Neural underpinnings of individual differences in emotion regulation: A systematic review

This review synthesises individual differences in neural processes related to emotion regulation (ER). It comprises individual differences in self-reported and physiological regulation success, self-reported ER-related traits, and demographic variables, to assess their correlation with brain activation during ER tasks. Considering region-of-interest (ROI) and whole-brain analyses, the review incorporated data from 52 functional magnetic resonance imaging studies. Results can be summarized as follows: (1) Self-reported regulation success (assessed by emotional state ratings after regulation) and self-reported ER-related traits (assessed by questionnaires) correlated with brain activity in the lateral prefrontal cortex. (2) Amygdala activation correlated with ER-related traits only in ROI analyses, while it was associated with regulation success in whole-brain analyses. (3) For demographic and physiological measures, there was no systematic overlap in effects reported across studies. In showing that individual differences in regulation success and ER-related traits can be traced back to differences in the neural activity of brain regions associated with emotional reactivity (amygdala) and cognitive control (lateral prefrontal cortex), our findings can inform prospective personalised intervention models.


Introduction
With the rising prevalence of affective disorders, such as major depression, detecting, treating or preventing mental disorders promptly is becoming increasingly important (McGrath et al., 2023).One way to detect disorders or predispositions for developing disorders at an early stage is by identifying biological markers in the brain (Gao et al., 2018;Horwitz and Rowe, 2011;Kang and Cho, 2020;Lener and Iosifescu, 2015).Since difficulties in emotion regulation represent a transdiagnostic feature in the development of affective disorders (Aldao et al., 2016;Cludius et al., 2020;Kring and Sloan, 2010;Sloan et al., 2017), individual differences in emotion regulation could serve as a proxy to identify biological markers indicating or predisposing for psychopathology (Strawbridge et al., 2017;Williams, 2017).In this sense, individual differences in the neural mechanisms underlying emotion regulation can be understood as risk vs. resilience factors for the development of stress-related mental disorders.
Emotion regulation describes how people control and modulate the intensity of their emotionseither by using implicit or explicit strategies, such as distraction, reappraisal, or suppression (Gross, 1998).For over twenty years, emotion regulation has been studied extensively with functional magnetic resonance imaging (fMRI).However, most of the previous fMRI literature has been concerned with identifying the neural underpinnings of emotion regulation as they are observed as common mechanisms across people.To this end, neural activation observed during emotion regulation tasks was typically averaged across subjects to improve the signal-to-noise ratio.Given the wealth of previous studies, the literature on the neural basis of emotion regulation has been summarized by several previous meta-analyses and reviews (Berboth and Morawetz, 2021;Buhle et al., 2014;Diekhof et al., 2011;Frank et al., 2014;Kalisch, 2009;Kohn et al., 2014;Messina et al., 2021;Morawetz et al., 2017Morawetz et al., , 2020;;Ochsner et al., 2012;Picó-Pérez et al., 2017).These previous synopses of findings across studies revealed large-scale neuronal networks commonly involved in emotion generation and regulation in the healthy population, including cortical regions in the prefrontal, parietal, temporal, and insular cortex, as well as subcortical regions such as the amygdala, hippocampus, and the thalamus (Morawetz et al., 2020).Two cortical networks (typically activated more strongly during emotion regulation than during a 'no regulation' control condition) have been associated with working memory, attention, and language processes involved in the cognitive control of emotions.On the other hand, a subcortical network (typically activated more strongly for unregulated processing of emotional stimuli during a control condition than for emotion regulation) has been related to basic emotional reactivity.These findings suggest that emotion regulation can be regarded as a multi-component process where sub-processes of emotional reactivity and cognitive control interact through a fine-tuned orchestration of the identified neuronal networks (Feldman Barrett and Satpute, 2013).
The importance of characterising brain function at the individual level has been acknowledged for several years (Dubois and Adolphs, 2016;Van Horn et al., 2008).During this time, the field of research in emotion regulation started to move from describing population averages to describing and understanding person-specific processes (Dore et al., 2016).Individual differences in emotion regulationdefined as between-subject differencesplay a central role in psychological well-being and mental health (Rammensee et al., 2023) and many psychological disorders are associated with emotion dysregulation, the predominant use of certain strategies, or rigid responses to the environment (Lincoln et al., 2022).Consequently, interventions addressing dysfunctional emotion regulation and related characteristics of underlying neural systems also need to be effective at the individual level.It is thus crucial to better understand how individuals vary in their brain responses to emotionally challenging stimuli and situations (emotional reactivity) and in brain activity shown during the effective regulation of emotions (cognitive control).Addressing these questions is essential for constructing basic science models that accurately reflect individual differences in neuronal responses and for translational research concerned with developing appropriate interventions for individuals who suffer emotion dysregulation in specific situations.
