Developmental Psychobiology
Michael J. Sulik1
Nancy Eisenberg2
Tracy L. Spinrad3
Kassondra M. Silva3
1
Institute of Human Development and
Social Change
New York University
New York, NY
2
3
Department of Psychology
Arizona State University
Tempe, AZ, 85287-1104
T. Denny Sanford School of Social and
Family Dynamics
Arizona State University
Tempe, AZ
Associations Between
Respiratory Sinus Arrhythmia
(RSA) Reactivity and Effortful
Control in Preschool-Age
Children
ABSTRACT: We tested whether respiratory sinus arrhythmia (RSA) reactivity
in response to each of three self-regulation tasks (bird and dragon; knock-tap;
and gift wrap) would predict self-regulation performance in a sample of 101
preschool-age children (M age ¼ 4.49, SD ¼ .64). While controlling for baseline
RSA, decreases in RSA from bird and dragon to knock-tap (but not from baseline
to bird and dragon) predicted a latent variable measuring self-regulation.
Furthermore, increases in RSA from the knock-tap to gift wrap—the only task
involving delay of gratification—were related to concurrent task performance
while controlling for the relation between RSA reactivity and the latent selfregulation variable. Results suggest that the relations between RSA reactivity
and self-regulatory ability are influenced by task-specific demands and possibly
by task order. Furthermore, RSA reactivity appears to relate differently to
performance on motivationally salient self-regulation tasks such as delay of
gratification relative to cool executive function tasks. ß Dev Psychobiol
Keywords: effortful control; executive function; respiratory sinus arrhythmia
(RSA); vagal tone; self-regulation
INTRODUCTION
Effortful control (EC)—the self-regulatory aspect of
temperament that supports volitional control of attention,
emotion, and behavior—is associated with a number of
important socioemotional outcomes, including social
competence, internalizing and externalizing symptoms,
and academic achievement (Eisenberg, Spinrad, &
Eggum, 2010; Liew, 2012). Temperament is believed to
have an underlying biological basis, and respiratory
sinus arrhythmia (RSA) has been proposed as a physioManuscript Received: 2 September 2014
Manuscript Accepted: 30 March 2015
Correspondence to: Nancy Eisenberg
E-mail: [email protected]
Contract grant sponsor: Arizona State University Graduate and
Professional Student Association
Contract grant sponsor: National Institute of Mental Health
Article first published online in Wiley Online Library
(wileyonlinelibrary.com).
DOI 10.1002/dev.21315 ß 2015 Wiley Periodicals, Inc.
logical marker of EC. RSA is a cardiac measure of
parasympathetic nervous system activation that is commonly referred to as vagal tone because this effect is
mediated by the vagus nerve (Porges et al., 1981). Our
study was motivated by a desire to better understand
how biological and behavioral measures of selfregulation are related to one another (Porges, 2007).
Previous research has indicated that resting RSA
is positively related to executive function (Hansen,
Johnsen, & Thayer, 2003; Marcovitch et al., 2010;
Staton, El-Sheikh, & Buckhalt, 2009), a construct that
overlaps considerably with EC (Zhou, Chen, & Main,
2012). Investigators have also studied RSA reactivity,
the increase or decrease in RSA relative to baseline or
resting levels in response to changing environmental
demands, as a correlate of self-regulatory processes.
Given that resting measures of RSA have been found to
predict self-regulation (Hansen et al., 2003; Marcovitch
et al., 2010; Staton et al., 2009) and that RSA reactivity
2
Sulik et al.
is often found to overlap with resting measures of RSA
(Burt & Obradovic, 2013), one unresolved question is
whether measures of RSA reactivity predict the
capacity for self-regulation over and above the relation
between resting RSA and self-regulation.
Porges (2007) theorized that RSA suppression (i.e.,
decreases in RSA relative to baseline RSA), even
more so than baseline RSA, is an index of social and
emotional regulation. However, empirical evidence for
relations between RSA reactivity in response to selfregulation tasks and performance on those tasks is
limited in early and middle childhood. In one study of
42 low-income preschool children, RSA reactivity in
response to two self-regulation tasks was unrelated to a
composite measure of performance on those tasks
(Blair & Peters, 2003). Similarly, no relation between
RSA reactivity in response to a reaction time task and
performance on an executive function task was found
in a sample of 41 school-age children (Staton et al.,
2009). However, the small sample size in these studies
translates into low statistical power, which limits the
conclusions that can be drawn from these studies. In
contrast to those null results, Marcovitch et al. (2010)
reported a quadratic relation between RSA reactivity
and executive functioning (measured using two laboratory tasks) in a larger sample (N ¼ 220) of 3.5-year
olds. In that study, moderate RSA reactivity during
these tasks was associated with high executive function,
whereas both low and high RSA reactivity were
associated with low executive function. In these studies,
measures of RSA reactivity have typically been based
on a single task (e.g., Staton et al., 2009) or averaged
across tasks to compute a single RSA reactivity score
(e.g., Blair & Peters, 2003; Marcovitch et al., 2010). It
is unknown whether multiple consecutive measures of
RSA reactivity, when considered independently, would
each relate to self-regulation in a similar way.
Miyake et al. (2000) have differentiated between
three distinct but interrelated aspects of executive
function: working memory, attention shifting, and
inhibitory control. Many of the studies relating RSA to
executive function include tasks that assess working
memory—the ability to keep information in short-term
memory and manipulate or update this information—or
attentional control (e.g., Hansen et al., 2003; Marcovitch et al., 2010). This is particularly true for studies
of adolescents and adults, perhaps because inhibitory
control—the ability to suppress a dominant behavioral
response—improves rapidly in early childhood, and
some laboratory measures assessing inhibitory control
show ceiling effects by age 6–7 (Lagattuta, Sayfan, &
Monsour, 2011). However, working memory is not
typically included in the construct of EC (Eisenberg &
Zhou, in press). It is, therefore, not clear whether RSA
Developmental Psychobiology
reactivity to cognitively demanding tasks predicts
performance on self-regulatory or executive function
tasks that measure primarily inhibitory control and
attention and have low working memory demands.
