Stressor Anticipation and Subsequent Affective Well-Being: A Link Potentially Explained by Perseverative Cognitions

Abstract

Anticipatory stress can prospectively and negatively influence diverse outcomes, including cognitive performance and emotional well-being. It has been suggested that perseverative cognitions (e.g., worry, rumination) during the anticipation period constitute a key mechanism driving these effects. The present study investigated the temporal dynamics among stressor anticipation, perseverative cognitions and affective well-being. To accurately test the suggested mechanism, we focused on how these dynamics unfold within individuals over time. To that end, we analyzed data from an ecological momentary assessment study in an ethnically diverse sample (N = 243, 25-65 year olds, 68.7% Hispanic or non-Hispanic Black; 14 days, 5 measurement occasions per day) using dynamic structural equation modeling. Anticipating an upcoming stressor was linked to higher levels of perseverative cognitions approximately three hours later. At times when individuals reported higher levels of recent perseverative cognitions than typical for them, they also reported higher levels of negative affect and lower levels of positive affect. Mediational modeling indicated that perseverative cognitions accounted for the persistent effects of previous stressor anticipation on negative as well as positive affect several hours later. These findings suggest that perseverative cognitions may play an important role in explaining the detrimental effects of stress anticipation on subsequent emotional well-being.

Research has consistently shown that daily stressors are not only related to health symptoms (Almeida, Wethington, & Kessler, 2002; Neupert, Almeida, & Charles, 2007) but also to impaired cognitive performance (Neupert, Almeida, Mroczek, & Spiro, 2006; Rickenbach, Almeida, Seeman, & Lachman, 2014; Sliwinski, Smyth, Hofer, & Stawski, 2006) as well as negative mood (Bolger, DeLongis, Kessler, & Schilling, 1989; Mroczek & Almeida, 2004; Suls, Green, & Hillis, 1998; van Eck, Nicolson, & Berkhof, 1998). However, not only events that have occurred or that are currently happening have an impact on people’s thoughts, feelings, and experience, but also things that (might) happen in the future. For example, pressing deadlines at work, upcoming examinations, important meetings or appointments, or simply being on time despite rush-hour traffic can also leave their marks. Previous research has suggested that the anticipation of a stressful event is associated with higher subsequent negative affect, even when no stressor has (yet) occurred (Neubauer, Smyth, & Sliwinski, 2018; Scott, Kim, Smyth, Almeida, & Sliwinski, 2019). Theoretical considerations have attributed this effect to perseverative cognitions, which are assumed to temporally extend the stress response to both before onset and after offset of the stressor (Brosschot, Gerin, & Thayer, 2006). The temporal extension of the stress response is a process that should unfold within an individual over time. However, this process has not yet been empirically tested in individuals’ daily lives. The present work fills this gap and investigates the persistent effects of anticipatory stress on affective well-being and specifically targets the mediating mechanism of this effect – the temporal extension of the stress response via perseverative cognitions.

Effects of Anticipatory Stress

Consistent with the account of Neupert, Neubauer, Scott, Hyun, and Sliwinski (2019), we refer to the expectation of a future stressful event as stressor forecasting or stressor anticipation. This should be distinguished from predictions about future states of stressfulness (i.e., feeling stressed), which are termed stress anticipation. Both concepts can be summarized under the term anticipatory stress. Growing interest in the time frame prior to stressor occurrence has stimulated research exploring the effects of anticipatory stress on physiological, cognitive, and emotional outcomes.

This research has shown that effects of stressor anticipation are sometimes similar to those of experienced stressors. On the physiological level, various studies found evidence for heightened arousal, such as activation of the hypothalamic-pituitary-adrenal axis, increased heart rate, or elevated blood pressure following the expectation of a stressful event (e.g., Ditzen et al., 2007; Engert et al., 2013; Kirschbaum, Pirke, & Hellhammer, 1993; Spacapan & Cohen, 1983). For example, during the anticipation phase of the Trier Social Stress Test (TSST; Kirschbaum et al., 1993) during which participants expect to give a speech in front of several judges, participants showed increases in heart rate (Ditzen et al., 2007) as well as an activation of the HPA-axis (Engert et al., 2013; Kirschbaum et al., 1993). In addition, Spacapen and Cohen (1983) reported increased levels of blood pressure in participants who expected to immerse their hand in ice water. Investigating the effects of stressor anticipation in daily life and thus in a more ecologically valid setting, Smyth et al. (1998) found that the expectation of a stressor was linked to higher levels of cortisol in saliva. Taken together, these results illustrate that effects of stressor anticipation correspond to effects of experienced stress on the physiological level. However, when discussing these findings, it is important to acknowledge the complexity of (anticipatory) stress processes, especially in the context of the TSST (Kirschbaum et al., 1993). Clearly distinguishing the start and end of the anticipation from the complex array of stressor characteristics or components is often difficult.

Stressor anticipation influences not only physiological parameters but also psychological outcomes such as cognition or emotion. For example, in the study by Spacapen and Cohen (1983) participants anticipating a stressor demonstrated lower levels of frustration tolerance than individuals who did not expect a stressor. Further, associations between stressor anticipation and mood correspond to those of experienced stress and mood. When individuals prospectively forecasted a stressful event, they also reported elevated levels of concurrent negative affect (Neubauer et al., 2018; Smyth et al., 1998). Moreover, these effects seem to last for some time: Anticipating a stressful event was also linked to higher levels of negative affect a few hours later (Neubauer et al., 2018). Both studies (Neubauer et al., 2018; Smyth et al., 1998) examined associations in daily life using ambulatory assessment. Ambulatory assessment refers to methods such as ecological momentary assessment (EMA) or experience sampling that enable psychological, behavioral and/or physiological data collection in real life and in (or at least close to) real time (e.g., Mehl & Conner, 2012), thus allowing for investigating processes in natural contexts (e.g., Smyth, Juth, Ma, & Sliwinski, 2017).

The persistent effects of anticipatory stress also extend to the realm of cognition. In an EMA study, Hyun, Sliwinski, and Smyth (2019)¹ asked participants in the morning how stressful they expected the upcoming day to be and measured their working memory performance later that day. They found that the anticipation of a stressful day was linked to lower working memory performance across the same day – even after controlling for the occurrence of stressors later on this same day. These results indicate that stressor anticipation produces stress regardless of – or in some cases above and beyond – whether or not a stress(or) occurred (see Smyth, Zawadzki, & Gerin, 2013). Altogether, these findings highlight that merely anticipating a future stressful event or state can impact our body as well as our cognitive performance and our feelings. In addition, anticipatory stress has both immediate and more persistent effects on mood and cognition. Consequently, these findings raise the question how these lagged effects might arise.