At the moment, the dominant approach to study emotion regulation with fMRI is to instruct participants to down-regulate negative emotions using specific regulation strategies such as reappraisal (i.e., the reinterpretation of an emotional stimulus to alter its emotional impact) while viewing an aversive image (Morawetz et al., 2017).In this study design, each experimental trial concludes with a subjective rating of the participants' emotional state.Usually, a Likert scale is used to capture these ratings, indicating the participant's regulation success.
Brain activity recorded during emotion regulation tasks has been related to online and offline measures of differences between participants (see Fig. 1).Online measures acquired during the regulation task inside the MRI scanner are used to assess individual differences in regulation success and include self-reported emotional state ratings or physiological measurements such as heart rate, skin conductance, or pupil diameter as objective indices of emotional arousal.These measures capture state-like differences in emotional experience and are highly context-dependent: The ability to regulate one's emotions might vary between individuals due to stable differences between individuals but also due to transient differences in motivation, cognitive effort, mood, or level of fatigue.The latter aspect is covered by studies implementing standard, well-established measures of the habitual use of emotion regulation strategies better suited to capture trait-like differences in emotion regulation.These trait-like differences have been assessed using offline measures, usually collected with questionnaires before or after the C. Morawetz and U. Basten fMRI experiment.A prominent example of a questionnaire measure reflecting trait-like differences in the habitual use of emotion regulation strategies is the Emotion Regulation Questionnaire (ERQ), which assesses the use of reappraisal and suppression (Gross and John, 2003).Demographic variables are also assessed by questionnaires and relate to questions about age, sex and cultural background.
Analyses of individual differences in fMRI studies on emotion regulation have mainly been correlational.Most studies relate their variable of interest to individual differences in BOLD signal changes during emotion regulation to establish a brain-behaviour association.This is either done (a) by using one or more regions of interest (ROI) or (b) by considering activation changes in the whole brain.Analyses using ROI analyses are strongly hypothesis-driven.The search space for an association between individual differences in brain activation and emotion regulation is a priori restricted to a single brain region or a small set of brain regions.Usually, analyses are restricted to brain regions that show increases or decreases in activity during emotion regulation at a group level, i.e., across participants.Technically, these regions are predetermined anatomically or functionally.This mainly concerns limbic structures such as the amygdala and prefrontal brain regions associated with cognitive control exerted to support emotion regulation.
On the other hand, whole-brain analyses are more exploratory, as researchers do not restrict their analyses based on specific hypotheses about where differences in activity might be expected in the brain.In our view, studies on inter-individual differences in brain activity must consider the whole brain, i.e., brain regions outside the areas that show activation effects at the group level.It could be precisely those brain regions that do not show group-level effects due to great variability between individuals that are particularly important for understanding inter-individual differences in emotion regulation.In addition to the correlational approach to individual differences, an alternative approach divides the sample into groups to compare activity in brain regions during emotion regulation according to, e.g., sex or age.Unpaired t-tests are used to determine group-related differences in brain activity.
To determine the neural underpinnings of individual differences in emotion regulation, we systematically reviewed all available fMRI studies that have related individual levels of brain activity shown during emotion regulation with other online or offline measures of the same individuals.For our review, we grouped findings for online measures into (a) self-reported regulation success and (b) physiological regulation success and findings for offline measures into (c) ER-related traits and (d) demographic variables.Subsequent analyses further differentiated between (i) hypothesis-driven ROI analyses and (ii) exploratory whole-brain analyses.
During the review preparation, it became clear that there are far more studies on self-reported regulation success and ER-related traits than on physiological regulation success and demographic measures.Any outcome of integrating findings for correlates of physiological and demographic measures in this review must be interpreted with caution.For completeness, we decided to include these studies in our review.When interpreting our findings, however, we will focus on the neuronal correlates of self-reported regulation success and ER-related traits, for which a more comprehensive database was available.
Our summary of findings across measures aims to determine in which brain regions the individual differences in activation are associated with regulation success and ER-related traits and could thus serve as targets for personalized intervention programs.