A second important theoretical distinction made in
research on self-regulation is between “cool” tasks that
have primarily cognitive demands and “hot” tasks such
as delay of gratification that involve a strong motivational component (Metcalfe & Mischel, 1999). Cool
self-regulation is believed to rely more on the dorsolateral prefrontal cortex and the anterior cingulate
cortex, whereas hot self-regulation is believed to rely
more on the ventromedial prefrontal cortex and orbitofrontal cortex (Happaney, Zelazo, & Stuss, 2004). In
some studies, cool self-regulation tasks have been
found to differentially predict children’s academic outcomes, whereas hot self-regulation tasks have been
found to differentially predict children’s socioemotional
outcomes (Kim, Nordling, Yoon, Boldt, & Kochanska,
2012; Willoughby, Kupersmidt, Voegler-Lee, & Bryant,
2011). However, little is known about whether RSA
reactivity relates differently to performance on hot and
cool self-regulation tasks.
The Present Investigation
Relations between baseline RSA and EC have already
been examined in this sample (Sulik, Eisenberg, Silva,
Spinrad, & Kupfer, 2013). Consequently, in this study,
we focus on the relations between RSA reactivity
during a series of self-regulation tasks and performance
on the self-regulation tasks. We address three research
questions.
One goal of this study was to describe the pattern of
changes in RSA over the course of each task and
between tasks. Previous studies have frequently aggregated RSA reactivity scores across multiple tasks without
examining whether patterns of change differ across tasks,
or how change scores across multiple tasks are related.
A second goal was to test whether RSA reactivity
across self-regulation tasks predicted children’s latent
self-regulation ability (based on task performance). We
hypothesized that children with greater RSA suppression (i.e., larger decreases from baseline RSA) would
have greater EC because task RSA has been found to
negatively predict performance on cognitively demanding tasks in older children, adolescents, and young
adults (Chapman, Woltering, & Lewis, 2010; Duschek,
Muckenthaler, Werner, & Reyes del Paso, 2009).
However, based on the findings reported by Marcovitch
et al. (2010) using a preschool-age sample, we also
tested for quadratic relations between RSA reactivity
and self-regulation.
RSA Reactivity and Effortful Control
Developmental Psychobiology
A third goal was to evaluate whether there were
unexplained relations between RSA reactivity in
response to each task and performance on that task that
was independent of the relation between RSA reactivity
and self-regulation performance across all tasks. We
tested this by estimating the residual covariance
between concurrently measured RSA reactivity and
performance for each of the three tasks that included
physiological recording (i.e., bird and dragon, knocktap, and gift wrap). A significant residual covariance
would indicate that the relation between RSA reactivity
and a child’s self-regulation ability does not completely
explain the relation between RSA reactivity and
concurrently measured task performance. Because RSA
has been theorized to index responsivity to changing
environmental demands (Porges, 1995), the presence of
substantial residual covariances when controlling for
latent self-regulation could potentially indicate that
RSA is related to an individual’s effort or task engagement (Duschek et al., 2009). However, the presence of
residual covariances could also indicate that RSA is
differentially related to specific types of self-regulation
tasks. Specifically, we predicted that RSA might relate
differently to performance on hot and cool selfregulation tasks.
METHOD
In this study, preschool-age children participated in two
laboratory visits. At the first visit, heart rate and
respiration were recorded during a baseline film and
during three self-regulation tasks. A computerized
continuous performance (CPT) task, another laboratory
measure of self-regulation, was completed in a second
laboratory session (without physiological monitoring).
Participants
Parental consent and child assent for physiological
recording were obtained for 106 children (42 girls)
attending any of three preschools at a large university
campus in the southwestern United States. One child
was dropped from analysis due to concerns about that
child having a developmental delay (see Sulik et al.,
2013). An additional four children were missing all
physiological data due to experimenter error (e.g.,
incorrect placement of respiration bellows or electrodes). All analyses include the remaining 101 children
(40 girls) with some usable physiological data. Age
ranged from 3.31 to 5.88 years (M ¼ 4.49, SD ¼ .64).
A parent from 83 families (69 mothers, 17 fathers)
returned questionnaires that included demographic
information. Median parental education (averaged
3
across the educational attainment of both parents)
reported on a 7-point scale ranging from “did not
graduate high school” to “Ph.D. or professional degree”
was “4-year college graduate.” Median family income
was $75,000–100,000, and 6% of children were from
single-parent families. Racial composition was as
follows: 73% Caucasian; 2% African American; 9%
Asian; 4% Native American; and 12% other/multiracial. Eighteen percent of children were of Mexican
American/Hispanic ethnicity.
Procedure
Laboratory visits were video recorded and were conducted by an experimenter and a camera person. Upon
entering the laboratory, the experimenter obtained verbal
assent and attached three electrodes to the child’s chest
and abdomen in an inverted triangle configuration and
placed a respiration bellows around the child’s torso.
Children were seated in front of a laptop computer. The
experimenter explained that the child would first watch
a short movie, and that he or she (i.e., the experimenter)
had computer work to do and would not be able to talk
while the movie was playing. Baseline physiological
data were collected while children watched a 2 min and
38 s meditation video showing dolphins swimming
while relaxing music played. If children stopped attending to the video, fidgeted excessively, or attempted to
engage the experimenter or camera person in conversation, the experimenter instructed the child to continue
watching the film. Following the film, children participated in three self-regulation tasks, with no breaks
between tasks. For each task, the experimenter first
explained the rules and, for the first two tasks, practiced
with the child. The experimenter started physiological
recording immediately before starting the assessment
period of each task, which was designed to last approximately 1 min. Physiological recording for each task
was terminated when the task ended. In a separate
laboratory session without physiological recording,
children completed a CPT, another measure of selfregulation.
Laboratory Measures of Effortful Control
Laboratory measures that are commonly used to study
self-regulation in preschool-age children were used in
this study. The video recordings of each task were
coded by two undergraduate research assistants who
were not involved in data collection. All tapes were
scored by a primary coder and at least 25% of the tapes
were independently scored by a reliability coder. Intraclass correlations (ICCs), used to assess reliability
between the two coders, are reported below for each
task.
