Influences of oxytocin and respiratory sinus arrhythmia on emotions and social behavior in daily life

Abstract

The literature concerning biological influences on positive social behavior shows that, in non-threatening contexts, tonic oxytocin (OT) and respiratory sinus arrhythmia (RSA) each predict positive, affiliative behaviors toward certain others and are associated with positive health outcomes. The purpose of this investigation was to determine the degree to which the positive affiliative correlates of OT and RSA can be distinguished when observed at the level of everyday life events. A sample of midlife adults (N = 73) provided tonic indices of these biological characteristics, as well as perceptions of a variety of common life events alongside reports of their emotions during those events. OT and RSA each independently moderated the link between perceived event sociality and positive emotions, whereas only RSA predicted the probability of being with other people during an event. These findings suggest that OT and RSA may each be linked to positive social experiences in complementary yet distinct ways.

Positive social connections are enjoyable in and of themselves, adding to the richness of people’s daily lives. Beyond their inherent rewards, they are associated with a host of mental and physical health benefits. For decades, scientists have touted the importance of social relationships, showing that richer social lives predict lower mortality risk, less chronic illness, greater resistance to infectious disease, and improved mental health, among other salubrious effects (for reviews, see House, Landis, & Umberson, 1988; Cohen, 2004; Holt-Lunstad, Uchino, Smith, & Hicks, 2007). Within individuals, day-to-day fluctuations in social connection have been robustly associated with day-to-day fluctuations in positive affect, even when health complaints are present (Clark & Watson, 1988). Moreover, evidence from evolutionary psychology suggests that social integration promoted survival for human ancestors (Brown & Brown, 2006), and that the human brain has adapted to the metabolically efficient state of existing within social environments (Coan & Sbarra, 2015). Given the rewarding and adaptive qualities of sociality, the human body likely evolved physiological systems that undergird the predisposition for affiliation. However, particular biological mechanisms and how they interrelate remain poorly understood (Cohen & Janicki-Deverts, 2009). Nonetheless, certain candidates have been identified: Oxytocin (OT), a neuropeptide, and respiratory sinus arrhythmia (RSA), an index of parasympathetic activity, are components of two biological systems believed to encourage high-quality relationships (see Uvnas-Moberg, 2003, and Porges, 2011, respectively). Although these biological systems serve multiple functions, their involvement in quotidian social experiences has rarely been studied (for exceptions, see Feldman, Weller, Zagoory-Sharon, & Levine, 2007; Gouin et al., 2010; Schwerdtfeger & Freidrich-Mai, 2009). Thus, through the lens of commonly-experienced life events, we explore the effects of OT and RSA on affect and behavior in the benign, frequently social situations that compose the majority of people’s lived experience. Examination of the behavioral correlates of these two physiological systems may contribute to scientific and medical understanding of how social connections influence illness and health.

Oxytocin (OT) serves a variety of functions relevant to social behavior, both peripherally and centrally (Landgraf & Neumann, 2004). Although implicated in positive social processes, OT is not purely prosocial; OT has also been implicated in social avoidance and antisocial behaviors (for a review, see Bartz, Zaki, Bolger, & Ochsner, 2011). The recently articulated social salience hypothesis of oxytocin (Shamay-Tsoory & Abu-Akel, 2016) offers a plausible account for OT’s variable effects on social behavior. Through OT’s modulation of the dopaminergic system, the authors posit, OT enhances attention to and salience of social stimuli, the construal of which depends on environmental and individual factors. Thus, in social contexts perceived as safe or non-threatening, OT may amplify attention to affiliative information, rendering social interactions more rewarding. For instance, neurobiological studies in rodents have observed that the natural reward of social interaction requires coordinated activity of OT and serotonin in the nucleus accumbens (Dolen, Darvishzadeh, Huang, & Malenka, 2013). Additionally, in human males with romantic partners, OT administration enhanced the reward value of their partner’s face together with activation of mesolimbic dopaminergic brain regions implicated in reward (e.g., ventral tegmental area, nucleus accumbens; Scheele et al., 2013). Because everyday experiences (Cacioppo, Gardner, & Berntson, 1999) and social interactions (Dimotakis, Scott, & Koopman, 2011; Vittengl & Holt, 1998) are typically perceived as benign, we predict that quotidian social interactions will be more enjoyable for individuals with high endogenous OT.¹