Study selection
A systematic literature search has been performed on emotion regulation studies using functional magnetic resonance imaging (fMRI) and focusing on individual differences.The search was done in PubMed and covered studies published between 01.01.2001 and 31.12.2022.After the analyses, manuscript preparation, and peer review, our review is expected to be published more than a year after the completion of the literature search.With regard to the question of whether it will still be up to date at the time of publication, we would like to point out that, especially for systematic reviews focussing on a qualitative integration of findings, it is unlikely that results will be substantially changed by the inclusion of limited new evidence (Stokes et al., 2023).The search term for the literature search consisted of a term referring to the paradigm ("emotion regulation" OR "affective regulation" OR "implicit emotion regulation" OR "explicit emotion regulation" OR "interpersonal emotion regulation" OR "extrinsic emotion regulation" OR "intrinsic emotion regulation") combined with a term related to regulation strategy ("reappraisal" OR "suppression" OR "distraction" OR "detachment" OR "labelling" OR "affective labelling" OR "reinterpretation" OR "rumination") and a term referring to the method ("fMRI" OR "neuroimaging" OR "functional magnetic resonance imaging" OR "functional MRI").This search resulted in a total of 353 studies considered for inclusion in the review.Studies meeting the following criteria were included: 1. Studies published in English in a peer-reviewed journal.2. Studies on healthy adults with no prior report of neurological, medical, or psychiatric disorders.We only included data reported separately for the control group from studies that compared patients and healthy control subjects.3. Studies using fMRI compared BOLD signal changes during an emotion regulation task versus a 'no regulation' control condition.4. Studies that used visual stimuli (pictures) in the emotion regulation task.

