The deleterious effects of air pollution on physical disease have been examined extensively in existing literature. For instance, air pollution is associated with chronic obstructive pulmonary disease (Arbex et al., 2009), higher blood pressure (Baumgartner et al., 2014), cardiovascular and cerebrovascular disease (Zhang et al., 2011) and lung cancer (Pope et al., 2011). More strikingly, using a quasi-experimental empirical approach, Chen et al. (2013) found that ambient concentrations of total suspended particulates are about 55% higher in northern China due to winter heating; as a result, incidence of cardiorespiratory mortality is significantly higher in the north leading to 5.5 years lower of life expectancies for northern citizens. This is in accordance with the finding by Zhang et al. (2011) that long-term exposure to ambient air pollution is associated with mortality due to cardiovascular and cerebrovascular diseases. These authors drew the conclusions by taking into account the time course of air pollution, cardiovascular as well as cerebrovascular diseases and mortality. Considering that smog in China is an extremely severe case of air pollution with PM exceeding 10 times of the number proposed by air quality guidelines, it is not surprising that some of the aforementioned findings are based on studies using Chinese database and samples.

In spite of the great attention paid to air pollution and its influences on physical health, the impact of air pollution on psychological health is less frequently investigated thereby remaining unclear. To date, some studies in western societies have suggested that toxic air pollution leads to higher level of psychopathological symptoms especially externalizing psychopathology (for a review, see Bell, 1993). It is very important to keep in mind, nevertheless, that most of the findings cited in the review relied on correlational analysis using epidemiological data in several consecutive years, making it impossible to determine whether air pollution is indeed causally linked to psychopathology. For instance, it is probable that a third variable, such as increased unemployment rate due to the promotion of large-scale machine industry, is playing a role. Moreover, the nature of air pollution is not necessarily the same in China and western countries given that the major pollutants in China are PM. As such, it seems critically important to test the causal relationship between air pollution and psychopathology in China.

In lay options, severe air pollution in China is thought to endanger not only physical but also psychological health. Nevertheless, relatively few empirical studies examine such an association. One notable exception, a longitudinal nationwide study, found that air pollution reduced hedonic happiness and increased the rate of depressive symptoms (Zhang et al., 2017). This study provides us with the first empirical evidence linking air pollution in China to depression, yet there are still at least two questions that remain unclear.

First, whether air pollution is uniquely associated with depression or a broader range of symptomologies is still unknown. It is possible that acute exposure to air pollution would lead to an increase in not only depressive symptoms but also other symptomologies such as anxiety and obsessive-compulsive symptoms. Indeed, it is possible that the influence of air pollution is not limited to depression considering the large amount of comorbidity between depression and other disorders (Kessler et al., 2005). However, given the discrepancies over the cause of psychopathology, symptoms induced by smog might not necessarily share the same mechanisms as those examined in previous literature. As such, it would be beneficial to assess a wide range of psychopathological symptoms to see if pollution induces a wide range of psychopathology or just depression.

Second, living in smog is a significant day-to-day stressor evoking negative emotions (Zhang et al., 2017). Under such circumstances, effective self-regulation of affect might be critical to cope with such a stressor and especially important for individuals faced with severe air pollution. Indeed, the experience of negative events plays a critical role in the onset of psychopathology especially depression (Brown and Harris, 2012), but only a minority of those who experience negative life events would develop a full-blown depressive episode. It has been suggested that self-regulation of emotion might play an important role in disordered affect following negative life events (for a review, see Joormann and Stanton, 2016). However, to date no study has examined whether individual differences in self-regulation of affect would be a factor influencing the development of pollution-induced psychopathology. Why some people's affect recovers easily even when exposed to smog whereas other people struggle in the same situation is a fascinating question with important implications for the onset and possible prevention of not only pollution-induced depression but also other forms of psychopathology. Therefore, it would be informative to examine the link between use frequencies of emotion regulation strategies and psychopathological symptoms following acute exposure to air pollution.

Given the critical role of emotion regulation in the etiology of psychopathology (Aldao and Nolen-Hoeksema, 2012; Joormann and Stanton, 2016), it is likely that more frequent use of adaptive strategies and less frequent use of maladaptive strategies is linked to less psychopathological symptoms following exposure to smog.