4
Sulik et al.
Bird and Dragon. For the bird and dragon task
(Kochanska, Murray, Jacques, Koenig, & Vandegeest,
1996), the experimenter presented the child with two
puppets, a “nice” bird and a “mean” dragon. The
experimenter instructed the child to “Do what the nice
bird says” but “Don’t do what the mean dragon says.”
After completing practice trials to verify that the child
understood the rules of the game, these puppets were
used to give the child commands (e.g., “touch your
nose”). The actions used in the instructions for this task
were adapted to minimize the amount of movement
required, although some movement was required (e.g.,
“wiggle your fingers,” “touch your nose,” “touch your
hair”). Each of the 6 bird trials and 10 dragon trials
was scored as correct (2), partially correct (1), or
incorrect (0). Within the bird trials and the dragon
trials, scores were averaged. A composite measure of
performance on this task was computed as the product
of the bird (ICC ¼ 1.00) and the dragon (ICC ¼ .99)
scores to ensure that children had to perform well on
both types of trials to receive a high score. Although
we are primarily interested in responses to the inhibitory control trials, using the product score prevents a
child who failed to respond correctly to either the
inhibition or activation trials (e.g., due to impulsivity or
behavioral inhibition) from receiving a high score for
this task.
Knock-Tap. For the knock-tap task (Luria, 1966),
children were first instructed to imitate the experimenter when he or she knocked (with a closed fist) or
tapped (with an open palm) on a table for eight trials.
Then the experimenter instructed the child to knock
when the experimenter tapped, and to tap when the
experimenter knocked. Following practice trials to
ensure that children understood the rules, the experimenter performed 24 test trials that were scored
as correct or incorrect. Performance on this task
was computed as the proportion of correct trials
(ICC ¼ .98).
Gift Wrap. In the gift wrap task (Kochanska et al.,
1996), the experimenter explained that he or she had a
surprise for the child, and that the child should look
straight ahead at the wall while the experimenter
wrapped a gift behind the child. The experimenter
reminded the child not to peek and noisily “wrapped”
the gift for 1 min, after which the experimenter gave
the gift (an animal finger puppet) to the child. Higher
scores indicated better performance: 5 ¼ child does not
peek; 4 ¼ Child peeks, but does not turn body and does
not turn head over shoulder; 3 ¼ Child peeks, but does
not turn body; 2 ¼ Child turns body while peeking in
last 10 s, or child turns body while peeking for 3 s or
Developmental Psychobiology
less; 1 ¼ Child turns body while peeking for more than
3 s (ICC ¼ .81).
Continuous Performance Task. In a separate session
(without physiological recording), children completed a
CPT adapted from the NICHD Study of Early Child
Care and Youth Development (NICHD Early Child
Care Research Network, 2003). In this task, a series of
cartoon pictures were displayed on a laptop computer
with all keys covered except for the space bar. Children
were asked to press the space bar each time they saw a
fish, but not to press it when they saw other pictures
(e.g., a beach ball). Signal detection theory (Wickens,
2002) was used to score the CPT: the proportion of
errors of commission (i.e., pressing the space bar when
a fish was not presented) and errors of omission (i.e.,
failing to press the space bar when a fish was
presented) were used to generate a single detectability
(d 0 ) score that indicates how well a child performed
across both types of trials.
Physiological Data
Heart rate and respiration data were recorded using
James Long Company (Caroga Lake, New York) equipment and the data acquisition program SnapMaster for
Windows. Interbeat interval (IBI) data were scored
using James Long Company (2008) software and
visually inspected for errors. Missing IBI data, although
rare, were prorated based on the surrounding IBIs. RSA
(in ms2) was calculated with James Long Company
(2008) software using the peak-to-trough method
(Grossman, Karemaker, & Wieling, 1991) and was
subsequently natural log transformed to reduce skewness (Riniolo & Porges, 2000).
RESULTS
Missing Data and Preliminary Analyses
Age and sex were not missing for any participants.
Missing data percentages for other study variables were
as follows: bird and dragon ¼ 3%, knock-tap ¼ 5%, gift
wrap ¼ 1%, CPT ¼ 12%, baseline RSA ¼ 1%, bird and
dragon RSA ¼ 10%, knock-tap RSA ¼ 13%, gift wrap
RSA ¼ 17%. Child refusal resulted in a small amount
of missing physiological data for each task: n ¼ 2 for
bird and dragon; n ¼ 1 for knock-tap; and n ¼ 1 for gift
wrap. Artifact in the EKG also resulted in some
missing physiological data: n ¼ 3 for bird and dragon,
n ¼ 2 for knock-tap, and n ¼ 6 for gift wrap. The
remaining missing physiological data were attributable
to experimenter error or equipment failure (n ¼ 1 for
baseline RSA; n ¼ 6 for bird and dragon; n ¼ 4 for
knock-tap; and n ¼ 12 for gift wrap).
RSA Reactivity and Effortful Control
Developmental Psychobiology
5
Table 1. Correlations and Descriptive Statistics
1
2
3
4
5
6
7
8
9
10
Female
Age
Baseline RSA
Bird and Dragon RSA
Knock-tap RSA
Gift wrap RSA
Bird and Dragon Performance
Knock-tap Performance
Gift wrap performance
CPT performance
1
2
–
.14
.09
.11
.21
.04
.06
.09
.06
.16
–
.13
.22
.08
.14
.31
.49
.35
.42
3
–
.81
.76
.68
.11
.15
.24
.14
4
–
.79
.71
.20
.26
.23
.11
5
–
.69
.04
.08
.11
.03
6
–
.05
.12
.32
.09
7
–
.34
.40
.46
8
–
.31
.30
9
M
SD
–
.42
.40
4.49
6.39
6.42
5.52
5.89
0.66
0.53
0.84
2.86
.64
1.11
1.27
1.19
1.17
0.36
0.25
0.27
0.96
Note. p < .05 is bold.