Studies that reported whole-brain or region-of-interest (ROI) analysis
with a covariate from one of the following categories: (a) online self-reported regulation success collected during fMRI scanning based on emotional state ratings; (b) online physiological regulation success collected during fMRI scanning or in direct temporal proximity, including, e.g., salivary cortisol, skin conductance, pupil dilatation, etc., (c) offline self-reported ER-related traits assessed outside the scanner, including, e.g., personality traits or the habitual use of regulation strategies; or (d) offline demographic measures such as age, sex, or cultural background.

Data extraction
The studies were reviewed to determine whether they met the inclusion criteria.All fMRI studies that reported a whole-brain or ROI-based analysis of a cognitive emotion regulation task (i.e., up-or downregulation of emotions using a cognitive strategy) with a covariate were included.For all studies included, we summarized the number of participants, types of covariates, stimuli, statistical method (correlation, regression, etc.), and spatial focus (ROI vs. whole-brain analyses) in Supplemental Table S1.
To evaluate consistent findings regarding individual differences in brain activity during fMRI, we report all brain regions for which significant task x covariate interactions have been reported for a regulation versus a no regulation control condition (usually maintaining the current emotion while viewing the stimulus).To integrate findings, we differentiated between (i) ROI-based and (ii) whole-brain analyses.To interpret the direction of the associations between BOLD signal changes and the covariates (positive vs. negative correlation), we additionally considered the regulation goal (up-vs.down-regulation).

Results
Of the 353 studies that met the search criteria and were screened for an analysis of individual differences in brain activation during emotion regulation (Fig. 1), we included 52 studies in our review.Sample sizes of C. Morawetz and U. Basten the included studies ranged from N = 10-112 participants (M = 24.5,SD = 25.6; for median sample sizes by publication year, see Supplemental Figure S1).The 52 studies included 68 separate analyses comprising 29 covariates, 299 foci of effects, and data from 2.156 participants.Eight studies reported analyses for individual differences falling into two different categories of covariates, e.g., self-reported regulation success and ER-related traits (for details, see Table S1).Thirty-two analyses investigated individual differences in online self-reported regulation success (e.g., reappraisal success, emotional state ratings).Nine analyses associated individual differences in brain activation with online physiological regulation success (e.g., skin conductance).Twenty analyses examined associations with offline self-reported traits (e.g., habitual use of emotion regulation strategies).Seven analyses used offline demographic variables (i.e., age, sex, and cultural background) as covariates.
Regarding the statistical method applied (Fig. 2A), the majority of studies (66%) used a correlation analysis (i.e., correlating individual BOLD signal changes extracted from a ROI with the covariate of interest).In comparison, 22% used a regression analysis (usually using the covariate of interest in a whole-brain regression).Only a minority of studies (12%) used other methods, such as structural equation modelling (SEM) and ANOVA.Regarding the spatial focus of analyses in the brain, the most prominent approach was the hypothesis-driven approach, restricting analyses to specific regions of interest (ROI approach, 71%; Fig. 2B).In contrast, only 20% of the experiments implemented an exploratory approach considering the whole brain.9% of analyses used a combination of both approaches.
Most studies (n ≥ 20) related individual differences in brain activation to self-reported regulation success and/or traits.Only a few studies considered physiological regulation success or demographic variables.

Self-reported regulation success
Of the 32 analyses relating individual differences in brain activity during emotion regulation to self-reported regulation success assessed for the same task, n = 24 (75%) used an ROI approach, n = 5 (15%) implemented a whole-brain approach, and n = 3 (9%) used both approaches.The most prevalent self-reported regulation success used in 47% (n = 15) of the analyses was a simple absolute affective state rating assessed during a regulation task.In contrast, 41% of the analyses (n = 13) used a relative measure of reappraisal success for which the absolute mean rating during a control condition is subtracted from the mean rating during a regulation condition.(Fig. 3A).The other measures of self-reported regulation success (12% of the analyses, n = 8) included anxiety reduction, loss aversion, cognitive functioning, dietary selfcontrol, and executive functioning.
In both ROI and whole-brain analyses, the self-reported regulation success correlated with activity in the bilateral prefrontal cortex (PFC), parietal, and temporal cortex (Figs.4A and B, Supplemental Table S2).However, an association with activation in the bilateral amygdalae was more pronounced in the ROI than in the whole-brain approach.
Overall, the direction of the association between the ability to downregulate emotions and brain activity was mainly positive -62% of all foci in this domain demonstrated a positive relation.In comparison, 32% of the foci showed a negative association.Positive correlations were observed within the PFC, temporal, and parietal cortex (Fig. 5), indicating that in more than half of the analyses, an increase in activity in those regions was related to an increase in self-reported regulation success (i.e., feeling less negative).In contrast, the ventromedial and ventrolateral PFC and the bilateral amygdalae demonstrated a negative association between brain activity during regulation and self-reported regulation success, indicating that increased activity in these regions was linked to greater negative feelings and reduced reappraisal success.In addition to these patterns of spatial clustering for positive and negative associations between self-reported regulation success and ERrelated brain activation, Fig. 5 also shows that in a couple of regions positive and negative associations were in close spatial proximity (see orange and yellow spheres in Fig. 5).This particularly affects the PFC (lateral and medial, esp. in the left hemisphere).Here, due to the divergence in the directions of the associations (positive vs. negative), no general statement can be made about the relation between selfreported regulation success and ER-related brain activation.Finally, for the up-regulation of emotions, too few studies were available to draw reliable conclusions regarding the neural correlates.

