Quantifying the dynamic nature of vagal responsivity in infancy: Methodological innovations and theoretical implications

Abstract

According to polyvagal theory, rapid modulation of the vagal brake develops early in infancy and supports social interactions. Despite being viewed as a dynamic system, researchers typically assess vagal regulation using global measures of respiratory sinus arrhythmia (RSA; an index of vagal tone). This study sought to capture the dynamic property of RSA and evaluate individual differences in within-infant RSA responsivity during mother-infant interaction. RSA was evaluated in a sample of 135 6-month-old Mexican-American infants during a 5-min free play task. Mothers reported on their children’s behavioral problems and competence at 18 months using the Brief Infant-Toddler Social and Emotional Assessment. Time-varying estimates of infant RSA during the interaction were obtained using a multiple window technique and spectrogram analysis. Using structural equation modeling, we evaluated whether within-infant SD of RSA predicted infants’ behavioral problems and competence at 18 months, after adjusting for infants’ mean RSA and covariates. Greater within-infant SD of RSA predicted more behavior problems at 18 months. This study demonstrates that assessing intra-individual variability in RSA, or the extent to which infants fluctuate around their average level of RSA during a task, enhances our ability to test polyvagal theory’s central tenet: vagal regulation supports well-regulated social interaction.

1 |. INTRODUCTION

Infants’ autonomic functioning may contribute to individual differences in emotional reactivity and regulation (Beauchaine, 2001; Calkins, Dedmon, Gill, Lomax, & Johnson, 2002). Research on the contributions of autonomic nervous system (ANS) functioning to emotion regulation has focused primarily on the role of the para-sympathetic nervous system (PNS; Perry et al., 2012), and more specifically, on vagal tone, which reflects the ebb and flow in heart rate during a respiratory cycle that is mediated by the vagus nerve. According to Porges’ polyvagal theory, rapid modulation of the vagal brake is an important component of an integrated biobehavioral social engagement system, which promotes adaptive, context-appropriate social interactions (Porges, 2007). Assessment of vagal functioning during infancy is especially important, as this is a period of rapid organization of various regulatory systems (Quigley & Moore, 2018). A growing body of research highlights relations between vagal functioning and diverse socioemotional outcomes in childhood, thus establishing vagal functioning as an important biomarker of regulation (Graziano & Derefinko, 2013). Furthermore, examining vagal responsivity (withdrawal and activation) among low-income, ethnic minority infants may offer insight into early life biomarkers that either increase risk for poor socioemotional well-being or confer resilience among high-risk populations (Cicchetti & Toth, 2009; Sroufe, 2007).

Although polyvagal theory posits that dynamic, moment-to-moment changes in vagal functioning support social engagement and behavioral regulation, the majority of empirical work testing these predictions has focused on global measures of vagal responsivity during stressful situations. According to polyvagal theory, withdrawal of the vagal brake modulates heart rate, allowing for increases in cardiometabolic output to support emotional and attentional demands from environmental stimuli. For example, the vagal brake is released during periods that require active regulatory coping, such as when one experiences pain or frustration. In contrast, activation of the vagal brake supports behaviorally calm states and social engagement during less stressful times. Despite theory-driven expectations about vagal regulation as it occurs within a task, on the time scale during which social processes unfold, vagal reactivity is typically calculated by sub-tracting average respiratory sinus arrhythmia (RSA; the degree of change in heart rate during the respiratory cycle) during a resting task from average RSA during a challenge task, which precludes directly examining dynamic within-task vagal responsivity. Thus, common methodical approaches are ill-equipped to test a central tenet of polyvagal theory: rapid changes in vagal influences on cardiac output in response to environmental demands support adaptive behavioral responding.

