Emotion regulation (ER) is defined by the activation of a goal to change an unfolding emotional response and can be described as any process by which individuals modify their emotional experiences, expressions, and physiology (Gross, 1998, 2015). Being able to effectively regulate one’s emotional reactions is of crucial importance for appropriate social interactions and an essential feature of mental and physical health (Gross and Munoz, 1995; English et al., 2012; Kanske et al., 2012; Hu et al., 2014; Johnstone and Walter, 2014; Kret and Ploeger, 2015). The influential Process Model of Emotion Regulation and its extension (Gross, 1998, 2015) categorizes strategies of emotion regulation according to the time point in the emotion generation process, at which they are being implemented. Cognitive change (e.g., reappraisal) appears early in the process (antecedent-focused) and refers to altering the value of the emotion eliciting stimulus, whereas response modulation (e.g., expressive suppression) takes effect later and aims at altering the emotional response. The most studied reappraisal strategy is reinterpretation, which implies changing the meaning of a stimulus (Ochsner et al., 2002). Another reappraisal strategy is detachment (distancing) where one is taking the perspective of an uninvolved observer to reduce the subjective relevance of the stimuli (Kalisch et al., 2005; Walter et al., 2009). It has been assumed that cognitive reappraisal (reinterpretation and detachment) as antecedent-focused strategies are most effective because the emotional response has not fully unfolded and the negativity of an event itself is altered, whereas response-focused strategies such as expressive suppression often fail in fully modifying the emotional response since they are initiated later in the emotion-generative process (for a review see Gross, 2002). Nevertheless, people implement both strategies in their daily lives and results are pointing to expressive suppression being advantageous in some contexts (Bonanno et al., 2004; Bonanno and Burton, 2013), while reappraisal may also turn out unsuccessful (Aldao et al., 2015). Research that investigates the underlying mechanisms influencing these long and short-term outcomes of both strategies is still ongoing.

Until recently, this research has roughly followed two approaches (Tull and Aldao, 2015): A task-related, experimental approach (hereinafter referred to as task-related ER) and an approach investigating individual differences in ER abilities and dispositional use of strategies, respectively (hereinafter referred to as dispositional ER). The task-related approach uses experimental tasks, in which participants are instructed to use one or more ER strategies to decrease or increase (mostly negative) emotions and investigates effects on different emotional (experiential), behavioral and psychophysiological outcomes. The dispositional approach frequently relies on self-report questionnaires to evaluate ER abilities and dispositional use of ER strategies. One of the most widely used self-report measures of reappraisal (with an emphasis on reinterpretation) and expressive suppression is the Emotion Regulation Questionnaire (ERQ, Gross and John, 2003). Based on the Process Model of Emotion Regulation, the ERQ evaluates the dispositional use of these two strategies.

In experimental settings, task-related reappraisal is effective in changing emotional experiences, behavior, and physiological responses (Webb et al., 2012). The authors report detachment (d+ = 0.45) being significantly more advantageous than reinterpretation (d+ = 0.36). Expressive suppression proved to be effective in the regulation of emotional experiences and behavioral, but not physiological responses (Webb et al., 2012). In contrast, a recent meta-analysis on psychophysiological outcomes of task-related ER reports mixed findings with low to medium effect-sizes for both reappraisal and expressive suppression and mostly non-significant meta-analytical effects (Zaehringer et al., 2020). Concerning dispositional ER, the dispositional use of reappraisal, measured with the ERQ, has been linked to interpersonal functioning and psychological as well as physical well-being (Gross and John, 2003). Moreover, Aldao et al. (2010) could meta-analytically show that dispositional reappraisal is negatively associated with symptoms of psychopathology. In contrast, dispositional suppression was positively associated with psychopathology with medium to large effect sizes (Aldao et al., 2010; Hu et al., 2014), worse interpersonal functioning, and greater risk of depression (Gross and John, 2003). Hence, concerning emotional experiences, both strategies have shown to be successful in the short-term regulation of emotions, while there are mixed findings of short-term physiological outcomes. Reappraisal is advantageous concerning long-term (self-reported) psychological outcomes.

