Does suppressing negative emotion impair subsequent emotions? Two experience sampling studies

Abstract

Suppression is one of the most commonly studied emotion-regulation strategies and a variety of studies have shown that suppression of emotions is associated with adverse affective outcomes. Most of the evidence for this conclusion comes from laboratory manipulations in which people enact experimentally induced suppression or from survey-based recollections. In the present pair of studies (468 participants total), we used real-time experience sampling data to examine the effect of naturally occurring suppression of negative emotion at one moment on subsequent reports of both negative and positive emotion. Results demonstrated that suppression led to later increases in both high-activation and low-activation negative emotions, over and above the level of negative emotion being suppressed. These findings add ecologically valid support to the growing body of evidence showing that emotional suppression is not only an ineffective emotion-regulation strategy, but also a costly one.

Full text

Vol.:(0123456789)
1 3
Motivation and Emotion (2020) 44:427–435
https://doi.org/10.1007/s11031-019-09774-w
ORIGINAL PAPER
Does suppressing negative emotion impair subsequent emotions?
Two experience sampling studies
YanRuan1,3 · HarryT.Reis1· WojciechZareba1· RichardD.Lane2
Published online: 18 June 2019
© Springer Science+Business Media, LLC, part of Springer Nature 2019
Abstract
Suppression is one of the most commonly studied emotion-regulation strategies and a variety of studies have shown that
suppression of emotions is associated with adverse affective outcomes. Most of the evidence for this conclusion comes from
laboratory manipulations in which people enact experimentally induced suppression or from survey-based recollections.
In the present pair of studies (468 participants total), we used real-time experience sampling data to examine the effect of
naturally occurring suppression of negative emotion at one moment on subsequent reports of both negative and positive
emotion. Results demonstrated that suppression led to later increases in both high-activation and low-activation negative
emotions, over and above the level of negative emotion being suppressed. These findings add ecologically valid support to
the growing body of evidence showing that emotional suppression is not only an ineffective emotion-regulation strategy,
but also a costly one.
Keywords Emotion suppression· Experience sampling· Emotion regulation· Negative affect
Introduction
Suppression is one of the most commonly studied emotion-
regulation strategies. Most research associates suppression
of emotions with adverse affective outcomes. For example,
in laboratory studies, experimentally induced suppression
has been shown to lead to worse mood (Campbell-Sills etal.
2006; Gross 1998; Gross and Levenson 1995), increased
sympathetic activation (Demaree etal. 2006; Gross and
Levenson 1993; Roberts etal. 2008), and negative feelings
about social interactions (Butler etal. 2003). In correlational
studies, expressive suppression has been related to continued
or worsened negative emotions (Ehring etal. 2010; John and
Gross 2004; Moore etal. 2008; Naragon-Gainey etal. 2017;
Nezlek and Kuppens 2008) and diminished positive emo-
tions (Wood etal. 2003). In short, the weight of evidence
clearly indicates that suppression generally is a taxing, inef-
fective, and often maladaptive emotion-regulation strategy
(Gross 2002).
For the most part, existing research has documented these
effects either in the context of laboratory manipulations or in
people’s survey-based self-reports of their experiences with
emotion regulation and its affective correlates. Rare in this
literature is real-time evidence about the temporal effects of
naturally occurring suppression; that is, when people spon-
taneously choose to suppress their emotions at one moment,
are there deleterious effects on subsequent emotional expe-
rience? In this report we provide such evidence from two
experience-sampling studies.
Experience sampling represents a useful approach for
examining spontaneous emotion-regulation in everyday life
(Augustine and Larsen 2012). Experimental studies have
value for their ability to support causal inferences but they
may not tap the emotion-regulation strategies that people
actually use in daily living (Heiy and Cheavens 2014). Sur-
veys are limited by people’s ability to recall and accurately
describe their experiences regulating emotions. In contrast
to these two approaches, experience sampling collects real-
time reports of what people are doing or feeling at any given
moment, reflecting their personal choices and predilections
about how to behave in each context and circumstance,
* Yan Ruan
yan.ruan@rochester.edu
1 University ofRochester, Rochester, NY, USA
2 University ofArizona, Tucson, AZ, USA
3 Department ofClinical andSocial Sciences inPsychology,
The University ofRochester, River Campus Box270266,
Rochester, NY14627-0266, USA

428 Motivation and Emotion (2020) 44:427–435
1 3
without instructions from experimenters about what to do.
These real-time reports are useful for describing spontane-
ous emotional experience, while minimizing retrospection
bias and theory-based retrospections (that is, beliefs about
the effects of suppression, instead of their actual invivo
experience of suppression) (Reis etal. 2014; Robinson and
Clore 2002). This approach to research thereby provides an
informative complement to existing laboratory experiments
and retrospective surveys.
A few experience-sampling studies have investigated
emotional suppression. Heiy and Cheavens (2014) found that
participants rated suppression the least effective of 20 strate-
gies for reducing negative emotions. Catterson etal. (2017)
showed that suppression was associated with lower momen-
tary well-being (although this effect was attenuated when
people felt low in social power). Le and Impett (2013) found
that highly interdependent people reported better personal
and relational well-being if they suppressed negative emo-
tions associated with sacrificing for their partner’s benefit.