Variables were inspected for high skewness and
kurtosis (West, Finch, & Curran, 1995) and the
presence of outliers. To improve the distributional
properties of the EC measures, a square-root transformation was applied to the knock-tap scores and an
inverse transformation was applied to the bird and
dragon and gift wrap scores (Tabachnik & Fidell,
2006). All analyses were conducted using Mplus 7.11
(Muthen & Muthen, 2012) using the robust maximum
likelihood estimator and full information maximum
likelihood as a missing data treatment. Descriptive
statistics and correlations among study variables (following transformation, as described above) are presented in Table 1.
The minimum and maximum log transformed RSA
values for each task were as follows: baseline ¼ 3.71
and 8.97; bird and dragon ¼ 3.76 and 9.62; knocktap ¼ 3.13 and 8.49; and gift wrap ¼ 3.42 and 8.96. We
computed the difference between each successive task
and calculated the percentage of individuals who
exhibited RSA suppression (i.e., a decrease in RSA
from the previous task). These percentages were as
follows: from baseline to bird and dragon ¼ 40%; from
bird and dragon to knock-tap ¼ 86%; and from knocktap to gift wrap ¼ 24%.
Analysis Plan
We began by presenting measurement models for EC
and for RSA and then examined the relations among
these constructs in a combined model that also included
children’s age and sex as predictors of latent EC. For
structural equation models, we report the model x2 and
the following fit indices: root mean square error of
approximation (RMSEA) with 90% confidence interval,
comparative fit index (CFI), and standardized root
mean square residual (SRMR).
Measurement Model for Effortful Control
Latent EC was defined by three tasks (bird and dragon,
knock-tap, and gift wrap) measured concurrently with
RSA and one additional task (CPT) completed in a
separate laboratory session. A single factor model for
EC fit the data well, x2(2) ¼ .456, p ¼ .796, RMSEA
¼ .000 (90% CI [.000, .125]), CFI ¼ 1.000, SRMR
¼ .012. Standardized factor loadings were as follows:
bird and dragon l ¼ .67 (SE ¼ .10), knock-tap l ¼ .49
(SE ¼ .10), gift wrap l ¼ .61 (SE ¼ .10), and CPT
l ¼ .65 (SE ¼ .11). All standardized factor loadings
were significant at p < .001. The reliability of the EC
factor was estimated to be .66, 95% CI [.49, .83]
(Raykov, 2009).
Measurement Model for Baseline RSA and RSA
Reactivity
We used latent difference score modeling (McArdle,
2009) to examine mean changes in RSA across the
four episodes (i.e., baseline, bird and dragon, knocktap, and gift wrap) and latent growth curve modeling
(Bollen & Curran, 2006) to examine mean changes
within each of the four episodes. The simultaneous use
of these two approaches in a single measurement model
allows us to describe (1) mean changes in RSA from
each task to the next; and (2) the mean changes in RSA
over the course of each individual task.
We created three RSA indicators for each episode,
with each indicator corresponding to one third of the
recording time for that task: total recording time was
2 min and 38 s for the baseline film and 60 s for each of
the EC tasks. As a consequence, RSA scores during
each of the EC tasks were based on 20 s of data, which
is comparable to prior research (Blair & Peters, 2003;
Marcovitch et al., 2010). The measurement model for
RSA is depicted in Figure 1. We estimated a linear
Sulik et al.
growth model with slope loadings fixed to 1, 0, and 1
so that the latent intercept for each task could be
interpreted as the average RSA score across that task.
In preliminary models, the slope variances were not
significantly different from zero, so we fixed all four
slope variances to zero. The intercepts for each episode
were used as the basis for latent difference score
modeling. This model fit well, x2(60) ¼ 54.005,
p ¼ .693, RMSEA ¼ .000 (90% CI [.000, .049]), CFI1
¼ 1.000, SRMR ¼ .051. The variance explained in each
of the indicators ranged from .74 to .97.
Mean Changes in RSA Within Each Task. There
were no changes in RSA over time within the baseline
period, slope M ¼ .03, SE ¼ .04, p ¼ .526. However,
RSA declined over time (indicating RSA suppression)
within the bird and dragon task and within the knocktap task, slope Ms ¼ .09 and .18, SEs ¼ .03 and .04,
p ¼ .005 and p < .001. There was no change in RSA
over time within the gift wrap task, slope M ¼ .01,
SE ¼ .05, p ¼ .813.
Mean Changes in RSA Across Tasks. With respect to
changes across tasks, there was no mean change from
baseline to bird and dragon, M ¼ .05, SE ¼ .08,
p ¼ .528. This was followed by a substantial decrease
(i.e., RSA suppression) from bird and dragon to knocktap, M ¼ .93, SE ¼ .09, p < .001, and a recovery (i.e.,
RSA augmentation) from knock-tap to gift wrap,
M ¼ .40, SE ¼ .11, p < .001. The cumulative change in
RSA from baseline to gift wrap was negative and
significantly different from zero, M ¼ .48, SE ¼ .09,
p < .001, indicating that RSA levels during the gift
wrap task had not fully recovered to baseline levels.
The absence of a complete recovery is consistent with
prior research, which suggests that watching a film is
associated with elevated RSA (Bush, Alkon, Obradovic,
Stamperdahl, & Boyce, 2011).
All correlations among baseline RSA and the latent
difference scores were freely estimated in this model.
The intercept for baseline RSA was unrelated to the
first (bird and dragon), second (knock-tap), and
third (gift wrap) latent difference scores: fs ¼ .06,
.16, and .14, SEs ¼ .11, .12, and .13, and ps ¼ .602,
.173, and .281. The lack of significant correlations
between the baseline RSA scores and the RSA
reactivity scores indicates that subsequent changes in
1
We used an alternative null model in which the repeated
measures were modeled using a random intercept model (i.e., no
growth over time) when calculating the CFI in all models that
include latent growth curves because the standard null model is
not nested with a growth curve model (Widaman & Thompson,
2003).