Neuroscientific studies of healthy but also impaired ER can contribute to the understanding of the mechanisms and underlying processes leading to these outcomes. Mostly, this research is investigating neuronal activity in brain regions implicated in cognitive control (i.e., the prefrontal cortex, PFC) and brain regions implicated in emotional processing (i.e., amygdala and insula) as well as the coupling between these structures. Functional brain imaging studies repeatedly showed that activation in the PFC [i.e., the anterior cingulate cortex (ACC), medial (m)PFC, and dorsolateral (dl) PFC] and reduction of amygdala activation is associated with a detachment in ER tasks (Kalisch et al., 2005; Walter et al., 2009; Erk et al., 2010; Koenigsberg et al., 2010; Schardt et al., 2010; Ochsner et al., 2012; Dörfel et al., 2014. During expressive suppression, activity within similar PFC regions and the supplementary motor area (SMA) has been reported (Phillips et al., 2008; Vrtička et al., 2011; Dörfel et al., 2014). In contrast, suppression has been associated with significant increases in the amygdala and insula activity (Goldin et al., 2008; Hayes et al., 2010; Vanderhasselt et al., 2013; Dörfel et al., 2014). Therefore, it can be assumed that the interaction between PFC regions and regions of emotional processing is different for the two strategies.

For successful ER, it is assumed that dorsal PFC regions exert an inhibitory effect on regions of emotional processing via ventral PFC regions (Ochsner et al., 2002; Wager et al., 2008; Lee et al., 2012; Buhle et al., 2014). Consequently, in reappraisal tasks, task-related functional connectivity has been reported between the amygdala and the PFC (Banks et al., 2007; Erk et al., 2010; Schardt et al., 2010; Winecoff et al., 2011; Sripada et al., 2014; Paschke et al., 2016). Moreover, the functional coupling of the amygdala with ventral and dorsal PFC regions were significantly correlated to experiential (self-reported) emotion regulation success (Banks et al., 2007; Paschke et al., 2016). Lee et al. (2012) suggested that functional coupling between the amygdala and prefrontal regions as well as pregenual ACC during cognitive reappraisal depends on individual differences in the capacity for reducing negative emotion. In line with this, studies including patients with psychological disorders with reduced ability to regulate emotions (e.g., depression and anxiety) have found deficits in functional and effective connectivity between the amygdala and frontal brain regions (Erk et al., 2010; Cullen et al., 2011; Niedtfeld et al., 2012; Clauss et al., 2014; Mochcovitch et al., 2014; Radaelli et al., 2015; Picó-Pérez et al., 2017). Additionally, there is evidence that not only task-related connectivity but also alterations in resting-state functional connectivity of (among others) PFC, amygdala, and insula are associated with depression and anxiety (Menon, 2011; Zhang et al., 2014; Barch, 2017). Resting-state functional brain connectivity (rsFC) reflects intrinsic connectivity, which is correlated temporal patterns among brain regions during rest. The resting-state networks closely match networks that have been continuously reported by different task conditions pointing to an intrinsic functional brain architecture important for task-specific brain activation (Smith et al., 2009). This also applies to ER-related brain networks (i.e., default mode network, the executive control network, and the salience network, see Beckmann et al., 2005; Damoiseaux et al., 2006; Seeley et al., 2007). Hence, it can be suggested that activity in task-related ER networks and resting-state connectivity between PFC and the amygdala show associations.

Gabard-Durnam et al. (2016) propose that experiences of stimulus-elicited coactivations in ER brain regions form these resting-state connectivity patterns (long-term phasic molding hypothesis), particularly during development in childhood and adolescence. The authors found that in a sample of children and adolescents, stimulus-elicited amygdala-mPFC connectivity predicted rsFC 2 years later. Likely, the dispositional, daily use of a specific ER strategy and the experience of (un)successful ER alters the functional architecture of these brain networks, which is represented in rsFC. In turn, it can be assumed that functional connectivity influences dispositional, daily strategy choice as well as strategy implementation.