In end-of-day reports, people high in trait-level tendencies
to suppress emotion indicated that their experiences of fear
lasted longer (Verduyn etal. 2009). None of these studies
examined with real-time reports the effects of suppressing
emotion at one moment on subsequent reports of emotion.
The present research provides the first such report of which
we are aware.
A pair of experience-sampling studies by Brans and
colleagues did look at affective changes associated with
emotion suppression (Brans etal. 2013). They found that
reports of suppression at one moment were associated with
more negative and less positive affective states at that same
moment, relative to prior emotional states. Because these
reports of suppression and affective state were contempo-
raneous, these studies did not establish whether emotion
suppression at one time leads to later emotional afteref-
fects. To be sure, contemporaneous accounts offer valuable
information about the affects associated with suppression,
but they do not speak to temporally later effects, something
that most theorizing predicts (Gross 2015). Evidence of this
link would provide the most direct evidence yet available
about the deleterious effects of suppression on subsequent
emotions.
We note here two experience-sampling studies that
yielded seemingly contrary findings. In these studies, sup-
pression led to positive outcomes: specifically, positive emo-
tions for participants with borderline personality disorder
(Chapman etal. 2009) and less negative emotions among
older Chinese insurance workers (Yeung and Fung 2012).
However, in the former study, participants were instructed
to suppress their emotions on the third day of the study, and
therefore this study did not examine spontaneous (i.e., self-
chosen) suppression. The latter study looked at contempora-
neous reports of suppression and negative emotions and not
downstream effects. Nevertheless, these studies point to the
utility of examining suppression in different populations, as
well as potential moderators for the effects of suppression,
both of which the present research did.
The present pair of studies examined the effect of sup-
pressing negative emotions on subsequent emotional experi-
ence in natural contexts. Adult participants were randomly
signaled 10 times a day for 3days, and asked their emotional
state at that moment, as well as whether they had suppressed
any negative emotions since the prior signal. The design
of these studies allowed us to investigate whether reports
of having suppressed negative emotion at one signal were
systematically related to changes in negative emotion at the
following signal (usually, between 30min and 3h later).
The short sampling interval between signals has the advan-
tage of capturing suppression as close as is feasible to the
moment when participants voluntarily and spontaneously
choose to suppress. Moreover, because the exact time of
response was recorded, the use of relatively brief intervals
makes it possible to examine the duration of these effects;
that is, whether the affective consequences of suppression
tend to subside during an identifiable interval after the report
of suppression, if at all.
Method
Participants
Participants were 468 individuals drawn from two exist-
ing studies.1 Study 1 included 167 patients (121 women)
with the congenital long QT syndrome from the Interna-
tional Long QT Syndrome Registry (located at the Univer-
sity of Rochester Medical Center, Rochester, NY). These
patients ranged in age from 16 to 50years (M = 34.9). Study
2 included 301 participants, of whom 199 were patients (49
women) with coronary artery disease. Half of the patients
(n = 100) had no history of serious arrhythmias, and the rest
had implanted defibrillators for primary prevention of sud-
den cardiac death. These participants ranged in age from 28
to 86 (M = 62.6). In addition, 50 older healthy participants
(13 women) were randomly selected but matched by age and
gender (1:4 ratio) to patients, ranging in age from 33 to 85
(M = 59.8). Fifty-two healthy control subjects (28 women)
were also recruited, their age matching the patients with long
QT syndrome, their age range from 21 to 50 (M = 36.8).2
1 Data from these two studies were previously reported in Lane etal.
(2011) and (2018). These papers did not examine suppression.
2 The decision to include subsamples of individuals with long QT
syndrome and coronary artery disease in part reflects the possibility
that these groups might be particularly prone to demonstrate down-
stream effects of sympathetic activation (e.g., Zipes 1991), which
may occur when people suppress emotion. However, as we explain

429Motivation and Emotion (2020) 44:427–435
1 3
Participants in Study 1 resided throughout North Amer-
ica. All participants in Study 2 resided in the Tucson, Ari-
zona, and Rochester, New York, metropolitan areas. Details
about these samples are provided in Table1. Enrollment in
both studies was limited to individuals who were fluent in
English and who did not show signs of diminished cognitive
capacity or extensive psychopathology.
Procedure
Prior to data collection, participants were visited by a
research assistant in their homes for an intake session,
where they were familiarized with study procedures. They
also completed a preliminary survey, including the follow-
ing measures.
Trait anxiety
Trait anxiety was measured using the State-Trait Anxiety
Inventory (STAI; Spielberger etal. 1983), a 20-item self-
report assessing consistent predispositions for anxiety. A
sample item reads, “I worry too much over something that
really doesn’t matter.” Cronbach’s alpha = .92 and .91 in
Study 1 and Study 2, respectively.
Baseline positive andnegative aect
Participants’ baseline positive and negative affect was
assessed using the Positive and Negative Affect Scale
(PANAS; Watson etal. 1988). The baseline PANAS includes
20 adjectives, with ten each for positive affects (e.g. “enthu-
siastic”), and negative affects (e.g., “upset”). In Study 1,
Cronbach’s alpha was .86 for positive affect, and .86 for
negative affect. In Study 2, Cronbach’s alpha was .87 for
positive affects, and .88 for negative affects.
