A method for measuring dynamic respiratory sinus arrhythmia (RSA) in infants and mothers

https://doi.org/10.1016/j.infbeh.2021.101569Get rights and content

Highlights

  • Respiratory sinus arrythmia (RSA) is often used to study physiological regulation.

  • Measuring RSA is typically completed at a low resolution of 30 s or longer.

  • A new high-resolution measure of dynamic RSA is introduced.

  • A step-by-step tutorial and key properties of RSA time series are provided.

  • Dynamic RSA can be used to study physiological synchrony and regulation.

Abstract

The measurement of respiratory sinus arrythmia (RSA) in infants, children and adults is critical to the study of physiological regulation, and more recently, interpersonal physiological covariation, but it has been impeded by methods that limit its resolution to 30 s or longer. Recent analytical developments have suggested methods for studying dynamic RSA in adults, and we have extended this work to the study of infants and mothers. In the current paper, we describe a new analytical strategy for estimating RSA time series for infants and adults. Our new method provides a means for studying physiological synchrony in infant-mother dyads that offers some important advantages relative to existing methods that use inter-beat-intervals (e.g. Feldman, Magori-Cohen, Galili, Singer, & Louzoun, 2011). In the middle sections of this paper, we offer a brief tutorial on calculating RSA continuously with a sliding window and review the empirical evidence for determining the optimal window size. In order to confirm the reliability of our results, we briefly discuss testing synchrony by randomly shuffling the dyads to control for spurious correlations, and also by using a bootstrapping technique for calculating confidence intervals in the cross-correlation function. One important implication that emerges from applying this method is that it is possible to measure both positive and negative physiological synchrony and that these categorical measures are differentially predictive of future outcomes.

Introduction

Self-regulation is foundational to the healthy development of human infants as they learn to control their state, attention, emotion, and behavior (Belsky, Friedman, & Hsieh, 2001; Gross, 2013; Moore & Calkins, 2004, Porges, 1996, Richards, 1985; Rothbart, Poser, & Boylan, 1990). In the past two decades, there has been growing interest in the study of physiological regulation as a complement to these behavioral studies (e.g. Conradt & Ablow, 2010; Moore & Calkins, 2004; Perry, Calkins, & Bell, 2016). Much of this research has focused on parasympathetic functioning as a physiological substrate of emotional reactivity and regulation (Beauchaine, 2015; Berntson, Quigley, & Lozano, 2007; Blandon, Calkins, Keane, & Brien, 2010). A frequently used measure for studying the regulation of the parasympathetic nervous system (PNS) is respiratory sinus arrhythmia (RSA).

RSA measures cardiac vagal tone which is derived from heart rate variability at the frequency of breathing and serves as an indicator of the parasympathetic influence on heart rate (Porges, 1985, 1991). During periods of calm and quiescence the autonomic nervous system (ANS) via the vagus nerve services the needs of the internal viscera to enhance growth and restoration, but during environmental challenges the ANS responds to environmental challenges by increasing metabolic output via vagal withdrawal and sympathetic excitation.

It is important to appreciate that heart rate and RSA are related but are different measures of cardiac activity. Vagal stimulation of the sinoatrial (S-A) node delays the onset of the heart beat resulting in heart rate deceleration, whereas vagal withdrawal (i.e., delay of neural transmission) reduces the time between heart beats resulting in heart rate acceleration (Porges, Doussard-Roosevelt, & Maiti, 1994). Heart rate reflects the degree of physiological arousal experienced by the individual. By contrast, vagal tone reflects the regulation of this arousal by controlling the ‘vagal brake’ which is responsible for slowing down heart rate and metabolic output. In addition, heart rate or heart period (i.e., inter-beat-interval), as it is more typically measured, is a composite output reflecting autonomic input from the myelinated vagal fibers, the sympathetic nervous system, and the unmyelinated vagal fibers. It is thus difficult to establish whether one or more of these neural systems are responsible for changes in heart rate as a function of a specific event (cf., Abney et al., under review). By contrast, RSA is primarily a measure of the ventral vagal complex (i.e., the myelinated vagal fibers), and thus changes in RSA reflect changes in the activation of this specific neural circuit.

This distinction between heart rate and cardiac vagal tone becomes especially critical when studying interactional synchrony between infants and parents. Early on young infants demand a great deal of co-regulation from their parents which they receive through face-to-face interactions, close physical contact, and vocal and affective turn-taking (e.g. Brazelton, Koslowski, & Main, 1975; Papousek, 1995; Stern, 1974). These behaviors are all examples of interpersonal synchrony that involve a dynamic and continuous matching of behavioral, physiological, and neural responses that are reciprocally coordinated between parents and their infants (Feldman, 2017). Although evidence for behavioral synchrony has existed for decades, there are a number of technological challenges to studying physiological synchrony. One approach was utilized by Feldman and colleagues (Feldman et al., 2011) who measured the coordination of heart-rate rhythms (inter-beat-intervals or IBIs) within one second of each other. A problem with this measure is that there are significant differences in the frequencies of heart rates between infants and adults. These differences result in very large phase lags between the two signals that are highly variable, and thus significantly limit the period of time during which synchrony can be measured. By contrast, RSA changes at similar frequencies for infants and adults, and thus it is possible to measure synchrony for extended periods of time. In order to avoid any misunderstanding, the frequency of change in RSA should not be confused with the respiratory frequencies used to determine the portion of heart rate variability involved in calculating RSA. These respiratory frequencies differ for adults and infants and range between .12 and .40 Hz for adults and .30 and 1.30 for infants (Porges, 1985).