Perseverative Cognitions as a Potential Mechanism Underlying Persistent Effects

According to the perseverative cognition hypothesis (Brosschot et al., 2006) perseverative thoughts constitute a key mechanism explaining the persistent effects of anticipatory stress. The term perseverative cognition serves as an umbrella term for cognitive phenomena such as worry or rumination. Whereas rumination refers to thoughts about past stressors, past experiences and/or the underlying reasons for experienced feelings (e.g., Nolen-Hoeksema, 1991) worry refers to thoughts about things that are going to or may happen in the future (e.g., Borkovec, Robinson, Pruzinsky, & DePree, 1983). Accordingly, the term perseverative cognition generally refers to any kind of thought that repeatedly activates a mental representation of a stressor. By activating the cognitive representation of a stressor, perseverative cognitions are hypothesized to extend the physiological activation associated with the respective stressor both before and after the stressor has occurred. Brosschot et al. (2006) argue that this mechanism constitutes the primary pathogenic pathway leading to prolonged detrimental effects on psychological and somatic outcomes. Using a meta-analytic approach, Ottaviani et al. (2016) provide evidence consistent with this claim by showing that trait as well as state perseverative cognitions (such as worry or rumination) are related to physiological activation of various biological systems such as the autonomic and the cardiovascular system as well as the endocrinological system. Across experimental studies, state and trait perseverative cognitions were associated with higher blood pressure and elevated cortisol levels (Ottaviani et al., 2016). In addition, state perseverative cognitions were also related to higher heart rate and lower heart rate variability (Ottaviani et al., 2016). Overall, effects were small to medium in size and indicated that perseverative cognitions may impact physiological activity.

Some studies illustrate that perseverative cognitions are not only linked to physiological but also to psychological outcomes. Higher levels of trait perseverative cognitions were for example associated with lower levels of vitality, subjective and emotional well-being as well as higher levels of depressive symptoms (Zawadzki, Sliwinski, & Smyth, 2018). Despite significant bivariate correlations, stressor exposure was not related to levels of vitality, emotional well-being, or depressive symptoms in this study when controlling for perseverative cognitions. On the basis of an indirect effects model the authors suggested that stressor exposure may impact psychological health via perseverative cognitions (Zawadzki et al., 2018). Using a micro-longitudinal approach, Brose, Schmiedek, Lövdén, and Lindenberger (2011) showed that days on which stressors occurred were also characterized by higher levels of negative affect. Further, this effect was moderated by intrusive or perseverative thoughts: On days with higher levels of perseverative cognitions the effect of stressors on negative affect was stronger. Daily perseverative cognitions have also been reported to moderate the effect of stressor anticipation: In their ambulatory assessment study, Neubauer et al. (2018) showed that the intraindividual effect of stressor anticipation on negative affect was exacerbated on days with more perseverative thoughts. Taken together, these findings provide some support for the notion that perseverative cognitions may play a prominent role in explaining the detrimental effects of stress and anticipatory stress on negative affect.

Despite this initial evidence, it is important to emphasize that the proposed mechanism of the perseverative cognition hypothesis (Brosschot et al., 2006) operates on the within-person level: The extension of the stress response – either prior to or after the occurrence of the stressor – is assumed to unfold within a person over a certain time period. Consequently, testing this temporal process requires intensive longitudinal data with individuals completing assessments repeatedly (Molenaar, 2004) in order to capture within-person dynamics and draw conclusions about the direction of effects. Zawadzki et al. (2018) used cross-sectional data and conclusions about the direction of effects are therefore limited and results pertain to the between-person level only. In contrast, Brose et al. (2011) as well as Neubauer et al. (2018) investigated within-person (often also called intraindividual) effects thus focusing on the level at which processes are assumed to unfold. Nevertheless, perseverative thoughts were measured once per day in both studies targeting the occurrence of these thoughts throughout the particular day. This precludes the analysis of the temporal order of effects because it is not clear whether perseverative thoughts occurred prior to or after the stressor or whether they occurred prior or in reaction to elevated negative affect. However, examining the temporal resolution, that is, how effects unfold within persons over time would be an important step to provide more insight into the process underlying the persistent effects of anticipatory stress.

Investigating Temporal Dynamics Using Intensive Micro-Longitudinal Studies

In order to examine these temporal dynamics, multiple assessments per individual are needed. In this regard, a critical aspect is the choice of the interval between measurements. Of course, this choice is determined by the time scale across which the effects of interest are expected to occur. In laboratory-based experimental research, effects are usually examined across minutes whereas studies using daily diaries focus on effects across days. Regarding the persistent effects of anticipatory stress, prior studies in daily life have shown that effects can occur within a day or within a few hours (Hyun et al., 2019; Neubauer et al., 2018). These results illustrate the possibly fine-grained temporal resolution of the mechanism underlying the persistent effects of anticipatory stress. Therefore, we decided to focus on comparatively small measurement intervals in our study, similar to the time scale investigated by Neubauer et al. (2018), with several daily measurements spaced approximately 2-3 hours apart.

Next to examining the temporal sequence of effects, the EMA approach offers a number of additional advantages (e.g., Mehl & Conner, 2012; Smyth et al., 2017): First, EMA studies investigate effects in everyday life, that is, in a naturalistic context (as opposed to, for example, a rather artificial laboratory setting). Second, measurements in real-time (or close to real-time) reduce memory bias. Third, with individuals completing several surveys per day, EMA designs can be used to investigate how variables of interest vary within individuals across a certain time period (e.g., changes in affect from moment to moment or day to day). Further, this within-person variability can be separated from between-person variability (i.e., differences in affect between different individuals) using adequate modeling techniques such as multilevel modeling (Raudenbush & Bryk, 2002). This is important because within-person effects do not necessarily correspond to between-person effects (Molenaar, 2004). In the present research, we focus specifically on the within-person level as the mechanism proposed by the perseverative cognition hypothesis (Brosschot et al., 2006) is assumed to unfold within an individual over time.

The Present Study

The aim of this study was to investigate the temporal dynamics among the anticipation of a stressful event and subsequent levels of perseverative cognitions as well as affective experience. Our goal was to shed light on the mechanism underlying the persistent effects of stressor anticipation and enhance our understanding of the chronological sequence of effects unfolding naturally in daily life. To that end, we used a temporally dense EMA design in an ethnically diverse sample of adults. Based on previous research, we expected that the anticipation of a stressful event would be linked to higher levels of perseverative thoughts 2-3 hours later. In addition, we predicted that higher levels of perseverative thoughts would be associated with higher levels of concurrently measured negative affect. In line with the assumption of the perseverative cognition hypothesis (Brosschot et al., 2006) that perseverative thoughts mediate the persistent effects of stressor anticipation, we expected an indirect effect of stressor anticipation on negative affect 2-3 hours later through perseverative cognitions. Effects of stressor experience were controlled for by including a variable indicating whether or not a stressor was reported since the previous assessment. Because the bulk of previous research (as well as considerations put forth by the perseverative cognition hypothesis) has targeted effects on negative affect specifically, our hypotheses were targeted towards this dimension of affective well-being. Research on the association between stressor anticipation and positive affect is rather scarce. Smyth et al. (1998) reported that anticipating a stressor was linked to lower levels of concurrent positive affect. If these effects were extended by perseverative cognitions, it would be reasonable to assume the same pattern of results (lower positive affect after anticipation of a stressful event mediated via perseverative thoughts) for this dimension of affective well-being. However, given the small amount of prior research, analyses involving positive affect should be considered exploratory.

Method

The data were drawn from the first wave of the Effects of Stress on Cognitive Aging, Physiology and Emotion (ESCAPE) study, which was approved by the Institutional Review Board at Albert Einstein College of Medicine (#2010–353; Study ESCAPE). See Scott et al. (2015) for the entire study protocol. In the following, we provide the relevant details for the present research only.

