Introduction

Depressive disorders are a leading cause of morbidity and disability worldwide, being associated with decreased well-being, life satisfaction, and life expectancy (Kessler, 2012; Vos et al., 2016; World Health Organization, 2017). Treatments for depression are still lacking in effectiveness, as affected individuals often experience another depressive episode months or years later (Eaton et al., 2008; Ormel et al., 2022), and relapse occurs in approximately 50% of cases even after acute, maintenance, or preventive treatments (Bockting et al., 2018; Cuijpers, 2017). The majority of individuals with depressive disorders do not receive adequate treatment (Thornicroft et al., 2017), and the lack of early identification or prevention programs for subthreshold depressive symptoms poses serious challenges (Klein et al., 2013; Ormel et al., 2022). Given the wide-reaching implications due to the high prevalence and relapse rates, it is of great public health relevance to identify potential risk or protective factors for the maintenance or development of depressive symptoms (Cuijpers, 2017; Kuyken et al., 2016; WHO, 2017). The overarching aim of this article is, therefore, to examine the prospective relationship between depressive symptoms and emotion regulation over time.

Emotion regulation and depressive symptoms

Research indicates that impairments in emotion regulation could play a central role in the etiology of depressive symptoms (Garnefski & Kraaij, 2006; Joormann & Gotlib, 2010; Joormann & Stanton, 2016), and might persist even in remitted patients (Ehret et al., 2015). Emotion regulation refers to the modulation of the intensity and duration of emotional states and can be implemented using various strategies (Gross, 2015). Individuals with depressive symptoms tend to respond to their emotional states in a way that increases negative emotions or fails to successfully maintain or increase positive emotions (Gross & John, 2003; Joormann & Stanton, 2016; Liu & Thompson, 2017). This may be due to an overreliance on strategies such as rumination (Aldao et al., 2010; Hsu et al., 2015; Joormann & Stanton, 2016; Kraiss et al., 2020) and suppression (Ehring et al., 2010; Liverant et al., 2008; Visted et al., 2018; Webb et al., 2012). The repetitive and persistent nature of negative thinking sustained by these strategies could contribute to a detrimental cycle that hinders effective emotion regulation (Aldao & Nolen-Hoeksema, 2010), prolongs negative moods (Koster et al., 2011), and potentially exacerbates depressive symptoms over time. In addition, individuals with depression (or at risk for it) often display lower engagement in strategies that can effectively buffer against negative emotions, such as reappraisal (Ford et al., 2017; Kraiss et al., 2020; Liu & Thompson, 2017; Riepenhausen et al., 2022) and acceptance (Aldao et al., 2010; Kotsou et al., 2018). These strategies can reduce negative affect and arousal by allowing individuals to find alternative novel perspectives (Garnefski & Kraaij, 2006), or by endorsing a more mindful and accepting attitude towards emotions or situations without wanting to modify them (Gross & John, 2003). Indeed, the cognitive reinterpretation and acceptance of emotional experiences play an integral role in cognitive-behavioral and mindfulness-based interventions for depression (Goyal et al., 2014; Hayes & Feldman, 2004; Sloan et al., 2017).

Seasonal fluctuations in depressive symptoms and emotion regulation

Overall, many studies support an association between emotion regulation and depressive symptoms (see Aldao et al., 2010; Liu & Thompson, 2017; Visted et al., 2018). However, relatively little is known about the temporal associations between emotion regulation strategies and depressive symptoms in the long term (Berking et al., 2014; Crowell et al., 2015). Changes in depressive symptoms could be accompanied – or perhaps even predated – by changes in emotion regulation (de France et al., 2019; Masters et al., 2019). Past research suggests that depressive symptoms as such are not stable but can fluctuate over the seasons (see review by Øverland et al., 2020). Some studies indicate an increase in depressive symptoms in early (Harmatz et al., 2000) or late winter (Cobb et al., 2014; Huibers et al., 2010). However, other studies reported no (or only negligible; Kerr et al., 2013) associations between depressive symptoms and latitude or sunlight exposure (Hansen et al., 2008; Johnsen et al., 2012; Struijs et al., 2020; Traffanstedt et al., 2016). Based on these heterogeneous findings, the first objective of our study is to examine whether depressive symptoms increase in the winter months.