Neuroticism
In Study 1, the baseline neuroticism was measured using the
Revised NEO Personality Inventory (NEO-PI-R; Costa and
McCrae 1992). Cronbach’s alpha = .94. In Study 2, base-
line neuroticism was measured using the Big Five Inven-
tory (Hendriks etal. 1999). Cronbach’s alpha = .85. In all
analyses with this variable, we used a standardized score.
The experience sampling method (ESM) was used to
assess momentary suppression and affect. Procedures in both
studies were similar, except where noted, and are described
simultaneously. Participants were paged ten times a day for
3 days during their usual waking hours (typically, between
8 AM and 10 PM), following a modified random schedule.
The minimum gap between signals was 60min in Study 1
and 30min in Study 2. For each day of the ESM protocol,
participants in Study 1 received a stipend of $50 per day; in
Study 2, this stipend was $100 per day.
Table 1 Demographic information and descriptive data for study participants
Tabled values are means and in parentheses, standard deviations. Means sharing a subscript do not differ from each other (p < .05) in post-hoc
Fisher’s least significant difference tests. Neuroticism was assessed using different measures in Study 1 and Study 2, and therefore descriptive
data should not be compared across studies
Study 1 Study 2
Long QT Low-risk CAD High-risk CAD Young healthy Older healthy
N 167 100 99 52 50
Age 36.8 (9.0) 63.7 (10.0) 61.5 (11.7) 34.9 (10.4) 59.5 (10.4)
% Female 72.5 33.0 16.1 53.8 26.0
Baseline measures
Anxiety 38.02b (9.87) 35.34c (8.48) 34.44c (8.43) 41.25a (11.26) 33.66c (7.96)
Neuroticism 129.08 (25.39) 2.36 (.78) 2.30 (.85) 2.68 (.91) 2.07 (.73)
Negative affect 17.66b (5.74) 15.92c (5.75) 15.77c (5.35) 19.88a (5.91) 15.35c (4.27)
Positive affect 35.02a,b (6.16) 35.32a,b (5.78) 34.44b (5.87) 34.10b (7.03) 36.38a (5.43)
ESM measures
Suppression 1.46a (1.48) 1.21b (.61) 1.23b (.73) 1.48a (.96) 1.30a,b (.68)
High-arousal negative affect 1.38a (.55) 1.16b (.34) 1.12b (.29) 1.44a (.86) 1.17b (.31)
Low-arousal negative affect 1.39a (.55) 1.21b (.37) 1.27b (.39) 1.54c (.73) 1.23b (.35)
High-arousal positive affect 3.32b (1.42) 3.69a (1.15) 3.54a (1.26) 3.36b (.84) 3.71a (1.12)
Low-arousal positive affect 3.50b (1.42) 4.10a (1.32) 4.23a (1.37) 3.62b (.99) 3.99a (1.28)
later, there were no differences between these groups and the normal
controls.
Footnote 2 (continued)

430 Motivation and Emotion (2020) 44:427–435
1 3
Participants in Study 1 were paged on Palm Pilot personal
digital assistants programmed to vibrate when a report was
scheduled, using the Experience Sampling Program (Barrett
and Feldman Barrett 2000). Participants in Study 2 were
paged on Motorola Droid X or X2 smartphones, using a
homemade program that provided an auditory signal at the
scheduled time and that recorded the exact time at which
answers were entered. Participants in both studies were
instructed to turn to their device as soon as possible after
the page, to begin responding immediately, and to complete
the protocol without interruption. Responses were permitted
up to 20min after signal delivery.
The ESM protocols in both studies included the follow-
ing questions.
Emotion suppression
At each signal, participants were asked “Since the last sig-
nal, have you felt negative emotions that you’ve kept bottled
up?” We used only a single item because of the necessity
to minimize participant burden in EMA studies. This item
was chosen to be representative of items used on multi-item
scales (e.g., “when I’m upset, I bottle it up;” Courtauld
Emotional Control Total Score; Watson and Greer 1983).
Responses used a 1 (not at all) to 7 (extreme) scale. Because
74.9% of the responses were reported as 1, we treated this
variable as a dichotomy,3 namely none (1, recoded as 0)
vs. some (2–7, recoded as 1), representing whether or not
suppression had occurred. Some level of suppression was
reported by 309 (66%) participants during the study period.
Aect
Participants were asked to rate the extent to which they had
experienced affects from the PANAS (Watson etal. 1988)
during the 5min preceding the signal, using a 1 (not at all)
to 7 (extreme) scale. Sixteen affect items were included
in the present analyses4: (1) High arousal positive affect:
interested, attentive, excited (in a positive way), enthusiastic,
and alert (α = .84); (2) high arousal negative affect: guilty,
anxious, angry, hostile, jittery, and afraid (α = .73); (3) low
arousal positive affect: calm, and relaxed (α = .74), and (4)
low arousal negative affect: sad, depressed (substituted for
the PANAS-X term, blue), and lonely (α = .61).These Cron-
bach’s alphas were calculated by first computing the reliabil-
ity separately for each signal, and then averaging across the
30 signals, so that reliability is not confounded with within-
person dependencies.
Results
Compliance with the ESM protocol was computed by com-
paring scheduled signal times to the device’s internal record
of when responding began. Participants responded to 93.0%
and 91.6% of the signals in Studies 1 and 2, respectively.