Successful task-related ER (defined by a decrease in self-reported emotional experiences as well as a deactivation of brain regions engaged in emotion processing), as well as dispositional ER, should, therefore, be associated with rsFC between amygdala and PFC. However, few studies so far have directly investigated this. Picó-Pérez et al. (2018) found rsFC between the amygdala and PFC regions as well as the insula to be distinctly associated with dispositional use of suppression and reappraisal, respectively, as measured by the ERQ (Gross and John, 2003). In contrast, Uchida et al. (2015) could not find associations with self-reported dispositional ER (measured with the Difficulties in Emotion Regulation Scale, DERS, Gratz and Roemer, 2004). In a study by Morawetz et al. (2016), task-related ER (reinterpretation) success, defined by affect ratings, was positively correlated with rsFC between right amygdala and the left ventrolateral (vl)PFC as well as the insula. Uchida et al. (2015) demonstrated that greater reappraisal (reinterpretation) success, again measured by affect ratings, showed a significant negative correlation with rsFC of the right amygdala with mPFC. However, the two latter studies only focused on experiential ER success (affect ratings), and did not report results on neuronal ER success (defined by deactivations in regions of emotional processing). Additionally, to our knowledge, findings of these few existing studies have not been validated by replications.

Following this, the present study aimed at replicating and extending findings of associations between ER and rsFC of the amygdala and PFC. To define the replication attempts, we draw upon the definitions proposed by Zwaan et al. (2018), who differentiate between direct and conceptual replication studies. Direct replication is described as a study that attempts to recreate the critical elements (e.g., samples, procedures, and measures) of an original study. The authors underline that “a direct replication does not have to duplicate all aspects of an original study. Rather it must only duplicate those elements that are believed necessary for producing the original effect.” Conceptual replication is defined as a study with theoretically meaningful changes “to the original procedures that might make a difference concerning the observed effect size” (p. 3).

The present study specifically pursued three objectives: (1) we aimed at an investigation of whether individual differences in dispositional reappraisal and expressive suppression (defined by self-reported habitual use as measured with the ERQ) can explain variance in rsFC between amygdala and PFC. To do so, we reanalyzed own existing data (Diers et al., 2014; Scheffel et al., 2019) from three related ER experiments containing measures of dispositional use of reappraisal and suppression (via ERQ) and fMRI resting-state scans to replicate the findings by Picó-Pérez et al. (2018). We aimed at a direct replication according to the definition outlined above. Due to using existing data, there were methodological differences between our investigation and the Picó-Pérez et al.’s (2018) study which will be outlined in the “Materials and Methods” section and in Supplementary Table S23. However, these differences are mostly technical and do not lead to a different operationalization of the constructs.

(2) We aimed at investigating whether individual differences in task-related, experiential reappraisal success (as defined by a decrease in self-reported arousal during a reappraisal task) explain variance in rsFC between the amygdala and PFC. To this end, we reanalyzed existing data from the aforementioned experiments, which focused on detachment as a reappraisal strategy. This investigation is inspired by the study of Uchida et al. (2015). Because there are important differences concerning the operationalization of the constructs and the experimental procedure between our study and the Uchida study (see Supplementary Table S24), the current investigation can be a conceptual replication at best.

(3) Extending the findings of Uchida et al. (2015), we aimed at an investigation of associations between task-related, neuronal reappraisal success (as defined by a decrease of amygdala activity during emotion regulation) and rsFC between the amygdala and PFC, again using existing data of the aforementioned data sets.

This study is a reanalysis of data collected within a larger project on neural correlates and individual differences of ER and its aftereffects (SFB 940 Project A5). To achieve a reliable sample size, we combined three samples from three slightly different ER experiments. Please note that results regarding research questions on task-related effects as well as associations with genetic polymorphisms are published elsewhere (Diers et al., in preparation; Gärtner et al., 2019; Scheffel et al., 2019). Results on the research question of this publication have not been reported in any of these publications. We report how we determined our sample size, all data exclusions (if any), all manipulations, and all measures in the study (Simmons et al., 2012). Data, scripts, and analysis routines can be found at https://osf.io/8wsgu.