Whereas fluctuations in depressive symptoms have been frequently investigated, studies systematically testing whether daily emotion regulation strategy use varies across the seasons remain scarce (de France et al., 2019). Therefore, we also explore if emotion regulation strategy use fluctuates alongside depressive symptoms across the seasons. Seasonal changes represent a fundamental variation in everyday context (Øverland et al., 2020) in temperate regions at about 50° northern latitude (Wirz-Justice, 2018), where the current study was conducted. Previous research suggests that the use of emotion regulation strategies is context-sensitive (Blanke et al., 2022) and depends on situational demands and opportunities (Bonanno & Burton, 2013), such as the social context (Paul et al., 2023), different activities (McRae et al., 2011), nature exposure (Hedblom et al., 2019), and outdoor thermal conditions (Zhang et al., 2021). Thus, changes in environmental and psychosocial affordances and social factors (e.g., availability of social support) over the seasons could be linked to shifts in emotion regulation. As a consequence, the use of emotion regulation strategies may change over the seasons. Especially the dynamic, prospective associations between emotion regulation and depressive symptoms have not yet been systematically studied. Therefore, as a second objective, we investigate whether variations in emotion regulation strategy use are associated with depressive symptoms across the seasons.

Initial empirical evidence suggests that emotion regulation might serve as a potential precursor of depressive symptoms (Arnarson et al., 2016; Berking et al., 2014). For example, longitudinal studies found that rumination could predict the development of depressive symptoms or episodes (Arnarson et al., 2016), and that daily repetitive negative thinking measured with ecological momentary assessment (EMA) predicted depressive symptoms two weeks later (Rosenkranz et al., 2020). However, less is known about other emotion regulation strategies. Therefore, our third objective was to investigate how daily emotion regulation relates to depressive symptoms over time using EMA. This naturalistic approach allows us to disentangle between- and within-person factors of emotion regulation in everyday life (Sanchez-Lopez, 2021; Southward et al., 2019; Visted et al., 2018) and how they relate to depressive symptoms over time. Moreover, this design enables us to examine if emotion regulation in summer could be a precursor of depressive symptoms in autumn, early or late winter, which could present a promising avenue for future research and clinical interventions.

The present study

To address these three objectives, we employed a longitudinal approach and investigated participants across four waves in summer, autumn, early winter, and late winter. At each wave, we assessed depressive symptoms with self-report questionnaires and gathered real-time data on the participants’ use of various emotion regulation strategies (e.g., rumination, suppression, reappraisal, and acceptance) over seven days with EMA. As depressive disorders are defined by negative emotions and an impaired ability to recover from them (Joormann, 2010; Rottenberg, 2007; Vanderlind et al., 2020), our study focuses on the regulation of negative emotional episodes exclusively.

First, we expect depressive symptoms to increase from summer to winter, aligning with the most commonly reported seasonal patterns observed in the literature (e.g., Cobb et al., 2014; Øverland et al., 2020). Furthermore, the temporal variation of emotion regulation across the seasons has not yet been systematically explored, and we aim to investigate if – and how – the use of different emotion regulation strategies fluctuates alongside depressive symptoms. Second, as emotion regulation is closely linked to depressive symptoms (Joormann & Stanton, 2016), we hypothesize that variation in emotion regulation strategy use will be associated with depressive symptoms across the seasons on a between-person and within-person level. Third, we expect that emotion regulation strategy use in summer can predict depressive symptoms in later seasons, acting as a potential risk or protective factor.

Methodology

Transparency and openness

The anonymized dataset used in this investigation, and further details about the study protocol and psychometric measures can be found on the Open Science Framework (OSF; at www.osf.io/ycv2n). A freely available implementation of the EMA application used in this study can be accessed at https://osf.io/md9ka. All procedures performed in this study adhered to the ethical standards of the Declaration of Helsinki, and the study protocol was approved by the Ethics Commission of the Faculty of Behavioural and Cultural Sciences at Heidelberg University. This study was not preregistered.