Participants

Participants were 243 adults² (65% female) recruited using systematic probability sampling of New York City Registered Voter Lists in a specific zip code in the Bronx, NY. Eligible participants were 25-65 years old, fluent in English, ambulatory, and had no visual impairment interfering with operating the study smartphone. The sample is representative of the area from which it was systematically sampled (except for the slight over-representation of women [65% compared to 58% in the population]) and thus ethnically and economically diverse with a mean age of 46.2 years (SD = 11.1). Of all participants, 63.8% identified as non-Hispanic Black, 4.9% as Hispanic Black, 17.7% as Hispanic White, 9.1% as White, 0.4% as Asian, and 3.7% as Other. For a more detailed description of the sample, see Scott et al. (2019).

Procedure

After introductory letters had been sent, participants received a phone call to assess eligibility and to consent and enroll those interested in participation. Participants were invited to the lab to complete demographic questionnaires as well as 1.5 hours of training on the study protocol and the use of study smartphones. During the next two days, participants completed a “run-in” phase to practice and habituate to the EMA protocol. Participants who responded to more than 80% of the EMA surveys during this run-in period were invited to complete the 14-day EMA study.

During the 14-day study period, participants completed five beeped surveys at quasi-randomized time points throughout the day. These surveys were approximately 2-3 hours apart with beep times varying across days but being programmed according to an individual wake schedule. On average, it took 2 min 52 s to complete the surveys. Adherence to the study protocol was satisfactory with 80.9% (n = 13,764) of all surveys (N = 17,010; 243 participants x 14 days x 5 beeps) being filled in. On the individual level, 7.4% of participants (n = 18) completed all surveys (= 70 surveys per person, 5 beeps x 14 days), 41.6% (n = 101) completed between 90% and 99% of all surveys, 29.2 % (n = 71) participants completed between 70% and 89% of all surveys and 21.8% (n = 53) completed less than 70% of all surveys. Participants completing the entire study protocol received up to $160. The protocol of this study was approved by the Institutional Review Board at Albert Einstein College of Medicine (#2010–353; Study ESCAPE).

Measurements

Negative Affect

At each beep, participants rated their current negative affective states using five items. Specifically, they were asked “How unhappy [tense/anxious, angry/hostile, depressed/blue, frustrated] do you feel right now?”. Responses were given using a slider from not at all (0) to extremely (100). Current negative affect was calculated as the average of the five ratings. Within-person reliability (within-person McDonald’s ω; see Geldhof, Preacher, & Zyphur, 2014) was .84 for this scale.

Positive Affect

Current positive affect was calculated as the average of four ratings of current positive states. Participants responded to the questions “How happy [pleased, much enjoyment, joyful] do you feel right now?”. Again, responses were given using a slider from not at all (0) to extremely (100). Reliability (within-person McDonald’s ω; see Geldhof et al., 2014) was good in the present sample, ω = .89.

Perseverative Cognitions

Participants rated the thoughts they had experienced in the 5 minutes prior to the current survey using a continuous analogue scale (ranging from not at all [0] to very much [100]). Ratings referred to negative self-focus (“Were you thinking about personal problems or worries?”), and uncontrollability (“Were you experiencing a train of thought that you couldn’t get out of your head?” and “Were you preoccupied with thoughts of something about to happen or that might happen in the future?”). The items were averaged into one measure of perseverative cognitions. Within-person reliability (within-person McDonald’s ω; see Geldhof et al., 2014) was .73 for this scale.

Stressor Anticipation

At the end of each assessment, participants indicated by yes (coded as 1) or no (coded as 0) whether they expected the upcoming hours to be stressful. Specifically, participants were asked “Do you think anything stressful or unpleasant will happen in the next few hours?”

Stressor Exposure

At each beep, the experience of stressful events was assessed with a yes (coded as 1) or no (coded as 0) response to the question “Did anything stressful occur since the last survey? A stressful event is any event, even a minor one, which negatively affects you.”

Data Analysis

To answer our research questions, we used multilevel dynamic structural equation modeling (DSEM; e.g., Asparouhov, Hamaker, & Muthén, 2018; Hamaker, Asparouhov, Brose, Schmiedek, & Muthén, 2018; McNeish & Hamaker, 2018). This approach has recently been implemented in Mplus (Muthén & Muthén, 1998-2017) and combines features of time series analysis and multilevel modeling in a structural equation modeling framework. DSEM decomposes repeatedly measured variables into a within-person and a between-person part allowing to separate within- and between-person effects. Furthermore, unlike in traditional cross-lagged models, auto-regressive and cross-regressive effects can be modeled on the within-person level (Hamaker, Kuiper, & Grasman, 2015). It is implemented using Bayesian estimation, which allows for modeling more complex models than models estimated, for example, using maximum likelihood procedures. Furthermore, unequally spaced measurement intervals can be explicitly considered in these models: Whenever there is no data available within a specified time interval between subsequent assessments (which was specified as three hours in the present work) the model inserts missing values (i.e., the dependent variable is set to missing for this time point). The result is a data frame with approximately equidistant observations of three-hour intervals. This procedure ascertains a constant interpretation of the estimated parameters, which depends on the chosen time interval between measurement occasions. Hence, the resulting lagged effects in the present work can be interpreted as the effect across a time window of three hours. For more detailed information on this procedure, as well as the performance of this approach, please see Asparouhov et al. (2018) and Hamaker et al. (2018).

Following the notation of Hamaker et al. (2018), the decomposition of a variable Y for individual i at time t can be illustrated as follows:

$${\mathrm{Y}}_{\text{it}}={\mu }_{\mathrm{i}}+{\mathrm{Y}}_{\text{it}}^{(\mathrm{W})},$$ (1)

where μ_i represents the mean value of person i on variable Y and ${\mathrm{Y}}_{\text{it}}^{(\mathrm{W})}$ constitutes the individual deviations from μ_i at time t (note that the superscript “(W)” refers to “within”). Both the within-person component (${\mathrm{Y}}_{\text{it}}^{(\mathrm{W})}$) and the between-person component (μ_i) are then modeled simultaneously. For our analyses, we used a two-level model with observations (beeps, Level 1) nested within individuals (Level 2).

On the within-person level, we modeled the dynamic relations among the five variables negative affect, positive affect, stressor anticipation, perseverative cognition, and stressor exposure. Figure 1 illustrates the main pathways of the model on the within-person level (i.e., the pathways of central interest to the present research question). The within-person deviations of negative affect, ${\text{NA}}_{\text{it}}^{(\mathrm{W})}$, were modeled as a function of lagged negative affect, ${\text{NA}}_{\text{i(t}-1)}^{(\mathrm{W})}$, lagged stressor anticipation, ${\text{SA}}_{\text{i(t}-1)}^{(\mathrm{W})}$, current stressor anticipation, ${\text{SA}}_{\text{it}}^{(\mathrm{W})}$, recent stressor exposure, ${\text{SE}}_{\text{it}}^{(\mathrm{W})}$, and recent perseverative cognitions, ${\text{PC}}_{\text{it}}^{(\mathrm{W})}$. Lagged negative affect refers to negative affect at the previous measurement time point. While lagged stressor anticipation refers to the expectation at the previous assessment that anything stressful or unpleasant might happen within the upcoming hours, current stressor anticipation indicates whether the participant expects an upcoming stressor in the next few hours. Recent stressor exposure captures whether the participant has experienced a stressor since the last assessment and recent perseverative cognitions refer to the negative self-focus and uncontrollability of thoughts experienced 5 minutes prior to the current survey. Thus, the within-person equation for negative affect can also be expressed as