Recent methodological developments allow for better assessment of the dynamic nature of vagal functioning. Gates and colleagues recently developed an innovative program that calculates second-by-second RSA from an individual’s interbeat interval (IBI) series (RSAseconds; Gates, Gatze-Kopp, Sandsten, & Blandon, 2015). This study presents the first application of their methodology to examine the presence and sequelae of intra-individual variability in infant RSA during a task. Intra-individual variability in infant RSA can be viewed as moment-to-moment vagal responsivity (withdrawal and activation) to task demands. Greater intra-individual variability in infant RSA reflects greater deviations from the infant’s own mean during the task, whereas lower intra-individual variability in infant RSA reflects levels of RSA that are closer to infant’s own mean throughout the task. Research with adults supports the benefits of assessing intra-individual variability to gain new information beyond what is captured by average levels (Castro-Schilo & Ferrer, 2013). For example, adults’ intra-individual variability in RSA across a 6-day period predicted affective instability and distinguished between adults with and without bipolar I disorder (Gruber, Mennin, Fields, Purcell, & Murray, 2015), illustrating the utility of assessing instability in adult RSA over the course of a week. However, to our knowledge, no studies have examined the predictive utility of RSA responsivity within a task, or among younger populations. Individual variation in dynamic vagal functioning may be an important mechanism apparent early in life that shapes underlying trajectories toward maladjustment and resilience among at-risk children.

This study aimed to demonstrate that there exist meaningful between-infant differences in vagal responsivity. To do so, we evaluated whether the within-infant standard deviation (SD) of RSA during a free play task predicted later behavior problems and socioemotional competence among a sample of low-income, Mexican American infants, a population at risk for later self-regulation deficits (Galindo & Fuller, 2010). We theorized that infants with greater vagal responsivity during a task that should not require active coping or significant regulation may not be judicious in their expenditure of cardiometabolic resources, predisposing them toward regulatory dysfunction and subsequent problems. Consistent with Porges’ polyvagal theory (Porges, 2007), we expected infants who showed more RSA responsivity during free play, as indexed by a larger SD, to exhibit more behavior problems and less competence 12 months later (at age 18 months).

2 |. METHODS

2.1 |. Participants

The sample included 322 mother-child dyads participating in a broader examination of postpartum adjustment and infant development among very low-income, Mexican-origin women, Las Madres Nuevas. Women’s eligibility criteria included: (a) self-identification as Mexican or Mexican American, (b) fluency in English or Spanish, (c) 18 years of age or older, (d) low-income status (family income below $25,000 or eligibility for Medicaid or Federal Emergency Services coverage for childbirth), and (e) anticipated delivery of a singleton baby with no significant health or developmental problems. The Arizona State University Institutional Review Board and the Maricopa Integrated Health System IRB approved all study procedures prior to recruitment or data collection.

At the first, prenatal visit, women were 18–42 years old (mean 27.8, SD = 6.5). The majority of women (86.3%) were born in Mexico. On average, women had lived in the United States for 11.9 years prior to study enrollment (range 0–32 years). The modal family income was $10,001–$15,000 for an average household of four people. Approximately 77.3% of women were married or living with a partner. The majority of women were not first-time mothers (77.8%) and the number of biological children (not including the target child) ranged from zero to nine (mean 2.0; SD 1.7). The sample included 149 (46.3%) male infants and 173 (53.7%) female infants.

2.2 |. Recruitment

Women were recruited from hospital-based clinics during prenatal care visits. Bilingual female interviewers explained the study, evaluated eligibility, and obtained permission for a prenatal home visit (26–39 weeks gestation). The larger study conducted home visits prenatally and at 6, 12, 18, and 24 weeks postpartum. Data for the present analyses come from the prenatal and 24-week (approximately 6-month) postpartum visits, and a laboratory visit when children were 18 months of age. Interviews were conducted in the participant’s choice of Spanish (86%) or English (14%).

The study employed a “planned missing” design (Enders, 2010) to reduce participant burden during the postpartum home visits. Although the entire sample completed the prenatal home visit, participants were randomly assigned to miss either the 12-, 18-, or 24-week postpartum visit. This design produces data missing completely at random (MCAR), preventing bias in parameter estimates (Enders, 2010). Missing data were handled using Full Information Maximum Likelihood (FIML; Allison, 2003).