RSA, an index of parasympathetic activity assessed as heart rate variability (Berntson et al., 1994), has also been linked to social behavior, particularly through regulation of the vagus nerve. The vagus nerve is a prime candidate for a role in translating the predisposition for sociality into affiliative behavior, as it is anatomically linked to nerves controlling eye contact, tuning inner ear muscles to the frequency of the human voice, and regulating muscles involved in facial expressions (Porges, 2011). Cardiac vagal tone, indexed as RSA, has been associated with positive socioemotional experiences. People with higher RSA are more securely attached (Diamond & Hicks, 2005), feel closer to others (Kok & Fredrickson, 2010), and have more supportive friendships (Holt-Lunstad, et al., 2007). Empirical evidence suggests that the vagus may serve the adaptive function of promoting positive social relations in at least two ways. First, like OT, the vagus appears to increase positive affective responses to benign social situations. In a field experiment, RSA predicted greater day-to-day positive emotions as a result of learning loving-kindness meditation (Kok et al., 2013), an intervention known to heighten prosocial tendencies (Jazaieri et al., 2013; Klimecki, Leiberg, Lamm, & Singer, 2013; Leiberg, Klimecki, & Singer, 2011; Weng et al., 2013). Second, in contrast to OT, which appears to intensify social salience regardless of valence, the vagus appears to increase social approach (and not social avoidance). People with higher indices of cardiac vagal tone, for instance, pay greater attention to faces (Park, Van Bavel, Vasey, Egan, & Thayer, 2012), are more likely to seek out social support when under stress (Geisler, Kubiak, Siewert, & Weber, 2013), and are more prone to affiliate with new groups (Sahdra, Ciarrochi, & Parker, 2015). These studies, although correlational, suggest that cardiac vagal activity may be related to both a greater ability to extract natural rewards and a proclivity to seek out social interactions. In combination, these behavioral tendencies may increase the likelihood of forming and maintaining high-quality relationships (Gable & Impett, 2012; Gable, Reis, Impett, & Asher, 2004). We thus predict that higher RSA will be associated with greater affective rewards gained within the social contexts of everyday life, which parallels our prediction for OT. In addition, we predict that RSA will be associated with greater odds of being around others. Because the literature on OT’s behavioral effects suggests context-sensitivity, we make no parallel prediction for OT.

Although OT and RSA share psychological and physiological correlates, the relationship between the two remains unclear. Anatomical evidence suggests that OT and the vagus nerve may work together to influence heart rate (Tribollet, Dubois-Dauphin, Dreifuss, Barberis, & Jard, 1992; Higa, Mori, Viana, Morris, & Michelini, 2002), and indeed, exogenous manipulation of OT changes cardiac indices of parasympathetic activity (Gamer & Buchel, 2012; Kemp et al., 2012; Norman et al., 2011). However, in the animal literature, vagal activity did not differ between OT-knockout versus control mice (Michelini, Marcelo, Amico, & Morris, 2003), and there is no published evidence that endogenous OT is correlated with RSA in humans. Thus, the current study seeks to shed light on potential links between OT and RSA by examining their shared and distinct correlates in the emotions and social behavior of daily life. Specifically, we predict that within everyday life events, individual differences in tonic levels of OT and RSA will independently moderate the link between sociality and positive emotions (Hypothesis 1), and that tonic RSA will be linked to the probability of being with other people (Hypothesis 2). To test these hypotheses, we first measured (1) typical daily events and the emotions associated with them, using an online event recall survey; (2) tonic OT, one day after the survey, using urine samples collected by participants over a 24-hour period; (3) tonic RSA, two days after the survey, using echocardiogram and respiratory data obtained in the laboratory during a 5-minute resting period.