ER-Related traits
Of those analyses relating individual differences in brain activity during emotion regulation to self-reported traits (n = 20), 70% used an ROI approach (n = 14), and 30% used a whole-brain approach (n = 6).The most popular questionnaires used in neuroimaging studies on emotion regulation were questionnaires assessing the habitual use of emotion regulation strategies, such as the Emotion Regulation Questionnaire (ERQ; REF) and the Cognitive Emotion Regulation Questionnaire (CERQ; REF) -used in n = 9 analyses (45%; Fig. 3C).The remaining analyses (n = 11; 55%) considered questionnaires assessing personality traits (e.g., Big Five, impulsivity, alexithymia, attachment style, mindfulness, sleep quality, and well-being).Regarding the directions of associations with brain activity during the down-regulation of emotions, most ER-related traits correlated positively with brain activity (46%; Supplemental Fig. S2).A negative association with regulation-related activity in 37% of the analyses was found mainly in dorsomedial PFC and amygdala.
For analyses using an ROI approach, convergence across studies was observed in bilateral amygdalae (indicated in yellow in Fig. 4C, Supplemental Table S2) and dorsomedial PFC (indicated in pink in Fig. 4C, Supplemental Table S2).In contrast, the few findings from whole-brain analyses did not show any apparent clustering across studies (Fig. 4D, Supplemental Table S2).

Physiological Regulation Success
Only nine original analyses looked at individual differences in brain activity related to physiological regulation success.These analyses preferentially used the ROI approach (n = 7, 78%) as opposed to a wholebrain approach (n = 2, 22%).The most prominent physiological regulation success measures studied in association with brain activity during the down-regulation of emotions were skin conductance (n = 4, 44% of all analyses; Fig. 3B) and heart-rate variability (n = 2, 22%).Other measures, including cortisol, endocrine activity, pupil diameter, have In ROI analyses, physiological regulation success was associated with brain activity during the down-regulation of emotions in bilateral amygdalae (indicated in yellow in Fig. 6A, Supplemental Table S2) and left lateral PFC (indicated in pink in Fig. 6A, Supplemental Table S2).While the peaks from the ROI analyses of the amygdalae were very close together (as can be expected for an ROI approach targeting an anatomically small structure), it must be noted that the peaks in the lateral PFC were quite widely scattered.We did not observe an apparent spatial clustering across analyses.In whole-brain analyses, physiological regulation success did not correlate with activity in the lateral PFC -but with activity in the occipital lobe (indicated in purple in Fig. 6B, Supplemental Table S2) and parietal cortex (indicated in green in Fig. 6B, Supplemental Table S2).Again, within the occipital and parietal cortex, we did not observe a solid spatial clustering.
The directions reported for the associations between the physiological measures and brain activity were mixedwith 47% of the analyses reporting a positive and 47% a negative association between the covariate of interest and brain activity during the down-regulation of emotions.The directions of these correlations follow an anteriorposterior gradient (Supplemental Fig. S3).Activity in the prefrontal cortex was positively linked to physiological regulation success during emotion down-regulation.In contrast, activity in parietal and occipital cortex regions was negatively related to physiological regulation success during down-regulation.Mixed findings were observed in the temporal lobes.

Demographic variables
Only seven analyses investigated how individual differences in brain activity during emotion regulation were related to demographic variables.Of these, n = 3 (43%) used an ROI approach, n = 3 (43%) implemented a whole-brain approach in addition to ROI analyses, and n = 1 (14%) only used a whole-brain analysis.Overall, the demographic variables age and sex have been studied equally often (both in n = 3 analyses; Fig. 3D).Beyond that, a single study investigated how regulation-related activity depended on cultural background, i.e., American vs. Chinese cultural background.
When looking at studies that used an ROI approach to study sex-and age-related differences in brain activity during emotion regulation, we observed clustering of effects across analyses in the left amygdala (indicated in yellow in Fig. 6C, Supplemental Table S2) and bilateral dorsolateral PFC (indicated in pink in Fig. 6C, Supplemental Table S2).Further analyses differentiating between the effects of age and sex on brain activation during regulation revealed that differences in amygdala activity were observed in association with sex.In contrast, differences in activation of lateral PFC have been associated with age (Supplemental Fig. S4).
Studies using whole-brain analyses reported widespread effects across prefrontal, parietal, temporal, and occipital cortex regions and limbic areas (Fig. 6D, Supplemental Table S2).Differentiating between age, sex, and culture revealed that these effects were selectively driven by sex as a covariate.However, the many peak coordinates reported for correlations between sex and regulation-related brain activity in wholebrain analyses were so widely spread that there was no apparent clustering across analyses.