Research design

The current investigation was part of a more extensive longitudinal study examining seasonal fluctuations in emotion regulation, cognitive control, and depressive symptoms across six wavesFootnote 1 (from August 2018 to August 2019) in a mixed student and community sample. The present investigation focused on assessment waves for which seasonal variations could be anticipated based on existing literature (Cobb et al., 2014; Øverland et al., 2020): summer (August), autumn (October), early winter (December), and late winter (February). All waves were eight weeks apart and spanned two weeks (see Fig. 1). In the first week of each wave, participants (N = 194) used a smartphone-based EMA application to indicate the emotions they experienced throughout the day and how they regulated them. Participants were asked to use the app at least three times daily for five out of seven days. In week two, participants completed an online questionnaire that assessed, among other constructs, emotion regulation, social relationships, well-being, and psychopathology (an overview of all measures and instruments can be accessed at www.osf.io/ycv2n). Participants were tiered into five cohorts that started every wave on a Monday using the EMA app.

Fig. 1
figure 1

Timeline and participant flow for all four assessment waves. Participants who met the inclusion criteria during Screening were informed about the study, consented to participate during Enrolment, and completed questionnaires assessing sociodemographic information and trait measures during Pre-assessment. EMA = Ecological Momentary Assessment. BDI-II = Beck Depression Inventory-II

Full size image

A required minimum sample size of N ≧110 was estimated a priori based on medium effect sizes for mixed within- and between-person analyses and 80% power while considering a drop-out rate of 20%. To determine the post hoc power for the multilevel analyses conducted in this study, we used the R package EMAtools (Kleiman, 2021). The power curve calculations for our final models were performed with the following criteria: participants at baseline = 194, max. number of experience sampling days (across all four waves) = 28, max. number of responses per day = 5. A minimum of approximately 10 EMA entries in total per person was required to detect medium-sized effects (d = 0.50), and around 20 for small-sized effects (d = 0.20). The original dataset (NEMA = 20,011) was filtered to only include situations involving the regulation of negative emotions from participants with more than 10 EMA entries per wave, which reduced the total number of entries across all four waves to N = 9,191 (Nsummer = 3,269; Nautumn = 2,114; Nearly winter = 1,693; Nlate winter = 2,115). Nevertheless, the calculated power curves (see Supplemental Material S2) confirmed that the number of responses included in the present analyses (mean number of filtered entries per participant = 21.2) was sufficient to detect medium-sized effects.

Participants

Figure 1 shows a detailed flowchart and timeline for all waves included in this investigation, also depicting participant attrition rates from summer (August) to late winter (February). The sample’s sociodemographic characteristics at baseline are presented in Table 1. A total of 194 German-speaking adults from a mixed student and community sample participated in this study (Mage = 27.26 ± 6.10; age range 18–45; 74.5% female). Apart from education level, no systematic differences in age, gender, nationality, psychotherapy status, psychotropic medication or baseline depressive symptoms between study completers and drop-outs (all ps > 0.05) were observed. For these analyses, participants who filled in the BDI-II at each wave were considered to have completed participation (N = 141), irrespective of their EMA completion after Wave 1. An overview of these comparisons can be found in the Supplemental Material S3.

Table 1 Sociodemographic characteristics of participants at baseline
Full size table

Procedure

Participants were recruited from the German-speaking community in Germany and Austria via electronic and print advertisements, using a URL or QR-code to SoSciSurvey (www.soscisurvey.de) to access more information and the initial screening. Inclusion criteria were a) age 18 to 45, b) fluency in German (level C1 and above), c) access to a computer and smartphone (Android or iOS) with internet access, and d) commitment to participate in all assessment waves. Participants were excluded if they had severe neurologic disorders, dyslexia, dyscalculia, or color blindness because the more extensive study also investigated cognitive function. If participants met the inclusion criteria and provided informed consent, they filled in a baseline questionnaire of sociodemographic characteristics and trait measures (Enrolment/Pre-assessment; see Fig. 1). Then, participants were instructed to download and use the EMA application EmoTrack2 (Pruessner et al., 2023) starting the following Monday at least three times a day for a minimum of five out of seven days (in line with similar EMA study protocols by Haines et al., 2016; Southward et al., 2019). All participants received five pseudo-random push notifications during their waking hours, which they specified at the beginning of the study, to report their emotions and the emotion regulation strategies they employed. Participants could also use the app unprompted anytime during the EMA week.

