ORIGINAL RESEARCH

published: 16 June 2020

doi: 10.3389/fpsyg.2020.01152

Heightened Stress in Employed

Individuals Is Linked to Altered
Variability and Inertia in Emotions
Diana Wang 1* , Stefan Schneider 1 , Joseph E. Schwartz 2 and Arthur A. Stone 1
1
Center for Self-Report Science, University of Southern California, Los Angeles, CA, United States, 2 Stony Brook Medicine,
Stony Brook, NY, United States

Stress has been widely recognized as a key factor contributing to health outcomes and
psychological well-being. While some growing evidence points to stress as having an
effect on emotion dynamics characteristics, there has yet to be a test of how global
perceptions of stress are associated with not only average levels of emotions but also
the variability in the intensity of the emotions, as well as how emotions linger (inertia),
and whether these characteristics differ by age. In an effort to better understand how
stress influences the emotional experiences of individuals, we examined associations
between perceived stress levels and emotion dynamics indices in a sample of 859
working individuals over 24 h. Participants ranged in age from 21 to 81 years. Each
Edited by:
participant was prompted at approximately 28 min intervals throughout a 24 h period
Changiz Mohiyeddini, to report intensity of emotional states. Overall, individuals who were more stressed
Oakland University William Beaumont
experienced lower mean levels of positive emotions (with the exception of higher levels
School of Medicine, United States
of excitement) and higher mean levels of negative emotions. They also experienced
Reviewed by:
Peter Koval, more pronounced variability in both positive and negative emotions, and greater inertia
The University of Melbourne, Australia in negative emotions. We also found some evidence for age-related differences in mean
Desirée Colombo,
Jaume I University, Spain
levels and variability in certain emotions. The relationship of emotion dynamics indices
*Correspondence:
to stress levels was not moderated by age. Many of the stress–emotion dynamics
Diana Wang associations did not remain statistically significant upon controlling for the mean level of
[email protected]
momentary emotions, indicating that the mean is a large component in the association.
Specialty section: Keywords: emotion dynamics, perceived stress, aging, variability, affect, inertia
This article was submitted to
Health Psychology,
a section of the journal INTRODUCTION
Frontiers in Psychology
Received: 25 February 2020 The overall levels of emotions that an individual experiences are important indicators of
Accepted: 05 May 2020 psychological health status. However, emotions can change over time at the scale of seconds to
Published: 16 June 2020
hours, and a large body of literature has demonstrated that characteristics of emotion fluctuations
Citation: (emotion dynamics) are important indicators of psychological health (Houben et al., 2015).
Wang D, Schneider S, Therefore, it is important to understand the various psychological and behavioral factors that
Schwartz JE and Stone AA (2020)
contribute to emotion dynamics. One’s overall perception of stress has been identified as a
Heightened Stress in Employed
Individuals Is Linked to Altered
key psychological factor that alters emotion dynamics (Koval and Kuppens, 2012). However,
Variability and Inertia in Emotions. evidence of the associations between stress and emotion dynamics to date has been limited to
Front. Psychol. 11:1152. laboratory-induced stressors, which may suffer from a lack of ecological validity in terms of their
doi: 10.3389/fpsyg.2020.01152 generalizability to individuals’ real-world perception of stressors (Dickerson and Kemeny, 2004).

Frontiers in Psychology | www.frontiersin.org 1 June 2020 | Volume 11 | Article 1152