One significant impediment to implementing the standard measure of RSA as an index of dynamic changes in self-regulation is that the periodicity of RSA is fairly slow ranging from a little less than once per second to once every 8 s. The Nyquist sampling theorem recommends a minimum of two complete cycles when analyzing the variability or power in a time series. As a consequence, RSA is typically derived as an aggregate measure from periods of time that range between 30 and 120 s (e.g. Ham & Tronick, 2006; Moore et al., 2009; Waters, West, & Mendes, 2014). This is problematic because an average RSA estimate across these relatively long periods will likely lose out on critical information about the individuals’ reactivity and dynamic regulation of emotion (Miller et al., 2013). Given that the timescale of most social behaviors, such as mutual gaze or smiling or orienting, is only a few seconds (Kaye & Fogel, 1980; Van Egeren, Barratt, & Roach, 2001), estimates of RSA that aggregate responses over 30 s or longer are likely averaging out critical information about the multidirectional relationships between physiological and behavioral regulation, and how these processes are constrained by and also influence the social interactions between infants and their caregivers.

As a solution to this problem, we developed a continuous measure of RSA that can be sampled multiple times per second. In the current paper, we provide a step-by-step description of how to estimate continuous RSA for infants and their adult caregivers, and also demonstrate how this measure offers new insights into the dyadic variability and interpersonal co-regulatory dynamics of infants and their mothers during a face-to-face play session.

Recent work has led to the development of novel analytic tools for estimating second-to-second RSA dynamics in adult populations. Gates, Gatzke-Kopp, Sandsten, & Blandon (2015) used a combination of spectral analysis and a sliding window to measure RSA at 1 Hz in order to study the physiological linkage of husbands and wives during a family play time session. One limitation of this technique is that spectral analysis assumes stationarity (i.e., mean and standard deviation is constant across the entire time series), but this condition is rarely, if ever, maintained with human responses. Fisher, Reeves, and Chi (2016) developed a measure of Dynamic RSA (dRSA), which provides second-to-second variation in RSA by modeling the relationships among RSA, inter-beat intervals, and respiration rate using vector autoregression. It is difficult, however, to obtain direct measures of respiration in young infants (for an exception see McFarland, Fortin, & Polka, 2020). Both analytic techniques provide researchers with useful ways to index RSA dynamics and physiological synchrony in adult populations. Neither of these techniques is sufficient, however, for addressing the unique challenges created when measuring RSA synchrony across different age groups.

The contribution of the current paper is to describe a novel analytic technique that improves on the limitations of the above-mentioned methods, and provides developmental researchers with a time-based continuous measure of RSA that is commensurate across developmental populations. Our first goal is to describe this method for estimating the RSA time series continuously for infants and adults. Although there are alternative methods for estimating RSA (e.g., frequency-based methods), we limit our discussion to the Porges and Bohrer (1990) time-based method that is frequently used in the study of infants and toddlers. One of the principal advantages of this method is that it includes a band-pass polynomial filter that controls for the stationarity of the times series and removes the linear dependencies (i.e., autocovariance function) in the signal. In other words, the polynomial filter removes linear dependencies up to the cutoff frequency. After describing our method, we discuss some properties of time-based estimates of RSA and how to use this measure for studying physiological synchrony. This will include the significance testing of cross-correlation coefficients when measuring physiological synchrony. Specifically, we will discuss both phase shuffling and bootstrapping techniques to control for spurious effects. Lastly, we will compare the results of measuring dynamic RSA and inter-beat intervals (IBIs) to demonstrate that they are very different measures.

Section snippets

Transforming RSA from a discrete to a continuous measure

Our analytical contribution extends the method described in Porges and Bohrer (1990) to estimate RSA using a time-based approach. The Porges and Bohrer (1990) method for estimating RSA involves four main steps:

  • 1

    Electrocardiograms are recorded at a high frequency (e.g., 1 KHz) and then visually inspected to remove artifacts (e.g., movements, ectopic beats, etc.).

  • 2

    Inter-beat-intervals are then estimated and filtered for periodic and aperiodic components.