Discussion
Overall, inter-individual differences in brain activity during emotion regulation have been reported for the whole brain (frontal, parietal, temporal, and occipital cortex) and limbic regions.These differences in activity were associated with online and offline measures of regulation success and ER-related traits.Most studies investigating individual differences in brain activation during emotion regulation correlated activity measures extracted from predefined ROIs with self-reported regulation success, i.e., emotional state ratings acquired during an fMRI emotion regulation experiment.In contrast, regression analyses were often applied in whole-brain analyses when looking for activity differences depending on ER-related traits like the habitual use of certain strategies.The current findings not only summarize the state of the current literature but also reveal which measures are underrepresented and should be investigated in future studies.

Differences in brain activity related to self-reported regulation success and ER-related traits
The reviewed studies provide evidence for a reliable association between brain activity during emotion regulation and self-reported regulation success on the one hand and ER-related traits on the other.A core pattern observed during emotion regulation was the link between the self-reported regulation success and the activity in the lateral PFC and the amygdala: Higher levels of regulation success (individual regulation capacity) were related to higher levels of activity in the lateral PFC and lower levels of activity in the amygdala.This pattern was observed in ROI and whole-brain analyses, underscoring the robustness of this association.In addition, in ROI analyses, the association of activity in lateral PFC and amygdala with regulation success was further corroborated by the link with ER-related traits, particularly the habitual use of certain regulation strategies (individual regulation tendency): People who show a higher tendency to use reappraisal in their daily lives (as assessed, e.g., with the ERQ), also show higher levels of regulationrelated activity in the lateral PFC and lower levels of activity in the amygdala.
This interaction of individual fronto-amygdalar activity with individual regulation capacity and tendency can be related to top-down models of emotion regulation (Johnstone et al., 2007;McRae et al., 2011;Ochsner et al., 2012;Phillips et al., 2008;Wager et al., 2008).These models suggest that during emotion regulation, frontal cortex regions inhibit the activity of the amygdala.Apart from these two key players in emotion regulation, we also found that activity in the temporal and parietal cortex is consistently related to differences in regulation success, suggesting that activation in these regions might contribute to an individual's emotion regulation capacity.
It should be mentioned that the findings regarding the direction of association between self-reported regulation success and brain activation during the down-regulation of negative emotions (cf.Fig. 5) showed some inconsistency.While the majority of findings suggest that greater regulatory success is associated with stronger activation of the lateral and medial PFC, some studies have shown the opposite, i.e., stronger PFC activity going along with lower regulatory success.We could not identify any systematic factor that would have determined the direction of association for these studies.Of note, the specific covariates we pooled into the category of 'self-reported regulation success' were rather homogeneous, usually reflecting a rating of affect right after the emotion regulation task.Some studies used simple affect ratings; others used a relative measure of regulation success, i.e., subtracting ratings after emotion regulation from a no-regulation control condition (cf.Supplemental Table S1).Theoretically, it is possible that the association between regulation success and brain activation is moderated by other factors, such as the difficulty of the regulation task or the regulation capacity of the participants.For example, in some studies, a positive association may be explained by stronger PFC activation making regulation success more likely.At the same time, in other studies, higher PFC activation might reflect a compensatory but ineffective (and inefficient) increase in neuronal effort shown in response to low success.It is important to emphasise that we cannot determine the mechanisms underlying the observed associations for the studies we reviewed.The fact that findings on individual differences in brain activation are often subject to post hoc interpretations in terms of effort and efficiency is a problem well-known from other fields of neuroimaging (e.g., Basten et al., 2015; for a discussion, see Poldrack, 2015).For now, our review shows that it is an important topic for future research to explain the inconsistencies in the direction of associations between PFC activation and regulation success across studies.By ensuring sufficient statistical power, future research can ensure that inconsistencies are not introduced by spurious correlations resulting from the study of too small samples (Yarkoni, 2009) Furthermore, potential moderation effects, such as those outlined above, should be investigated systematically.
In line with the multiple network model of emotion processing (Feldman Barrett and Satpute, 2013), our findings show that activity in most brain regions implicated in emotion generation and regulation are related to individual differences in self-reported regulation success and ER-related traits (Fig. 7).The lateral PFC together with parietal and temporal regions, has been implicated in processes of working memory and attention as well as with language and semantic processes involved in emotion regulationprocesses that help sustain the regulation goal and support the implementation of a potential semantic re-interpretation of the emotional stimulus (Messina et al., 2015(Messina et al., , 2016;;Morawetz et al., 2017Morawetz et al., , 2020)).These regions are part of large-scale neural networks supporting diverse cognitive functions (like working memory, attention, and language/semantics) that interact in cognitive emotion regulation (green regions in Fig. 7).In contrast, the occipital cortex and the amygdala are part of an emotional reactivity network (red regions in Fig. 7) and have been implicated in the perception and processing of emotion stimuli and the generation of an emotional response (Morawetz et al., 2020;Zald, 2003).Especially for associations between brain activity and self-reported regulation success there is robust evidence from whole-brain analyses (represented in black in Fig. 7).In contrast, an association with ER-related traits yielded by ROI analyses might be less generalizable (indicated in blue in Fig. 7).
As, generally, our findings demonstrate that regulation success (individual regulation capacity) and ER-related traits (individual regulation tendency) modulate activity in brain regions functionally associated with emotional reactivity and cognitive emotion regulation, studies investigating the interplay between these regions during emotion regulation should integrate measures of regulation capacity and tendency as predictors in their models, as they might also explain individual differences in functional and effective connectivity between these regions (Berboth and Morawetz, 2021).
Our review demonstrates that the lateral PFC is intricately linked to self-reported regulation success (individual regulation capacity) and ERrelated traits (individual regulation tendency).The observed individual characteristics of the neural processes underlying emotion regulation might be understood as individual risk or resilience factors with regard to the development of stress-related mental disorderswhere higher PFC activity and lower amygdala activity with their link to greater regulation success might support higher resilience.In the future, these findings may help to identify people who have a rather weak basis for successful emotion regulation at the neural level (i.e., an individual pattern of strong amygdala activity along with low recruitment of the PFC) and who may, therefore, particularly benefit from interventions to train and increase their regulation capacity.With different approaches, researchers have aimed to enhance an individual's capacity to regulate emotions by strengthening PFC function through some sort of training.These approaches comprise neurofeedback (Herwig et al., 2019;Linhartová et al., 2019;Lorenzetti et al., 2018;Yu et al., 2023;Zaehringer et al., 2019;Zotev et al., 2020Zotev et al., , 2013Zotev et al., , 2011)), reappraisal training (Denny et al., 2014(Denny et al., , 2013)), and transcranial direct current stimulation (Chen et al., 2023;De Smet et al., 2023;Sanchez-Lopez et al., 2021;Vanderhasselt et al., 2013).These studies showed that training programs that effectively engaged and modulated activity in the lateral PFC improved emotion regulation skills.
Moreover, investigating the link between training and lateral PFC function holds promise for identifying biomarkers associated with improved or impoverished regulatory abilities (Barkus, 2020;Denny, 2020;Iwakabe et al., 2023;LeBlanc et al., 2020;Schweizer et al., 2013).Such biomarkers could aid in assessing the efficacy of training interventions and contribute to a better understanding of the neurobiological underpinnings of individual emotion regulation capacity as a transdiagnostic risk of resilience factor for developing psychopathology (Aldao et al., 2016;Kring and Sloan, 2010).Effective training strategies targeting the lateral PFC may impact the predisposition and course of various affective disorders, highlighting their potential for applications in prevention and therapy (Kang and Cho, 2020;Lener and Iosifescu, 2015).Ultimately, understanding the role of the lateral PFC in resilience the ability to adapt to and bounce back from adversityadds another layer of significance to the prevention aspect of training, as enhanced levels of emotion regulation capacity may contribute to adaptive strategy choices (Rammensee et al., 2023) and increased resilience in individuals facing various life challenges (Kalisch et al., 2019 ;Troy et al., 2023;Ungar and Theron, 2020).Overall, focusing on training emotion regulation with lateral PFC activity as an outcome measure holds promise for unravelling biomarkers, addressing transdiagnostic features (McTeague et al., 2017;Phillips et al., 2008;Zilverstand et al., 2017), and fostering resilience, thereby advancing our understanding and application of emotion regulation interventions.