Wang et al. Perceived Stress and Emotion Dynamics

Higher levels of perceived stress have been shown to be a strong Hypotheses advanced in this study are largely based on
predictor of worse mental health (Keller et al., 2012), accelerated findings from studies that have found that variability and inertia
cellular aging (Epel et al., 2004), increased risk of disease (Cohen of emotions are altered at various stages of a stressful experience.
et al., 2007), and premature mortality (Keller et al., 2012). For example, depressed adolescents experienced higher inertia
As part of an effort to understand the potential pathways in emotional behavior during experimentally induced stressful
through which perceived stress can influence health and well- interactions with their parents (Kuppens et al., 2010). In
being, it is of interest to know whether an elevated level of another study, individuals who were exposed to a laboratory
perceived stress is related to the emotional experiences in the stressor experienced a decrease in inertia of emotional states
daily lives of individuals. Advances in ecological momentary in anticipation of the stressor. Paradoxically, those who are
assessment (EMA) methodology allow for the collection of high- particularly sensitive to negative evaluation by others generally
resolution data in naturalistic settings as well as reduction in experience higher levels of inertia overall, but experience a larger
recall bias (Stone and Shiffman, 1994). To date, the role of drop in inertia in anticipation of a stressor (Koval and Kuppens,
global perceptions of stress has yet to be examined in association 2012). These studies demonstrate that short-term exposure to
with short-term emotion dynamics during everyday life. Further, laboratory stressors alters emotional inertia, though the direction
aging-related theory and empirical evidence of reactivity to in which inertia changes seems to vary based on individual traits,
daily stressful events suggest that perceptions of stress might be as well as timing. There is also evidence that stress can impact
differentially associated with emotion dynamics over the adult life variability in emotions. In younger adults who experienced
course (Charles, 2010; Scott et al., 2013). In the present study, a breakup, variability in both positive and negative emotions
we examined associations of perceived stress levels and age with was heightened in the week following the breakup (Sbarra and
emotion dynamics, and whether age moderates the associations Emery, 2005). While these studies show that emotion fluctuations
between levels of perceived stress and emotion dynamics. may be altered before, during, and after the experience of a
Finally, a recent meta-analysis reported that psychological well- stressor, it is also important to understand how stress as it
being states did not predict emotion dynamics indices above is generally perceived in one’s life – that is, people who are
and beyond the mean levels of momentary emotional states generally feeling high levels of stress versus those who do not –
(Dejonckheere et al., 2019a). Thus, we set out to examine whether is associated with emotion dynamics. We hypothesize that those
our hypothesized associations hold after controlling for the mean who perceive a higher level of global stress compared to those
levels of emotions, as this would aid in our understanding of the with lower levels will have higher levels of variability and inertia
extent to which the emotion dynamics indices are independent of in emotional states.
mean levels of an individual’s emotions. Furthermore, other dimensions of emotional experience
While there is a wide array of emotion dynamics indices, we have been shown to change under stress. Empirical evidence
focus on two primary measures: variability (innovation variance) demonstrates a reduction in the amount with which positive
and inertia (autoregressive parameter). The level of an emotional and negative emotions are differentiated on stressful days and
state depends in part on the previous level of an emotion as following a stressful event (Zautra et al., 2002, 2005). More
well as a number of other factors that occur, such as negative recently, it was found that during what may be considered
events, or social interactions (Jongerling et al., 2015). Innovation an acute stressful life event of receiving college examination
variance reflects the proportion of emotion fluctuations that results, individuals’ positive and negative affect shifted from
are not predicted from the previous emotional states, and are having a weaker negative correlation to having greater bipolarity
therefore considered to be due to exposure and/or reactivity to (Dejonckheere et al., 2019b). Finally, the fluctuations in the
events. Emotional inertia refers to the ability of the intensity differentiation between emotions differ such that differentiation
of an emotional state at one moment to predict the intensity of negative emotions (the ability to identify emotions with
measured at a subsequent moment. This is measured by a specificity) was lower when individuals were experiencing higher
first-order autoregressive parameter (AR) of emotions across levels of stress (Erbas et al., 2018). Taken together, these findings
time, with a higher inertia value indicating that an emotion demonstrate that the experience of stressful events is associated
lingers longer (Houben et al., 2015). In these analyses, we use with shifts in the differentiation between negative and positive
dynamic structural equation modeling (DSEM), which allows affect; stress not only influences processes that underlie temporal
for between-person variability in innovation variances through dynamics, but also the space in which emotions vary.
including a random effect for the parameters. It is important to As demonstrated by recent empirical findings, emotion
note that many previous studies have used summary statistics dynamics also change with age, and several theoretical
such as standard deviation to examine net variability, which is perspectives suggest developmental changes in emotion
a function of both the innovation variance and autoregressive regulation that may underlie these trends. In one study in which
parameter, whereas the DSEM modeling approach offers a individuals’ emotions were assessed daily over 45 days, older
novel method to examine these constructs. Temporal dynamics adults (ages 70–80) had less variability (lower intraindividual
indices of emotional variability and inertia have been linked standard deviations) in both positive and negative emotional
to psychological health: a recent meta-analysis found greater states compared to younger adults (ages 20–30) after controlling
variability and greater inertia in emotions in individuals with low for mean affect level differences (Röcke et al., 2009). In another
psychological well-being and in those with psychiatric disorders study, older adults (ages 65–80) were found to have lower
(Houben et al., 2015). variability in emotional states that were measured once a day

Frontiers in Psychology | www.frontiersin.org 2 June 2020 | Volume 11 | Article 1152

Wang et al. Perceived Stress and Emotion Dynamics

over 100 days, compared to younger adults (ages 20–31) (Brose from the Masked Hypertension Study (MHTS), which assessed
et al., 2013). In a 7-day ecological momentary assessment (EMA) momentary emotions at a high density (every 28 min), allowing
study in which emotional states were measured five times a day, for the analysis of short-term fluctuations in emotional states.
this negative association between age and emotion variability With this dataset, we set out to examine whether perceived stress
was found for both positive and negative emotions within a and age were associated with more or less variability and inertia
day (Carstensen et al., 2000, 2011). Based on these findings, in emotions, and whether age and stress interact to predict these
we predicted that older individuals would have lower levels emotion dynamics.
of variability in emotions. There is much less evidence for the Our study also utilizes an innovative statistical approach
association between age and inertia of emotions, though when to model indices of emotion dynamics – dynamic structural
compared with younger adults (aged 20–31), older adults (aged equation modeling (DSEM) (Hamaker et al., 2018). This method
65–80) have higher inertia in positive emotions, and lower inertia allows us to model individual differences in innovation variance,
in negative emotions (Hamaker et al., 2018). Given the limited autoregressive parameter (inertia), and mean affect levels, as
evidence for age differences in inertia, our analyses of age–inertia latent (i.e., random) variables in multilevel models. Thereby, the
associations were exploratory in nature. DSEM method sampling error in these indices (which results
Theoretical perspectives also suggest that age may moderate from the obtained momentary reports being only a random
the association between stress and emotion dynamics indices. sample of all possible reports that could have been obtained from
The Strength and Vulnerability Integration (SAVI) model posits each participant) further provides an elegant solution for unequal
that there are age-related improvements in strategies to avoid spacing between measures that often occurs in EMA studies due
or reduce exposure to distress, such that older adults respond to missing data (Hamaker et al., 2018).
better emotionally than younger adults to distressing situations
(Charles and Carstensen, 2008; Charles, 2010). However, the
age-related advantage in regulating emotions is reduced under MATERIALS AND METHODS
circumstances in which people cannot easily use these skills, such
as the continued exposure to chronic unrelenting stressors. The Sample
advantage that older adults seem to have in emotion regulation Data from the MHTS, a multi-site study conducted at Stony
is hypothesized to exist only in the context of lower levels of Brook University and Columbia University in 2005–2012, were
perceived stress. Based on the SAVI model, we also examined the used in this study. The primary goal of the MHTS was to
interaction between age and stress in predicting variability and examine the phenomenon of masked hypertension and related
inertia of emotions. We hypothesized that while older age will psychosocial factors. To be eligible for the study, participants had
be associated with lower variability of emotions, this relationship to be age 21 or older and employed at either of the universities
will be less pronounced for individuals who report higher levels or a financial institution in the New York City metropolitan
of perceived stress. As inertia (in particular, of negative emotions) area. Recruitment criteria included a screening blood pressure
has been thought to reflect poor emotion regulation capacity, we of below 160/105 mmHg, and those who were using medication
hypothesize that there will also be an interaction between age and that lowered blood pressure were excluded from the study.
perceived stress levels, such that older adults would have greater Participants were also excluded if they had evidence of secondary
inertia of negative emotions in the context of higher levels of hypertension, a history of overt cardiovascular disease, chronic
stress, and less inertia when stress levels are low. renal, liver, thyroid, or adrenal disease, or cancer not in remission
The majority of studies on emotion dynamics were based for at least 6 months, active substance abuse, or a serious
on end-of-day reports or measurements taken a few times mental health illness. A total of 1,011 participants were consented
throughout a day, which yield data with recall periods over and enrolled in the MHTS. The study was approved by the
24 h, or at best, over several hours. Since shifts in emotional Institutional Review Boards (IRBs) of Stony Brook University and
states can occur quickly (on the scale of seconds to minutes), Columbia University. The University of Southern California IRB
shorter intervals between measurements may provide an entirely approved the secondary analyses reported here.
different picture of the emotional lives of individuals. In In the MHTS, 903 participants completed EMAs of emotional
EMA research, there is often a balance in the frequency of states throughout a 24 h period. Participants were provided
measurements throughout the day and the number of days with a pre-programmed electronic diary (Palm Pilot Tungsten
sampled, in efforts to manage the participant burden. Given the 3), on which they were prompted to answer EMA questions
existing evidence that short-term emotion fluctuation processes about their situation, activities, emotional states, and social
are appropriately measured using 15 or 30 min measurement interactions immediately prior to ambulatory blood pressure
increments (Ebner-Priemer and Sawitzki, 2007), our research measurements that were taken approximately 28 min apart over
question was better suited to examine emotions measured a 24 h period that included one full or parts of two workdays
in higher frequency under a shorter measurement period, as (Schwartz et al., 2016).
opposed to emotions measured every 2–4 h, or at the end of
the day. While acknowledging that a 24 h period may be less Measures
ideal than having a sample of multiple days, the processes that Emotional States
we are aiming to capture with the emotion dynamics indices are Ecological momentary assessments of emotional states included
best examined in this frequency. The current study utilized data items that measured positive and negative emotions on a