Developmental Psychobiology
FIGURE 1 Baseline RSA and RSA reactivity: the within-task latent growth curve and between-task latent difference score model. BD, Bird and Dragon; KT,
Knock-tap; GW, Gift wrap.
6
RSA Reactivity and Effortful Control
Developmental Psychobiology
RSA did not depend on baseline values. In contrast, the
first latent difference score was negatively associated
with the second latent difference score, f ¼ .54,
SE ¼ .09, p < .001, and the second latent difference
score was, in turn, negatively associated with the third
latent difference score, f ¼ .44, SE ¼ .12, p < .001.
The first latent difference score was not significantly
related to the third latent difference score, f ¼ .08,
SE ¼ .19, p ¼ .676. These correlations among adjacent
latent difference scores (i.e., from the first to second
task, and from the second to third task) indicate that
changes in RSA were negatively related to the immediately preceding change score. This means that individuals with more positive than average DRSA scores
in response to one task were likely to have more
negative than average DRSA change scores in response
to the next task. The nonsignificant correlation between
the first and third RSA change score suggests that this
dependency on prior changes does not extend beyond a
single task.
Combined Structural Models
To examine the relations between baseline RSA, RSA
reactivity across the three tasks, and EC, we switched
to a latent residualized change score model for RSA
because this model is ideally suited to test hypotheses
about RSA reactivity independent of prior RSA scores
(Burt & Obradovic, 2013). This change was accomplished by replacing all (nondirectional) correlations
among the baseline RSA intercept and the latent
difference scores with directional paths in which later
scores are predicted by earlier scores. In this parameterization, the residualized latent difference (DRSA)
scores can be interpreted as the change in RSA that is
orthogonal to baseline RSA and any prior DRSA
scores. In this model, latent EC was predicted by age
and sex of the child, baseline RSA intercept, and all
three DRSA scores. In addition, we included residual
covariances between concurrently measured DRSA
scores and task performance. This model fit the data
well, x2(133) ¼ 135.281, p ¼ .429, RMSEA ¼ .013
(90% CI [.000, .050], CFI ¼ .986, SRMR ¼ .079.
To test whether each subsequent measure of RSA
uniquely contributed to prediction, we adopted a
hierarchical regression approach. We first constrained
all the relations between the RSA scores and EC to
zero, and successively released these constraints
to estimate the unique contributions of each of
these predictors over and above all prior RSA scores.
These results are presented in Table 2. Although
baseline RSA and the knock-tap DRSA score both
added to prediction, the only physiological variable
that uniquely predicted EC in the final model was the
7
knock-tap DRSA score. In the final model, there was
a significant standardized residual covariance between
performance on the gift-wrap task and gift wrap
DRSA, f ¼ .34, SE ¼ .14, p ¼ .015. This indicates that
performance on the gift wrap task that was independent of the overall EC factor was related to concurrently measured DRSA. This residual covariance
was positive, which means that individuals who
exhibited more positive changes in RSA from the
knock-tap task to the gift wrap task performed better
than expected on the gift wrap task relative to their
latent self-regulation ability, as measured by the EC
factor. The corresponding standardized residual covariances for bird and dragon and for knock-tap were
nonsignificant, fs ¼ .21 and .18, SEs ¼ .12 and .25,
ps ¼ .091 and .487.3
Given the complexity of testing quadratic terms as
predictors in a latent difference score model, we
adopted a simpler approach for testing quadratic
relations between RSA reactivity and the EC factor.
Estimating a separate model for each task, we tested
task RSA and the square of task RSA as predictors of
latent EC while controlling for age and sex of the child
and baseline RSA. None of the quadratic terms
significantly predicted latent EC, indicating the absence
of a quadratic relation between RSA reactivity and
self-regulation.
2
DISCUSSION
To our knowledge, this is the first manuscript that has
attempted to differentiate within-task RSA reactivity
and between-task RSA reactivity. Separating these two
types of change yielded an interesting pattern of
descriptive and predictive results that would have been
obscured by aggregating the RSA reactivity scores
across tasks.
Describing Patterns of RSA Reactivity
One goal of this investigation was to describe the
pattern of changes in RSA during and across a baseline
period and a series of self-regulation tasks. Mean
within-task changes in RSA were not consistent for
When each DRSA score was examined as a predictor while
controlling for baseline RSA but not the other DRSA scores, the
second DRSA score (from bird and dragon to knock-tap) was the
only DRSA score to significantly predict the EC factor.
3
For comparison, we also estimated a model combining
RSA reactivity across all three self-regulation tasks. In this
model, the combined RSA reactivity latent variable was unrelated
to the EC factor while controlling for age, sex, and baseline
RSA.
2
8
Sulik et al.
Developmental Psychobiology
Table 2. Hierarchical Regression Results
Predictor Added
DRSA BD
Baseline RSA
b
Age
Female
Baseline RSA
DRSA BD
DRSA KT
DRSA GW
R2
DR2
Dx2(df ¼ 1)
.61
.26
.13
SE
.13***
.15†
.05*
42.2%
2.3%
4.744*
b
b
SE
.61
.62
.26
.13
.07
.13***
.15†
.05*
.12
.21
42.7%
.5%
.305
DRSA KT
b
.61
.21
.06
b
.57
.15
.10
.29
.41
SE
.12***
.15
.06†
.18†
.19*
54.7%
12.0%
4.286*
DRSA GW
b
.59
.16
.30
.49
b
SE
.56
.16
.08
.36
.52
.12
56.0%
1.3%
.595
.12***
.15
.06
.20†
.23*
.15
b
.58
.13
.36
.61
.17
Note. Satorra-Bentler adjustment was applied to x2 difference tests (Satorra & Bentler, 2001). BD, Bird and dragon; KT, Knock-tap; GW, Gift
wrap.
***p < .001; **p < 01; * p < .05; † ¼ p < .10.
different tasks: RSA was stable over the course of the
baseline period, declined over the course of the bird
and dragon and knock-tap tasks, and remained stable
over the course of the gift wrap task. It is notable that
the pattern of within-task changes differed for the more
cognitive self-regulation tasks (bird and dragon; knocktap) and the delay of gratification task (gift wrap). RSA
declined over the course of the more cognitive tasks,
but was stable during the delay of gratification task.