2.2.1 |. Participant attrition

Of the 322 women who participated in the prenatal home visit, 210 (93% of the randomly assigned 226 women) completed the 24-week assessment, and 237 dyads (73.6% of the full sample) completed the 18-month interview.

2.2.2 |. Analyses of attrition

Missingness on infant RSA data was not related to maternal age, maternal country of birth, parity, birth outcomes, child sex, or 18-month child behavior problems (all p’s > .05). Attrition at 18 months was not related to parity, birth outcomes, child sex, or infant RSA (all p’s > .05). Mothers born in the United States and younger mothers were more likely to have missing data at 18 months, p’s ≤ .001. Therefore, maternal age and country of birth were included as covariates in statistical models.

2.3 |. Procedures

At the prenatal home visit, mothers completed questionnaires. Women were compensated $75 and given small gifts at the prenatal interview. At the 24-week home visit, mothers completed questionnaires and interaction tasks with their infants, including a free play task. During the free play task, mothers were given a small basket of toys and objects and told to play with their infants as they normally would if alone. Women were compensated $50 and small gifts for the child for the 24-week visit. The 18-month visit was conducted in the laboratory at the Arizona State University. Women were compensated $100 for the laboratory visit and either free transportation to/from the laboratory or $50 for travel costs.

2.4 |. Measures

2.4.1 |. Infant RSA

At the 24-week visit, infants were seated upright at rest and a research assistant placed electrodes on the infants’ left shoulder and right and left waist in a standard lead configuration. Heart rate data were recorded at 256 Hz with electrocardiography (ECG) equipment from Forest Medical, LLC (Trillium 5000; East Syracuse) during a 5-min free play period. QRSTool software 1.2.2 (Allen, Chambers, & Towers, 2007) was used to process the data and automatically obtain R-spikes from the ECG data. Coders then used the QRSTool software to manually correct misidentified or unidentified R-spikes, and obtain R-R interval data. Using CardioBatch software (Brain-Body Center, 2007), a moving polynomial filter was applied to the R-R interval data to extract heart rate variability in the frequency band of RSA (for infants, 0.3–1.3 Hz; Brain-Body Center, 2007). The RSA estimates were log-transformed, and a mean RSA value averaged from 30-s epochs during the free play period was obtained. RSA data were unusable for 31.0% of the infants. Estimates of RSA derived from the aforementioned Porges method (Porges, 2007) were used in preliminary analyses as a validity check for the time-varying estimates, which were obtained using the procedures described below.

We estimated time-varying RSA for the entire 5-min free play period using the MATLAB toolbox RSAseconds (Gates et al., 2015). Each of the cleaned infant IBI series was interpolated at 4 Hz using a cubic spline to create equal data intervals. The data were then tapered using Peak Matched Multiple Windows (PM MW), which is the most optimal way to identify changes in RSA over time, as it yields RSA estimates with lower variance and less bias than the Porges method (Hansson-Sandsten & Jönsson, 2007). A short-time Fourier transform (STFT) was applied to 32-s IBI windows in order to produce second-by-second RSA. Infants’ mean and within-person standard deviation of these RSA estimates were obtained and used in primary analyses.

The values from the STFT approach are always lower than the values from the Porges approach due to the scaling that is introduced via the PM MW technique (K. Gates, pers. commu., April 29, 2019). Eleven infants had negative estimates of RSA during one or more seconds of the free play task; data from these infants were removed prior to analysis.