$${\text{NA}}_{\text{it}}^{(\mathrm{W})}={\varphi }_{01}{\text{NA}}_{\text{i(t}-1)}^{(\mathrm{W})}+{\varphi }_{02}{\text{SA}}_{\text{i(t}-1)}^{(\mathrm{W})}+{\beta }_{11}{\text{SA}}_{\text{it}}^{(\mathrm{W})}+{\beta }_{12}{\text{SE}}_{\text{it}}^{(\mathrm{W})}+{\beta }_{13\mathrm{i}}{\text{PC}}_{\text{it}}^{(\mathrm{W})}+{\zeta }_{1,\text{it}},$$ (2)

where ϕ₀₁ is the first-order autoregressive parameter, ϕ₀₂ represents the cross-lagged effect of stressor anticipation and ζ_1,it is the residual term of person i at time t. β₁₁ to β_13i represent concurrent effects (i.e., effects of variables assessed at the measurement occasion t). The additional subscript i for the regression coefficient β indicates that the effect of recent perseverative cognitions was modeled as a random effect, meaning that the respective parameter is allowed to vary between persons. Thus, it is further modelled on the between-person level (see below).

Figure 1. Illustration of the Mediation Model on the Within-Person Level.

Note. Within-person effects not involved in the mediational pathways (e.g., auto-regressive effects, effects of stressor anticipation or stressor exposure at time t) are not depicted. Coefficients represent standardized estimates with 95% credible intervals in brackets.

To model the within-person deviations of positive affect, ${\text{PA}}_{\text{it}}^{(\mathrm{W})}$, the same equation was estimated with negative affect being replaced by positive affect:

$${\text{PA}}_{\text{it}}^{(\mathrm{W})}={\varphi }_{03}{\text{PA}}_{\text{i(t}-1)}^{(\mathrm{W})}+{\varphi }_{04}{\text{SA}}_{\text{i(t}-1)}^{(\mathrm{W})}+{\beta }_{14}{\text{SA}}_{\text{it}}^{(\mathrm{W})}+{\beta }_{15}{\text{SE}}_{\text{it}}^{(\mathrm{W})}+{\beta }_{16\mathrm{i}}{\text{PC}}_{\text{it}}^{(\mathrm{W})}+{\zeta }_{2,\text{it}},$$ (3)

with ϕ₀₃ representing the effect of positive affect at the previous assessment on positive affect at the current assessment. The parameter ϕ₀₄ also constitutes a lagged effect, that is, the effect of stressor anticipation at the previous measurement on positive affect at the current measurement. Parameters β₁₄ to β_16i are concurrent effects whereas ζ_2,it constitutes the residual term of person i at time t. As described above, the subscript i for the regression coefficient β indicates that the effect of recent perseverative cognitions was allowed to vary between persons and is therefore further modelled on the between-person level (see below).

The within-person deviations of recent perseverative cognitions, ${\text{PC}}_{\text{it}}^{(\mathrm{W})}$, were modelled as a function of lagged perseverative cognitions, ${\text{PC}}_{\text{i(t}-1)}^{(\mathrm{W})}$, lagged stressor anticipation, ${\text{SA}}_{\text{i(t}-1)}^{(\mathrm{W})}$, and recent stressor exposure, ${\text{SE}}_{\text{it}}^{(\mathrm{W})}$. Hence, the respective equation is as follows:

$${\text{PC}}_{\text{it}}^{(\mathrm{W})}={\varphi }_{05}{\text{PC}}_{\text{i(t}-1)}^{(\mathrm{W})}+{\varphi }_{06\mathrm{i}}{\text{SA}}_{\text{i(t}-1)}^{(\mathrm{W})}+{\beta }_{17}{\text{SE}}_{\text{it}}^{(\mathrm{W})}+{\zeta }_{3,\text{it}}.$$ (4)

Here, ϕ₀₅ represents the first-order autoregressive parameter, that is, the effect of PC at the previous measurement occasion on PC measured at the current occasion. ϕ_06i is the cross-lagged effect of SA on PC. The random effect of SA on PC is represented by the subscript i and ζ_3,it is the residual representing unexplained variations in PC at time t.

In addition, we added autoregressive effects for stressor anticipation and stressor exposure: A person’s current stressor anticipation, ${\text{SA}}_{\text{it}}^{(\mathrm{W})}$, was predicted by his or her stressor anticipation at the previous measurement occasion, ${\text{SA}}_{\text{i(t}-1)}^{(\mathrm{W})}$. Further, recent stressor exposure, ${\text{SE}}_{\text{it}}^{(\mathrm{W})}$, was predicted by stressor exposure at the previous assessment, ${\text{SE}}_{\text{i(t}-1)}^{(\mathrm{W})}$, as well as lagged stressor anticipation, ${\text{SA}}_{\text{i(t}-1)}^{(\mathrm{W})}$. These effects can also be expressed as

$${\text{SA}}_{\text{it}}^{(\mathrm{W})}={\varphi }_{07}{\text{SA}}_{\text{i(t}-1)}^{(\mathrm{W})}+{\zeta }_{4,\text{it}},$$ (5)

$${\text{SE}}_{\text{it}}^{(\mathrm{W})}={\varphi }_{08}{\text{SE}}_{\text{i(t}-1)}^{(\mathrm{W})}+{\varphi }_{09}{\text{SA}}_{\text{i(t}-1)}^{(\mathrm{W})}+{\zeta }_{5,\text{it}}.$$ (6)

Finally, recent perseverative thoughts, ${\text{PC}}_{\text{it}}^{(\mathrm{W})}$, and current stressor anticipation, ${\text{SA}}_{\text{it}}^{(\mathrm{W})}$, were allowed to correlate on the within-person level. We also estimated the residual covariance between negative affect, ${\text{NA}}_{\text{it}}^{(\mathrm{W})}$, and positive affect, ${\text{PA}}_{\text{it}}^{(\mathrm{W})}$.

The mediating effects of previous stressor anticipation (at time t-1) on current negative affect or current positive affect, respectively, via perseverative cognitions were operationalized as the indirect effect according to Preacher, Zyphur and Zhang (2010). For an illustration of the mediation model, please see Figure 1.

To sum up, on the within-person level the model contained autoregressive effects for all five variables (i.e., negative affect, positive affect, stressor anticipation, perseverative cognitions, and stressor exposure). In addition, the model included the paths implicated by the proposed mediation models (see Figure 1)³. Further, we estimated the effect of stressor exposure on recent perseverative cognitions, concurrent positive affect, and concurrent negative affect to account for stressor exposure as a potential common cause for these three variables. Finally, we included the effect of previous stressor anticipation on current stressor exposure, as previous research has reported increased odds of experiencing a stressor when a stressor has been expected at the previous measurement occasion (Neubauer et al., 2018).