Frontiers in Psychology | www.frontiersin.org 3 June 2020 | Volume 11 | Article 1152

Wang et al. Perceived Stress and Emotion Dynamics

horizontal visual analog scale using anchors at 0 (Not at all) distribution (i.e., the within-person variance is implicitly log-
and 100 (Very much). Participants were presented with a transformed to normalize its distribution). In the models, the
question stem of “Just before [the] BP [reading]: How ____ random effects in means, variability, and inertia were regressed
were you feeling?” where BP referred to the ambulatory blood simultaneously on both perceived stress and age (Model 1 in
pressure reading that served as the signal to complete an Figure 1). Then, the interactions between perceived stress and age
electronic diary entry. Positive emotional state items included were examined in predicting each of the parameters (Model 2 in
“excited,” “happy,” and “relaxed.” Negative emotional state Figure 1).
items included “frustrated,” “angry/hostile,” “anxious/tense,” and The DSEM implementation in Mplus is based on Bayesian
“depressed/blue.” parameter estimation using a Markov chain Monte Carlo
(MCMC) algorithm. We used the potential scale reduction
Global Perceived Stress criterion (Gelman et al., 2013) to decide on the number of MCMC
Participants completed the Perceived Stress Scale (PSS), iterations needed for convergence of each model. We present
which addresses the extent to which in the last month, an regression estimates along with 95% credible intervals provided
individual perceives current life demands as uncontrollable or in Bayesian analysis (these can be interpreted analogous to 95%
overwhelming and how well they believe they can deal with it confidence intervals).
(PSS; Cohen et al., 1983). Participants completed the 14-item
PSS as part of the psychosocial questionnaire within 2 weeks Time Intervals
before the EMA assessments. Example items include: “In the last The DSEM method addresses the challenge presented by unequal
month, how often have you felt difficulties were piling up so high spacing between observations. In the present study, respondents
that you could not overcome them?” Participants were asked to were signaled every 28 min to complete the EMA ratings.
provide their responses on a 5-category Likert scale of “Never” to However, individuals did not provide measures of emotional
“Very often.” states overnight when they were asleep, in addition to missing
prompts for other reasons throughout the day, resulting in
unequal time intervals (gaps) between observations. These
Analytical Plan are especially problematic for the estimation of autoregressive
We used DSEM in Mplus version 8.3 (Muthén and Muthén, parameters (i.e., inertia) because the size of the parameter
2017) to examine the associations between both inter-individual depends on the length of the lag. In Mplus, this problem is
factors (age and stress levels) and the intra-individual mean approached by dividing the person’s day into 28 min segments
level, variability, and inertia of the emotional states (Mcneish and inserting a missing value into any time segment for which
and Hamaker, 2017). We examined each of the emotion items there is no observation (i.e., no EMA rating). The missing values
assessed in the study in separate models (as opposed to summary are treated as missing at random, and the method has been shown
scores of positive emotions and negative emotions, for example). to yield appropriate parameter estimates in Bayesian analysis
Multilevel DSEM is based on decomposing the data into a even when a large amount (80%) of missing values is inserted
within-person and between-person part (illustrated in Figure 1). (Asparouhov et al., 2018).
On the within-person level, the momentary emotional state for
person i at time point t (Emotioni,t ) is regressed on the preceding
state for the same emotion at time point t-1 (Emotioni,t−1 ) RESULTS
in a time-series model. The resulting autoregressive parameter
ϕi assumes values between −1 and 1, where more strongly The sample was restricted by excluding those who did not
positive values indicate that it takes a person longer to return have data on perceived stress levels or age. This resulted in
back to his or her “normal” state (i.e., the person shows more a total of 859 participants. The mean age for the sample was
inertia) after being perturbed; this “normal” state is represented 44.8 years (SD = 10.4, range 21–81 years). The sample is
by the person’s mean emotion level µi . The residual deviations 59% female, 7.4% Black/African American, and 12% Hispanic.
of emotional states from the person’s mean (ζi,t ) have a variance The mean perceived stress level was 21.74 (ranging from 0
of πi , where this variance represents the magnitude of emotion to 51), which is comparable to previously reported values
variability within the person. All parameters have a subject of 23 in two United States samples (Cohen et al., 1983;
index i to indicate that the within-person mean level (µi ), Cohen and Janicki-Deverts, 2012). The average compliance rate
variability (πi ), and inertia (ϕi ) of an emotion item can differ was 76.3% (SD = 18.0, median = 80.5%) for this sample.
from person to person. Momentary emotional state scores were divided by 10 causing
On the between-person level, individual differences in each the transformed scores to range from 0 to 10. For the positive
of these three parameters are represented as latent variables valence emotional states, participants reported an average of 2.1
(shown as circles on the between-person level in Figure 1) that in levels of excitement, 5.3 in happiness, and 5.3 in relaxation
are modeled as multivariate outcomes (i.e., they are allowed (see Table 1). They reported substantially lower mean levels
to correlate with each other). Random effects for the within- of negative emotions, with a mean level of 1.5 in anxiety,
person mean (µi ) and inertia (ϕi ) parameters are assumed 1.4 in frustration, 0.5 in depressed, and 0.8 in anger. The
normally distributed, whereas the random effect for within- average variability (log variance) of the emotional states ranged
person variability (πi ) is assumed to follow a log-normal from −2.3 (relaxation) to 1.1 (depressed), and average inertia

