Introduction
Social anxiety disorder (SAD)1 is characterized by an excessive and persistent fear of social or performance situations in which evaluations from other people may occur (American Psychiatric Association, 2013). SAD is linked to being unmarried and lower education, income, and social support (Furmark, 2002; Heimberg et al., 1990), and once the disorder develops, it tends to follow a chronic course (Wittchen & Fehm, 2003). One factor that increases the risk of developing and maintaining social anxiety is the experience of negative life events (Magee, 1999; Marteinsdottir et al., 2007; Wong & Rapee, 2016). For instance, Marteinsdottir et al. (2007) showed that individuals with social anxiety reported more negative life events during childhood and that interpersonal difficulties during the social anxiety debuting year could be related to the progress of social anxiety. Furthermore, negative life events are expected to increase threat value, which leads individuals to experience more frequent anxiety and higher levels of anxiety in social evaluative situations (Wong & Rapee, 2016). Additionally, the experience of negative life events may intensify maladaptive cognition in individuals with social anxiety (Siegel et al., 2018). Therefore, the experience of negative life events is assumed to significantly affect the development and maintenance of social anxiety.
In contrast to studies that examined the influence of the experience of life events on susceptibility to psychopathology, some emerging studies assume the opposite causal direction. The stress generation model postulates that the occurrence of negative life events is, at least partially, dependent on an individual’s traits, behaviors, and cognitive styles (Hammen, 1991). These events are referred to as dependent events because they occur, at least partially, through an individual’s behavior and characteristics (e.g., fights or conflicts with family members). In contrast, independent events occur randomly and are uncontrollable (e.g., parents divorcing). Socially anxious individuals are more likely to experience significant anxiety and fear during interactions with others (Clark & Wells, 1995), this characteristic may generate negative life events by creating problematic regulatory patterns (Kashdan et al., 2011). A small body of research offers preliminary support for stress generation in social anxiety. Undergraduate students with higher social anxiety exhibited greater reliance on positive emotion suppression and reported a greater number of negative social events and fewer positive emotions on the subsequent day over a 14-day period (Farmer & Kashdan, 2012). Furthermore, adults with SAD experience more frequent stressful social events than psychologically healthy adults and display greater stress sensitivity, particularly in negative emotional reactions to stressful social events (Farmer & Kashdan, 2015). In a survey study, undergraduate students with higher social anxiety symptoms were significant predictors of negative interpersonal dependent events after accounting for depressive symptoms in the past 6 weeks compared to those with lower social anxiety, and levels of interpersonal distress (perceived burdensomeness) moderated the association between social anxiety symptoms and negative interpersonal dependent events (Siegel et al., 2018). Furthermore, adults with more severe social anxiety symptoms and SAD report more dependent stressful life events than those with lower social anxiety (Goodman et al., 2023). In conjunction, people with social anxiety may be more likely to experience negative interpersonal dependent events.
Although people with social anxiety may be more likely to experience negative interpersonal events, the individual factors that influence the experience of negative interpersonal dependent events have not been adequately examined. Modifiable cognitive and behavioral factors that contribute to the generation of stressors may be useful targets for intervention (Santee et al., 2023). Therefore, it may be important to examine the cognitive and behavioral factors of social anxiety that influence the generation of interpersonal dependent events. Although prior studies have examined emotion regulation strategies, stress sensitivity, and interpersonal distress as factors that may influence the experience of negative interpersonal events (Farmer & Kashdan, 2012, 2015; Siegel et al., 2018), these studies are cross-sectional; thus, causal relationships lack clarity. Furthermore, only Siegel et al. (2018) and Goodman et al. (2023) distinguished independent and dependent events. Therefore, to infer the process of generating negative interpersonal dependent events in social anxiety, it is necessary to distinguish between dependent and independent events, and longitudinally examine the association between cognitive and behavioral factors and negative interpersonal dependent events.