On the between-person level, there were eight variables: these were the latent means/thresholds of the five measures (negative affect, positive affect, stressor anticipation, perseverative cognition, stressor exposure; μ_NA,i, μ_PA,i, μ_PC,i, μ_SA,i, and μ_SE,i), and the means of the parameters β_3i, β_6i, and ϕ_6i (= the average or fixed effects of recent PC on current NA or current PA, respectively, and the fixed effect of lagged SA on recent PC). Since our research questions are exclusively focused on the within-person level, we fitted a fully saturated model on the between-person level (i.e., freely estimating the correlations among all eight variables on the between-person level). Consequently, the model on the between-person level can be expressed using following equations:

$${\mu }_{\text{NA,i}}={\gamma }_{\text{NA}}+{\upsilon }_{\text{NA,i}},$$ (6)

$${\mu }_{\text{PA,i}}={\gamma }_{\text{PA}}+{\upsilon }_{\text{PA,i}},$$ (7)

$${\mu }_{\text{PC,i}}={\gamma }_{\text{PC}}+{\upsilon }_{\text{PC,i}},$$ (8)

$${\mu }_{\text{SA,i}}={\gamma }_{\text{SA}}+{\upsilon }_{\text{SA,i}},$$ (9)

$${\mu }_{\text{SE,i}}={\gamma }_{\text{SE}}+{\upsilon }_{\text{SE,i}},$$ (10)

$${\beta }_{13\mathrm{i}}={\gamma }_{13}+{\upsilon }_{13\mathrm{i}},$$ (11)

$${\beta }_{16\mathrm{i}}={\gamma }_{16}+{\upsilon }_{16\mathrm{i}},$$ (12)

$${\varphi }_{06\mathrm{i}}={\gamma }_{06}+{\upsilon }_{06\mathrm{i}},$$ (13)

where the γ’s represent the average effects and the υ’s represent the deviation of person i from the respective average.

Analyses were performed using Mplus (Version 8.3; Muthén & Muthén, 1998-2017) with default settings for (non-informative) priors and convergence criteria. Variables were centered using the latent person-mean (default in Mplus). Results are based on an analysis using two chains with 25,000 iterations, 50% burn-in, and a thinning factor of 50. That is, the resulting point estimates and credible intervals are based on 25,000 data points. For all parameters that are presented below, we report the median of the posterior distribution as point estimate and the 95% credible interval (95% CI) as indicator of uncertainty. We interpret parameters whose 95% CI does not contain zero as statistically significant. We inspected model convergence using various indicators. The first criterion is the potential scale reduction (PSR; Gelman & Rubin, 1992), which represents the (square root of) the proportion of total variability of a parameter (the sum variability of parameters estimates within a chain and the variability between the chains) relative to the parameter variability within the chains. PSR estimates close to 1 indicate convergence, and values below 1.10 or 1.05 are typically chosen as cutoffs to indicate model convergence (Asparouhov & Muthén, 2010). Further, following Hamaker et al. (2015), we visually inspected trace plots for all parameters. The respective input and output files, as well as trace plots can be found in a repository on the open science framework (https://osf.io/uzcmt/?view_only=72f6bc042717401dba60529e050c2bbe). The data is currently not publicly available but can be requested from https://osf.io/4ctdv/.

Results

Descriptive Statistics

Table 1 depicts descriptive statistics of the study variables on the within- as well as the between-person level. On the between-person level, negative affect and perseverative cognitions were substantially correlated (r = .71, 95% CI: [.64; .77]) indicating that individuals reporting high negative affect also reported higher levels of perseverative cognitions. Further, persons reporting to expect more stressors also reported to experience more stressors (r = .73 [.64; .79]). On the within-person level, these correlations were statistically meaningful, but somewhat smaller (r = .44 [.42; .45] and r = .50 [.45; .54], respectively). Positive affect was negatively correlated with perseverative cognitions, stressor anticipation, and stressor experience on the within- as well as on the between-person level. Correlations ranged from r = −.29 [−.31; −.27] to r = −.41 [−.52; −.30].

Table 1.

Descriptive Statistics and Correlations of Study Variables

					Correlations [95% Credible Interval]
Variable	Mean	Range	SD between	SD within	1.	2.	3.	4.	5.	6.
1. Negative affect	21.94	0-100	15.53	14.04	-	−.59 [−.60; −.58]	.44 [.42; .45]	.38 [.36; .41]	.48 [.46; .50]
2. Positive affect	61.26	0-100	19.10	16.54	−.45 [−.56; −.35]	-	−.29 [−.31; −.27]	−.36 [−.38; −.33]	−.39 [−.41; −.36]
3. Perseverative cognitions	36.08	0-100	20.64	18.32	.71 [.64; .77]	−.31 [−.42; −.20]	-	.33 [.30; .36]	.41 [.39; .43]
4. Stressor anticipation	0.13	0-1	0.20		.38 [.26; .49]	−.41 [−.52; −.30]	.37 [.26; .50]	-	.50 [.45; .54]
5. Stressor exposure	0.17	0-1	0.16		.27 [.14; .39]	−.34 [−.46; −.21]	.29 [.16; .40]	.73 [.64; .79]	-
6. Age	46.24	25-65	11.08		−.05 [−.18; .09]	.14 [.01; .25]	−.06 [−.18; .06]	−.03 [−.17; .11]	.25 [.09; .37]	-
7. Sexa	0.65	0-1	0.48		.05 [−.09; .17]	−.08 [−.22; .05]	.00 [−.13; .15]	−.04 [−17; .10]	.03 [−.11; .17]	−.01 [−.15; .10]

Note. Between-person correlations are depicted below the diagonal, within-person correlations are shown above the diagonal. For better clarity, values are printed in bold if the credible interval does not contain zero.

^a0 = male; 1 = female.

Across all individuals and occasions, being exposed to a stressor during the past hours was reported at 17.3% of all measurement occasions, whereas participants were expecting a stressor in the upcoming hours at 13.6% of all assessments. At 7.4% of all occasions, participants reported both having experienced a stressor during the past hours and anticipating a stressful event in the next few hours. Twenty-one individuals (8.6%) did not show variability in stressor exposure and 73 individuals (30.0%) did not exhibit variation in stressor anticipation.

Multilevel Mediation Model

Because our research question specifically targets within-person processes, we report the fixed within-person effects and corresponding random effects in Table 2. Parameter estimates of between-person correlations can be found in Table A in the online supplement.

Table 2.

Unstandardized Point Estimates (Posterior Medians) and 95% Credible Intervals for Fixed Within-Person Effects, the Indirect Effects and Random Effects

Parameter		Point Estimate	95% Credible Interval
Within-person level: Fixed effects
Negative affect predicted by
lagged negative affect	ϕ01	0.24	[0.23; 0.26]
lagged stressor anticipation	ϕ02	−1.41	[−1.84; −1.00]
recent perseverative cognitions	γ13	0.17	[0.14; 0.20]
current stressor anticipation	β11	2.24	[1.93; 2.55]
recent stressor exposure	β12	4.06	[3.75; 4.38]
Positive affect predicted by
lagged positive affect	ϕ03	0.35	[0.33; 0.36]
lagged stressor anticipation	ϕ04	2.22	[1.70; 2.75]
recent perseverative cognitions	γ16	−0.08	[−0.12; −0.05]
current stressor anticipation	β14	−2.95	[−3.35; −2.55]
recent stressor exposure	β15	−4.25	[−4.68; −3.82]
Recent perseverative cognitions predicted by
lagged perseverative cognitions	ϕ05	0.08	[0.06; 0.10]
lagged stressor anticipation	γ06	2.72	[1.67; 3.80]
recent stressor exposure	β17	4.86	[4.42; 5.29]
Current stressor anticipation predicted by
lagged stressor anticipation	ϕ07	0.55	[0.51; 0.59]
Recent stressor exposure predicted by
lagged stressor exposure	ϕ08	0.14	[0.10; 0.19]
lagged stressor anticipation	ϕ09	0.34	[0.30; 0.39]
Covariance between recent perseverative cognitions and current stressor anticipation	cov1	2.24	[1.80; 2.69]
Covariance between current negative and current positive affect	cov2	−61.81	[−65.22; −58.45]
Indirect effect for negative affect	ind1	0.56	[0.28; 0.67]
Indirect effect for positive affect	ind2	−0.38	[−0.65; −0.12]
Between-person level: Random effects
Negative affect predicted by
recent perseverative cognitions	Var(υ13i)	0.04	[0.03; 0.05]
Positive affect predicted by
recent perseverative cognitions	Var(υ16i)	0.04	[0.03; 0.05]
Recent perseverative cognitions predicted by
lagged stressor anticipation	Var(υ06i)	54.90	[43.70; 68.69]