More than half the sample in both studies began all or all
but one of their reports within 10min, and only 19 partici-
pants in Study 1 and 2 participants in Study 2 began 4 or
more reports more than 15min after the page. On average,
participants in Studies 1 and 2 took 2.35 ± 1.36min and
4.00 ± 3.50min, respectively, to complete a report. Using the
VARCOMP procedure in SPSS, within-person effects were
estimated to account for the following variance percentages
in affect: high-arousal negative affect (55.9%), low-arousal
negative affect (45.8%), high-arousal positive affect (49.8%),
and low-arousal positive affect (42.4%).
Analyses were conducted using multilevel modeling
(MLM) within SPSS 24.0 MIXED. To maximize power,
the samples from Study 1 and 2 were combined in all analy-
ses.5 All predictors were entered as fixed effects, except that
the level-one variable, within-person suppression (Prior
suppressionwithin), and the intercept were modeled as ran-
dom effects. We followed Bolger and Laurenceau’s (2013)
recommendations strictly: level-one predictors were first
grand-mean centered to establish an interpretable zero point,
and then person-mean centered, so that coefficients reflect
momentary deviations from each person’s average. We also
partitioned each level-one predictor into two orthogonal
components: a between-subjects component, represented by
the person’s grand-mean-centered aggregated mean score
over the course of the full ESM period, and a within-subjects
component, represented by the person-mean-centered score
(i.e., the momentary deviation from the mean). Both within-
and between-subjects indicators of suppression were entered
the model simultaneously (marked by the subscript “within”
and “between”, respectively, in the tables).
High-arousal positive affect, low-arousal positive affect,
high-arousal negative affect, and low-arousal negative affect
served as outcome variables. First, concurrent models were
tested to examine the relationship between emotion suppres-
sion and affect within the same signal. Temporal artifacts
were controlled by including time of day, survey day (1–3),
and time elapsed since the prior survey (all centered) in
the model. We applied a first-order autoregressive [AR(1)]
3 All later analyses were run with suppression as a dichotomous vari-
able. The findings remain the same if suppression is treated as a con-
tinuous variable with log transformation.
4 Affect items in the two studies differed slightly; these 16 items were
selected because they were used with all participants in both studies.
5 Study was treated as a potential moderator in preliminary analyses.
Study group did not moderate any of the results reported here.

431Motivation and Emotion (2020) 44:427–435
1 3
model to correct for autocorrelation of residuals and an
unstructured covariance matrix for the random effects.
Consistent with Brans and colleagues’ findings (2013),
suppression was associated with significantly lower posi-
tive affect and higher negative affect, as shown in Table2.
Emotion suppression was significantly associated with
higher levels of high arousal negative affect at the same
signal [B = .26, t(257.08) = 12.36, p < .001] and low arousal
negative affect at the same signal [B = .15, t(232.60) = 8.45,
p < .001]. Similarly, suppression was significantly associ-
ated with lower levels of high arousal positive affect at the
same signal [B = − .14, t(953.02) = −3.78, p < .001] and
low arousal positive affect at the same signal [B = −.43,
t(236.42) = −10.57, p < .001].
Next, we tested our main question, about the effect of
suppression at one signal predicting mood at the subsequent
signal. Here, we computed lagged models where affect at
signal t was predicted by suppression at signal t-1, control-
ling for affect at signal t-1. The first signal of each day was
dropped to exclude overnight lags. Results are presented in
Table3. Reporting emotion suppression at one signal (Prior
Suppressionwithin) significantly predicted higher levels of
both high-arousal and low-arousal negative affect in the
next signal [B = .04, t(270.17) = 3.26, p = .001] and [B = .04,
t(250.50) = 2.58, p = .010], respectively.6 Similar analyses
were conducted for predicting positive affect, but were
not significant for high-arousal positive affect [B = −.04,
t(1009.85) = −1.21, p = .23] or low-arousal positive affect
[B = .01, t(226.37) = .29, p = .77]. In short, suppressing neg-
ative emotions predicted higher levels of negative affect at
Table 2 Results of concurrent
analysis predicting affect from
suppression
B(SE) t p95% CI
High arousal negative affect
Intercept .91(.02) t(2564.77) = 43.73 <.001 [.87, .95]
Suppressionwithin .26(.02) t(257.08) = 12.36 <.001 [.22, .30]
Suppressionbetween .50(.03) t(352.01) = 14.69 <.001 [.43, .56]
Prior high arousal negative affect .28(.01) t(7790.75) = 32.55 <.001 [.26, .30]
Time of day −.00(.00) t(3704.37) = −2.06 .039 [−.00, .00]
Day of the survey .01(.00) t(2158.96) = 3.38 .001 [.01, .02]
Time elapsed since last signal .00(.00) t(7242.88) = 5.47 <.001 [.00, .00]
Low arousal negative affect
Intercept .85(.02) t(2041.91) = 40.37 <.001 [.80, .89]
Suppressionwithin .15(.02) t(232.60) = 8.45 <.001 [.11, .18]
Suppressionbetween .47(.04) t(295.83) = 13.10 <.001 [.40, .54]