This study focused on PEP, a cognitive factor specific to social anxiety that involves repetitive thinking about social situations after leaving or escaping them (Clark & Wells, 1995), to examine the stress generation model in social anxiety. Rumination, a concept similar to PEP, may lead to interpersonal stress generation (e.g., Flynn et al., 2010). Rumination may play a role in initiating and maintaining negative interpersonal processes that lead to negative interpersonal dependent events. For instance, ruminators generate less effective solutions to interpersonal problems and inhibit instrumental behavior by increasing uncertainty (Lyubomirsky & Nolen-Hoeksema, 1995; Ward et al., 2003; Watkins & Baracaia, 2002). Furthermore, rumination is associated with maladaptive interpersonal behaviors, such as excessive reassurance behavior and poor perception of social support (Nolen-Hoeksema & Davis, 1999; Weinstock & Whisman, 2007). Thus, cognitions and behaviors associated with rumination may ultimately lead to interpersonal tension and conflict. Furthermore, PEP often focuses on feelings of anxiety and negative self-perceptions, causing the person to recall interactions as more negative than they actually were (Hofmann, 2007). In fact, PEP contributes to the tendency to recall negative self-related information and affect biased interpretations (Makkar & Grisham, 2013; Mellings & Alden, 2000). Additionally, in depression, repetitive thoughts after events often focus on one’s internal (depressive) symptoms (Nolen-Hoeksema, 1991), whereas in social anxiety, repetitive thoughts after events center around self-appraisals and external evaluations of social partners and other details involving social events (Brozovich & Heimberg, 2008). Thus, PEP may predict the generation of negative interpersonal dependent events by facilitating the development of negative self-related information in social situations and maladaptive interpersonal relationships. This study aimed to longitudinally examine whether PEP predicted the experience of negative interpersonal dependent events. We used a two-timepoint (T1 and T2) cross-lagged panel model to control for the effects of baseline social anxiety symptoms, depressive symptoms, dependent events (interpersonal and non-interpersonal), and independent events. We hypothesized that PEP at T1 would predict the experience of negative interpersonal dependent events at T2, even after controlling for the effects of negative interpersonal dependent events, social anxiety symptoms, depressive symptoms, negative non-interpersonal dependent events, and negative independent events at T1.
Methods
This study was pre-registered. For more information and details on the study, see: https://osf.io/3anmk.
Participants
Participants were recruited via Rakuten Insight Inc., an online market research company with approximately 2.2 million Japanese enrollments that has been utilized for academic purposes. Rakuten Insight Inc. has established a quality control statement and ensures quality control for registrants. Specifically, Rakuten Insight Inc. regularly checks the response behavior of registrants, screens new registrants using bot countermeasures and other technologies, and regularly excludes registrants who do not meet the standards due to reasons such as fraudulent responses. Participants were deemed eligible if they (a) belonged to a junior college, university, or graduate school and (b) committed to answering the survey seriously (Masuda et al., 2019) before participating. Initially, 800 students (initial sample; 508 women, 292 men, mean age = 21.8, SD = 2.2 years) participated in the study via the Internet and completed questionnaires at T1. The final sample consisted of 500 participants (303 women, 197 men, mean age = 21.9, SD = 2.2 years). This sample size provided sufficient power to identify the anticipated large effect size found in prior studies (Flynn et al., 2010; Shapero et al., 2013; Tuna, 2020). The participants were given Rakuten Super Points™ (the actual number of points was not disclosed owing to company policy). Informed consent was obtained from all the participants. This study was approved by the Research Ethics Committee of Graduate School of Education, Tohoku University (no. 23-1-035), and was conducted in accordance with the Declaration of Helsinki.
Measurements
The Trait version of the Post-Event Processing Inventory scale.
The Trait Post-Event Processing Inventory (PEPI-T; Blackie & Kocovski, 2017) was used to assess PEP. The PEPI-T is a self-report measure consisting of 12 items rated on a 5-point Likert scale that assess the general tendency to engage in post-event thought in anxiety-provoking social situations (range: 12–60). We used the Japanese version of the PEPI-T (Maeda et al., 2022). The PEPI-T has demonstrated good psychometric properties; Cronbach’s α was 0.94 and 0.95 at T1 and T2, respectively.