Mostly due to the analysis of existing data, our procedures deviated from the methodological procedures of the original studies, differences can be found in detail in Supplementary Tables S23, S24.

The first aim of this investigation was to directly replicate the study by Picó-Pérez et al. (2018). We analyzed associations of dispositional emotion regulation (ER), which is the habitual use of reappraisal and suppression measured via self-report (Abler and Kessler, 2009), with functional resting-state connectivity (rsFC) between the amygdalae and PFC by reanalyzing data from 107 participants of an ER study. None of the hypotheses could be confirmed, that is, we could not statistically confirm associations of dispositional reappraisal and suppression use with rsFC between left and right basolateral and centromedial amygdala, respectively, and regions in the PFC (ACC and SMA) and the insula. Thus, we failed to replicate the results of Picó-Pérez et al. (2018).

Second, we extended the investigation of resting-state functional networks and ER to associations with experiential reappraisal success. This investigation was based on findings by Uchida et al. (2015). Following the recommendations of Zwaan et al. (2018), we aimed at a conceptual replication and hypothesized that experiential reappraisal success (measured via arousal ratings) is associated with rsFC between left and right amygdala and left insula, with rsFC between left and right amygdala and dmPFC, with rsFC between the amygdalae and vmPFC, as well as with rsFC between the amygdala and dlPFC. Again, none of our hypotheses could be confirmed.

Lastly and to extend the research question to neuronal reappraisal success, we added a third analysis. Data of the same sample was analyzed to examine the hypotheses of associations between neuronal reappraisal success (defined by amygdala downregulation during reappraisal in an ER task) with rsFC between the amygdalae and insula, rsFC between the amygdalae and dmPFC, and rsFC between the amygdalae and dlPFC. We were not able to find any significant correlations here either.

To clarify whether there were any basic problems in detecting resting-state networks in our sample, we conducted a whole-brain functional connectivity analysis without any additional predictors and the covariate. This revealed that left and right basolateral and centromedial amygdala were negatively coupled with dlPFC, vlPFC, dmPFC, and dorsal ACC regions, and positively coupled with vmPFC, SMA, subgenual ACC as well as posterior cingulate gyrus and insula/vlPFC (see Supplementary Figure S5 and Supplementary Table S15). Overall, there is an overlap of these regions with the regions reported by Picó-Pérez et al. (2018), albeit not in the same direction. Similar connectivity maps have been reported by others (Roy et al., 2009; Weis et al., 2019 and Tetereva et al., 2020). We, therefore, assume that our resting state measurement has been successful in principle. However, when we included the number of Experiment as a covariate in this basic functional connectivity analysis, none of the clusters showed significant coupling with the amygdala anymore. Separate analyses of basic, whole-brain rsFC of the amygdala nuclei revealed variability between the three different sub-samples of our study (see Supplementary Figures S2–S4). However, this variability points to differences in strength rather than in the composition of the network.

There are several differences in methodology between the study of Picó-Pérez et al. (2018) and our study. Mainly, the differences refer to acquisition parameters of the functional MR images resulting, for instance, in a much lower spatial while slightly higher temporal resolution in Picó-Pérez et al. (2018). There were also differences in the preprocessing of fMRI data and statistical procedures (see Supplementary Table S23). Most importantly, we were not able to directly replicate the size of the PFC ROI for small volume correction. Although we followed the procedure laid out in the original study by Picó-Pérez et al. (2018), which resulted in an ROI of 17,391 voxels (2 × 2 × 2 mm³) in the original study, our mask contained 56,833 voxels (2 × 2 × 2 mm³). A visual comparison further points to some differences in coverage of the PFC, although the regions targeted by our hypotheses were included. Nevertheless, corrections for multiple comparisons had to be performed for a much smaller ROI in the Picó-Pérez study, which might have led to a higher possibility for smaller effects to reach statistical significance (see Figure 1). Moreover, differences are obvious regarding sample size and composition. The original study sample comprised 48 participants with a mean age of 39.6 years, while the participants in our study (N = 107) were much younger with a mean age of 24.4 years. Since emotion control, motives, as well as the choice of strategies change with age (e.g., Scheibe and Carstensen, 2010), this difference in mean age between the samples certainly plays a role. Because of the larger sample size, our study offers greater statistical power, which could have led to a reduced likelihood of false-positive findings.