Prior low arousal negative affect .36(.01) t(5595.72) = 45.41 <.001 [.34, .37]
Time of day −.01(.00) t(2093.51) = −1.92 .055 [−.01, .00]
Day of the survey .00(.00) t(3707.57) = .75 .453 [.00, .00]
Time elapsed since last signal .00(.00) t(6763.21) = 2.43 .015 [.00, .00]
High arousal positive affect
Intercept 2.08(.05) t(2958.85) = 42.15 <.001 [1.98, 2.17]
Suppressionwithin −.14(.04) t(953.02) = −3.78 <.001 [−.21, −.07]
Suppressionbetween −.13(.08) t(366.90) = −1.59 .113 [−.29, .03]
Prior low arousal negative affect .47(.01) t(5693.66) = 55.95 <.001 [.45, .48]
Time of day −.00(.00) t(4810.77) = −6.21 <.001 [−.00, −.00]
Day of the survey −.01(.01) t(3169.11) = −1.63 .103 [−.03, .00]
Time elapsed since last signal .00(.00) t(8668.36) = 3.35 <.001 [.00, .00]
Low arousal positive affect
Intercept 2.44(.06) t(2302.43) = 38.87 <.001 [2.31, 2.56]
Suppressionwithin −.43(.04) t(236.42) = −10.57 <.001 [−.50, −.35]
Suppressionbetween −.58(.11) t(306.23) = −5.26 <.001 [−.80, −.36]
Prior low arousal negative affect .37(.01) t(6533.56) = 42.38 <.001 [.35, .38]
Time of day .00(.00) t(3954.22) = 3.83 <.001 [.00, .00]
Day of the survey −.08(.01) t(2326.70) = −7.51 <.001 [−.10, −.06]
Time elapsed since last signal .00(.00) t(7307.83) = .76 .450 [−.00, .00]
6 The findings remained significant after removing the three temporal
control variables from the models.

432 Motivation and Emotion (2020) 44:427–435
1 3
the next signal, controlling for levels of negative affect at the
time suppression was reported, but it did not predict change
in positive affect at the next signal.7
We then examined whether the time that elapsed between
two successive signals moderated the association between
prior suppression and negative affect. Following the afore-
mentioned plan, we tested models that included prior sup-
pression, time elapsed since the prior signal, and their
interaction, as well as time of day, survey day, negative
affect at the prior signal, and between-persons components
of suppression as covariates. As before, suppression was
first grand-mean centered, and then person-mean centered.
As Table4 shows, the prior suppression × time elapsed
interaction was significant for low-arousal negative affect
(p = .022) but not for high-arousal negative affect (p = .518).
For post hoc probing of this significant interaction, we com-
puted regions of significance, which indicate the specific
value of the moderator (time elapsed since the prior signal)
at which the effect of prior suppression on negative affect
becomes nonsignificant (i.e., p = .05; Preacher etal. 2006).
This analysis revealed that the effect of suppression on low-
arousal negative affect became nonsignificant at 1.37h (1h,
22min) after the previous signal.
Finally, we also explored several other potential mod-
erators, including gender, age, study (health status/age)
groups, as well as the trait measures mentioned above. Each
potential moderator (contrast codes were created for study
groups, trait measures were centered) and its interaction
with momentary suppression were added to the analyses
described earlier. None of these 28 analyses found significant
Table 3 Results of multilevel
models predicting affect from
suppression at the prior signal
B(SE) t p95% CI
High arousal negative affect
Intercept .81(.02) t(2608.23) = 39.25 <.001 [.77, .85]
Prior suppressionwithin .04(.01) t(270.18) = 3.26 .001 [.02, .07]
Suppressionbetween .45(.03) t(322.61) = 14.14 <.001 [.38, .51]
Prior high arousal negative affect .36(.01) t(6333.06) = 39.66 <.001 [.34, .38]
Time of day −.00(.00) t(3720.70) = −1.99 .046 [−.00, .00]
Day of the survey .01(.00) t(2221.19) = 3.19 .001 [.00, .02]
Time elapsed since prior signal .01(.00) t(7695.35) = 5.45 <.001 [.01, .02]
Low arousal negative affect
Intercept .82(.02) t(2035.27) = 39.30 <.001 [.78, .86]
Prior suppressionwithin .04(.01) t(250.50) = 2.58 .010 [.01, .06]
Suppressionbetween .47(.04) t(285.04) = 13.24 <.001 [.40, .54]
Prior low arousal negative affect .37(.01) t(5273.26) = 47.11 <.001 [.36, .39]
Time of day .00(.00) t(3716.12) = .64 .526 [.00, .00]
Day of the survey −.01(.00) t(2081.19) = −1.77 .077 [−.01, .00]
Time elapsed since prior signal .00(.00) t(6746.31) = 1.42 .156 [−.00, .01]
High arousal positive affect
Intercept 2.01(.05) t(3049.73) = 41.25 <.001 [1.92, 2.11]
Prior suppressionwithin −.04(.04) t(1009.85) = −1.21 .225 [−.11, .03]
Suppressionbetween −.13(.08) t(366.02) = −1.67 .095 [−.28, .02]
Prior low arousal negative affect .48(.01) t(5297.55) = 58.05 <.001 [.47, .50]
Time of day −.00(.00) t(4886.36) = −6.20 <.001 [.00, .00]
Day of the survey −.01(.01) t(3268.18) = −1.46 .144 [−.03, .00]
Time elapsed since prior signal .01(.00) t(8793.77) = 2.84 .005 [.00, .02]
Low arousal positive affect
Intercept 2.28(.06) t(2428.95) = 36.64 <.001 [2.15, 2.40]
Prior suppressionwithin .01(.03) t(226.36) = .29 .769 [−.06, .08]
Suppressionbetween −.57(.10) t(296.41) = −5.44 <.001 [−.78, −.36]
Prior low arousal negative affect .40(.01) t(4999.09) = 46.19 <.001 [.39, .42]
Time of day −.00(.00) t(4127.71) = 4.00 <.001 [−.00, −.00]
Day of the survey −.07(.01) t(2499.33) = −7.16 <.001 [−.09, −.05]
Time elapsed since prior signal .00(.01) t(7705.36) = .14 .887 [−.01, .01]
7 The findings remained significant after adding baseline positive
affect and baseline negative affect as control variables.

