Wilk’s Λ = 0.978. However, a significant difference was detected between adolescents and young adults on the performance on the EST, F(3, 85) = 7.14, p < .001, Wilk’s Λ = 0.799. There was no significant interaction effect between the clinical groups and the age groups F(3, 85) = 0.06, p = .98, Wilk’s Λ = 0.998, and no significant effect of sex, F(3, 85) = 0.063, p = .94, Wilk’s Λ = 0.998.

Fig. 2
figure 2

Half violin plots of the congruence effect in the EST of the CLIN and HC groups.

Full size image

Dimensional approaches to psychopathology and their effect on affective control

Trait and state anxiety on affective control

When looking at anxiety as a dimensional variable as scores of state and trait anxiety on the STAI, neither trait anxiety (B = − 0.21, SE = 0.41, ß = − 0.06, p = .614) nor state anxiety (B = − 0.17, SE = 0.51, ß = − 0.04, p = .747) could predict the congruency effect R2 = 871.12, F(2,89) = 0.22, p = .800. Similarly, neither trait anxiety (B = 0.23, SE = 0.2, ß = 0.05, p = .251) nor state anxiety (B = − 0.2, SE = 0.25, ß = − 0.09, p = .414) could predict the error rate on the EST R2 = 897.15, F(3,88) = 0.65, p = .587.

Depressive symptoms on affective control

The depressive symptoms along the BDI-II (B = − 0.1, SE = 0.47, ß = − 0.02, p = .829) did not predict the congruency effect R2 = 90.08, F(1,91) = 0.05, p = .829. The BDI-II score (B = 0.21, SE = 0.09, ß = 0.23, p = .029) did however predict the error rate R2 = 0.21, F(1,90) = 4.71, p = .029 as demonstrated in this secondary analysis.

Effects of ER strategies and difficulties on affective control

To analyse the effects of ER strategies (hypothesis 2) and ER difficulties (hypothesis 3) on the congruency effect and error rate in the EST as two measures of affective control, we computed two multiple linear regression models using the enter method. Both models testing the effect of the ER strategies failed to reach significance (congruency effect: F[3, 87] = 1.18, p = .32, R2 = 0.039, adjusted R2 = 0.01; error rate: F[3, 87] = 0.02, p = 1.00, R2 = 0.001, adjusted R2 =  − 0.03). Similarly, both models testing the effect of ER difficulties failed to reach significance (congruency effect: F[3, 87] = 1.90, p = .16, R2 = 0.06, adjusted R2 = 0.03; error rate: F[4, 86] = 0.16, p = .96, R2 = 0.07, adjusted R2 = − 0.04).

Developmental trajectory of affective control

To explore if age would moderate the relation between ER and affective control, we had planned to compute a moderator analysis. We tested the assumptions of moderator analysis (i.e., if age would moderate the relationship between affective control and ER.) by computing the correlation between trait ER and the two affective control measures. As there were no significant correlations between trait ER and the congruency effect (r[91] = 0.13, p = .2) or the error rate (r[91] = 0.14, p = .18), we instead compared adolescents and young adults on the two affective control measures within the MANOVA presented before (see section "Affective control between internalising disorders and healthy participants"). To check for a strong disordinal interaction, the moderation analysis itself was still carried out but revealed non-significant results (see supplementary materials).

Exploratory analyses on links between ER measures and depressive symptoms

As we had found the CLIN group to show not surprisingly significantly higher levels of depression symptoms than the HC group (see Table 1), we exploratorily investigated which factors related to ER and affective control would predict depressive symptoms in our sample. We therefore computed a multiple linear regression analysis using the entry method. The overall model was significant, F(7,83) = 5.70, p < .001, R2 = 0.325, adjusted R2 = 0.26. The regression showed that the EST error rate, the ER strategy Acceptance, ER strategy repertoire and ER success were all predictive of depressive symptoms (see Table 2), while the congruency effect and the ER strategies Distraction and Cognitive reappraisal were not significant predictors. Higher error rates and higher ER strategy repertoire, respectively, were associated with higher levels of depressive symptoms (see Fig. 3a,c). On the contrary, more frequent use of the ER strategy Acceptance and higher perceived ER success, respectively, were associated with lower levels of depressive symptoms (see Fig. 3b).

Table 1 Descriptive statistics of participant age and psychometric data (mean ± standard deviation).
Full size table
Table 2 Linear regression analysis of affective control and ER measures on depressive symptoms.
Full size table
Fig. 3
figure 3

(a) Relationship between scores on the BDI-II and the error rate in percentage as a scatterplot. (b) Relationship between BDI-II scores and the ER success along a numerical analogue scale. (c) Relationship between the BDI-II score and the ER flexibility, with 0 = only one strategy was used and 1 = more than one strategy were used.