However, we cannot rule out the possibility that the
fixed ordering of the tasks or different task demands
(e.g., the need for motor activity during the bird and
dragon and knock-tap tasks), rather than the difference
in motivational salience between the “cool” and “hot”
self-regulation tasks, influenced the within-task pattern
of changes in RSA.
Unlike Brooker and Buss (2010), who estimated
growth in RSA across a fear-eliciting task in a sample
of 2-year olds, we did not find significant variability for
the within-task slopes. It is unclear whether this
indicates that RSA reactivity over the course of
cognitively demanding self-regulation tasks is less
variable than responses to emotion-eliciting tasks, or
whether the selected nature of Brooker and Buss’
(2010) sample (which included a mix of typically
developing and high-fear toddlers) contributed to variability in physiological reactivity to the fear-eliciting
episode. The shorter length of time in which RSA was
recorded could also have contributed to differences
between our results and the results reported by Brooker
and Buss (2010). Additional research is needed to
understand how task demands and individual differences in emotional reactivity and self-regulation affect
variability in the slope of RSA over the course of a
single task.
With respect to changes in mean RSA across tasks,
we also found a differentiated pattern of changes. RSA
did not change from baseline to bird and dragon,
decreased substantially from bird and dragon to knock
tap, and then showed a partial recovery toward baseline
values from knock tap to gift wrap. This pattern of
mean-level change was somewhat surprising because
bird and dragon and knock-tap were characterized by
similar cognitive and motor demands, so differences in
mean RSA were not hypothesized for these tasks (Bush
et al., 2011). One possibility is that children were not
fully relaxed during the baseline film, which could
potentially limit initial RSA reactivity. Ordering effects
and differences in task difficulty, the amount of gross
motor movement required by bird and dragon and
knock-tap, the slower pace of the bird and dragon trials
relative to the knock-tap trials, or other task differences
such as greater language processing during bird and
dragon could also have contributed to differences in
average RSA reactivity for bird and dragon and knocktap. In future research, the use of matched control tasks
(Bush et al., 2011), counterbalanced task order, and
experimentally varying task characteristics are suggested to better understand how task characteristics
affect RSA reactivity. Because investigators have drawn
somewhat conflicting conclusions about the effect of
small amounts of movement on RSA reactivity (Bush
et al., 2011; Porges et al., 2007), investigators might
consider using actigraphs to quantify movement in
future studies of RSA reactivity.
Correlations among the latent difference scores in
the RSA measurement model indicated that changes in
RSA were negatively associated across tasks, such that
individuals with more positive than average RSA
change scores in response to a given task tended to
Developmental Psychobiology
have more negative than average RSA change scores in
response to the subsequent task. This serial dependence
suggests that earlier RSA reactivity can potentially
constrain subsequent RSA reactivity (Thayer & Lane,
2000). In future investigations in which multiple tasks
are used, we recommend that researchers allow a
sufficient amount of time between each task to prevent
serial dependency in RSA reactivity scores. For young
children, who have a limited attention span, this
suggestion should be balanced against the time required
to administer multiple tasks. At present, it is difficult to
judge how much of a problem serial dependency is in
the published literature because investigators using
multiple RSA reactivity tasks do not typically report
the correlations among change scores. We suggest that
even when multiple baselines are used (e.g., Bush
et al., 2011), investigators should report whether serial
dependence of RSA scores extends beyond a single
task and adjust appropriately for such effects (Burt &
Obradovic, 2013).
Relations Between Task Performance and RSA
Reactivity
One goal of this study was to determine whether RSA
reactivity in response to a battery of self-regulation
tasks would predict children’s self-regulatory performance over and above baseline RSA. The inclusion
of RSA reactivity scores across three different selfregulation tasks allowed to us to examine the unique
contributions of baseline RSA and RSA reactivity to
each task in the prediction of children’s self-regulation.
RSA during the the knock-tap task was the only RSA
score that uniquely predicted self-regulation when
baseline RSA and all three DRSA scores were included
as predictors of latent EC. Consistent with other studies
in older populations (Chapman et al., 2010; Duschek
et al., 2009), RSA reactivity (in this case, only for the
knock-tap task) was negatively related to EC: greater
than expected decreases in RSA predicted better
performance on the self-regulation tasks. This differential pattern of results is noteworthy because the first
two tasks (bird and dragon, knock-tap) had similar
cognitive and motor demands (Bush et al., 2011). Our
results suggest that initial reactivity to self-regulation
tasks could be less informative (relative to baseline
values) than subsequent reactivity in response to
sustained demands on self-regulation. Furthermore, this
differential pattern of relations between measures of
RSA reactivity and EC would have been obscured if
RSA reactivity had been aggregated across tasks rather
than examined separately.
Unlike Marcovitch et al. (2010), we did not find
evidence of a quadratic relation between RSA reactivity
RSA Reactivity and Effortful Control
9
and self-regulation. Differences in the measures
could potentially explain this difference in findings.
Marcovitch et al.’s (2010) measure of self-regulation
included working memory, which is considered an
aspect of executive function but not an aspect of
effortful control (Eisenberg & Zhou, in press); however,
baseline RSA was positively related to self-regulation
in both studies. In the present investigation, children
generally performed at a high level on the EC
tasks. The relations between RSA reactivity and selfregulation could potentially be different for children
who, on average, find the tasks to be more challenging
(Marcovitch et al., 2010).