Method

Participants

Eighty-four mid-life adults (56% female; M_age = 48 years, SD = 8) were enrolled in the study. Fifty-one participants self-identified as Caucasian, 26 as African-American, 4 as Asian and 3 as Hispanic. Participants were recruited using postings to the UNC Research Studies listserv as well as flyers distributed throughout the local community. Participants indicated their willingness to join the study by reading and completing a consent form available on the study website. Stratified sampling was used to select participants with gender and racial characteristics that reflected the demographics of the regional population. After indicating willingness to participate, interested individuals who fit the sampling criteria received a phone call from the study coordinator to confirm the details and logistic elements of the study and to schedule a laboratory session.²

Measures

Event Reconstruction Method (ERM) Survey

The ERM (Grube, Schroer, Hentzschel, & Hertel, 2008) is an efficient approach to gather data on participants’ recollections of certain events and bears similarity to the Day Reconstruction Method, which has been shown empirically to minimize respondent retrospection bias without the burden and distraction associated with ecological momentary assessment (Grube, et al., 2008; Kahneman, Krueger, Schkade, Schwarz, & Stone, 2004; Dockray et al., 2010). In particular, the ERM selects for commonly experienced events, many of which may not be captured with focus on a single day. Our ERM survey, which was implemented online, asked participants to recall the last time they had experienced each of 14 specific events and describe what they felt and thought during that event. The 14 events included: eating a nutritious meal, interacting with others, getting ready for the day, engaging in vigorous physical activity, spending time outdoors, hugging, learning something new, working, travel/commuting, helping someone, having sex, eating an unhealthy meal, drinking alcohol, and smoking/using tobacco. Each event was rated on several dimensions, including sociality, affective experiences, how frequently the person engaged in the activity, when the person had last engaged in that activity, and whether another person or persons were physically present during the event (the full list of questions is available in the online supplementary material). If respondents reported that they did not generally engage in an activity, they were not asked to complete any assessments for that activity.

Participants rated the sociality (“To what extent would you consider this particular event social?”) and stressfulness (“What is the most stressed, nervous, or overwhelmed you felt [during this event]?”) of each activity using a 5-point scale (0 = “not at all” to 4 = “extremely”), and reported the presence of others (“Was there at least one other person engaging in this activity at the same time as you?”) with a binary answer (yes or no). Participants also described the emotions they felt during each event using the modified Differential Emotions Scale (mDES; Fredrickson, 2013). Participants rated their strongest experiences during the event of 20 different emotions using a 5-point scale (0 = “not at all” to 4 = “extremely”). For each event, participants’ mean positive emotions score was computed from their ratings of amusement, awe, gratitude, hope, inspiration, interest, joy, love, pride and serenity (mean event α = .934, SD = .017). Participants’ mean negative emotions score was computed from ratings of anger, boredom, contempt, disgust, embarrassment, fear, guilt, hate, sadness and shame (mean event α = .899, SD = .053). The ERM survey was completed two days prior to the laboratory session.

Health indices and intimacy checklist

Participants’ height and weight were instrument-recorded to calculate body mass index (BMI). Participants reported the amount of physical activity engaged in per week, and female participants were asked to indicate their reproductive status (e.g. menopausal, pregnant/nursing). All participants were also asked about steroid use and completed an “intimacy checklist,” modified from Light, Grewen, & Amico (2005), which asked about the frequency of activities over a 24-hour period, such as holding hands, hugging, massage, sexual activity, and orgasm, that could possibly affect urinary oxytocin levels during the 24 hours of urine collection (see below). The health questions and intimacy checklist are available in the online supplementary material.