Negative Independent/Dependent Events Scale
The Negative Independent/Dependent Events Scale (Hasegawa et al., 2023) was used to assess experiences of negative interpersonal dependent events, negative non-interpersonal dependent events, and negative independent events. The Negative Independent/Dependent Events Scale is a self-report measure that contains 25 items on the negative interpersonal dependent events (NIDE) subscale, 14 items on the negative non-interpersonal dependent events (NNIDE) subscale, and 20 items on the negative independent events (NIE) subscale, rated on a 4-point Likert scale to assess each event experienced in the past 8 weeks. In this study, Item 26 of the NNIDE, “I was arrested for breaking the law including traffic violations and unpaid rides.” was excluded from the survey because it was likely to lead to an invasion of privacy. The Negative Independent/Dependent Events Scale has demonstrated good psychometric properties; Cronbach’s α was 0.96 (NIDE) and 0.91 (NNIDE) at both T1 and T2, and 0.94 and 0.95 at T1 and T2, respectively (NIE).
Social Interaction Anxiety Scale
The Social Interaction Anxiety Scale (SIAS; Mattick & Clarke, 1998) was used to assess the symptoms of social anxiety. The SIAS is a self-report measure consisting of 20 items rated on a 5-point Likert scale to assess anxiety regarding social interaction (range: 0–80). We used the Japanese version of SIAS (Kanai et al., 2004). The SIAS has demonstrated good psychometric properties; Cronbach’s α was 0.93 at both T1 and T2.
Patient Health Questionnaire-8
The Patient Health Questionnaire-8 (PHQ-8; Kroenke et al., 2009) was used to assess depressive symptoms. The PHQ-8 is a self-report measure consisting of eight items rated on a 4-point Likert scale to assess depressive symptoms during the last two weeks (range: 0–24); created by omitted item 9 from the PHQ-9. We used the Japanese version of the PHQ-8 (Muramatsu et al., 2007). The PHQ-8 has demonstrated good psychometric properties; Cronbach’s α was 0.90 and 0.92 at T1 and T2, respectively.
Procedure
The survey was conducted October 2023 to December 2023. An email containing a link to the online questionnaire was sent to individuals randomly by the research company. The participants completed the PEPI-T, Negative Independent/Dependent Events Scale, SIAS, and PHQ-8 at T1 and T2, respectively. The interval between T1 and T2 was two months. Geyer et al. (1994) reported difficulty in recalling events that occurred more than six months previously. Furthermore, prior studies examining the association between life events and mental health have used a 6-week (Riskind et al., 2013) to 2-month (Dentale et al., 2020) survey interval between the two time points and have shown no significant challenges with the recall of life events. Based on these studies, a 2-month survey interval was deemed appropriate for this study.
Statistical Analyses
To examine whether sample selection bias occurred, we conducted a paired-samples t-test to determine whether there were significant differences in each T1 measurement and sex ratio between the initial (only T1 responders) and final samples. For the preliminary analyses, we computed the means, standard deviations, internal consistencies, and correlations for all variables.
Before conducting a cross-lagged analysis, the measurement invariances (configural, metric, scalar, and residual) were examined. Measurement invariance was checked to ensure that the nature of the construct did not change substantially over time (Mackinnon et al., 2022). Because total scores were used in this study, each scale was treated as a single factor.
We conducted a cross-lagged analysis to test the hypotheses. A cross-lagged analysis identifies direct influences in longitudinal data by controlling for within-timepoint correlations and autoregressive effects, making them useful for understanding longitudinal associations and potential bidirectional impacts between variables (Selig & Little, 2012). We tested four competing models (the stability, normal causal, reversed causal, and reciprocal models) to compare causal relations between the variables of interest. In all models, we controlled for the effect of sex on each variable at T1. We obtained the following fit indices for each model (Hu & Bentler, 1999): comparative fit index (should be 0.95 or higher for a good fit), Tucker–Lewis Index (should be 0.95 or higher for a good fit), root-mean-square error of approximation (should be 0.06 or lower), and standard root-mean-square residual (should be 0.08 or lower). We compared the competing nested models using the chi-square difference test to confirm that the reciprocal model fit the data better than the stability, normal causal, and reversed causal models. We then examined whether the path from PEP at T1 to NIDE at T2 was significant. If the model did not fit the data well, we modified the model using a modification index. We added theoretically explainable paths in order of the largest modification index and examined whether the goodness of fit improved. We used the “lavaan” package for R software to examine each model. The significance level was set at p < .05 (two-tailed). All analyses were conducted using R 4.3.2.