Afterward, participants were invited to complete an online questionnaire battery within seven days. This part took approximately 60–90 min, and participants could pause and continue the questionnaire at their leisure. The research protocol was identical at all waves. Participants were reimbursed 30€ to 40€ for completing both study parts after each wave, with the variable sum depending on the frequency of EMA app usage (from > 15 entries required per wave to a maximum of 35). In cases where participants only filled in the online questionnaire but did not complete the EMA, they received a reimbursement of 15€. To reduce attrition, participants were reminded via email or telephone to participate in the ongoing assessment. Individuals who could not participate on time within seven days due to personal circumstances (e.g., lack of time, being on vacation) were allowed to complete their assessment before the next wave started.

Measures

Depressive symptoms

The Beck Depression Inventory-II (BDI-II; German version: Hautzinger et al., 2006) was administered at each wave to measure the severity of depressive symptoms. The questionnaire consists of 21 items covering the main symptoms of depression during the last two weeks, such as “I feel my future is hopeless and will only get worse.” Each item is scored from 0 to 3, with higher scores demonstrating greater severity. Sum scores range between 0 and 63, where higher scores indicate higher depressive symptomatology. Scores between 0 and 13 suggest no or almost no depressive symptoms, scores from 14 to 19 suggest mild symptoms, scores from 20 to 28 suggest moderate, and scores above 29 suggest severe depressive symptoms. The internal consistency of the BDI-II was excellent in this sample, with Cronbach’s α ranging between 0.93 and 0.94 over all four waves.

Emotion regulation

Emotion regulation strategy use in everyday life was assessed at each wave using the smartphone-based EMA application EmoTrack2, which was indicated to be a reliable and valid measure of emotion regulation in everyday life (Pruessner et al., 2023). The selection of emotion regulation strategies measured with this application was based on the individual stages of the emotion regulation process described by Gross’ extended process model (2015) and the existing literature about the most frequently investigated strategies (e.g., Grommisch et al., 2020). The items referred to the emotion regulation strategies used by the participant after the strongest emotion that occurred since the last assessment and were phrased as follows: “I thought over and over again about my feelings or the situation” (rumination), “I have not shown my emotions” (expressive suppression), “I suppressed my feelings” (experiential suppression), “I changed my feelings by thinking about the situation in a different way” (reappraisal; also see Izadpanah et al., 2019; Southward et al., 2019), “I allowed and accepted my feelings as they were” (emotional acceptance), and “I accepted the situation as it is” (situational acceptance). The extent (i.e., the intensity) of emotion regulation strategy use was rated on an 11-point Likert scale ranging from 0 (not at all) to 10 (very much). A list of the EMA items relevant to the present investigation is shown in the Supplemental Material S1, and a more detailed overview of all EMA items and their wording can be found on the OSF (https://osf.io/md9ka; Pruessner et al., 2023).

Experience sampling studies frequently use a compliance rate of 50% as a threshold for data inclusion (Grommisch et al., 2020; van Berkel et al., 2017). Thus, only EMA data from participants who had completed at least ten entries (per wave) were included in this investigation. The number of (applicable) individual EMA entries at each wave was as follows: summer = 7,485; autumn = 4,833; early winter = 4,108; and late winter = 3,585 (a total of 20,011 entries). These data were then filtered to include only entries from emotion-eliciting events with a negative valence (< 5 on the item “How did you feel in the situation?” with a scale ranging from 0 = unwell/discontent to 10 = well/content), yielding a total of 9,191 entries. A total of 1,153 entries consisting of duplicates, invalid data, and data from participants who did not complete the BDI-II at Wave 1 were excluded. Therefore, the filtered data used for the present analyses were: Nsummer = 2,826; Nautumn = 1,820; Nearly winter = 1,620; and Nlate winter = 1,772 (8,038 entries total). These filtered entries were nested into the following number of participants at each wave: Nsummer = 194; Nautumn = 139; Nearly winter = 136, and Nlate winter = 132.

Data analysis

Data were analyzed using R (version 4.3.2). EMA data was first processed and cleaned using the packages dplyr and tidyr to remove duplicates and filter data according to wave. We screened for potential outliers in the data, and values with z-scores that surpassed ± 3.29 were winsorized (Leys et al., 2019) to the nearest outlier (in total < 20 values across all four waves). The significance tests were two-tailed, and the alpha level was 5% for all reported analyses.