Nevertheless, concerning aim 1, we did not achieve an exact but direct replication following Zwaan et al. (2018). A different definition of replications offers Brandt et al. (2014). They define close replications as studies that “aim to recreate a study as closely as possible so that ideally the only differences between the two are the inevitable ones” (p. 218). Concerning this strict definition, we did not achieve a close replication of the Pidó-Pérez study. However, we do not consider the differences in data acquisition and preprocessing to produce the failed replication, but the differences between the samples might at least partly explain the divergent results. However, we think that a replication of the same results should be rather independent of the detailed methodology.

Concerning aim 2, the methodological differences between our study and Uchida et al. (2015) are more pronounced. While the samples’ mean age is very similar (see Supplementary Table S24), the sample size is larger in our study. Additionally, Uchida et al. (2015) selected their participants according to their ER abilities to achieve an equal distribution of their abilities. This was not the case in our study. Thus, the original study ensured a higher variability of their main construct, which might have increased the possibility of finding an effect. Concerning the operationalization of experiential reappraisal success, the original study instructed reinterpretation as ER strategy and measured trial-by-trial affect ratings (Uchida et al., 2015), whereas in our replication attempt distancing was used as ER strategy and trial-by-trial ratings were only implemented in one of our data sets. Additionally, all of our experiments used arousal ratings. With these different operationalizations of a central construct, our replication attempt could be considered as conceptual (Zwaan et al., 2018) at best. In other words, “on a continuum from ‘close’ to ‘conceptual’ ” (Brandt et al., 2014) our replication attempt might be placed at the very end of the continuum. Thus, we can only conclude that the findings of the original study could not conceptually be replicated, the results do not extend to a different reappraisal strategy nor arousal instead of affect outcomes. Minor differences between the original and the replication can be found in data acquisition and processing, however, we consider these negligible (see Supplementary Tables S23, S24 for a detailed comparison).

Our findings not only contrast with the two studies on which we based our a priori hypotheses (Uchida et al., 2015; Picó-Pérez et al., 2018), they also contradict several other studies that have identified patterns of intrinsic functional connectivity that differ between the dispositional use of ER strategies or are associated with experiential and neuronal reappraisal success (Morawetz et al., 2016; Pan et al., 2018; Burr et al., 2020). Common to these studies is that regions in the default mode network have been identified. Particularly, the latest study by Burr et al. (2020) used the largest sample up to date (N = 1,316) in a data-driven, theory-free approach and found that intrinsic connectivity of the default mode network was associated with dispositional use of suppression (but not reappraisal). Critically, the authors used general functional connectivity (GFC, Elliott et al., 2019) to leverage shared features of task and resting-state fMRI and circumvent reported reliability issues of resting-state measures (e.g., Noble et al., 2019). Thus, instead of focusing on connectivity in a priori regions of interest between cortical and subcortical areas, distributed networks of brain regions might be a more promising target in future studies as they take into account the complexity of the underlying neuronal processes.

In conclusion, the present replication study calls into question the reported findings on individual differences in resting-state cortico-limbic functional connectivity related to dispositional use of ER strategies and even task-related, experiential, and neuronal reappraisal success. The most parsimonious explanation for the lack of replication is that these differences are either small or non-existent, and/or swamped by sample effects and methodological differences. However, we remain optimistic that continued developments towards improving methodology in resting-state measurement (enhancing reliability) and distributed network approaches will help to eventually reveal reliable patterns of functional connectivity underlying successful emotion regulation.

The datasets generated and analyzed during this study can be found at the Open Science Framework (https://osf.io/p7hb5/).

The studies involving human participants were reviewed and approved by Ethics committee of the TU Dresden (EK 10012012). The patients/participants provided their written informed consent to participate in this study.

AG and DD contributed to the design of the study. AG and CS organized the database and performed the statistical analysis. DD, CS, and AG wrote the first draft of the manuscript. All authors contributed to manuscript revision, read and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.