Differences in brain activity related to physiological regulation success and demographic variables
A handful of studies have also linked individual activity levels in amygdala and lateral PFC to physiological regulation success in ROI analyses.Given the small number of studies (n = 6) and the diversity of physiological measures used as objective markers for emotional arousal (heart rate, skin conductance, pupil dilation, etc.), this finding must be understood as preliminary.In principle, physiological measures allow recording of automatic, relatively unconscious, and fast responses to emotional stimuli (Bradley et al., 2008;Edelmann and Baker, 2002;Lapate et al., 2014;Olsson and Phelps, 2004).However, it is important to note that some physiological measures can also be influenced by other factors, such as cognitive effort, which may confound their association with emotional arousal.For example, the electrodermal response, linked to emotional arousal, typically decreases with successful emotion regulation.However, other measures, like pupil dilation, which can reflect both emotional arousal and cognitive effort, might increase during emotion regulation tasks due to the cognitive demands of controlling emotions (Maier and Grueschow, 2021;Scheffel et al., 2021;Wang et al., 2018).Though emotion regulation does have an effect on, e. g., cardiovascular, electrodermal, respiratory, and pupillometric responses, a recent meta-analysis revealed that physiological measures assessed during emotion regulation provided inconsistent results across studies with relatively small mean effect sizes (Zaehringer et al., 2020).Consequently, it has been suggested that in fMRI studies, physiological measures might be more important to control for physiological noise in the data (i.e., intra-subject variance) than to indicate individual regulation success (Hutton et al., 2011;Krüger and Glover, 2001;Triantafyllou et al., 2005).
Previous research on sex differences in emotion regulation suggests that sex-related differences in emotion regulation can explain the empirically observed association between sex and psychopathology, esp.higher rates of affective disorders in women (Nolen-Hoeksema, 2012).Research into sex differences in general emotion processing in the amygdala (i.e., emotional reactivity) provided mixed results, and meta-analyses could not support the idea of sex-related differences in emotion processing in the amygdala (Eliot et al., 2021;Sergerie et al., 2008).Most studies investigating neural underpinnings of emotion regulation have not studied the effects of sex -either because samples consisted of only one sex or, in the case of mixed-sex samples, because samples were too small to address sex differences adequately.For our review, we found only three studies that reported sex differences in brain activity during emotion regulation.All effects referred to differences in amygdala activity, although the direction of the effects was inconsistent across studies.However, due to the very small number of studies, this preliminary convergence of findings must be interpreted carefully.Sex differences in brain activity during emotion regulation still remain largely unexplored.More studies are needed to address this issue, particularly in relation to individual differences in strategy use and emotion regulation flexibility, where sex differences have been previously reported (Goubet and Chrysikou, 2019;Nolen-Hoeksema, 2012;Nolen-Hoeksema et al., 1994;Nolen-Hoeksema and Aldao, 2011).

Methodological limitations
The main limitation of the present study is that the overall number of studies that have investigated individual differences in brain activity during emotion regulation is still relatively small.Though more and more studies are considering individual differences, there were not enough studies available to perform a quantitative meta-analysis.Current state-of-the-art simulations recommend at least 20 studies to run a quantitative meta-analysis to examine which brain regions activity converges across studies (Eickhoff et al., 2016).Therefore, for this review, we have opted for a qualitative integration of findings available to date.The review provides an important overview into which brain regions have been related to individual differences in emotion regulation assessed by tasks or questionnaires.We hope that our review stimulates further research on individual differences in brain activation during emotion regulation.When more studies on this topic will be available in the future, it will be an important goal to perform a quantitative integration of findings in coordinate-based meta-analyses (e.g., by activation likelihood estimation, ALE; Eickhoff et al., 2009).
A second limitation are the small sample sizes and consequently low statistical power of some of the original studies.With the rise of a new "big data" era, a shift in data collection towards increased sample sizes can be expected (Elam et al., 2021;Harms et al., 2018;Littlejohns et al., 2020;Somerville et al., 2018;Taylor et al., 2017;Thompson et al., 2020).The possibility of pooled funding sources and data-sharing infrastructure opens the way to collect data on very large samples and, thus, sets the stage for investigating more complex questions about individual differences in emotion processing (e.g., Cambridge Centre for Ageing and Neuroscience, Cam-CAN) (Shafto et al., 2014;Taylor et al., 2017).Single studies aiming to examine individual differences need to follow the recommendations of increasing sample size to address this issue adequately (Yarkoni, 2009;Yarkoni and Braver, 2010).
A third limitation is that many of the original studies used ROI analyses, restricting their search space of individual differences to a small set of a priori defined brain regions (esp.amygdala and lateral PFC) and thus potentially missing effects in other parts of the brain (Poldrack, 2007).Future studies should focus on conducting whole-brain analyses and ensure sufficient statistical power by studying sufficiently large samples rather than by restricting the search space (Soares et al., 2016;Yarkoni and Braver, 2010).
Finally, all previously published analyses are potentially limited due to publication bias that emphasizes positive findings and potentially overlooks negative or null results (Ioannidis, 2011;Jennings and Van Horn, 2012).This selection bias may distort our understanding, causing Fig. 7. Individual Differences in Brain Activity Related to Self-Reported Regulation Success and ER-Related Traits, Note.Schematic illustration highlighting effects that showed convergence across studies.Red and green are brain regions associated with emotional reactivity and cognitive emotion regulation, respectively.Icons represent the categories of covariates that have been observed to be related to activity differences in the illustrated brain regions across the reviewed studies.WB: whole-brain analyses; ROI: region of interest analyses.
an excess of significance in the published literature (David et al., 2013).