To address the first objective of investigating seasonal fluctuations in depressive symptoms and emotion regulation, we performed two-level mixed-effects linear models with individual assessments (Level 1) nested within participants (Level 2), testing for temporal change in all study variables from summer to late winter with the lmer function of the lme4 package. Seven models were calculated: one for every emotion regulation strategy and one for depressive symptoms, with time (i.e., assessment wave) predicting each variable (depressive symptoms, six emotion regulation strategies) while accounting for participants’ random effects. In addition, we fitted different polynomial terms (linear, quadratic, cubic) to determine the exact nature of change using the stats function, and chose the most appropriate polynomial trajectory based on the model fit and the significance level of the respective b estimates. Degrees of freedom and significance levels for all two-level mixed-effects models were calculated with the Satterthwaite method.

To address the second objective, three-level mixed effects linear models were used to investigate the within-person and between-person associations between depressive symptoms and emotion regulation throughout the four waves while accounting for the nested data structure. In line with Enders and Tofighi’s (2007) recommendations, predictors were centered around the grand mean and the person mean. The Level 1 consists of individual EMA observations nested within Level 2, denoted by the four assessment waves during each of the seasons, which are nested within Level 3, the participant.

Null models were calculated for all variables to check the appropriateness of using a multilevel approach, and to determine intra-class correlation coefficients (ICC). Then, fixed parameters and random coefficients were added as appropriate (Snijders & Bosker, 2011). We calculated multiple models using both Restricted Maximal Likelihood (REML) and Maximum Likelihood (ML) approaches and different covariance structures (Diagonal Covariance and Compound Symmetry), but they did not yield any different results or notable changes in model fit than the final REML models. We used fixed slopes, as adding random slopes did not improve the model fit. All model comparisons were based on Akaike’s information criteria (AIC) and Bayesian information criteria (BIC). Model assumption checks were conducted using the performance package and were confirmed visually.

Finally, to address the third objective, hierarchical linear regression models were performed, testing whether emotion regulation strategy use in the summer could predict depressive symptoms at subsequent waves in autumn, early winter, and late winter. Depressive symptoms at Wave 1 were included in the first step of the regression to control for baseline depressive symptoms. In the second step, the emotion regulation strategy was entered into the respective regression model (one model for every emotion regulation strategy). The final models did not include any other control variables besides baseline depressive symptoms, as the inclusion of additional control variables (age, gender, nationality, education, emotion regulation goal) did not produce different results, nor did they significantly improve model fit. Moreover, to test the reverse direction, we also investigated whether depressive symptoms in the summer could predict the use of emotion regulation strategies in subsequent seasons. We controlled for baseline emotion regulation strategy use in the first step of the regression, followed by depressive symptoms in autumn, early winter, and late winter (i.e., three separate regressions for each strategy). We performed bias-corrected and accelerated (BCa) bootstrapping using 1,000 samples for confidence intervals (CIs) due to the skewed nature of the BDI-II data (examined with histograms and the Shapiro–Wilk test, W = 0.86, p < 0.001).

Results

Objective 1: Seasonal fluctuations in emotion regulation and depressive symptoms

The means, standard deviations, and changes in study variables across all four waves are presented in Table 2. We hypothesized that depressive symptoms would increase from summer to winter, and that the use of emotion regulation strategies would also fluctuate across the seasons. Mixed effects models showed that depressive symptoms significantly fluctuated across the seasons, b = -21.07, p = 0.001. The quadratic trajectory showed an increase in depressive symptoms from summer to early winter, followed by a decrease in late winter, where depressive symptom severity was lower than the baseline assessed in summer. Regarding emotion regulation strategy use, there were significant seasonal changes in rumination, b = 7.53, p = 0.009, following a cubic trajectory with an increase from summer to autumn, and again from early to late winter. Moreover, situational acceptance linearly decreased from summer to winter, b = -9.34, p = 0.001, while emotional acceptance showed a decrease with a cubic trajectory, b = -8.99, p = 0.001 (see Fig. 2). The mean use of suppression and reappraisal did not differ significantly across the waves, even when fitting different polynomial trajectories (ps > 0.082). Figure 2 illustrates the significant seasonal changes (temporal trajectories) for depressive symptoms, rumination, and acceptance from summer to winter.