Frontiers in Psychology | www.frontiersin.org 4 June 2020 | Volume 11 | Article 1152

Wang et al. Perceived Stress and Emotion Dynamics

FIGURE 1 | Multilevel DSEM model.

(autocorrelations over 28 min) ranged from 0.27 (excitement) to stress was associated with higher levels of the positive emotional
0.41 (depressed). state of excitement. Higher levels of perceived stress were also
Results from analyses examining perceived stress and age significantly associated with greater variability in both negative
as predictors of the mean level, variability, and inertia of each and positive emotional states (with the exception of a non-
emotional state are presented in Table 2. We first summarize the significant association with the variability in relaxation). Higher
perceived stress associations. We found that those who reported levels of stress were also significantly associated with higher
higher levels of global perceived stress experienced significantly inertia in negative emotional states. No significant association
higher mean levels of negative emotions and lower mean levels of was found between stress and inertia in the three positive
positive emotions; an exception, however, was that higher global emotional states (excitement, happiness, and relaxation).
Turning to age and emotions, we found that older age
was significantly, yet weakly, associated with lower mean levels
TABLE 1 | Descriptive statistics (mean and SD) of individual differences in mean
of excitement and higher mean levels of relaxation. Older
levels, variability, and inertia of each emotional state. age was also significantly but weakly associated with less
variability in frustration, excitement, and happiness. There were
Emotional state Mean emotion levela Variabilityb Inertiac no associations between age and inertia of emotions. Finally, we
Frustration 1.360 (0.916) 0.113 (1.856) 0.308 (0.240) observed no significant interactions of age with perceived stress
Anxiety 1.528 (1.114) −0.072 (1.849) 0.348 (0.246) when predicting mean levels, variability, or inertia of emotions
Depression 0.505 (0.437) −2.262 (2.530) 0.321 (0.307) (results not shown; available on request). Together, age and stress
Anger 0.758 (0.601) −1.130 (2.356) 0.270 (0.246) accounted for between 5.8 and 7.5% of the variation in the mean
Excitement 2.086 (1.446) 0.351 (1.601) 0.291 (0.211) levels of frustration, anxiety, depression, anger, happiness, and
Happiness 5.253 (1.646) 0.364 (0.967) 0.386 (0.222) relaxation, and 2.8% of the variation in mean excitement. Of
Relaxation 5.322 (1.421) 1.091 (0.733) 0.413 (0.217) the emotion dynamics indices, they accounted for 4–5% of the
a Person
variation in the variability of the negative emotional states, but
mean level on a 0–10 scale. b Log intraindividual variance. c First order
autocorrelation; Estimates come from an empty DSEM model which does not
under 2% in that of negative emotional states. Age and perceived
include any predictors. stress, primarily the latter, accounted for approximately 3% of

Frontiers in Psychology | www.frontiersin.org 5 June 2020 | Volume 11 | Article 1152

Wang et al. Perceived Stress and Emotion Dynamics

TABLE 2 | Effect sizes of perceived stress and age on mean, variability, and inertia of each emotional state.