Perspectives for future research
To build upon the standard individual difference approach (i.e., correlations) and capture more fine-grained individual differences in brain activity, the field of emotion regulation could relate psychometric theory to statistically advanced methods (Cooper et al., 2019).One such approach is statistical mediation analysis, which indicates whether a set of correlations between, e.g., three variables are consistent with a causal model postulating that the effect of the first variable on the third variable is at least partly explained through the mediating role of the second variable (MacKinnon et al., 2002).So far, only one of the studies identified for the review used mediation analysis to identify prefrontal-subcortical pathways that mediate regulation success (Wager et al., 2008).Another statistical approach widely used in psychometric research is structural equation modelling, which examines relationships between latent constructs inferred from multiple measured indicators and other directly observable variables (Bollen, 1989).So far, only one study has implemented this approach predicting reappraisal success from ER-related trait measures and changes in brain activity from the left PFC and the amygdala (Morawetz et al., 2017).
An alternative multivariate technique that could examine individual differences in emotion regulation concerning brain activity is partial least squares, where the focus is on network-level activation rather than circumscribed brain regions (McIntosh and Lobaugh, 2004).To the best of our knowledge, this approach has yet to be used to examine the brain-behaviour relationship in emotion regulation.However, it could help to examine, e.g., how functional network configurations relate to mood profiles over time and to profiles of emotion dysregulation (Mirchi et al., 2019).To determine biomarkers related to emotion dysregulation, it is also of great importance to shift from a correlational to a predictive framework to ensure generalizability, for example, by using machine learning to predict emotion regulation success (Morawetz et al., 2020) integrating cross-validation of training, validation, and test datasets (Yarkoni and Westfall, 2017).
In sum, many studies examined individual differences in emotion regulation using a standard correlational approach.In future, this could be extended by implementing more complex and flexible analytical allowing for more detailed insights into the complexity of individual differences in emotion perception, generation, and regulation (Cooper et al., 2019;Dubois and Adolphs, 2016).

Conclusions
Understanding individual differences in brain function related to emotion regulation is critical for developing personalised interventions for treating affective disorders.Looking back at more than two decades of imaging research on emotion regulation, the convergence across studies between self-reported regulation success and higher levels of activity in the lateral PFC along with lower levels of activity in the amygdala points towards reliable differences between people.Our descriptive integration of findingsalso considering specific measures and directions of effectswill help to build neurobiological models of emotion regulation that consider individual differences and can be used to understand emotion regulation as a transdiagnostic dimension of psychopathology.To capture inter-individual differences in emotion processing, the critical challenges for future studies are to increase sample sizes of original studies, integrate more advanced statistical approaches, and work towards standardized procedures to enhance reproducibility.

Fig. 1 .
Fig. 1.Procedure of the Systematic Review, Note.Flow chart outlining the study selection process.n[a]=number of analyses; n[f]=number of foci; n[p]=number of participants.

Fig. 2 .
Fig. 2. Methodological Approaches of the Original Studies Included in the Review, Note.Number of fMRI analyses reporting individual difference analyses related to an emotion regulation task with regard to methodological approach: (A) by statistical method and (B) by spatial focus.

Fig. 3 .
Fig. 3. Individual Differences Analysed by the Original Studies Included in the Review, Note.Number of fMRI analyses relating brain activity to specific (A) self-reported regulation success, (B) physiological regulation success, (C) self-reported traits, and (D) demographic variables.

Fig. 4 .
Fig. 4. Effects of Self-Reported Regulation Success and ER-Related Traits, Note.Individual differences in brain activation during emotion regulation illustrated in relation to self-reported regulation success (top row: A & B) and self-reported ER-related traits (bottom row: C & D) separated by analysis approach (left panel: ROI-analysis; right panel: whole-brain analysis).All peak coordinates of brain regions are represented in MNI space and colour-coded for brain area (purple: frontal areas; green: parietal areas; blue: temporal areas; violet: occipital areas; yellow: limbic areas).The sizes of spheres represent the sample sizes of the original studies.Spheres (partly) covered by the transparent illustration of the brain represent coordinates beneath the surface of the brain.LH: left hemisphere; RH: right hemisphere.

Fig. 5 ,
Fig. 5,.Directions of Correlations between Self-Reported Regulation Success and ER-Related Brain Activation, Note.Individual differences in brain activation during emotion regulation in relation to self-reported regulation success.Peak coordinates are represented in MNI space and colour-coded for regulation goal (down-vs.up-regulation) and direction of correlation (positive vs. negative).

Fig. 6 .
Fig. 6.Effects of Physiological Regulation Success and Demographic Variables, Note.Individual differences in brain activation during emotion regulation illustrated in relation to physiological regulation success (top row: A & B) and demographic variables (bottom row: C & D) separated by analysis approach (left panel: ROI-analysis; right panel: whole-brain analysis).All peak coordinates of brain regions are represented in MNI space and colour-coded for brain area (purple: frontal areas; green: parietal areas; blue: temporal areas; violet: occipital areas; yellow: limbic areas).The sizes of spheres represent the sample sizes of the original studies.Spheres (partly) covered by the transparent illustration of the brain represent coordinates beneath the surface of the brain.LH: left hemisphere; RH: right hemisphere.