After 25,000 iterations, the highest PSR value was estimated as 1.001. Visual inspection of the trace plots of all parameters did not reveal any spikes, trends, or other irregularities. We therefore deemed convergence satisfactory.

The model revealed a positive effect of lagged stressor anticipation on perseverative thoughts (γ₀₆ = 2.72 [1.67; 3.80]), meaning that when participants anticipated a stressor at the previous measurement occasion, their levels of perseverative cognitions at the next assessment were higher. In addition, when participants reported higher levels of recent perseverative cognitions than usual, they also experienced higher current negative affect (γ₁₃ = 0.17 [0.14; 0.20]). The corresponding random variances indicated between-person variability in these effects: Var(υ_06i) = 54.90 [43.70; 68.69] and Var(υ_13i) = 0.04 [0.03; 0.05], respectively (see Figure 2, Panels A and B). Individuals displayed substantial differences in the effect of lagged stressor anticipation on recent perseverative cognitions: Some individuals reported stronger changes in levels of perseverative cognitions than others after having anticipated a stressor at the previous measurement occasion (vs. not). However, the pattern differed for recent perseverative thoughts on current negative affect. For this effect, variability between individuals was smaller implying that people were more similar to each other in the effect of recent perseverative cognitions on negative affect. In line with the hypothesized mediation, the indirect effect was larger than zero to a statistically meaningful degree (ind₁ = 0.56 [0.28; 0.87]). In addition, the model revealed evidence for a remaining direct effect of previous stressor anticipation on current negative affect after controlling for perseverative thoughts (ϕ₀₂ = −1.41 [−1.84; −1.00]) indicating that (after controlling for all other predictors in the model) when participants anticipated a stressor at the previous assessment, they reported lower levels of negative affect at the next assessment.

Figure 2. Between-Person Differences in Within-Person Effects.

Note. Panel A: Between-person differences in within-person effect of lagged stressor anticipation on perseverative cognitions. Panel B: Between-person differences in within-person effect of perseverative cognitions on negative affect. Panel C: Between-person differences in within-person effect of perseverative cognitions on positive affect. Each black dot represents the estimate of the effect for the respective person. Black vertical lines indicate the 95% credible interval of the respective effect. The solid horizontal line represents the average within-person effect and the shaded area around the solid line indicates the 95% credible interval for the average within-person effect.

For positive affect, there was a negative association with perseverative cognitions, that is, at occasions on which individuals reported higher levels of recent perseverative thoughts than usual they also reported lower levels of current positive affect (γ₁₆ = −0.08 [−0.12; −0.05]). Again, the corresponding variance estimate indicated that participants varied in the strength of this effect: Var(υ_16i) = 0.04 [0.03; 0.05] (see Figure 2, Panel C). In addition, the indirect effect of lagged stressor anticipation on current positive affect through perseverative thoughts was statistically different from zero (ind₂ = −0.38 [−0.65; −0.12]). Further, there was a positive direct effect of previous stressor anticipation on current positive affect (ϕ₀₄ = 2.22 [1.70; 2.75]) demonstrating that participants having expected a stressor reported higher levels of positive affect at the next measurement occasion (controlling for all other predictors in the model).

Exploratory Analyses on Between-Person Variability in Within-Person Effects

The findings from our model revealed interindividual differences in the strength of the associations between lagged stressor anticipation and perseverative cognitions, as well as perseverative cognitions and negative affect and positive affect, respectively (depicted in Figure 2). We conducted additional exploratory analyses to test whether demographic variables (age and gender), or mean levels of experiences across the study period might be associated with these differences. To that end, we estimated correlations between the estimated within-person associations (see Figure 2) and age, gender, and the person-means of all study variables (i.e., the person-means of negative affect, positive affect, perseverative cognitions, stressor anticipation, and stressor experience). Given the exploratory nature of these analyses, the level of significance for these correlations was adjusted according to the Holm correction (Holm, 1979). Analyses were performed in R (version 3.5.1 for Windows; R Core Team, 2018). All correlations can be found in Table B in the online supplement. Neither age nor gender were associated with any of the estimated within-person associations (r = −.14 – .10, p ≥ .42, for all). Yet, mean levels of experiences across the study period were to some extent related to the heterogeneity of effects across individuals: The association between perseverative cognitions and negative affect was stronger for individuals with higher mean levels of negative affect (r = .36, p < .001), lower mean levels of positive affect (r = −.25, p < .001), higher mean levels of stressor anticipation (r = .36, p < .001) as well as higher mean levels of stressor experience (r = .37, p < .001). The association between perseverative cognitions and positive affect was stronger (more negative) for individuals who anticipated more stressors (r = −.33, p < .001), who experienced more stressors (r = −.38, p < .001) and who generally reported lower levels of perseverative cognitions (r = .24, p < .001). Interindividual differences in the link between lagged stressor anticipation and perseverative cognitions were not meaningfully related to all these person-level variables, p > .07 for all.

Discussion

The goal of the present study was to examine the temporal dynamics between stressor anticipation and later affective well-being in daily life and to test a potential underlying mechanism of these dynamics in an ethnically diverse sample. Theoretical considerations as well as empirical findings allude to the idea that perseverative cognitions may mediate the persistent effects of stressor anticipation on affective well-being. This mechanism has been examined between persons, however, it is theorized to unfold within individuals over time. To our knowledge, no study to date has targeted the temporal within-person dynamics among stressor anticipation, perseverative cognitions and affective well-being. To address this gap, we used a temporally dense EMA design allowing us to investigate how effects unfold in individuals’ daily lives over a few hours.

A Mechanism Potentially Explaining Persistent Effects of Stressor Anticipation

Consistent with our hypothesis, findings revealed that the anticipation of a stressor was linked to higher levels of perseverative cognitions approximately 3 hours later. As predicted, higher levels of perseverative cognitions were associated with higher levels of negative affect. In line with the idea of perseverative cognitions mediating the effect of prior stressor anticipation on negative affect, our results indicated an indirect effect of stressor anticipation on negative affect through perseverative thoughts. These findings corroborate and add to previous findings that have shown that effects of stressor experience or stressor anticipation were more pronounced on days with higher levels of perseverative cognitions (Brose et al., 2011; Neubauer et al., 2018). By explicating the temporal order of effects, the present study provides insights into the temporal direction of the effects of stressor anticipation and perseverative cognitions. Additionally, the present work suggests that the detrimental effects of anticipatory stress may be extended via perseverative cognitions. Notably, the indirect effect was obtained after controlling for current stress anticipation and therefore constitutes the prolonged effect of previous stress anticipation on current negative affect. Accordingly, our findings are in line with the perseverative cognition hypothesis (Brosschot et al., 2006).