The two most frequently examined strategies, cognitive reappraisal and expressive suppression, have been shown as transdiagnostic factors for psychopathology (Aldao et al., 2010; Webb et al., 2012). Reappraisal refers to changing a situation's meaning so as to alter one's emotional response to the situation (Gross and John, 2003). Several meta-analysis articles suggest a beneficial effect of reappraisal on mental health, but it should be noted that the effect size is relatively small (Aldao et al., 2010; Webb et al., 2012). Moreover, other authors have found that reappraisal can be ineffective when used in the wrong context (Troy et al., 2013). These authors showed that reappraisal is adaptive when using in face with uncontrollable stressful life events while maladaptive when using in relatively controllable stressors. Similarly, using both self-report and experimental assessment of reappraisal, Troy et al. (2017) found that effectively regulating one's emotions using cognitive reappraisal could be more beneficial in lower vs. social economic status contexts. These authors further discussed this finding by suggesting that people with lower social economic status might have less resources to control their surroundings thereby leaving them no option but to self-regulate. Being exposed to air pollution is equal and almost unchangeable for people living in cities with air pollution. Considering this relatively uncontrollable nature of severe air pollution, it is highly possible that higher use frequency of reappraisal would buffer the undesirable impact of air pollution on psychopathology.

Suppression, on the other hand, involves attempts to inhibit the effects of external cues on internal and external states of emotion (Gross and John, 2003). Whereas some researchers suggest that it is maladaptive (Gross and John, 2003), other researchers find no such link (D'Avanzato et al., 2013), or even beneficial effect of suppression on reducing psychological distress following extreme life events (Dunn et al., 2009). These mixed findings suggest that empirical study is needed if we were to determine the adaptiveness of suppression in face with smog.

On the surface, it seems as if reappraisal and suppression work separately, but in fact they are largely interactive. For instance, Aldao and Nolen-Hoeksema (2012) found that the composite score of the use frequencies of several adaptive strategies is negatively associated with psychopathological symptoms only when the composite score of the use frequencies of several maladaptive strategies is high. The authors further discussed this result and suggested that adaptive strategies play its protective role by functioning as a buffer to the harmful effects of maladaptive strategies. It is important to keep in mind, therefore, reappraisal and suppression might interact to predict pollution-induced symptoms in addition to their independent effects. To be specific, it is hypothesized in the present study that reappraisal is more adaptive when the use frequencies of suppression is high for the individuals.

In the present study, we followed 120 university students at a 1 month interval and measured their symptoms of psychopathology at the two time points. The first time point was at the end of a week that is relatively free of pollution, while the second time point followed a severely polluted week. By such longitudinal design, we hope to compare symptoms across these two conditions so as to determine if acute exposure to smog does increase psychopathological symptoms. In addition, habitual use of reappraisal and suppression, which have been identified as stable traits over time (Gross and John, 2003), is assessed during the baseline assessment so that we could test the possible effects of emotion regulation on pollution-induced symptoms.

One hundred and twenty undergraduate students completed online questionnaires at two time points for course credits. They were recruited from a university in Beijing. Eighteen participants only participated in the study at Time 1, which left 102 participants for the final analysis. The participants' age ranged from 17 to 22 (M = 18.03, SD = 1.03), and 58.8% of the sample were females. At Time 1, participants provided demographical information and completed instruments measuring psychopathological symptoms as well as trait emotion regulation. One month later (Time 2), psychopathological symptoms were reassessed. The Institutional Review Board of Tsinghua University approved the research protocol, and prior completing the Time 1 assessment all participants signed informed consent. After the Time 2 assessment, participants were debriefed.

The week prior to the Time 1 assessment was relatively free of air pollution, with a mean PM_2.5 of 16.1 μg/m³ (average by hour). In contrast, the week prior to the Time 2 assessment was moderately to severely polluted, with a mean PM_2.5 of 173.5 μg/m³, which was significantly higher than that of the week before Time 1 [t₍₂₃₈₎ = 17.61, p < 0.001]. The temperature at Time 1 was similar to that at Time 2, both having an average temperature of 12°C. Moreover, the measures at Time 1 took place on a cloudy day. In this way, we managed to control both the temperature and light, which have been found to be influential factors for seasonal affective disorders (SAD; Miller, 2005).