Full size image

Discussion

The present study examined the relationship between different components of ER and affective control, operationalized as the congruency effect and the number of errors on the EST, and investigated, if this relationship differed in youth with internalising disorders and healthy counterparts.

First, we had hypothesized that participants with internalising mental disorders would exhibit poorer affective control, manifesting as longer RTs and a larger congruency effect, compared to healthy participants. Contrary to our expectation, we found that the two groups did not differ on the two affective control measures, which diverges from earlier findings suggesting affective control deficits in individuals with internalising mental disorders31,38. Although we did not find the expected group differences, which might be explained by inadequate power given the sample size, the EST did capture facets of psychopathology in our sample. Our exploratory findings showed that error rates in the EST positively predicted the level of depressive symptoms across groups, but not symptoms of anxiety. On one hand, the findings regarding depressive symptoms are in line with previous studies findings more errors and slower RTs in depressed subjects53. On the other hand, our findings regarding anxiety symptoms contradict those of previous studies suggesting that higher levels of trait anxiety were related to slower RTs and more errors on a facial Stroop task31. In light of these discrepancies, several potential factors are to be discussed. It is possible that psychopathology-related effects are obscured by methodological differences between our study and previous studies, such as the choice of visual stimuli. Meta-analytic findings revealed greater congruency effects in adults with depression compared to healthy controls when presented with mood-congruent negative stimuli38. Although the meta-analysis did not differentiate further within the negative-valence category (i.e., angry versus sad faces), it has been hypothesized that only specific stimuli elicit attentional biases specific to psychopathology, in that depressed individuals exhibit an attentional bias when they are viewing sad faces54. However, the series of experiments conducted by54 revealed no significant attentional biases regardless of the valence of the stimulus in adults with depression despite adequate statistical power, which is consistent with our findings. It is thus unlikely that the choice of emotional stimulus alone camouflaged potential psychopathology-related effects. Instead, the age of our clinical group might play a role, as the vast majority of prior studies that explored psychopathology-related effects on affective control recruited adult samples (see meta-analysis by31). Since our study is one of the first to experimentally explore affective control and ER in youth with only a few studies having looked at this association before55,56,57,58, with a previous review suggesting that ER success in adolescence relies on the maturation of affective control4, it is unclear to which extent other findings from adult samples can be generalized to younger individuals still in development. One explanation for a difference between adolescents and adults might be that this attentional bias to mood-congruent words for depressed adult subjects correlates with fMRI data of increased activity in an area of the prefrontal cortex, the rostral anterior cingulate cortex28, which has been found to develop at a different pace in depressed compared to healthy adolescents59. Due to the lack of longitudinal data, it remains unknown whether the effects of psychopathology on affective processing are immediate or whether they manifest gradually over time, which would explain why we did not observe them in our young sample.

Second, we had investigated whether ER strategies and subjective ER success would predict affective control following a psychosocial stress induction. We had hypothesized that more cognitive reappraisal, distraction and a higher subjective ER success in implementing those strategies would positively predict affective control across groups. Contrary to our hypothesis, we found that the choice of ER strategy and ER success did not predict affective control. This contradicts Pruessner and colleagues19 notion that ER is closely linked to cognitive control, although one should note that their perspective focuses on the relevance of cognitive control for ER flexibility in particular. However, there is initial evidence that cognitive control should not only influence ER flexibility but also ER strategies. For instance, individuals with a higher working memory capacity were found to use cognitive reappraisal more often than those with lower working memory capacity60. However, the relationship between cognitive control and different ER facets such as strategy repertoire or subjective ER success remains unknown, highlighting the need for further research in this area.

Third, we had analysed what difficulties in ER would predict affective control and had hypothesized that higher levels of trait ER difficulties, impulse control difficulties, higher lack of emotional awareness and poorer access to ER strategies would inversely predict affective control. In our study, ER difficulties did not predict affective control, which diverges from our hypothesis and the established literature. For instance, O’Bryan and colleagues61 found links between affective control and better emotional clarity and a higher ability to engage in goal-directed behaviour during negative affect. At the same time, they failed to find links with impulse control difficulties, emotional acceptance and access to ER strategies, suggesting that only specific aspects of ER difficulties might show associations with affective control. The authors noted that better attentional control might allow for more emotional clarity by reducing threat-related attentional biases and increasing one’s engagement in goal-directed behaviour when upset, thus preventing getting absorbed by negative emotions. As the relationship between ER difficulties and affective control remains poorly understood, further research should investigate potential links in more detail.