Results
Preliminary Analyses
The T1 measurements (ts < 0.64, ps > 0.525) did not differ significantly between the final (N = 500) and initial (N = 300) samples. However, there was a significant difference in the sex ratio between the initial and final samples (χ2 = 4.51, p = .034). Therefore, the selection bias due to attrition was considered minimal. Table 1 presents the means, standard deviations, and correlation coefficients for the scales.
Table 1
Means, standard deviations, and correlation coefficients for the scales
Variables | Means (± SD) | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1. Age | 21.9 (2.2) | — | — | — | — | — | — | — | — | — | — | — | — |
2. T1 PEPI-T | 37.17 (10.87) | 0.07 | — | — | — | — | — | — | — | — | — | — | — |
3. T1 NIDE | 40.93 (15.70) | 0.17*** | 0.28**** | — | — | — | — | — | — | — | — | — | — |
4. T1 NNIDE | 22.70 (8.35) | 0.19**** | 0.30**** | 0.86**** | — | — | — | — | — | — | — | — | — |
5. T1 NIE | 31.84 (11.74) | 0.16*** | 0.22**** | 0.90**** | 0.85**** | — | — | — | — | — | — | — | — |
6. T1 SIAS | 38.77 (16.12) | 0.04 | 0.52**** | 0.30**** | 0.32**** | 0.25**** | — | — | — | — | — | — | — |
7. T1 PHQ-8 | 6.84 (5.68) | 0.14** | 0.48**** | 0.60**** | 0.62**** | 0.57**** | 0.49**** | — | — | — | — | — | — |
8. T2 PEPI-T | 35.96 (11.15) | 0.11* | 0.63**** | 0.24**** | 0.26**** | 0.19**** | 0.43**** | 0.36**** | — | — | — | — | — |
9. T2 NIDE | 41.20 (15.95) | 0.14** | 0.21**** | 0.69**** | 0.63**** | 0.66**** | 0.27**** | 0.51**** | 0.28**** | — | — | — | — |
10. T2 NNIDE | 23.01 (8.50) | 0.15** | 0.25**** | 0.59**** | 0.67**** | 0.60**** | 0.28**** | 0.50**** | 0.34**** | 0.83**** | — | — | — |
11. T2 NIE | 32.09 (12.29) | 0.13** | 0.16*** | 0.65**** | 0.61**** | 0.69**** | 0.23**** | 0.48**** | 0.22**** | 0.92**** | 0.85**** | — | — |
12. T2 SIAS | 38.09 (15.78) | 0.03 | 0.49**** | 0.25**** | 0.27**** | 0.21**** | 0.78**** | 0.42**** | 0.48**** | 0.31**** | 0.34**** | 0.25**** | — |
13. T2 PHQ-8 | 6.61 (5.90) | 0.11* | 0.41**** | 0.52**** | 0.51**** | 0.48**** | 0.44**** | 0.76**** | 0.42**** | 0.57**** | 0.60**** | 0.53**** | 0.49**** |
Hypothesis Testing
Before testing the hypotheses, measurement invariances were tested for each scale. The validation process for each scale is provided in Supplementary material 1. The NIDE displayed configural invariance, SIAS and PEPI-T displayed metric invariance, PHQ-8 displayed partial residual invariance, and NNIDE and NIE displayed scalar invariance. These results indicate that the nature of each scale was generally unchanged between T1 and T2.
The goodness-of-fit indices of the competing models and model comparisons are listed in Table 2. The chi-square difference test showed that the reciprocal model (M4) fit the data well compared to the stability model (M1; Δχ2 (30) = 103.18, p < .001), the normal causal model (M2; Δχ2 = 37.22 (15), p = .001), and the reversed causal model (M3; Δχ2 = 45.43 (15), p < .001). Therefore, the reciprocal model (M4) was accepted as the final model, which modeled the temporal stability, synchronous correlation, and cross-lagged relations of each scale. The results of the path model are shown in Fig. 1. PEP-T at T1 was not significantly predictive of NIDE at T2 (β = −0.04, p = .266). Furthermore, SIAS at T1 was not significantly predictive of NIDE at T2 (β = 0.04, p = .256). PHQ-8 at T1 was significantly predictive of an increase in NIDE at T2 (β = 0.14, p = .002), as well as increases in NNIDE (β = 0.12, p = .014) and NIE (β = 0.13, p = .004).