Psychopathology was assessed using the Symptom Check List-90 (SCL-90; Derogatis and Cleary, 1977). The Chinese version of the SCL-90 we used was translated and published in China (Wang, 1984). The SCL-90 is comprised of 90 items, each measured on a 5-point scale of distress from “not at all” (0) to “extremely” (4). Current psychopathology is reflected in terms of nine primary symptom dimensions: somatization (SOM), obsessive-compulsive (O-C), interpersonal sensitivity (INS), depression (DEP), anxiety (ANX), hostility (HOS), phobic anxiety (PHO), paranoid ideation (PAR) and psychoticism (PSY). Moreover, scores were averaged among positive symptom dimensions to reflect the level of positive symptom distress (PSD). In this study, all subscales of SCL-90 were internally consistent (all α.s. > 0.70 in the present study).

Additionally, we used the Emotion Regulation Questionnaire (ERQ; Gross and John, 2003) to assess the frequency with which people use cognitive reappraisal or expressive suppression to regulate affect. The ERQ consists of two subscales that measure the habitual use of reappraisal and suppression using a 7-point likert scale. The reappraisal subscale (e.g., I control my emotions by changing the way I think about the situation I'm in) consists of 6 items (α = 0.85 in the present study) and the suppression subscale (e.g., I keep my emotion to myself) consists of 4 items (α = 0.80 in the present study). The Chinese revised version of the scale was utilized in the current study (Wang et al., 2007).

Within-group comparisons showed that participants reported significantly higher levels of obsessive-compulsive symptoms on the day with air pollution (M = 2.00, SD = 0.60), compared with the day without air pollution (M = 1.87, SD = 0.60), t₍₁₀₁₎ = 3.11, p < 0.01, d = 0.123, 95% CI [0.04, 0.20]. Participants also reported significantly higher levels of interpersonal sensitivity on the day with air pollution (M = 1.83, SD = 0.68), compared with the day without air pollution (M = 1.68, SD = 0.69), t₍₁₀₁₎ = 3.20, p < 0.01, d = 0.144, 95% CI [0.05, 0.23]. In addition, the depression level of participants on the day with air pollution (M = 1.69, SD = 0.66) was significantly higher than on the day without air pollution (M = 1.62, SD = 0.61), t₍₁₀₁₎ = 2.04, p < 0.05, d = 0.076, 95% CI [0.00, 0.15]. They also reported significantly higher levels of psychoticism on the day with air pollution (M = 1.49, SD = 0.46), compared with the day without air pollution (M = 1.38, SD = 0.40), t₍₁₀₁₎ = 3.49, p < 0.01, d = 0.112, 95% CI [0.05, 0.18]. The PSD level on the day with air pollution was also significantly higher (M = 2.50, SD = 0.45), compared with the day without air pollution (M = 2.43, SD = 0.44), t₍₁₀₁₎ = 2.93, p < 0.01, d = 0.079, 95% CI [0.03, 0.13]. Together, the results indicate that air pollution could influence individuals' psychopathological symptoms in multiple ways (see Table 1).

We generated 10 items of change scores from SCL-90 scales by subtracting the scores on the day without air pollution from those on the day with pollution. We then executed zero-order correlations between these items (SOM change score, O-C change score, INS change score, DEP change score, anxiety change score, HOS change score, PHO change score, PAR change score, PSY change score, and PSD level change score) and another two items generated from ERQ (reappraisal and suppression) (see Table 2). The PSD level change score (r = −0.23, p < 0.05), the SOM change score (r = −0.38, p < 0.01), the INS change score (r = −0.29, p < 0.01), the DEP change score (r = −0.21, p < 0.05), the ANX change score (r = −0.28, p < 0.01) and the HOS change score (r = −0.22, p < 0.05) were negatively correlated with the suppression strategy score. However, none of the change scores was correlated with the reappraisal strategy score.