Forth, we had assessed the developmental trajectory of affective control and ER and had hypothesised that age would moderate the relationship between the two variables. In our study, participant age did not appear to moderate the relationship between affective control and ER. This might suggest that the two constructs do not develop linearly throughout adolescence but at different rates. However, age did have an effect on performance on the EST, with more errors in adolescents than in young adults. This finding is in line with previous studies describing higher error rates in adolescents compared to young adults on affective control tasks34,62, suggesting that neural circuits involved in the processing of emotional stimuli undergo significant changes throughout adolescence (for an overview, see63). This could underline the vulnerability of adolescence as a developmental stage given it is often the onset time of mental disorders that carry on into adulthood with a peak age of onset at 14.5 years64.

Exploratory analyses revealed that the error rate, the ER strategy Acceptance, ER success and ER strategy repertoire significantly predicted depression symptoms. Higher error rates and higher ER strategy repertoire, respectively, were associated with higher levels of depressive symptoms. Higher error rates in a spectrum of cognitive tasks have previously been linked to higher levels of depression in adults65,66, possibly reflecting psychopathology-related alterations in brain function that have their onset during adolescence (for a review, see67). Regarding ER strategy repertoire, our finding is surprising, as higher strategy repertoire and therefore higher ER flexibility has previously been associated with lower but not higher levels of depressive symptoms68,69. Those studies differed from our investigation in that they used a self-report questionnaire to determine trait ER flexibility. In contrast, our study assessed only the strategy repertoire component of ER flexibility by asking participants to report which strategies they had actually applied after the psychosocial stress induction. As a result, both measures capture different facets of ER flexibility. Another interpretation might be that subjects that had higher depressive scores were not as effective in either implementing the strategy that they had initially picked or that that strategy was not sufficient for their self-regulation hence they switched, while those without depressive symptoms were successful with their single strategy and did not require switching. While trait ER refers more to the breadth of an individual’s ER strategy repertoire, state ER refers to the ability to evaluate contextual demands and the ability to choose and modify ER strategies accordingly18. As a result, ER flexibility can promote stress resilience and indirectly protect against mental health problems70, but only when strategy switching is actually effective for ER. Increased strategy switching in adolescent girls has been associated with them experiencing the interpersonal stressor to be more uncontrollable, the intensity of the elicited negative emotion and having the possibility to co-regulate with a peer71. The same authors discussed whether this increased likelihood to switch strategies in older adolescents compared to younger adolescents might reflect age-related increases in underlying processes like affective control.

Regarding ER strategy use and ER success, we found that more frequent use of the ER strategy Acceptance and higher perceived ER success, respectively, were associated with lower levels of depressive symptoms. Previous studies in depressed youth have shown that they relied less on adaptive ER strategies such as Acceptance and more on maladaptive strategies such as Rumination72. Although this might suggest that specific ER strategies act as protective or risk factors and are thus prognostic for depression in youth, longitudinal studies are needed to determine the causal relationship between ER strategies and psychopathology, as depression onset could trigger a shift in affective processing and thus influence facets of ER such as strategy use and repertoire. Regarding ER success, we found individuals with higher levels of depressive symptoms to be less likely to perceive their ER as successful, suggesting lower ER self-efficacy. This observation corresponds with findings from a longitudinal study in young adults that linked lower ER self-efficacy to higher levels of depression six months later72. These findings specific to ER mirror longitudinal data that linked lower self-efficacy, conceptualized on a broader level, to higher levels of depression in youth, young adults as well as across the lifespan73. Although this might hint at a particular significance of ER self-efficacy for depression risk, more research on this matter is needed. To sum up, the role of different ER strategies, ER strategy repertoire and ER success in psychopathology should be further investigated, since deeper insights into this matter might help to identify potential targets of psychotherapeutic interventions.

While our study provides novel insights into affective control and ER in youth with internalising disorders, our findings are subject to methodological limitations that should be addressed in future research. First, we did not systematically manipulate which ER strategies participants applied but rather assessed spontaneous strategy use. While this approach offers higher ecological validity and simultaneously allows for measuring the strategy repertoire component as a facet of ER flexibility, it does not control for interindividual variability in other factors that could influence spontaneous strategy use. For instance, research suggests that subjective stressor intensity plays a part in the selection of specific ER strategies74,75. Although all participants in our study underwent a standardizes psychosocial stress induction, it is possible that they perceived the paradigm differently and hence chose different ER strategies. Thus, future research is needed to systematically explore which factors determine ER strategy choice and which differential effects specific ER strategies have. Second, we measured ER difficulties and ER success using self-reports. Until this day, it is unclear to what extent self-reports reflect actual biological processes associated with ER, especially in youth, as previous studies yielded mixed results on this matter76,77. Fruitful insights could thus come from incorporating psychophysiological or neural measures of ER to better understand the psychobiological links between ER and psychopathology.

Taken together, our findings emphasize the importance of different implicit facets of ER for mental health during adolescence and young adulthood. Gaining a better understanding of the cognitive and neural processes involved in affective processing is crucial for developing targeted prevention programmes and psychotherapeutical treatments for youth at high risk for or affected by depression and anxiety.