2.4.2 |. Child behavior problems and competence

Women reported on their infant’s behavior problems and competence at the 18-month visit using the 42-item Brief Infant-Toddler Social and Emotional Assessment (BITSEA; Briggs-Gowan, Carter, Irwin, Wachtel, & Cichetti, 2004). Eleven items comprised the competence subscale (e.g., “Plays well with other children”; α = 0.59) and 31 items comprised the behavioral problem subscale (atypical, maladaptive, internalizing, and externalizing problems, and dysregulation subscales; e.g., “Has trouble adjusting to changes”; α = 0.84). Items are rated from 0 (Not true/rarely) to 2 (Very true/often). Higher scores indicate higher competence or more behavioral problems. The BITSEA has been validated among low-income Hispanic/Latinx families (Hungerford, Garcia, & Bagner, 2015) and normative community samples (Briggs-Gowan et al., 2004). Low internal consistency of the competence subscale is expected given that the behaviors captured in the scale may not co-occur (Briggs-Gowan et al., 2004).

2.4.3 |. Potential covariates

Child sex and birth outcomes (gestational age, birthweight, APGAR score) were obtained through medical record review. At the prenatal visit, women reported their country of birth and age, and completed the 20-item Economic Hardship Scale (Barrera, Caples, & Tein, 2001; e.g., “we fell far behind in paying bills”; α = 0.72).

2.5 |. Data analysis

2.5.1 |. Preliminary analyses

Pearson correlation was conducted with the estimates of average RSA obtained from the STFT approach and from the Porges method. Next, we evaluated a two-level multilevel model, in which Level 1 consisted of the second-by-second RSA estimates and Level 2 comprised of infants, to determine whether there was a linear trend in infant RSA over time. The presence of a linear trend would indicate that infant RSA did not fluctuate around a constant mean, and the RSA data would have to be detrended to distinguish variability due to fluctuating RSA from variability due to a systematic trend (Jahng, Wood, & Trull, 2008). The multilevel model was conducted with Proc Mixed in SAS 9.4, which uses a restricted maximum likelihood approach to estimate the variance-covariance components. The denominator degrees of freedom in t tests was computed using Satterthwaite degrees of freedom. The covariance structure of the residuals at Level 1 was modeled as un-structured. Random effects were included at Level 2.

2.5.2 |. Primary analyses

A structural equation model was conducted with child behavior problems and competence at 18 months regressed on the within-infant standard deviation (SD) in RSA at 24-weeks, adjusting for average RSA (obtained from the STFT approach), child sex, maternal age, and maternal country of birth. Primary analyses were conducted with MPlus v.7.2 (Muthén & Muthén, 1998–2012) using all available values and maximum likelihood estimation.

3 |. RESULTS

3.1 |. Preliminary analysis

3.1.1 |. Comparison of methods

Consistent with expectations, average RSA levels computed using the STFT approach and the Porges method were highly correlated (r = .991, p ≤ .001), supporting the validity of the STFT approach. However, only the STFT approach provides information about second-by-second fluctuation of infant RSA over time, which allows for estimates of the within-infant SD in RSA during a task. Thus, results presented hereafter are only for RSA derived from the STFT approach.

3.1.2 |. Trend in RSA

As expected, infants did not exhibit a linear trend in RSA across the 5-min free play task, Est = 0.000, t(124) = 0.40, p = n.s. Furthermore, there was not significant variability in the linear growth in RSA across the task, Est = 0.000, p = n.s.

3.1.3 |. Descriptive statistics

Table 1 presents descriptive statistics and zero-order correlations for the primary study variables. All primary study variables had adequate skew and kurtosis (West, Finch, & Curran, 1995). One outlier (>3 SD from the sample mean) was identified on the within-infant SD of RSA. When this outlier was removed from the analyses, the pattern of results remained unchanged; therefore, the results presented include all available data.

TABLE 1.

Descriptive statistics and zero-order correlations among primary study variables

	Mean (SD)	1	2	3
1. Average infant RSA at 24 weeks	1.86 (0.74)	–
2. Within-infant SD of RSA at 24 weeks	0.45 (0.16)	0.26	–
3. Child behavior problems at 18 months	13.93 (7.4)	0.04	0.22	–
4. Child behavioral competence at 18 months	15.95 (3.0)	0.15	0.08	−0.05

Note: All ages refer to child age. Infant RSA was evaluated during free play and derived with the STFT approach. Correlation coefficients in bold are statistically significant, p < .05.