Table 2
Goodness-of-fit indices and model comparisons
Model | χ2 (df) | p | RMSEA | CFI | TLI | SRMR | Comparison | Δχ2 (Δdf) | p |
|---|---|---|---|---|---|---|---|---|---|
Stability (M1) | 122.29 (36) | < 0.001 | 0.069 | 0.985 | 0.968 | 0.086 | |||
Normal (M2) | 56.33 (21) | < 0.001 | 0.058 | 0.994 | 0.977 | 0.034 | M1 vs. M2 | 65.96 (15) | < 0.001 |
Reversed (M3) | 64.53 (21) | < 0.001 | 0.064 | 0.992 | 0.972 | 0.035 | M1 vs. M3 | 57.75 (15) | < 0.001 |
Reciprocal (M4) | 19.11 (6) | 0.004 | 0.066 | 0.998 | 0.971 | 0.013 | M1 vs. M4 | 103.18 (30) | < 0.001 |
M2 vs. M4 | 37.22 (15) | 0.001 | |||||||
M3 vs. M4 | 45.43 (15) | < 0.001 |
Fig. 1
Standardized path coefficients for the reciprocal model. Non-significant paths, error variables, and covariances are omitted. PEPI-T: Trait Post-Event Processing Inventory; NIDE: Negative interpersonal dependent events; NNIDE: Negative non-interpersonal dependent events; NIE: Negative independent events; SIAS: Social Interaction Anxiety Scale; PHQ-8: Patient Health Questionnaire-8. ***p < .001, **p < .01, *p < .05
×
![]()
Discussion
This longitudinal study examined whether PEP predicted the experience of negative interpersonal dependent events. Contrary to our hypothesis, the PEPI-T score at T1 did not significantly predict NNID at T2. This finding suggests that PEP may not predict negative interpersonal dependent events in social anxiety. One possible explanation is that PEP may have affected the avoidance of negative interpersonal-dependent events. The present study assumed that engaging in PEP would increase the experience of negative interpersonal-dependent events as rumination promotes negative interpersonal styles (e.g., Flynn et al., 2010). Furthermore, PEP contributes to the tendency to recall negative self-related information and affects biased interpretations (Makkar & Grisham, 2013; Mellings & Alden, 2000). However, Flynn et al. (2010) suggested that chronic experiences of negative cognition and emotions associated with rumination may lead to interpersonal isolation. Furthermore, Rowa et al. (2016) suggested that negative memories distorted by PEP may lead to the avoidance of social situations, as they showed that the more PEP participants experienced after giving their speech, the less willing they were to complete another similar task. Thus, if social situation avoidance occurred as a result of engaging in PEP, the effect of PEP on the experience of negative interpersonal dependent events may have been less likely.
In this study, social anxiety symptoms did not predict the experience of negative interpersonal dependent events, while depressive symptoms predicted the experience of all negative events. These findings suggest that social anxiety symptoms are not directly related to stress generation. Our results are consistent with a finding of Goodman et al. (2023) that social anxiety symptoms did not significantly predict the experience of dependent events to a greater degree than depression. The possibility that social anxiety symptoms do not directly predict stress generation has been supported by existing research. For instance, anxiety has been suggested to promote avoidance and withdrawal from potentially stressful situations (Meyer & Curry, 2017). Farmer and Kashdan (2015) showed that, following days of negative emotions, the control group experienced an increase in negative social events, but the socially anxious group did not. Furthermore, Rnic et al. (2023) showed that depression exhibits larger effects on both dependent interpersonal and non-interpersonal stress generation, suggesting that depression is associated with more pronounced stress generation. Conway et al. (2012) found that, after accounting for the influence of the internalizing spectrum, major depressive disorder predicted dependent interpersonal stress. Depression-specific vulnerability factors (e.g., excessive reassurance seeking, negative feedback seeking, hopelessness, and co-rumination) may promote interpersonal conflicts, which may explain the significant effect of depression (Conway et al., 2012). While some of these vulnerability factors have also been suggested as predictors of stress generation in anxiety (e.g., excessive reassurance seeking: Meyer & Curry, 2017), many have been theorized to be particularly central to depression (e.g., negative feedback seeking; Joiner, 1995). Furthermore, Rnic et al. (2023) suggested that stress generation may appear more transdiagnostic because of the ubiquity of comorbid depression, as greater associations were observed for internalization, externalization, depression, and anxiety with dependent stress than with independent stress. Because the present study controlled for the effects of baseline depressive symptoms, we examined the effects of social anxiety symptoms after excluding the effects of depression on the generation of dependent events. Our findings suggest that social anxiety symptoms alone cannot predict stress generation.