Mean emotion levela Variabilityb Inertiac

B 95% CI R2 B 95% CI R2 B 95% CI R2

Frustration
Stress 0.222 0.175, 0.264 0.075 0.329 0.249, 0.410 0.044 0.039 0.026, 0.053 0.034
Age −0.069 −0.129, −0.001 −0.191 −0.304, −0.070 0.006 −0.013, 0.026
Anxiety
Stress 0.274 0.218, 0.325 0.072 0.365 0.283, 0.446 0.047 0.039 0.025, 0.053 0.031
Age −0.008 −0.079, 0.070 0.017 −0.094, 0.137 0.005 −0.016, 0.023
Depression
Stress 0.100 0.079, 0.121 0.062 0.534 0.426, 0.642 0.054 0.066 0.051, 0.080 0.060
Age 0.020 −0.007, 0.051 0.029 −0.122, 0.190 0.001 −0.020, 0.023
Anger
Stress 0.232 0.180, 0.279 0.058 0.200 0.155, 0.247 0.043 0.152 0.093, 0.210 0.027
Age −0.029 −0.077, 0.025 −0.042 −0.087, 0.006 −0.036 −0.097, 0.023
Excitement
Stress 0.091 0.023, 0.156 0.029 0.133 0.063, 0.204 0.019 0.002 −0.012, 0.017 0.005
Age −0.296 −0.386, −0.197 −0.212 −0.321, −0.113 −0.017 −0.036, 0.001
Happiness
Stress −0.361 −0.438, −0.285 0.058 0.066 0.023, 0.111 0.016 0.001 −0.013, 0.015 0.002
Age 0.043 −0.063, 0.148 −0.128 −0.192, −0.067 −0.003 −0.022, 0.017
Relaxation
Stress −0.316 −0.380, −0.254 0.065 0.032 −0.001, 0.068 0.003 0.007 −0.006, 0.021 0.003
Age 0.125 0.024, 0.229 −0.012 −0.063, 0.041 −0.007 −0.025, 0.012
a Person mean level of emotion is on a 0–10 scale. b Log intraindividual variance. c First order autocorrelation. In these models, to allow for convenient interpretability and

convergence of DSEMs, we transformed independent variables by dividing age by 10, and perceived stress scores by 5.

the variation in inertia of frustration, anxiety, and anger, 6% of independent effect of perceived stress on the variability of
the variation in inertia of depression, and less than 1% of that in affective states, after controlling for mean levels, became non-
positive emotional states. significant for frustration, depression, anger, and happiness, but
A question that is inevitably raised in investigations of not for anxiety and excitement (see Supplementary Table 1).
emotion dynamics is whether the associations of indices of Similarly, the effect of independent perceived stress on the
variability and inertia with the variables of interest (in our inertia of negative affective states became non-significant for
case, age, and perceived stress) might be spurious, attributable frustration and anxiety, but not for depression and anger, after
to their shared associations with the mean emotion level. controlling for mean levels. The indirect effects of stress on
Indeed, in our sample, mean levels of emotions exhibit sizeable variability and inertia via mean levels of affective states were
correlations with the variability and inertia indices, especially significant in all instances, suggesting that the mean level of
for the negative emotions (see Table 3). We ran supplementary momentary emotions is a large underlying component in the
DSEM analyses in which the mean levels of emotion were association between perceived stress and emotion dynamics. An
included as a potential mediator in the equations predicting additional way to examine the associations of intra-individual
the variability and inertia of emotions. We found that the factors (age, perceived stress levels) with mean levels, variability,
and inertia while controlling for the correlations among the
emotion dynamics indices is simultaneously regressing the
TABLE 3 | Correlations among mean levels, variability, and inertia for each
emotion dynamics indices onto age and perceived stress.
emotional state. These results mirror those from the models controlling
for the mean, demonstrating remaining associations between
Emotional Correlation Correlation Correlation
higher levels of stress and greater variability in anxiety and
state mean – variability mean – inertia variability – inertia
excitement, and greater inertia in depressive symptoms (see
Frustration 0.688 0.623 0.372 Supplementary Table 2).
Anxiety 0.664 0.533 0.352
Depression 0.885 0.783 0.661
Anger 0.759 0.704 0.454 DISCUSSION
Excitement 0.563 0.439 0.325
Happiness −0.294 −0.001 −0.108 Emotions fluctuate and the fluctuations are thought of as outputs
Relaxation −0.206 −0.008 −0.314 from an affective system that responds to both external events

Frontiers in Psychology | www.frontiersin.org 6 June 2020 | Volume 11 | Article 1152