Another goal of this study was to determine whether,
after accounting for the relations between the latent EC
factor and RSA reactivity, there would be residual
associations between RSA reactivity in response to
each task and concurrent performance on each selfregulation task. Although RSA reactivity to the third
task did not predict the latent factor, which measured
EC performance across all tasks, it was positively
related to the residual of performance on the third task
after accounting for the EC factor. The third task, gift
wrap, was unique in that it involved delay of gratification and may have measured temperamental surgency
(impulsivity or approach behavior) as well as EC. One
possibility is that RSA reactivity to the third task was
related to surgency over and above the self-regulatory
demands of this task, although this finding will need to
be replicated to verify that it was not idiosyncratic to
the ordering of the tasks or to other characteristics of
the tasks such as movement, which was required by the
more cognitive tasks (bird and dragon, knock-tap) but
not the delay task. Some investigators have reported
that tasks involving “hot” and “cool” self-regulation are
differentially related to children’s academic and socioemotional outcomes (Kim et al., 2012; Willoughby
et al., 2011). The possibility that measures of autonomic function differentially relate to children’s EC
and more reactive aspects of temperament such as
surgency is an important avenue for future research.
Limitations and Strengths
One limitation was that our sample was composed of
primarily middle-class, nonminority children, all of
whom were attending preschool, and it is unclear
whether these results would generalize to other populations. This was balanced by considerable strengths,
including the high-quality measurement of EC (which
included four structured laboratory tasks) and the use
of latent variable modeling, which provides more
accurate estimates of relations among constructs by
accounting for measurement error (Bollen, 1989).
10
Sulik et al.
Conclusion
This study draws attention to the need to carefully
consider the measurement of RSA reactivity when
multiple tasks are used. Our results suggest that simply
aggregating across multiple tasks when calculating
RSA reactivity is inappropriate, and we provide evidence that more sophisticated methods such as modeling RSA reactivity over time within a single task
(Brooker & Buss, 2010) and using latent variable
methods for analyzing change (Burt & Obradovic,
2013) can provide a more nuanced understanding of
RSA reactivity and that these methods can be applied
even in relatively small samples. Finally, we find that
more positive change in RSA in response to some selfregulation tasks predicted preschoolers’ worse selfregulation performance (even while controlling for
baseline RSA), although more needs to be done to
understand how this relation is influenced by specific
task characteristics.
NOTES
This research was supported by a grant from the Arizona State
University Graduate and Professional Student Association to
Michael J. Sulik and a grant from the National Institute of
Mental Health to Nancy Eisenberg and Tracy L. Spinrad. We
express our appreciation to the parents and children who
participated in the study and to the many research assistants
who contributed to this project. We also gratefully acknowledge Anne Kupfer and Snjezana Huerta for their assistance
with data collection.
REFERENCES
Blair, C., & Peters, R. (2003). Physiological and neurocognitive correlates of adaptive behavior in preschool
among children in Head Start. Developmental Neuropsychology,
42,
479–497.
doi:
10.1207/
S15326942DN2401_04
Bollen, K. A. (1989). Structural equations with latent
variables. New York, NY: Wiley.
Bollen, K. A., & Curran, P. J. (2006). Latent Curve Models.
Hoboken, NJ: John Wiley & Sons.
Brooker, R. J., & Buss, K. A. (2010). Dynamic measures of
RSA predict distress and regulation in toddlers. Developmental Psychobiology, 52, 372–382. doi: 10.1002/
dev.20432
Burt, K. B., & Obradovic, J. (2013). The construct of
psychophysiological reactivity: Statistical and psychometric issues. Developmental Review, 33, 29–57. doi:
10.1016/j.dr.2012.10.002
Bush, N. R., Alkon, A., Obradovic, J., Stamperdahl, J., &
Boyce, W. T. (2011). Differentiating challenge reactivity
Developmental Psychobiology
from psychomotor activity in studies of children’s psychophysiology: Considerations for theory and measurement.
Journal of Experimental Child Psychology, 110, 62–79.
doi: 10.1016/j.jecp.2011.03.004
Chapman, H. A., Woltering, S., & Lewis, M. D. (2010).
Hearts and minds: Coordination of neurocognitive and
cardiovascular regulation in children and adolescents.
Biological Psychology, 84, 296–303. doi: 10.1016/j.biopsycho.2010.03.001
Duschek, S., Muckenthaler, M., Werner, N., & Reyes del
Paso, G. A. (2009). Relationships between features of
autonomic cardiovascular control and cognitive performance. Biological Psychology, 81, 110–117. doi: 10.1016j.
biopsycho.2009.03.003
Eisenberg, N., Spinrad, T. L., & Eggum, N. D. (2010).
Emotion-related self-regulation and its relation to
children’s maladjustment. Annual Review of Clinical
Psychology, 6, 495–525. doi: 10.1146/annurev.clinpsy.
121208.131208
Eisenberg, N., & Zhou, Q., (in press). Conceptions of
executive functioning and regulation: When and to what
degree do they overlap? In J. A. Griffin, L. S. Freund, &
P. McCardle (Eds.), Executive function in preschool age
children: Integrating measurement, neurodevelopment and
translational research. Washington, DC: American Psychological Association.
Grossman, P., Karemaker, J., & Wieling, W. (1991). Prediction of tonic parasympathetic cardiac control using respiratory sinus arrhythmia: The need for respiratory control.
Psychophysiology, 28, 201–216. doi: 10.1111/j.14698986.1991.tb00412.x
Hansen, A. L., Johnsen, B. H., & Thayer, J. F. (2003). Vagal
influence on working memory and attention. International
Journal of Psychophysiology, 48, 263–274. doi: 10.1016/
S0167-8760(03) 00073-4
Happaney, K., Zelazo, P. D., & Stuss, D. T. (2004). Development of orbitofrontal function: Current themes and future
directions. Brain and Cognition, 55, 1–10. doi: 10.1016/j.
bandc.2004.01.001
James Long Company. (2008). IBI analysis system. Retrieved
from http://www.jameslong.net/#IBI%20Analysis%20System
Kim, S., Nordling, J. K., Yoon, J. E., Boldt, L. J., &
Kochanska, G. (2012). Effortful control in “hot” and
“cool” tasks differentially predicts children’s behavior
problems and academic performance. Journal of Abnormal
Child Psychology. doi: 10.1007/s10802-012-9661-4
Kochanska, G., Murray, K., Jacques, T. Y., Koenig, A. L., &
Vandegeest, K. A. (1996). Inhibitory control in young
children and its role in emerging internalization. Child
Development, 67, 490–507. doi: 10.1111/j.14678624.1996.tb01747.x
Lagattuta, K. H., Sayfan, L., & Monsour, M. (2011). A new
measure for assessing executive function across a wide
age range: Children and adults find happy-sad more
difficult than day-night. Developmental Science, 14, 481–
489. doi: 10.1111/j.1467-7687.2010.00994.x
Liew, J. (2012). Effortful control, executive functions, and
education: Bringing self-regulatory and social-emotional
Developmental Psychobiology
competencies to the table. Child Development Perspectives, 6, 105–111. doi: 10.1111/j.1750-8606.2011.00196.x
Luria A. R. (1966). Higher cortical functions in man. New
York, NY: Basic Books.