433Motivation and Emotion (2020) 44:427–435
1 3
moderation of the association between prior suppression and
negative affect.8
Discussion
Drawing on two experience-sampling studies, our findings
provide the most direct evidence yet available for the tem-
poral effect of naturally occurring emotion suppression.
We found that suppressing negative emotions at one signal
significantly predicted more negative emotions at the next
signal, for both high-arousal and low-arousal negative emo-
tion. These effects remained significant when controlling
for several potential moderators. Further, we found that the
effect of suppression on low-arousal negative affect faded
after 1h 22min, whereas time elapsed did not influence the
effect of suppression on high-arousal negative affect, sug-
gesting that, at least within the maximum 2–3h time frame
of our ESM protocol, time did not alleviate the high activa-
tion associated with negative affect brought by suppression.
The present research demonstrates with real-time reports
that naturally occurring suppression adversely affects sub-
sequent emotions, a finding that aligns with research exam-
ining the sequelae of suppression in laboratory sessions
(Campbell-Sills etal. 2006; Gross and Levenson 1993)
as well as research on individual differences in emotion
regulation tendencies (see review by Webb etal. 2012).
Taken together, our findings support the view that when
negative emotions arise, suppression is not only an ineffec-
tive emotion-regulation strategy, but also a costly one, bring-
ing with it lingering negative emotional experience. This
effect, although small in the moment, may accumulate over
multiple events and time, potentially undermining psycho-
logical well-being and relationship functioning in the long
run (Cameron and Overall 2018).
Note that in the current research, suppression was associ-
ated only with subsequent negative affect, but not positive
affect, which may at first glance seem to contradict prior
research that found suppression to be associated with both
higher negative affect and lower positive affect at the end
of the day (Brans etal. 2013). Although we did not find an
association between suppression and positive affect at the
next signal, we did replicate their finding that suppression
was associated with lower positive affect in the same signal.
This could indicate that while participants experience higher
negative affect and lower positive affect during suppression,
this dip in positive affect recovers soon afterwards while the
rise in negative affect persists. Additional studies are needed
to shed light on this possibility.
It is noteworthy that we did not find moderation of the
effects of suppression by neuroticism or baseline affect, traits
that have been linked to prolonged negative affect in prior
research (e.g., Gross etal. 1998). One potential explanation
is that the current studies examined voluntary, self-chosen
suppression—that is, these personality traits may influence
people’s choice of an emotion regulation strategy in daily
life (e.g., Heiy and Cheavens 2014), but not their reaction to
Table 4 Results of multilevel
models predicting negative
affect from suppression at the
prior signal as moderated by
time elapsed since prior signal
B(SE) t p95% CI
High arousal negative affect
Intercept .81(.02) t(2608.97) = 39.24 <.001 [.77, .85]
Prior suppressionwithin .04(.01) t(270.02) = 3.27 .001 [.02, .07]
Suppressionbetween .45(.03) t(322.77) = 14.14 <.001 [.39, .51]
Prior high arousal negative affect .36(.01) t(6337.80) = 39.65 <.001 [.34, .37]
Time of day .00(.00) t(3721.47) = −2.00 .046 [−.00, .00]
Day of the survey .01(.00) t(2221.35) = 3.18 .001 [.00, .02]
Time elapsed since prior signal .01(.00) t(7694.90) = 5.44 <.001 [.01, .02]
Prior suppressionwithin × time elapsed .00(.01) t(8068.52) = −0.65 .518 [−.02, .01]
Low arousal negative affect
Intercept .82(.02) t(2046.88) = 39.28 <.001 [.78, .86]
Prior suppressionwithin .04(.01) t(251.11) = 2.62 .009 [.01, .06]
Suppressionbetween .47(.04) t(285.95) = 13.23 <.001 [.40, .54]
Prior low arousal negative affect .37(.01) t(5279.71) = 47.24 <.001 [.36, .39]
Time of day .00(.00) t(3734.18) = .63 .527 [.00, .00]
Day of the survey −.01(.00) t(2096.97) = −1.77 .077 [−.01, .00]
Time elapsed since prior signal .00(.00) t(6783.04) = 1.41 .159 [−.00, .01]
Prior suppressionwithin × time elapsed −.02(.01) t(8403.73) = −2.28 .022 [−.03, −.00]
8 There also were no significant moderation effects for positive
affect. Full results can be found in the online supplemental material,
https ://doi.org/10.7910/DVN/PCVIW 6.