Wang et al. Perceived Stress and Emotion Dynamics

and internal regulatory processes (Kuppens et al., 2010). In research is needed to clarify whether chronically elevated levels of
this study we examined whether perceived stress and age were perceived stress might be related to altered emotion fluctuations
independently predictive of mean, variability, and inertia of during daily life.
several emotional states. As expected, individuals with higher
levels of global perceived stress experienced higher mean levels
of negative emotions and excitement, and lower mean level of
Associations Between Age and Mean,
happiness and relaxation. In addition, we found that those with Variability, and Inertia of Emotional
higher perceived stress exhibited greater variability in emotions States
throughout the day, and that negative emotions tended to linger We found some evidence that age was associated with mean
for longer in those with higher perceived stress levels. There levels and variability of emotional states. To aid in the
was no consistent evidence that the variability or inertia differed interpretation of the results, we refer to the Circumplex Model
across age, but our data do suggest that variability in frustration, of Affect (Posner et al., 2005). According to this model, the
excitement, and happiness was lower in older individuals. Finally, emotional states measured in this study can be categorized into
we hypothesized that there would be an interaction between age positive valence high arousal (excitement, happiness), positive
and stress in predicting emotion dynamics, but did not find valence low arousal (relaxed), negative valence high arousal
evidence for this. (anger, frustration, anxiety), or negative valence low arousal
(depression). While the strengths of the associations are weak,
our findings suggest that older adults experience relatively
Associations Between Stress and Mean, lower mean levels of the high-arousal state of excitement,
Variability, and Inertia of Emotional and higher mean levels of the low-arousal relaxation. Older
States age also predicted lower variability, particularly in the high-
While it may be a reasonable expectation that individuals with arousal emotions of frustration, excitement, and happiness. Our
more perceived stress experience all positive affective states at findings demonstrate the importance of examining individual
lower levels, we did find higher mean levels of excitement in emotional state items, and pursuing further research on age-
those who are more stressed. A potential explanation for this may related shifts in the arousal dimension of emotional states.
be that both stress and excitement are high activation or high Aging and developmental theories suggest a decrease in
arousal emotions, and as some theoretical perspectives would negative valence emotions, and increase in positive valence
suggest, they are likely to occur together (Watson and Tellegen, emotions as people age (Carstensen and Mikels, 2005). The
1985; Posner et al., 2005). Consistent with our hypotheses, average age of participants was relatively young at 45 years
we also found greater variability in emotions across both (range 21–81 years), and thus, it will be important to
positive and negative valence emotions in those who are more extend these analyses to investigate these associations across
stressed. Greater variability in emotions has been consistently a sample of individuals that is more evenly distributed
found in those with psychopathology and poorer psychological across the age range.
well-being in nonclinical populations (Houben et al., 2015). Changes in emotion dynamics with older age may be
Importantly, this is in contrast to the interpretation of findings attributed to several factors, including changes in social and
from much of the existing literature which has used more environmental contexts. For example, changes in status in
traditional calculation methods of SD as variability. By using the workplace, more familiar routines, or changes in the
the DSEM method, we have been able to model the innovation composition of social networks may lead to greater stability
variance, thereby explicitly allowing for individual differences in emotions (Röcke et al., 2018). It is possible that older
in unobserved shocks to the emotion regulation system beyond adults are simply exposed to fewer negative events or differing
what can be predicted by the level of intensity of a preceding intensity of events compared with middle-aged and younger
emotional state. adults (Almeida and Horn, 2004). Indeed, previous studies
We now turn to the last emotion dynamics index that we have found that those who were older reported fewer stressors,
examined, inertia. Some have suggested that greater inertia which were less heterogeneous in nature and less disruptive,
of emotions reflects ineffective emotion regulation capacity in suggesting that age-related trends in emotion dynamics are
response to events, although it is inconclusive whether inertia related to context (Brose et al., 2013). Although the current
predominantly reflects internal emotion regulation processes or study did not examine external factors that may influence
more prolonged exposure to external circumstances (Kuppens the interindividual differences in emotion dynamics indices, it
and Verduyn, 2015). On the one hand, greater inertia in negative will be important for future research to include environmental
emotions has been linked to the tendency to suppress the factors to determine the extent to which these indices represent
expression of feelings (Koval et al., 2015b) as well as rumination endogenous factors. For example, we suggest future studies
(Koval and Kuppens, 2012). Some evidence suggests that inertia to collect comprehensive data on the momentary level (event
reflects exposure to more intense events, but not the frequency exposure frequency and intensity) as well as global assessments
of events (Koval et al., 2015a). On the other hand, higher of contextual factors that may account for interindividual
inertia has also been associated with impaired recovery from differences in emotion dynamics.
emotional stimuli in the lab, though this association was weak The findings from our supplementary analyses are consistent
and requires replication (Koval et al., 2015a). Longitudinal with those of a recent meta-analysis, which shows that

Frontiers in Psychology | www.frontiersin.org 7 June 2020 | Volume 11 | Article 1152

Wang et al. Perceived Stress and Emotion Dynamics

many indices do not have independent associations with SUMMARY

psychological well-being variables above and beyond mean
levels (Dejonckheere et al., 2019a). This calls for caution These results show that feeling higher levels of stress is associated
when examining psychological well-being states and emotion not only with higher average levels of negative emotions and
dynamics indices. However, our findings might also suggest lower average levels of most positive emotions, but also with
that stress affects multiple aspects of emotion dynamics in the ways in which these emotions fluctuate over the course
concert with each other, and more strongly so for some of a day. Individuals who generally feel more stressed are
emotions than others. likely to experience greater variability in both positive and
negative emotions, and their negative emotions linger for longer.
Strengths and Limitations We also observed a tendency for older individuals to exhibit
Some strengths of this study are that participants reported lower mean levels and variability in high-arousal emotions,
their experience of emotions at a high frequency for 24 h, which warrants further research into developmental processes
approximately every half hour while participants are awake. The using EMA methods. The findings from this study add to
instructions for momentary reporting of emotional states are our understanding of how perceived stress is associated with
thought to increase the accuracy of reported information and altered emotional functioning. While there have been strides
minimize the biases in retrieval or reconstruction of memory. toward improving the field’s understanding of the mechanisms
Much of the existing emotion dynamics research uses end- through which stress influences physiological functioning as
of-day reports on emotional states, which cannot provide as well as emotional reactivity to stressors, this study contributes
fine-grained a portrayal of the shifts in emotions. We also toward filling a gap in our understanding of how perceived
chose not to collapse emotions by valence in our analyses, and stress may be linked with everyday profiles of emotions,
instead examined each emotional state individually. Our age- yielding a high-resolution snapshot of the ongoing processes
related associations illustrate that when considering potential that have an accumulative impact on our health and well-
developmental processes, creating sum scores by positive or being over time.
negative valence may obscure important age differences by
high or low arousal. The use of DSEM in modeling the
random innovation variance allows for more explicit modeling DATA AVAILABILITY STATEMENT
of inter-individual differences that were previously unexplained
with more traditional variability measures. This is particularly The raw data supporting the conclusions of this article will be
important in examining emotional dynamics in relation to made available by the authors, without undue reservation.
perceived stress; we believe that there are individual differences
in the unobserved factors (events that occur, behaviors) related to
higher perceptions of stress, as well as emotional responsiveness ETHICS STATEMENT
or sensitivity to these factors.
While this study included participants with a wide age The study was approved by the Institutional Review Boards
range, the average age was relatively young, and all were (IRBs) of Stony Brook University and Columbia University. The
employed and healthy. The associations that we find are University of Southern California IRB approved the secondary
relevant for the working population, but not necessarily analyses reported here.
generalizable to other populations, such as unemployed or
retired individuals. Furthermore, we speculate that the types of
stressors that participants were exposed to over the workday AUTHOR CONTRIBUTIONS
and that contributed to perceptions of global stress would
more frequently have been related to events at work (compared DW, AS, and SS contributed to the conception and design of the
to the general population). It remains a possibility that age- analyses and wrote sections of the manuscript. JS developed and
related changes in emotion fluctuations are due to contextual oversaw the collection of the data. DW and SS performed the
differences such as changes in social networks, and more statistical analysis. DW wrote the first draft of the manuscript.
research is needed to examine the possibility of contextual All authors contributed to the manuscript revision, read and
differences that occur with aging, or age-related difference in approved the submitted version.
the type/content, frequency, intensity, or appraisal of external
events. A systematic examination in a sample which includes
more older individuals, including nonworking individuals as SUPPLEMENTARY MATERIAL
well as more data on individuals’ environmental and social
contexts would be helpful in teasing apart exogenous (exposure- The Supplementary Material for this article can be found
related) or endogenous (regulation-related) factors that underlie online at: https://www.frontiersin.org/articles/10.3389/fpsyg.
emotion dynamics. 2020.01152/full#supplementary-material