In the present study, depressive symptoms significantly predicted experiences of independent and dependent events as well as dependent events. Prior studies have shown that the somatic factors of the Beck Depression Inventory significantly predict independent events (Harkness & Stewart, 2009), and there is a small but significant relationship between psychopathology and independent events (Rnic et al., 2023). Harkness and Stewart (2009) suggested that a depressive context may increase the risk of exposure to environments in which particular events are more likely to occur. Hammen et al. (2011) suggested that individuals with psychopathology are more likely to actively drive themselves into challenging contexts featuring more independent stressors. Future studies should focus on the context of the individual’s environment, considering the impact of depressive symptoms on the occurrence of independent and dependent events.
This study had several limitations. First, participants were not screened for social anxiety and depression and the clinical group was not included, which may have led to heterogeneity in the effect of PEP on stress generation. Future studies should control for diagnoses and comorbidities among participants to more accurately examine the effects of PEP on stress generation. The clinical group experienced more dependent events than the control group (Goodman et al., 2023) and comorbidities may influence stress generation (Rnic et al., 2023). Hence, future studies need to check for the presence of other diagnoses and comorbidities. Second, we used a self-report scale to measure negative events. However, as self-reported measures may be influenced by cognitive bias (Krackow & Rudolph, 2008), the possible influence of participant bias cannot be ignored. In the future, it will be necessary to verify whether similar results can be obtained using interviews that are less prone to bias. Third, we did not examine the subjective experience of each negative event. Goodman et al. (2023) examined the subjective experience of events and found no difference in the impact of dependent and independent events in the clinical group. That is, the clinical group may feel averse to stressful events regardless of whether they are dependent or independent. In addition, differences in individuals’ subjective experiences of stressful life events have consequences on how they cope and respond to stress (Goodman et al., 2023). Future studies should examine the presence or absence of negative events, as well as the subjective experience, to clarify the influence of negative events on the onset and maintenance of social anxiety. Fourth, our study did not measure whether participants avoided the experience of negative life events. However, considering the possibility that engaging in PEP may lead to the avoidance of the experience of events, it is better to measure avoidance behavior in future studies. Finally, there are some limitations in the survey design and analysis of our study. The relatively short period of two months may not have been enough to detect the effects of PEP on stress generation, and future studies should design a longer-term survey. Furthermore, our study used cross-lagged analysis to examine stress generation. Several studies have used cross-lagged analysis to examine stress generation (Allen et al., 2020; Skobic et al., 2024). However, it has been indicated that cross-lagged analysis is only an inference of causality, and there are statistical limitations to conducting a cross-lag analysis at two time points, as in our study. Moreover, cross-lagged analysis has been criticized for its inability to distinguish between within-person and between-person effects. However, incorporating random intercepts can address this issue, although it requires more than three timepoints (Hamaker et al., 2015). Future studies that take measurements at multiple timepoints will enable analyses that are more suitable for inferring causality.
Despite these limitations, this study is significant because it longitudinally examined the stress generation model in the context of social anxiety. Our results suggest that PEP and social anxiety symptoms may not be predictors of stress generation in social anxiety. Future studies should examine the role of comorbid depression in stress generation. Furthermore, by examining the factors specific to depression, we may be able to propose targets for interventions aimed at stress generation.
Acknowledgements
We would like to thank Editage (www.editage.jp) for English language editing.
Declarations
Ethical Approval
This study was approved by the Ethics Committees of Tohoku University (no. 23-1-035), and was conducted in accordance with the Declaration of Helsinki.
Research Involving Human and Animal Participants
No animal studies were carried out by the authors for this article.
Informed Consent
Informed consent was obtained from all the participants.
Competing Interests
None.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.