434 Motivation and Emotion (2020) 44:427–435
1 3
the strategy once it is self-chosen. If confirmed by additional
research, this finding would have important implications
for our understanding of individual differences in emotion
regulation. It is also possible that suppression of episodes of
more intense negative emotions, which were relatively rare
due to our emphasis on ordinary, everyday emotions, would
reveal moderation by individual differences.
The current study has several limitations that bear note.
First, the samples included individuals with specific medi-
cal conditions along with healthy control individuals, which
makes generalizability unclear. Our moderation analyses
help address this limitation, showing that study group did
not moderate the obtained results. However, additional
research is needed to see if similar results obtain in more
diverse samples. Another limitation lies in our measurement
of emotion expression. Because of the necessity to minimize
participant burden, emotion suppression was examined with
only a single item. In addition, we did not examine effects of
suppressing positive emotions, nor did we assess the motiva-
tion behind suppression, which has been shown to moderate
the association between suppression and well-being in the
specific case of sacrifice (Le and Impett 2013). Indeed, it
is reasonable to expect that the effects of emotional sup-
pression may be context-dependent to some extent, and it
will be useful in future research to identify those contexts
in which emotional suppression may be less detrimental.
Finally, our ESM protocol featured 10 signals per day, which
created relatively brief intervals designed to capture what-
ever short-lived effects emotional suppression might have.
Nevertheless, because the ESM reports asked about suppres-
sion “since the last signal,” it is possible that some diminish-
ment in negative affect had already taken place by the time
of the report of suppression. However, it seems reasonable
to expect that this would likely have lessened the changes
observed in the current analyses.
Conclusion
The current research utilized two experience-sampling stud-
ies and found that naturally occurring suppression of nega-
tive emotions at one signal led to higher negative affect in
the next signal. That emotion suppression may be ineffec-
tive and potentially costly is a robust idea in the existing
literature, as well as in popular ideas about emotion regula-
tion. These two studies provide the most direct evidence yet
available about this association: In everyday life, suppress-
ing one’s negative emotions is likely to yield a more negative
emotional experience that lingers afterwards.
Acknowledgement This research was funded by RO1 HL68764 and
RO1 HL103692 from the National Institutes of Health.
References
Augustine, A. A., & Larsen, R. J. (2012). Emotion research. In M.
R. Mehl & T. S. Conner (Eds.), Handbook of research methods
for studying daily life (pp. 497–510). New York: Guilford Press.
Barrett, D. J., & Feldman Barrett, L. (2000). The experience-sampling
program (ESP). http://www2.bc.edu/~barre tli/esp/.
Bolger, N., & Laurenceau, J. P. (2013). Intensive longitudinal methods:
An introduction to diary and experience sampling research. New
York: Guilford.
Brans, K., Koval, P., Verduyn, P., Lim, Y. L., & Kuppens, P. (2013).
The regulation of negative and positive affect in daily life. Emo-
tion, 13(5), 926–939.
Butler, E. A., Egloff, B., Wlhelm, F. H., Smith, N. C., Erickson, E.
A., & Gross, J. J. (2003). The social consequences of expressive
suppression. Emotion, 3(1), 48–67.
Cameron, L. D., & Overall, N. C. (2018). Suppression and expression
as distinct emotion-regulation processes in daily interactions: Lon-
gitudinal and meta-analyses. Emotion, 18(4), 465–480.
Campbell-Sills, L., Barlow, D. H., Brown, T. A., & Hofmann, S. G.
(2006). Effects of suppression and acceptance on emotional
responses of individuals with anxiety and mood disorders. Behav-
iour Research and Therapy, 44(9), 1251–1263.
Catterson, A. D., Eldesouky, L., & John, O. P. (2017). An experience
sampling approach to emotion regulation: Situational suppression
use and social hierarchy. Journal of Research in Personality, 69,
33–43.
Chapman, A. L., Rosenthal, M. Z., & Leung, D. W. (2009). Emotion
suppression in borderline personality disorder: An experience
sampling study. Journal of Personality Disorders, 23(1), 29–47.
Costa, P. T., & McCrae, R. R. (1992). Normal personality assessment
in clinical practice: The NEO personality inventory. Psychological
Assessment, 4(1), 5–13.
Demaree, H. A., Schmeichel, B. J., Robinson, J. L., Pu, J., Everhart, D.
E., & Berntson, G. G. (2006). Up- and down-regulating facial dis-
gust: Affective, vagal, sympathetic, and respiratory consequences.
Biological Psychology, 71(1), 90–99.
Ehring, T., Tuschen-Caffier, B., Schnülle, J., Fischer, S., & Gross, J.
J. (2010). Emotion regulation and vulnerability to depression:
Spontaneous versus instructed use of emotion suppression and
reappraisal. Emotion, 10(4), 563–572.
Gross, J. J. (1998). The emerging field of emotion regulation: An inte-
grative review. Review of General Psychology, 2(3), 271–299.
Gross, J. J. (2002). Emotion regulation: Affective, cognitive, and social
consequences. Psychophysiology, 39(3), 281–291.