Table 2 Objective 1: Changes in depressive symptoms and emotion regulation strategy use across the seasons
Full size table
Fig. 2
figure 2

Objective 1: Trajectories of depressive symptoms and emotion regulation strategies across the seasons. This analysis is based on data from N = 194 participants, encompassing a total of N = 8,038 Ecological Momentary Assessment (EMA) data points. It illustrates seasonal fluctuations in depressive symptoms (Panel A) alongside emotion regulation strategy use that exhibited changes from summer to winter: rumination (Panel B) and acceptance-based strategies (Panel C). The shaded error bands surrounding the trend lines represent 95% confidence intervals

Full size image

Objective 2: Temporal links between emotion regulation and depressive symptoms

Using three-level mixed effects models, we examined whether the extent of emotion regulation strategy use was associated with depressive symptoms on a within- and between-person level with season as a covariate.

As seen in Table 3, significant associations between emotion regulation and depressive symptoms were only observed on the between-person level. Specifically, there was a positive association between depressive symptoms and rumination, experiential suppression as well as reappraisal. That is, participants with greater use of these emotion regulation strategies reported more depressive symptoms overall. In contrast, a negative association was observed for emotional acceptance, suggesting that participants who reported greater use of emotional acceptance overall experienced lower levels of depressive symptoms.

Table 3 Objective 2: Multilevel models of associations between emotion regulation and depressive symptoms across the seasons
Full size table

To better understand the positive link between reappraisal and depressive symptoms, emotion regulation goals (i.e., maintaining, down- or upregulating emotion) were included as a covariate in additional exploratory analyses. This was done as using reappraisal to down- or upregulate emotions might be associated with different effectiveness (Hartmann et al., 2023; Huang et al., 2020; Tamir et al., 2019). Our analysis suggested that a higher within-person use of reappraisal was associated with wanting to decrease negative emotions, b = -0.11, p = 0.001, which is consistent with hedonic perspectives on emotion regulation (Kalokerinos et al., 2015; Ortner et al., 2018). However, the initial finding of a between-person link between reappraisal and depressive symptoms was not moderated by the type of regulatory goals that participants reported.

In contrast to the findings on the between-person level, we did not observe notable associations between depressive symptoms and emotion regulation on the within-person level (ps > 0.253).

Objective 3: Emotion regulation as a precursor of depressive symptoms

We expected that the extent of emotion regulation strategy use would predict higher symptom severity in later seasons (autumn, early winter, and late winter). We performed six hierarchical regressions, controlling for baseline depressive symptoms in summer in the first step before including rumination, experiential suppression, expressive suppression, reappraisal, emotional acceptance, and situational acceptance in the second step. As shown in Table 4, only rumination and reappraisal predicted depressive symptoms, with higher scores in the summer being associated with higher depressive symptoms in autumn for reappraisal, b = 0.97, p = 0.004, and in early winter for rumination, b = 0.78, p = 0.036. These findings remained significant when controlling for age, gender, nationality, education, and emotion regulation goals  (rumination: bs = 0.78–0.81, ps < 0.05; reappraisal: bs = 0.88–0.97, ps < 0.01). Neither suppression nor acceptance could predict depressive symptoms longitudinally (all ps > 0.083). When examining the other direction (see Table 4), i.e., if depressive symptoms in the summer could predict emotion regulation strategy use in the autumn and winter, no associations emerged except that depressive symptoms in summer could predict reappraisal use in autumn, b = 0.03, p = 0.024. These findings suggest that rumination has a unidirectional predictive association with depressive symptoms, while reappraisal shows a bidirectional relationship.

Table 4 Objective 3: Emotion regulation strategy use in the summer predicting depressive symptoms in later seasons, and vice versa
Full size table

Discussion

The overarching goal of this study was to examine the dynamic interplay between the use of emotion regulation strategies and depressive symptoms, with a particular focus on the role of everyday emotion regulation as a potential precursor of depressive symptoms from summer to winter. Understanding this relationship is crucial, given the substantial impact of depressive disorders on individual well-being and public health (Kessler, 2012; Vos et al., 2016; WHO, 2017).