Regarding positive affect, the present study revealed that persons reporting higher levels of perseverative thoughts than usual also reported lower levels of positive affect. Similar to negative affect, our findings indicated an indirect effect of stressor anticipation on subsequent positive affect through perseverative cognitions (here, too, after controlling for current stressor anticipation). This may indicate that perseverative cognitions could also play an important role in explaining detrimental effects of stressor anticipation on positive affect. However, we do not intend to draw strong conclusions on the potential underlying mechanism from these results as analyses were exploratory. Studies investigating associations between stressor anticipation and positive affect are scarce and thus, more research on dynamics of stressor anticipation, perseverative cognitions and positive affect is needed. On the most general level, however, our findings suggest that the pattern of effects of stressor anticipation on affective well-being might be comparable for the two dimensions of affective well-being targeted in the present work (positive affect and negative affect).

Perseverative cognitions are usually conceptualized in a rather broad way referring to any thoughts which repeatedly activate the mental representation of a stressor (Brosschot, Pieper, & Thayer, 2005). Thus, the construct comprises various phenomena such as worry or rumination (Brosschot et al., 2006). Although worry and rumination are related to one another they should be treated as distinct constructs differing in a central aspect: their temporal orientation (Nolen-Hoeksema, Wisco, & Lyubomirsky, 2008). Whereas worries are described as future-oriented, ruminative thoughts tend to be focused on past events or feelings. Accordingly, anticipating a stressful event should be linked to higher levels of worries and thus, worries may matter to a greater extent for prolonging the effects of anticipatory stress. The measure of perseverative thoughts used in the present study mainly assessed the negative self-focus and uncontrollability of thoughts. However, our measure also asked about future events and thereby captured the construct of worry more so than other types of perseverative cognitions.

In addition to the (predicted) indirect effect, the present study also revealed a negative direct effect of stressor anticipation on negative affect and a positive direct effect on positive affect (controlling for perseverative cognitions, stressor exposure, and previous affect). This means that – after controlling for the other covariates in the model – expecting a stressor to occur within the upcoming hours was linked to lower levels of negative affect and higher levels of positive affect at a later time point and as such, to better emotional well-being. Regarding negative affect, this finding contradicts previous results using the same dataset: Scott et al. (2019) found stressor anticipation to be associated with higher levels of negative affect at the next assessment. These deviations in results could be explained by different focus areas addressed in the work of Scott et al. (2019) and in the present work. In the present study perseverative cognitions were of central interest as we focused on a mechanism potentially explaining the extended response to anticipated stressors. However, Scott et al. (2019) disaggregated (extended) responses related to anticipated stressors from responses to occurred stressors. Thus, perseverative cognitions were not directly relevant to their research question and were not included in the analyses.

In interpreting the direct effects of stressor anticipation on affect in the present work, it is important to keep in mind that these effects represent the variance in negative and positive affect, respectively, not explained by perseverative thoughts, experienced stressors or any other covariate: At measurement occasions for which a stressor has been forecasted (vs. not), affective well-being was somewhat higher when previous affective well-being, odds of stressor exposure, and perseverative cognitions were at the person’s average (i.e., statistically controlled for) – and therefore somewhat higher than what could have been expected based on these predictors alone.

One possible explanation for these direct effects is that they might indicate relief effects: It might be that a stressor that has not occurred, even though it was expected (such as the unexpectedly smooth ride to work), yielded a slight boost in momentary affective well-being. Individuals might have successfully employed proactive coping strategies (see Aspinwall & Taylor, 1997) which aim at preventing the anticipated negative event from occurring and, in turn, might boost momentary well-being. Another explanation of this direct effect might be that participants employed anticipatory coping strategies (Folkman & Lazarus, 1985). The goal of these strategies is not to avoid but to prepare for the stressor and thereby to mitigate its impact on well-being. Thus, individuals might have been able to better cope with the stressful event so that the stressor impacted their emotional well-being to a lesser extent than it would have impacted their well-being if they had not seen it coming (Neupert et al., 2019). These findings suggest that effects of anticipatory stress might not be exclusively negative in all circumstances but could sometimes help individuals to prepare not only psychologically (Aspinwall & Taylor, 1997) but also physiologically (Kramer, Neubauer, Stoffel, Voss, & Ditzen, 2019) for the demands of upcoming stressful events. We note, however, that we had not predicted this effect, and this finding should therefore be considered exploratory. Accordingly, our interpretations of the direct effects need to be considered post-hoc and need to be targeted in future studies before strong conclusions on these direct effects can be drawn.

Between-person Differences in the Strength of Within-person Effects

The present research explored between-person variability in the within-person effect of stressor forecasting on perseverative cognitions, and results revealed substantial heterogeneity of this effect across individuals. Although some individuals reported stronger increases in levels of perseverative cognitions after having anticipated a stressor (compared to the average within-person effect), this effect was attenuated for other individuals (see Figure 2, Panel A). There were even some participants whose effect of stressor forecasting on subsequent perseverative cognitions was estimated as negative. This finding illustrates that while – on average – the effect of stressor anticipation on subsequent levels of perseverative cognitions is detrimental, from a person-specific point of view this is not always the case. In addition, participants differed in the strength of the effect of perseverative thoughts on negative and positive affect. While nearly all participants showed a positive association between perseverative cognitions and negative affect (see Figure 2, Panel B), the association between perseverative cognitions and positive affect was somewhat more variable across individuals (see Figure 2, Panel C). This might suggest that perseverative cognitions have more similar negative consequences in terms of increasing negative affect across individuals. Couplings of perseverative cognitions with positive affect, on the other hand seem to be more variable across individuals.

Additional exploratory analyses revealed that this between-person variability in within-person effects was associated with mean levels of experiences across the study period. Individuals reporting more negative and less positive affect and anticipating and experiencing more stressors during the study period were characterized by a stronger coupling between perseverative cognitions and negative affect. Further, participants who anticipated and experienced more stressors during the study period also showed a stronger (negative) coupling between perseverative cognitions and positive affect. Average levels of perseverative cognitions were only associated with the coupling between perseverative cognitions and positive affect, suggesting that individuals who engage more in perseverative cognitions on average might be less vulnerable for its negative consequences on positive mood. Interestingly, interindividual differences in the coupling between (lag) stress anticipation and perseverative cognitions (those couplings that exhibited the largest heterogeneity across participants) were unrelated to all variables investigated in our exploratory analyses. We note that these findings result from exploratory analyses and therefore, interpretations should be considered particularly cautiously. Potentially, constructs such as neuroticism, trait anxiety, or depression may be promising factors explaining the heterogeneity among individuals. For example, individuals high in neuroticism exhibit higher stress reactivity (e.g., Bolger & Schilling, 1991). Accordingly, these individuals may also be more responsive to anticipatory stress (but see Neubauer et al., 2018). Furthermore, perseverative cognitions (e.g., worry and rumination) represent core symptoms of depression and anxiety (e.g., Watkins, 2008) and thus, may be conceptualized as a response style to (past or future) distressing (internal or external) events. Consequently, depressed or anxious individuals may be more disposed to engage in perseverative cognitions in response to the anticipation of a stressor.