Gross, J. J. (2015). Emotion regulation: Current status and future pros-
pects. Psychological Inquiry, 26(1), 1–26.
Gross, J. J., & Levenson, R. W. (1993). Emotional suppression: Physi-
ology, self-report, and expressive behavior. Journal of Personality
and Social Psychology, 64(6), 970–986.
Gross, J. J., & Levenson, R. W. (1995). Emotion elicitation using films.
Cognition and Emotion, 9(1), 87–108.
Gross, J. J., Sutton, S. K., & Ketelaar, T. V. (1998). Relations between
affect and personality: Support for the affect-level and affective-
reactivity views. Personality and Social Psychology Bulletin, 24,
279–288.
Heiy, J. E., & Cheavens, J. S. (2014). Back to basics: A naturalistic
assessment of the experience and regulation of emotion. Emotion,
14(5), 878–891.
Hendriks, A. J., Hofstee, W. K., & De Raad, B. (1999). The five-factor
personality inventory (FFPI). Personality and Individual Differ-
ences, 27(2), 307–325.
John, O. P., & Gross, J. J. (2004). Healthy and unhealthy emotion regu-
lation: Personality processes, individual differences, and life span
development. Journal of Personality, 72(6), 1301–1334.

435Motivation and Emotion (2020) 44:427–435
1 3
Lane, R. D., Reis, H. T., Hsu, C. H., Kern, K. B., Couderc, J. P., Moss,
A. J., etal. (2018). Abnormal repolarization duration during eve-
ryday emotional arousal in long QT syndrome and coronary artery
disease. The American Journal of Medicine, 131(5), 565–572.
Lane, R. D., Zareba, W., Reis, H. T., Peterson, D. R., & Moss, A.
J. (2011). Changes in ventricular repolarization duration during
typical daily emotion in patients with long QT syndrome. Psycho-
somatic Medicine, 73(1), 98–105.
Le, B. M., & Impett, E. A. (2013). When holding back helps: Sup-
pressing negative emotions during sacrifice feels authentic and
is beneficial for highly interdependent people. Psychological Sci-
ence, 24(9), 1809–1815.
Moore, S. A., Zoellner, L. A., & Mollenholt, N. (2008). Are expres-
sive suppression and cognitive reappraisal associated with stress-
related symptoms? Behaviour Research and Therapy, 46(9),
993–1000.
Naragon-Gainey, K., McMahon, T. P., & Chacko, T. P. (2017). The
structure of common emotion regulation strategies: A meta-ana-
lytic examination. Psychological Bulletin, 143(4), 384–427.
Nezlek, J. B., & Kuppens, P. (2008). Regulating positive and negative
emotions in daily life. Journal of Personality, 76(3), 561–580.
Preacher, K. J., Curran, P. J., & Bauer, D. J. (2006). Computational
tools for probing interaction effects in multiple linear regression,
multilevel modeling, and latent curve analysis. Journal of Educa-
tional and Behavioral Statistics, 31, 437–448.
Reis, H. T., Gable, S. L., & Maniaci, M. R. (2014). Methods for study-
ing everyday experience in its natural context. In H. T. Reis &
C. M. Judd (Eds.), Handbook of research methods in social and
personality psychology (pp. 373–403). New York: Cambridge
University Press.
Roberts, N. A., Levenson, R. W., & Gross, J. J. (2008). Cardiovascu-
lar costs of emotion suppression cross ethnic lines. International
Journal of Psychophysiology, 70(1), 82–87.
Robinson, M. D., & Clore, G. L. (2002). Belief and feeling: Evidence
for an accessibility model of emotional self-report. Psychological
Bulletin, 128(6), 934–960.
Spielberger, C. D., Gorsuch, R. L., Lushene, R., Vagg, P. R., & Jacobs,
G. A. (1983). Manual for the state-trait anxiety inventory. Palo
Alto: Consulting Psychologists Press.
Verduyn, P., Delvaux, E., Van Coillie, H., Tuerlinckx, F., & Van
Mechelen, I. (2009). Predicting the duration of emotional experi-
ence: Two experience sampling studies. Emotion, 9(1), 83–91.
Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and
validation of brief measures of positive and negative affect: The
PANAS scales. Journal of Personality and Social Psychology,
54(6), 1063–1070.
Watson, M., & Greer, S. (1983). Development of a questionnaire meas-
ure of emotional control. Journal of Psychosomatic Research,
27(4), 299–305.
Webb, T. L., Miles, E., & Sheeran, P. (2012). Dealing with feeling:
A meta-analysis of the effectiveness of strategies derived from
the process model of emotion regulation. Psychological Bulletin,
138(4), 775–808.
Wood, J. V., Heimpel, S. A., & Michela, J. L. (2003). Savoring versus
dampening: Self-esteem differences in regulating positive affect.
Journal of Personality and Social Psychology, 85(3), 566–580.
Yeung, D. Y., & Fung, H. H. (2012). Impacts of suppression on emo-
tional responses and performance outcomes: An experience-
sampling study in younger and older workers. Journals of Ger-
ontology. Series B, Psychological Sciences and Social Sciences,
67(6), 666–676.
Zipes, D. P. (1991). The long QT interval syndrome: A Rosetta stone
for sympathetic related ventricular tachyarrhythmias. Circulation,
84(3), 1414–1419.
Publisher’s Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.