Marcovitch, S., Leigh, J., Calkins, S. D., Leerks, E. M.,
O’Brien, M., & Blankson, A. N. (2010). Moderate vagal
withdrawal in 3.5-year-old children is associated with
optimal performance on executive function tasks. Developmental Psychobiology, 52, 603–608. doi: 10.1002/
dev.20462
McArdle, J. J. (2009). Latent variable modeling of differences
and changes with longitudinal data. Annual Review of
Psychology,
60,
577–605.
doi:
0.1146/annurev.
psych.60.110707.163612
Metcalfe, J., & Mischel, W. (1999). A hot/cool-system
analysis of delay of gratification: Dynamics of willpower.
Psychological Review, 106, 3–19. doi: 10.1037/0033295X.106.1.3
Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H.,
Howerter, A., & Wager, T. D. (2000). The unity and
diversity of executive functions and their contributions to
complex “frontal lobe” tasks: A latent variable analysis.
Cognitive Psychology, 41, 49–100. doi: 10.1006/cogp.
1999.0734
Muthen, L. K., & Muthen, B. O. (2012). Mplus User’s Guide
(7th ed.). Los Angeles, CA: Muthen & Muthen. Retrieved
from http://www.statmodel.com
NICHD Early Child Care Research Network. (2003).
Do children’s attention processes mediate the link
between family predictors and school readiness? Developmental Psychology, 39, 581–593. doi: 10.1037/00121649.39.3.581
Porges, S. W. (1995). Cardiac vagal tone: A physiological
index of stress. Neuroscience and Biobehavioral Reviews,
19, 225–233. doi: 10.1016/0149-7634(94) 00066-A
Porges, S. W. (2007). The polyvagal perspective. Biological
Psychology, 74, 116–143. doi: 10.1016/j.biopsycho.2006.06.009
Porges, S. W., Bohrer, R. E., Keren, G., Cheung, M. N.,
Franks, G. F., & Drasgow, F. (1981). The influence of
methylphenidate on spontaneous autonomic activity and
behavior in children diagnosed as hyperactive. Psychophysiology, 18, 42–48. doi: 10.1111/j.1469-8986.1981.
tb01541.x
Porges, S. W., Heilman, K. J., Bazhenova, O. V., Bal, E.,
Doussard-Roosevelt, J. A., & Koledin, M. (2007).
Does motor activity during psychophysiological paradigms
confound the quantification and interpretation of heart
rate and heart rate variability measures in young children?
RSA Reactivity and Effortful Control
11
Developmental Psychobiology, 49, 485–494. doi: 10.1002/
dev.20228
Raykov, T. (2009). Evaluation of scale reliability for unidimensional measures using latent variable modeling. Measurement and Evaluation in Counseling and Development,
42, 223–232. doi: 10.1177/0748175609344096
Riniolo, T. C., & Porges, S. W. (2000). Evaluating group
distributional characteristics: Why psychophysiologists
should be interested in qualitative departures from the
normal distribution. Psychophysiology, 37, 21–28. doi:
10.1111/1469-8986.3710021
Satorra, A., & Bentler, P. M. (2001). A scaled difference chisquare test statistic for moment structure analysis. Psychometrika, 66, 507–514. doi: 10.1007/BF02296192
Staton, L., El-Sheikh, M., & Buckhalt, J. A. (2009).
Respiratory sinus arrhythmia and cognitive functioning in
children. Developmental Psychobiology, 51, 249–258. doi:
10.1002/dev.20361
Sulik, M. J., Eisenberg, N., Silva, K. M., Spinrad, T. L., &
Kupfer, A. (2013). Respiratory sinus arrhythmia, shyness,
and effortful control in preschool-age children. Biological
Psychology, 92, 241–248. doi: 10.1016/j.biopsycho.2012.10.009
Tabachnik, B. G., & Fidell, L. S. (2006). Using multivariate
statistics (5th ed.). Boston, MA: Pearson/Allyn & Bacon.
Thayer, J. F., & Lane, R. D. (2000). A model of neurovisceral
integration in emotion regulation and dysregulation. Journal of Affective Disorders, 61, 201–216. doi: 10.1016/
S0165-0327(00) 00338-4
West, S. G., Finch, J. F., & Curran, P. J., (1995). Structural
equation models with nonnormal variables: Problems and
remedies. In R. H. Hoyle (Ed.), Structural equation modeling:
Concepts, issues, and applications. Newbury Park, CA: Sage.
Wickens, T. D. (2002). Elementary signal detection theory.
Oxford: Oxford University Press.
Widaman, K. F., & Thompson, J. S. (2003). On specifying
the null model for incremental fit indices in structural
equation modeling. Psychological Methods, 8, 16–37. doi:
10.1037/1082-989X.8.1.16
Willoughby, M. T., Kupersmidt, J., Voegler-Lee, M., &
Bryant, D. (2011). Contributions of hot and cool selfregulation to preschool disruptive behavior and academic
achievement. Developmental Neuropsychology, 36, 162–
180. doi: 10.1080/87565641.2010.549980
Zhou, Q., Chen, S. H., & Main, A. (2012). Commonalities
and differences in the research on children’s effortful
control and executive function: A call for an integrated
model of selfregulation. Child Development Perspectives,
6, 112–121. doi: 10.1111/j.1750-8606.2011.00176.x