We predicted the PSD level change score using a hierarchical linear regression (see Table 3). In Step 1, we included gender and age variables to control gender/age-related differences. The model was insignificant, ΔR² = 0.02, ΔF_{(2, 99)} = 2.20, ns. In Step 2, we entered suppression and reappraisal strategies scores (which were centered to reduce multicollinearity; Tabachnick and Fidell, 2007). However, they did not significantly improve the model, ΔR² = 0.06, ΔF_{(2, 97)} = 2.96, ns. The suppression strategy score contributed significantly to the prediction of the PSD level change score, t₍₉₇₎ = −2.27, p < 0.05, β = −0.23. The reappraisal strategy score was not a significant predictor, t₍₉₇₎ = −0.55, ns, β = −0.05. In Step 3, we added the interaction term between the reappraisal and suppression strategies, ΔR² = 0.13, ΔF_{(1, 96)} = 8.88, p < 0.01. The suppression strategy was no longer a significant predictor, t₍₉₆₎ = −1.49, ns, β = −0.15, and the reappraisal strategy remained an insignificant predictor, t₍₉₆₎ = −0.93, ns, β = −0.09. However, the interaction term between the reappraisal and suppression strategies was significant, t₍₉₆₎ = −2.97, p < 0.01, β = −0.29. We further investigated the change scores of SOM, INS, DEP, ANX, and HOS using the same method. Except for the SOM change scores, similar results were found in the final model. For INS change scores [R² = 0.195, F_{(5, 96)} = 4.66, p < 0.01], reappraisal [t₍₉₆₎ = −0.70, ns, β = −0.07] and suppression [t₍₉₆₎ = −1.95, ns, β = −0.19] were insignificant while the interaction term was marginally significant, t₍₉₆₎ = −1.79, p = 0.08, β = −0.17. For DEP change scores [R² = 0.095, F_{(5, 96)} = 2.02, ns], reappraisal [t₍₉₆₎ = −0.17, ns, β = −0.02] and suppression [t₍₉₆₎ = −1.49, ns, β = −0.16] were insignificant while the interaction term was marginally significant, t₍₉₆₎ = −1.94, p = 0.06, β = −0.20. For ANX change scores [R² = 0.134, F_{(5, 96)} = 2.98, p < 0.05], reappraisal [t₍₉₆₎ = −0.76, ns, β = −0.07] and suppression [t₍₉₆₎ = −0.74, ns, β = −0.08] were insignificant while the interaction term was significant, t₍₉₆₎ = −2.68, p < 0.01, β = −0.27. For HOS change scores [R² = 0.074, F_{(5, 96)} = 1.54, ns], reappraisal [t₍₉₆₎ = −0.54, ns, β = −0.05] and suppression [t₍₉₆₎ = 0.44, ns, β = 0.05] were insignificant while the interaction term was marginally significant, t₍₉₆₎ = −1.96, p = 0.05, β = −0.20.

We followed any significant or marginally significant interaction with simple slopes analyses (Aiken and West, 1991). We wanted to examine the predictions about the relationship between the reappraisal strategy and the PSD level change score, the INS change score, the DEP change score, the ANX change score, and the HOS change score in the presence of high vs. low levels of suppression strategy. Thus, we evaluated the role of the reappraisal strategy as a predictor of these five change scores at high and low levels of suppression strategy (i.e., 1 SD above/below the mean). At low levels of suppression strategy, the reappraisal strategy was unrelated to the PSD level change score [t₍₉₈₎ = 1.64, ns, β = 0.192], the INS change score [t₍₉₈₎ = 0.88, ns, β = 0.104], the ANX change score [t₍₉₈₎ = 1.61, ns, β = 0.193], and the HOS change score [t₍₉₈₎ = 1.19, ns, β = 0.147]. However, at high levels of suppression strategy, greater levels of reappraisal strategy was related or marginally related to lower levels of the PSD level change score [t₍₉₈₎ = −2.64, p < 0.01, β = −0.353], the INS change score [t₍₉₈₎ = −1.73, p = 0.09, β = −0.235], the ANX change score [t₍₉₈₎ = −2.34, p < 0.05, β = −0.320], and the HOS change score [t₍₉₈₎ = −1.73, p = 0.09, β = −0.244]. There was no significant relation between the reappraisal strategy and the DEP change score. As an example, Figure 1 showed the interaction between reappraisal and suppression strategies in predicting change scores of PSD level.

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.