  • 3

    A bandpass filter is applied to the time

Data collection and measurement of RSA time series

We begin by describing the procedure we follow in estimating RSA continuously with a sliding window and also to explore the parameter space of the sliding window. As will become clear, the reliability of the RSA measure will vary across the parameterization of sliding window sizes selected for RSA estimation, and as such will provide informed suggestions for which window size lengths are best to use. The ECG data used for demonstrating our method was drawn from a larger study testing 4- to

Properties of the RSA time series as a function of window size

There are many techniques for measuring the patterns and structure of time series data. One simple question that is afforded by an estimation of a continuous measure of RSA is whether or not RSA increases, decreases, or does not change throughout the play phase. To measure trends in RSA time series, we submitted each RSA time series (for infants and mothers and across sliding window sizes) to linear regression. We categorized each time series as ‘positive’ if p < .001 and the regression beta

Testing the covariation between infants’ and mothers’ RSA time series

One of the major motivations for creating a continuous measure of RSA is to study physiological synchrony by calculating the cross-correlation function of the infants’ and mothers’ RSA time series. To determine whether or not these calculations of physiological synchrony go beyond patterns that could be explained by the natural frequency of the RSA time series, we compare the empirical cross-correlation coefficients to cross-correlation coefficients computed from shuffled RSA time series (i.e.,

Testing the significance of the cross-correlation coefficients with bootstrapping

Although it is important to assess whether empirical cross-correlation estimates are different from estimates of shuffled RSA time series between infants and mothers, it is also informative to assess which cross-correlation coefficients measured at different time lags differ from chance levels. To do so, we employed a bootstrapping analysis to estimate confidence intervals of the cross-correlation function. For current purposes, we performed a bootstrapping analysis only on the data processed

Similarities and differences between RSA and IBI synchrony

As we briefly mentioned in the Introduction, previous research with infants and mothers reveals that they sometimes synchronize their heart-rate rhythms based on IBIs (Feldman et al., 2011). This is clearly a different measure than RSA because it is associated with multiple sources of heart rate variability, whereas RSA is associated primarily with the myelinated vagal nerve. Although RSA and IBIs are generally correlated, it is an empirical question as to whether physiological synchrony based

Implications and future directions

First and foremost, we have demonstrated that it is possible to calculate a continuous measure of RSA that can be used for comparing dynamic changes in physiological regulation of infants and mothers. In order to appreciate the value of this measure, it is important to emphasize that we measured dynamic changes across a 120-sec time interval and correlated these changes in RSA between infants and their mothers. If we had used the standard method for calculating RSA (i.e., aggregating RSA over

Declaration of Competing Interest

There are no conflicts of interest

Acknowledgments

We gratefully acknowledge all of the families who participated in this research. An outstanding team of students assisted with video coding: Kelsey Blalock, Meghan Burmeister, Khushboo Chougule, Quinn Cox, Diandra Elsner, Deanne Hoaglund, Rebecca Hailperin-Lausch, Hannah Maluvac, Lela Minor, Pooja Pandita, Ellen Parrish, and Stephanie Younker. Karen Wilkie assisted with subject recruitment and testing. Dr. Stephen Porges provided valuable feedback regarding the development of the methodological

References (54)

  • B. Beebe et al.

    How does microanalysis of mother-infant communication inform maternal sensitivity and infant attachment?

    Attachment & Human Development

    (2013)
  • J. Belsky et al.

    Testing a core emotion-regulation prediction: Does early attentional persistence moderate the effect of infant negative emotionality on later development?

    Child Development

    (2001)
  • G.G. Berntson et al.

    Cardiovascular psychophysiology

  • A.Y. Blandon et al.

    Contributions of child’s physiology and maternal behavior to children’s trajectories of temperamental reactivity

    Developmental Psychology

    (2010)
  • M.H. Bornstein et al.

    Physiological self-regulation and information processing in infancy: Cardiac vagal tone and habituation

    Child Development

    (2000)
  • T.B. Brazelton et al.

    Early mother-infant reciprocity

    Parent-Infant Interaction

    (1975)
  • A. Busuito et al.

    In sync: Physiological correlates of behavioral synchrony in infants and mothers

    Developmental Psychology

    (2019)
  • K.L. Creavy et al.

    When you go low, I go high: Negative coordination of physiological synchrony among parents and children

    Developmental Psychobiology

    (2020)
  • A.J. Fisher et al.

    Dynamic RSA: Examining parasympathetic regulatory dynamics via vector-autoregressive modeling of time-varying RSA and heart period

    Psychophysiology

    (2016)
  • L. Galbusera et al.

    Interpersonal synchrony feels good but impedes self-regulation of affect

    Scientific Reports

    (2019)
  • K.M. Gates et al.

    Estimating time-varying RSA to examine psychophysiological linkage of marital dyads

    Psychophysiology

    (2015)
  • J.J. Gross

    Emotion regulation: Taking stock and moving forward

    Emotion

    (2013)
  • J. Ham et al.

    Infant resilience to the stress of the still-face: Infant and maternal psychophysiology are related

    Annals of the New York Academy of Sciences

    (2006)
  • R.A. Isabella et al.

    Interactional synchrony and the origins of infant-mother attachment: A replication study

    Child Development

    (1991)
  • J. Jaffe et al.

    Rhythms of dialogue in infancy: Coordinated timing in development

    Monographs of the Society for Research in Child Development

    (2001)
  • K. Kaye et al.

    The temporal structure of face-to-face communication between mothers and infants

    Developmental Psychology

    (1980)
  • E.M. Leerkes

    Predictors of maternal sensitivity to infant distress

    Parenting

    (2010)
  • Cited by (0)

    View full text