Changes in emotion regulation strategy use and depressive symptoms from summer to winter

Based on heterogeneous findings in the literature (Øverland et al., 2020; Traffanstedt et al., 2016), our first objective was to track the specific temporal pattern of depressive symptoms and emotion regulation strategy use from summer to winter – a time in which noticeable changes could be expected (Cobb et al., 2014; Patten et al., 2017).

Our investigation revealed an increase in depressive symptoms from summer to autumn and early winter, followed by a decrease in late winter, when depressive symptoms reverted to summer levels. This quadratic trajectory – characterized by an inverse U-shaped curve – follows some of the seasonal patterns observed in prior studies, with higher symptom severity or prevalence of depressive episodes in autumn and winter (Cobb et al., 2014; Harmatz et al., 2000; Lukmanji et al., 2019). This temporal trajectory of depressive symptoms highlights the need to develop more targeted interventions and prevention efforts to counteract the rise in symptom severity from summer to winter (Halfin, 2007; Øverland et al., 2020).

In contrast to the wealth of studies on the seasonality in depressive symptoms, it is still unknown if changes in seasonal context are associated with shifts in the use of emotion regulation strategies (Masters et al., 2019). Our findings confirm this assumption: Rumination and acceptance use varied significantly across the seasons. More specifically, rumination use followed a cubic trajectory, with symptoms increasing from summer to autumn, and again from early to late winter. The increase in rumination alongside depressive symptoms, especially from summer to autumn, is in line with our expectations. At the same time, we observed a decrease in acceptance, suggesting that people are less inclined to accept their negative emotions as the year moves into the colder months.

To our knowledge, these findings represent the first empirical evidence that seasonal context fluctuations are not only linked to depressive symptom severity, but are also reflected in altered emotion regulation. The observed changes in emotion regulation over time align with variations in multiple aspects of everyday contexts across the seasons. For example, individuals experience a change in social activities and interactions with others in the winter compared to the summer months (Paul et al., 2023), as well as changes in the amount of indoor and outdoor activities (Hedblom et al., 2019; McRae et al., 2011), and outdoor thermal conditions (Zhang et al., 2021). These seasonal fluctuations in emotion regulation further support theoretical models and empirical work highlighting the essential role of situational factors shaping the use and effectiveness of emotion regulation strategies (Bonanno & Burton, 2013; Gross, 2015; Pruessner et al., 2020).

While these findings suggest fluctuations that are associated with seasonal changes in depressive symptoms and emotion regulation strategy use, the observed trajectories of these changes are more complex than assumed: Both the quadratic shape ( ∩) of changes in depressive symptoms and the linear or cubic changes in emotion regulation emphasize that more sophisticated modeling approaches are required to understand how these dynamic processes unfold. Furthermore, our findings underline the importance of tracking the concurrent and temporal associations between these constructs.

Associations between emotion regulation and depressive symptoms over the seasons

The second objective of our study was to investigate the dynamic temporal interplay in more detail on a between- and within-person level, i.e., distinguishing between differences in a person’s overall severity of depressive symptoms and tendency to use certain regulatory strategies versus their fluctuations within persons over the seasons.

Our data revealed notable associations between emotion regulation strategy use and symptom severity only on a between-person level but not on a within-person level. This means that an individual's general tendency to use certain emotion regulation strategies is associated with their overall level of depressive symptoms. Specifically, we observed a positive association between depressive symptoms and emotion regulation for rumination, (experiential) suppression, and reappraisal, and a negative association for (emotional) acceptance. These results support the extensive literature showing that higher levels of rumination and suppression as well as lower levels of acceptance are associated with more depressive symptoms (Aldao et al., 2010; de France et al., 2019; Ehring et al., 2008; Hijne et al., 2020; Kraiss et al., 2020; Nolen-Hoeksema, 2000; Schäfer et al., 2017; Visted et al., 2018).