Abbreviations: RSA, respiratory sinus arrhythmia. SD, standard deviation.

Correlations between potential covariates (maternal country of birth, age, household economic hardship, birth outcomes, and child sex) and primary study variables were evaluated. Only maternal country of birth and child sex were significantly correlated with primary study variables. Infants whose mothers were born in the United States had higher within-infant SD of RSA, (M = 0.56, SD = 0.22) than those whose mothers were born in Mexico (M = 0.44, SD = 0.15; t(127) = 2.55, p = .012). Boys exhibited less behavioral competence at 18 months (M = 15.36, SD = 3.19) than did girls (M = 16.40, SD = 2.70), t(235) = −2.71, p = .007. Maternal country of birth and child sex, along with maternal age, were retained as covariates.

3.2 |. Primary analyses

We evaluated a structural equation model predicting child behavior problems and competence at 18 months from the within-infant SD of RSA during free play at 24-weeks, average infant RSA, child sex, maternal age, and maternal country of birth (see Figure 1). Within-infant SD of RSA was a statistically significant predictor of 18-month child behavior problems, Est = 10.537, SE Est = 4.369, p = .016, such that more RSA variability predicted more behavior problems. Average RSA, maternal country of birth, maternal age, and child sex were not statistically significant predictors of 18-month child behavior problems, p’s > .10. Only child sex was a statistically significant predictor of 18-month child behavioral competence, Est = 1.095, SE Est = 0.390, p = .005; girls exhibited greater behavioral competence than boys.

FIGURE 1.

Model results. RSA, respiratory sinus arrhythmia; SD, within-infant standard deviation. Unstandardized estimates of path coefficients are presented in the figure. Country of birth was coded 1 = United States and 2 = Mexico. Child sex was coded 1 = boy and 2 = girl. Correlations among exogenous variables not shown. *p < .05. Statistically significant paths are represented by solid lines; dashed lines are not statistically significant

4 |. DISCUSSION

This study demonstrated that a novel measure of infant vagal responsivity (indexed by the within-infant standard deviation of RSA) during a relatively calm mother-infant free play task at approximately 6 months of age was associated with meaningful differences in behavior at 18 months. Consistent with Porges’ polyvagal theory and our prediction, greater infant vagal responsivity (withdrawal and activation) during a free play task predicted more behavioral problems one year later when children were 18 months of age, suggesting that heightened infant RSA responsivity during a neutral task confers risk for subsequent behavior problems in toddlerhood. Whereas exertion of the vagal brake promotes calm behavioral states, activation and withdrawal of the vagal brake facilitates responding to environmental challenges (Beauchaine, 2001); when challenges are not present in the environment, heightened RSA responsivity may lead to context-inappropriate behavior. Heightened RSA responsivity in the absence of psychosocial challenge may also contribute to allostatic load through simultaneous vagal withdrawal and increased sympathetic influences to the heart, and overtaxation of physiological stress response systems may lead to behavior problems later in life (Beauchaine, 2001; Quigley & Moore, 2018). Overall, our results suggest that there are meaningful differences in infants’ moment-to-moment responsivity during mother-infant free play, which predict subsequent behavior problems.