Frontiers in Psychology | www.frontiersin.org 8 June 2020 | Volume 11 | Article 1152

Wang et al. Perceived Stress and Emotion Dynamics

REFERENCES Keller, A., Litzelman, K., Wisk, L. E., Maddox, T., Cheng, E. R., Creswell, P. D., et al.
(2012). Does the perception that stress affects health matter? The association
Almeida, D. M., and Horn, M. C. (2004). “Is daily life more stressful during middle with health and mortality. Health Psychol. 31, 677–684. doi: 10.1037/a0026743
adulthood?,” in How Healthy Are We?: A National Study of Well-being at Midlife, Koval, P., and Kuppens, P. (2012). Changing emotion dynamics: individual
ed. R. C. Kessler (Chicago, IL: University of Chicago Press), 425–451. differences in the effect of anticipatory social stress on emotional inertia.
Asparouhov, T., Hamaker, E. L., and Muthén, B. (2018). Dynamic structural Emotion 12, 256–267. doi: 10.1037/a0024756
equation models. Struct. Equ. Model. 25, 359–388. doi: 10.1080/10705511.2017. Koval, P., Brose, A., Pe, M. L., Houben, M., Erbas, Y., Champagne, D., et al. (2015a).
1406803 Emotional inertia and external events: The roles of exposure, reactivity, and
Brose, A., Scheibe, S., and Schmiedek, F. (2013). Life contexts make a difference: recovery. Emotion 15, 625–636. doi: 10.1037/emo0000059
Emotional stability in younger and older adults. Psychol. Aging 28, 148–159. Koval, P., Butler, E. A., Hollenstein, T., Lanteigne, D., and Kuppens, P. (2015b).
doi: 10.1037/a0030047 Emotion regulation and the temporal dynamics of emotions: effects of cognitive
Carstensen, L. L., and Mikels, J. A. (2005). At the intersection of emotion and reappraisal and expressive suppression on emotional inertia. Cogn. Emot. 29,
cognition: aging and the positivity effect. Curr. Direct. Psychol. Sci. 14, 117–121. 831–851. doi: 10.1080/02699931.2014.948388
doi: 10.1111/j.0963-7214.2005.00348.x Kuppens, P., Allen, N. B., and Sheeber, L. B. (2010). Emotional inertia
Carstensen, L. L., Pasupathi, M., Mayr, U., and Nesselroade, J. R. (2000). Emotional and psychological maladjustment. Psychol. Sci. 21, 984–991. doi: 10.1177/
experience in everyday life across the adult life span. J. Pers. Soc. Psychol. 79, 0956797610372634
644–655. doi: 10.1037/0022-3514.79.4.644 Kuppens, P., and Verduyn, P. (2015). Looking at emotion regulation through the
Carstensen, L. L., Turan, B., Scheibe, S., Ram, N., Ersner-Hershfield, H., Samanez- window of emotion dynamics. 26, 72–79. doi: 10.1080/1047840X.2015.960505
Larkin, G. R., et al. (2011). Emotional experience improves with age: evidence Mcneish, D., and Hamaker, E. L. (2017). A primer on two-level dynamic structural
based on over 10 years of experience sampling. Psychol. Aging 26, 21–33. doi: equation models for intensive longitudinal data in Mplus. Psychol. Methods
10.1037/a0021285 doi: 10.1037/met0000250 [Epub ahead of print].
Charles, S. T. (2010). Strength and Vulnerability Integration (SAVI): a model Muthén, L. K., Muthén, B. O. (2017). Mplus User’s Guide, Eighth Edn. Los Angeles,
of emotional well-being across adulthood. Psychol. Bull. 136, 1068–1091. doi: CA: Muthén & Muthén.
10.1037/a0021232 Posner, J., Russell, J., and Peterson, B. (2005). The circumplex model
Charles, S. T., and Carstensen, L. L. (2008). Unpleasant situations elicit different of affect: An integrative approach to affective neuroscience, cognitive
emotional responses in younger and older adults. Psychol. Aging 23, 495–504. development, and psychopathology. Dev. Psychopathol. 17, 715–734. doi: 10.
doi: 10.1037/a0013284 1017/S0954579405050340
Cohen, S., and Janicki-Deverts, D. (2012). Who’s Stressed? Distributions of Röcke, C., Brose, A., and Kuppens, P. (2018). “Emotion dynamics in older age,”
Psychological Stress in the United States in Probability Samples from 1983, in Emotion Regulation: A Matter of Time, Part II: Development of Emotion
2006, and 2009. J. Appl. Soc. Psychol. 42, 1320–1334. doi: 10.1111/j.1559-1816. Regulation As a Dynamic, Real-Time Process, eds P. M. Cole, and T. Hollenstein
2012.00900.x (Milton Park: Taylor & Francis Group), 179–207.
Cohen, S., Janicki-Deverts, D., and Miller, G. E. (2007). Psychological stress and Röcke, C., Li, S. C., and Smith, J. (2009). Intraindividual variability in positive and
disease. J. Am. Med. Assoc. 298, 1685–1687. doi: 10.1001/jama.298.14.1685 negative affect over 45 days: do older adults fluctuate less than young adults?
Cohen, S., Kamarck, T., and Mermelstein, R. (1983). A global measure of perceived Psychol. Aging 24, 863–878. doi: 10.1037/a0016276
stress. J. Health Soc. Behav. 24, 385–396. Sbarra, D. A., and Emery, R. E. (2005). The emotional sequelae of nonmarital