Our finding of participants with higher levels of reappraisal showing increased levels of depression required additional analyses. A potential explanation might be the contra-hedonic use of reappraisal (Huang et al., 2020; Ortner et al., 2018) that is, to increase negative emotions. In contrast, most prior studies investigating reappraisal in the context of decreasing negative emotions reported a negative association between reappraisal and depressive symptoms (Kalokerinos et al., 2015) – in particular for positive reappraisal (Brewer et al., 2016; Haga et al., 2009). However, while we found a notable negative association between regulatory goals and reappraisal use, we found no moderating effect of regulatory goals (maintenance, up- or down-regulation) on the relationship between reappraisal use and depressive symptoms. These additional exploratory analyses suggest that regulatory goals in this sample do not account for our divergent results. Therefore, these findings highlight the importance of considering further contextual factors when investigating reappraisal use to better understand its complex relationship with regulatory processes (e.g., regulatory goals; Tamir & Ford, 2009; Tamir et al., 2020; Webb et al., 2012) and outcomes such as negative affect or depressive symptoms (Uusberg et al., 2023). Taken together, these studies combined with our findings suggest a more heterogeneous, context-dependent relationship between reappraisal and depressive symptoms (Aldao et al., 2010; Hartmann et al., 2023; Mueller et al., 2024; Troy et al., 2017).

Emotion regulation in summer as a precursor of depressive symptoms in winter

The third objective was to investigate the prospective relationship between emotion regulation and depressive symptoms across the seasons – especially if emotion regulation in the summer can act as a precursor of depressive symptoms in autumn and winter. Although some existing studies showed that reappraisal could predict a decrease in depressive symptoms over time (Brewer et al., 2016; Haga et al., 2009), and that rumination or repetitive negative thinking may predict increases in depressive symptoms (Hankin, 2008; Spinhoven et al., 2018), the specific seasonal associations are still underexplored. Our study suggests that the extent of reappraisal and rumination in the summer could predict subsequent increases in depressive symptoms in autumn and early winter, respectively, when controlling for baseline symptom severity.

These results align with Sigmon et al. (2009), who reported a predictive effect of rumination on depressive symptoms across the seasons. Other longitudinal studies found that rumination could predict the development of depressive symptoms or episodes (Arnarson et al., 2016; Hijne et al., 2020; Visted et al., 2018), and that daily repetitive negative thinking measured with EMA predicted depressive symptoms two weeks later (Rosenkranz et al., 2020). Thus, these prior findings combined with our results support the predictive power of rumination for the increase of depressive symptoms, further emphasizing the role of rumination as a risk factor in the etiology of depressive disorders (Aldao et al., 2010; Hsu et al., 2015; Liu & Thompson, 2017; Spinhoven et al., 2018).

Our results thus indicate that both rumination and reappraisal could act as potential risk factors for the emergence or maintenance of depressive symptoms. Notably, only reappraisal exhibited bidirectional links, as it could also be predicted by depressive symptoms. This was not the case for rumination, where depressive symptoms in summer could not predict the use of rumination during the colder months. Therefore, our study is one of the first to allow causal conclusions for rumination preceding depressive symptoms, and for reappraisal demonstrating reciprocal associations with symptom severity across the seasons.

Limitations

First, our sample was primarily German, female, young, and received higher education, thus limiting the generalizability of these findings to other populations. Second, the associations between regulatory processes and depression may vary with the severity and duration of depressive symptoms, as well as across the lifespan, thereby necessitating studies that span more extended time periods and include a variety of relevant measures (Whisman et al., 2020). Third, the two-month span of each wave in this study invites further exploration into the effects of assessment length and timing on data comparability (Huffziger et al., 2009) since there are different operationalizations of the seasons in the literature. Going forward, the time lags between assessments should also be considered when interpreting results, as the observed relationships are not uniform in their temporal and reciprocal dynamics (e.g., Huffziger et al., 2009; De France et al., 2019).

Conclusion

In sum, our study observed that changes in seasonal context were associated with fluctuations in both depressive symptoms and emotion regulation. Our findings suggested an increase in depressive symptoms and rumination from summer to early winter, while acceptance decreased over the seasons. Participants with an overall higher use of acceptance demonstrated lower depressive symptoms. In contrast, participants showing a greater engagement in suppression, rumination, and reappraisal exhibited higher depressive symptoms, with the latter two strategies emerging as temporal precursors of symptom severity. By employing a prospective multi-wave design and EMA, the current study revealed that rumination could function as a risk factor for depressive symptoms, allowing preliminary conclusions about its role in preceding increases in symptom severity. These findings could, therefore, contribute to the clinical knowledge necessary for a more effective treatment and prevention of depressive symptoms.