Contrary to expectations, infant RSA responsivity did not predict subsequent behavioral competence. Prior work suggests the influence of infant RSA on subsequent behavioral competence may be contingent on the broader social environmental context, such as the availability of maternal social support (Somers, Jewell, Hannah Ibrahim, & Luecken, 2018). Given the current sample of low-acculturated Mexican American families, understanding the broader cultural context in which childrearing occurs may offer deeper understanding of developing socioemotional processes (García Coll et al., 1996). For example, poorly regulated vagal functioning may not have the same negative consequences for behavioral competence among this sample of families given traditional Mexican cultural values of bien educado which emphasize the importance of respectful behavior and displaying good manners (Fuller & García Coll, 2010). Culturally salient socialization practices may contribute to stronger reinforcement of regulated behavior, even among infants; these practices may buffer against the potentially adverse effects of physiological dysregulation on poor behavioral competence. Additionally, infants whose mothers were born in the United States had higher RSA responsivity, highlighting heterogeneity within Mexican-American families. Culturally based expectations of regulation may extend their influence to impact not only behavior, but also children’s physiological functioning. Future research should replicate and could extend the current results by evaluating the role of cultural values and acculturative experiences on infant vagal functioning.

The study has a number of strengths. Primarily, we leveraged a recent methodological innovation, RSAseconds (Gates et al., 2015), to yield time-varying estimates of infant RSA. We evaluated the existence and impact of intra-individual variability in RSA among a unique sample of low-income, Mexican-American children, a population at elevated risk for regulatory deficits (Galindo & Fuller, 2010). Whereas the common practice of summarizing or averaging RSA levels across an entire interaction masks within-task variability in RSA, our approach allowed us to evaluate moment-to-moment fluctuations in infants’ vagal functioning during social interaction. Our results demonstrate that variability in RSA during free play exists and is predictive of future behavioral problems. Failure to capture this variability limits researchers’ ability to evaluate the influence of rapid vagal modulation on subsequent emotional and behavioral outcomes, as posited by polyvagal theory.

The results should be considered in light of several limitations. We assessed vagal responsivity during a free play interaction that poses minimal demands for infant vagal withdrawal. Tasks designed to elicit frustration from infants may yield different patterns of results. For example, when assessed during tasks that require regulatory efforts, greater vagal responsivity may be associated with more competence and fewer behavior problems later in childhood. In addition, findings in this unique sample of low-income, Mexican-American infants may not generalize to higher SES or ethnic majority infants. Research is needed to assess the extent to which children from diverse backgrounds demonstrate vagal responsivity during resting, neutral, and stressful tasks, which may reveal context- or task-specific patterns of relations between vagal responsivity and subsequent outcomes.

Results from the study encourage generation of theory-driven research on infant vagal responsivity. One such application of the current methodology may include evaluating trajectories of vagal responsivity over time. Perhaps, as their vagal systems mature, infants may gain greater control over physiological processes and become better able to maintain homeostasis, resulting in less variability during calm states. While it is possible that 6-month infant vagal responsivity directly affected 18-month problems, it is also possible that the effects of early life vagal responsivity operate through responsivity later in development. These hypotheses suggest rank-order stability of vagal responsivity over time; however, it is also possible that, through allostatic processes, infants with heightened vagal responsivity may show less physiological flexibility later in life. Future work may also aim to connect infant vagal responsivity to specific infant behavioral regulatory strategies related to successful navigation in social contexts. For example, infants with greater vagal responsivity may rely on more immature regulatory strategies (e.g., self-soothing, gaze aversion), preventing them from developing the skills necessary to navigate more complex social interactions later in life. Another potentially fruitful area of future research involves transactions between infant vagal responsivity and maternal care-giving. Infants may elicit different responses from their mothers, or may differentially influence maternal well-being, depending on infant vagal functioning (Somers, Curci, & Luecken, 2019).

Overall, the results highlight the advantages of adopting a more dynamic approach to evaluating infant vagal responsivity. Evaluating intra-individual variability in infant RSA with time-varying estimates of RSA is an accessible means by which researchers can evaluate antecedents, correlates, and consequences of infant vagal responsivity. We demonstrated that heightened vagal responsivity during a relatively calm task, as indexed by larger within-infant variability in RSA during a mother-infant free play interaction, at approximately 6 months of age predicted more infant behavior problems at 18 months.