Dejonckheere, E., Mestdagh, M., Houben, M., Rutten, I., Sels, L., Kuppens, P., et al. relationship dissolution: analysis of change and intraindividual variability over
(2019a). Complex affect dynamics add limited information to the prediction of time. Person. Relationsh. 12, 213–232. doi: 10.1111/j.1350-4126.2005.00112.x
psychological well-being. Nat. Hum. Behav. 3, 478–491. doi: 10.1038/s41562- Schwartz, J., Burg, M., Shimbo, D., Broderick, J., Stone, A. A., Ishikawa,
019-0555-0 J., et al. (2016). Clinic blood pressure underestimates ambulatory blood
Dejonckheere, E., Mestdagh, M., Verdonck, S., Lafit, G., Ceulemans, E., Bastian, pressure in untreated employer-based US population: Results from the
B., et al. (2019b). The relation between positive and negative affect becomes masked hypertension study. Circulation 134, 1794–1807. doi: 10.1161/
more negative in response to personally relevant events. Emotion doi: 10.1037/ CIRCULATIONAHA.116.023404.Clinic
emo0000697 [Online ahead of print] Scott, S. B., Sliwinski, M. J., and Blanchard-Fields, F. (2013). Age differences in
Dickerson, S. S., and Kemeny, M. E. (2004). Acute stressors and cortisol responses: emotional responses to daily stress: The role of timing, severity, and global
a theoretical integration and synthesis of laboratory research. Psychol. Bull. 130, perceived stress. Psychol. Aging 28, 1076–1087. doi: 10.1037/a0034000
355–391. doi: 10.1037/0033-2909.130.3.355 Stone, A. A., and Shiffman, S. (1994). Ecological Momentary Assessment (EMA) in
Ebner-Priemer, U. W., and Sawitzki, G. (2007). Ambulatory assessment of affective behavioral medicine. Ann. Behav. Med. 16, 199–202. doi: 10.1093/abm/16.3.199
instability in borderline personality disorder: The effect of the sampling Watson, D., and Tellegen, A. (1985). Toward a Consensual Structure of Mood.
frequency. Eur. J. Psychol. Assess. 23, 238–247. doi: 10.1027/1015-5759.23.4.238 Psychol. Bull. 98, 219–235. doi: 10.1037/0033-2909.98.2.219
Epel, E. S., Blackburn, E. H., Lin, J., Dhabhar, F. S., Adler, N. E., Morrow, J. D., et al. Zautra, A. J., Affleck, G. G., Tennen, H., Reich, J. W., and Davis, M. C. (2005).
(2004). Accelerated telomere shortening in response to life stress. Proc. Natl. Dynamic approaches to emotions and stress in everyday life: Bolger and
Acad. Sci. U.S.A. 101, 17312–17315. doi: 10.1073/pnas.0407162101 zuckerman reloaded with positive as well as negative affects. J. Pers. 73, 1511–
Erbas, Y., Ceulemans, E., Kalokerinos, E. K., Houben, M., Koval, P., Pe, M. L., 1538. doi: 10.1111/j.0022-3506.2005.00357.x
et al. (2018). Why i don’t always know what i’m feeling: The role of stress in Zautra, A. J., Berkhof, J., and Nicolson, N. A. (2002). Changes in affect
within-person fluctuations in emotion differentiation. J. Pers. Soc. Psychol. 115, interrelations as a function of stressful events. Cogn. Emot. 16, 309–318. doi:
179–191. doi: 10.1037/pspa0000126 10.1080/02699930143000257
Gelman, A., Carlin, J. B., Stern, H. S., Dunson, D. B., Vehtari, A., and Rubin,
D. B. (2013). Bayesian Data Analysis, Third Edn. Boca Raton, FL: CRC Press, Conflict of Interest: AS was a Senior Scientist with the Gallup Organization and
doi: 10.1201/b16018 was a consultant for Adelphi Values.
Hamaker, E. L., Asparouhov, T., Brose, A., Schmiedek, F., and Muthén, B. (2018).
At the frontiers of modeling intensive longitudinal data: dynamic structural The remaining authors declare that the research was conducted in the absence of
equation models for the affective measurements from the COGITO study. any commercial or financial relationships that could be construed as a potential
Multivar. Behav. Res. 3171, 1–22. doi: 10.1080/00273171.2018.1446819 conflict of interest.
Houben, M., Van Den Noortgate, W., and Kuppens, P. (2015). The relation between
short-term emotion dynamics and psychological well-being: A meta-analysis. Copyright © 2020 Wang, Schneider, Schwartz and Stone. This is an open-access
Psychol. Bull. 141, 901–930. doi: 10.1037/a0038822 article distributed under the terms of the Creative Commons Attribution License
Jongerling, J., Laurenceau, J. P., and Hamaker, E. L. (2015). A Multilevel AR(1) (CC BY). The use, distribution or reproduction in other forums is permitted, provided
model: allowing for inter-individual differences in trait-scores, inertia, and the original author(s) and the copyright owner(s) are credited and that the original
innovation variance. Multivar. Behav. Res. 50, 334–349. doi: 10.1080/00273171. publication in this journal is cited, in accordance with accepted academic practice. No
2014.1003772 use, distribution or reproduction is permitted which does not comply with these terms.

Frontiers in Psychology | www.frontiersin.org 9 June 2020 | Volume 11 | Article 1152