With previous research showing differences between people of color and Whites in stress exposure (e.g., Boen, 2020; Turner & Avison, 2003; Sternthal, Slopen, & Williams, 2011), for example stemming from discrimination or stereotypes (Contrada et al., 2000), ethnicity constitutes an important aspect in the context of this study. In the present work, the sample investigated is ethnically more diverse than many studies that recruited volunteers from established participant lists or postings in the nearby university community. When comparing subjective stress reports of the present sample to national norms for the same measure around the same period (Cohen & Janicki-Deverts, 2012), the present sample reported significantly higher subjective stress scores than the national average (Sliwinski, Freed, Scott, Pasquini, & Smyth, 2020; see Table 1). Due to unequal cell sizes, we were not able to thoroughly test whether ethnicity might explain differences between persons in, for instance, the effect of (lag) stress anticipation on perseverative cognitions. However, as the majority of the participants in the present sample were people of color, the present findings suggest that theoretically postulated mechanisms underlying the temporal extension of the stress response seem to apply to a pre-dominantly non-White sample.

Our findings have important implications for future studies targeting the potentially harmful effects of stressor anticipation on affective well-being. First, results highlight that people differ in the dynamic, within-person interplay among stressor anticipation, perseverative cognitions, and affective well-being. This emphasizes the importance to carefully investigate differences between individuals in within-person processes in order to understand the full picture of the complex interplay of these variables. Second, the present findings provide a valuable starting point for interventions. With higher levels of perseverative thoughts being associated with poorer emotional well-being for most individuals in the sample, preventing these thoughts from occurring may constitute a beneficial intervention approach to disable the potential underlying mechanism that prolongs the detrimental effects of stressor anticipation. Hence, interventions keeping perseverative cognitions in check (or mindfulness interventions, e.g., Hilt & Pollak, 2012; such as worry reduction training, e.g., Borkovec, Wilkinson, Folensbee, & Lerman, 1983; Brosschot & van der Doef, 2006) might aid in ameliorating the potentially detrimental effects of stressor forecasting on subsequent affective well-being. Furthermore, these findings also underline the need for individualized or tailored interventions: For some persons anticipating a stressor was not as consistently linked to lower levels of perseverative cognitions indicating that the mediating mechanism proposed by the perseverative cognition hypothesis may not equally apply to all individuals. Therefore, these individuals may not profit from interventions reducing perseverative thoughts to the same extent as individuals who show a larger association between stressor anticipation and subsequent perseverative cognitions.

Limitations and Future Directions

A main limitation of the present study is the use of observational data. As a consequence, we cannot draw strong causal conclusions from the present findings. Utilizing mediation models for observational data requires additional assumptions. The most important assumption we made was that the residuals of the mediator (perseverative cognitions) and the outcome (positive or negative affect, respectively) were uncorrelated (indicating that they are not affected by a common cause). Accounting for variables influencing both is one way to make this assumption more realistic. In our analyses, we controlled for reported stressors because having experienced a stressor is likely to have an effect on perseverative thoughts as well as on affective states. Nevertheless, other variables not controlled for in our analysis may have similar effects and accordingly, the assumption of uncorrelated residuals may be violated. Experimental methods are ultimately required to rule out such third-variable explanations. We encourage research implementing experimental manipulations in daily life, for example via within-person encouragement designs (Schmiedek & Neubauer, 2020), which allow for testing the causality of within-person associations in a daily life context. Manipulating the anticipation of a stressor may be difficult (and unethical) in daily life, however, reducing perseverative cognitions using interventions (e.g., mindfulness interventions) would represent a promising attempt to elucidate the causality of the effects of perseverative cognitions on affective well-being.

Second, assessing all variables of interest several times per day enabled us to investigate lagged effects and thus – in contrast to previous studies – dismantle the temporal dynamics of the assumed process and, in turn, the temporal direction of effects. It needs to be noted, however, that perseverative cognitions as well as measures of affect were assessed at the same measurement occasion. Although participants were carefully instructed to pay attention to the different time frames referred to in the respective questions (i.e., “during the past 5 minutes” for perseverative cognitions vs. “right now” for negative and positive affect), we cannot draw firm conclusions about the direction of these effects. It is also possible that anticipating a stressor is linked to higher levels of negative affect (or lower levels of positive affect, respectively) at the next measurement occasion which, in turn, is associated with higher levels of perseverative cognitions. Consequently, it would be necessary to assess perseverative cognitions and affect at different time points. In the present work, we explicitly decided against using an alternative lag-2 model (i.e., stressor anticipation assessed at time t-2 predicting perseverative cognitions at time t-1, in turn predicting affect at time t), because we did not consider the resulting time frame with effects of (a) perseverative cognitions on affect over three hours and (b) stressor anticipation on affect over six hours a theoretically plausible time frame over which the effects should occur. Higher sampling frequencies (e.g., one assessment per hour) would be desirable in future studies to comprehensively capture the dynamics of the associations among stressor anticipation, perseverative cognitions, and affective well-being.

Third, we controlled for the experience of a stressor in our analyses, but we have no information about whether the stressor reported is actually the same stressor that has been anticipated at the previous time point. Further, we do not know in which domain the stressor was anticipated (e.g., anticipated stressor at work or in the family) or whether the content of perseverative thoughts was related to the previously anticipated stressor. Future designs should therefore examine these relations while considering both the domain of stressor anticipation, the content of perseverative thoughts and whether the experienced stressors were the ones participants had actually expected. This could be accomplished by having participants indicate at what time point they expect the respective stressor to occur. Using this information combined with event-contingent sampling would permit to schedule an assessment prior to the designated time point of stressor occurrence (e.g., 30 minutes prior to the anticipated onset of the stressor) and examine the specific content of perseverative thoughts.

Fourth, appraisals of anticipated stressors might also play an important role: for example, individuals’ expectations of duration, severity, or controllability of anticipated stressors could be related to the extent to which participants engaged in perseverative cognitions as well as their affective well-being. Subjective appraisals of stressors represent a key component influencing the reactivity to experienced stressors (see Almeida, 2005). It is therefore reasonable to assume that appraisals of anticipated stressors may have a similar impact. In the present work, we did not collect data on participants’ appraisals of anticipated stressors or on engagement in (proactive or anticipatory) coping strategies. Future studies might consider including this information to elucidate the role of appraisal and coping processes.

Conclusions

The present study adds to previous research that has demonstrated negative persistent effects of stressor anticipation on psychological outcomes. Our findings emphasize the previously hypothesized, but not directly tested critical role of perseverative cognitions in this association in individuals’ daily lives. Combining a temporally dense EMA design with a timely data analytic approach via dynamic structural equation modeling allowed us to investigate how effects of stressor anticipation on affective well-being unfold within persons over time. Our findings provide evidence in line with the perseverative cognition hypothesis (Brosschot et al., 2006), by revealing an indirect effect of previous stressor anticipation on current affective well-being (negative affect and positive affect) via perseverative cognitions. Furthermore, the present results emphasize the importance of examining differences between individuals in within-person processes. These findings might be informative for tailored interventions which could be helpful to ameliorate the detrimental effects of anticipatory stress while acknowledge the idiographic nature of within-person effects of stress anticipation, perseverative cognitions, and affective well-being.