Oxytocin (OT)

OT was extracted from a 24-hour urine sample to create an index of tonic endogenous OT for each participant. First, total volume (mL) was measured, and two 5 mL aliquots of each participant’s sample were centrifuged at 3000 rpm for 10 minutes. 1.5 mL aliquots of the centrifuged urine were transferred to four 2 mL cryo-tubes and immediately frozen at −80°C until batch assay. Levels of OT were assayed in an extraction and enzyme immunoassay (EIA) procedure previously described by Grewen, Davenport, and Light (2010). EIA was preceded by an extraction step, per instructions that accompanied the EIA kit available in 2011 (ADI-900-153, Enzo Life Science, Plymouth Meeting, PA, USA). In effect, the extraction concentrated the sample 3.2 times, which increases precision and reduces matrix interference. Extraction efficiency for urinary OT was 91%, determined by spiking a known amount of OT paired with urine samples. OT levels were then measured using the oxytocin EIA, in which the optical density from the resulting enzyme reaction was read using a Sunrise plate reader (Tecan, Research Triangle Park, NC, USA) at 405 nm. For a given sample, OT levels are inversely proportional to the intensity of the color. Thus, OT concentrations can be inferred from the optical density of a processed sample. Urinary creatinine was assayed by colormetric detection at UNC McClendan Clinical Laboratory. OT values are reported as a ratio of urinary OT to creatinine to adjust for differences in renal clearance (pg oxytocin/mg creatinine).

Respiratory sinus arrhythmia (RSA)

Disposable snap electrodes were placed in a bipolar configuration on lateral sides of the torso on the lowermost ribs to obtain the participant’s electrocardiogram (ECG). Respiration period and amplitude were collected with pneumatic bellows, placed around the participant’s torso at the bottom of the sternum. Continuous heart rate recordings were made at a sampling rate of 1000 Hz. From these recordings, second-by-second averages were computed. All data were inspected offline and corrected for artifacts; a section of data was excluded if neither a peak nor a valley could be identified through visual inspection or application of a peak-detection algorithm.

RSA was calculated offline based on changes in heart rate associated with respiration using custom software produced by James Long Company (Caroga Lake, NY, USA) that applies a modified Grossman peak-to-valley method (Grossman, 1983) with resampling every 125 ms. RSA across a 5-minute resting period was averaged to generate a tonic RSA score for each participant. Analyses were also conducted using an alternative method of indexing cardiac vagal tone, high-frequency heart rate variability (HF-HRV), calculated using spectral frequency analysis of heart rate. For all hypothesis tests reported below, the patterns of significance were identical for RSA and HF-HRV. For economy, only results for RSA are reported below. Results for HF-HRV are provided in the online supplementary material.

Procedure

For each participant, the study included an online event recall survey (completed outside the laboratory), a 24-hour urine collection (gathered outside the laboratory), and participation in a 90-minute laboratory session. First, participants completed the ERM survey two days prior to their laboratory session. Then, using a urine collection kit with instructions that had been delivered to the participant’s home or office, they were asked to collect all urine produced over the course of 24 hours, beginning in the morning, the day before the laboratory session. Urine was to be stored in the containers provided, with ice packs, then placed within a provided cooler for transport to the lab. Event reports, urine collection, and the laboratory session were completed at discrete, non-overlapping time points.

Upon arrival to the laboratory session, participants underwent several assessments unrelated to the focus of this investigation (e.g., other surveys, blood pressure measurement, a blood draw). Several minutes after the blood draw, the examiner attached ECG electrodes and respiration bellows, and proceeded to gather a physiological recording. Participants, seated in a large, padded chair, were instructed to “sit quietly in a comfortable position, with both feet on the floor, and try not to think about anything in particular.” No instructions were given to participants regarding respiratory rate or amplitude. Participants were left alone in the small and quiet laboratory room during the physiological recording. Near the end of the lab session, the examiner recorded participants’ instrument-measured height and weight to calculate body mass index (BMI). Finally, the participants were thanked and debriefed.

Results

Preliminary analyses

Two participants did not provide sufficient urinary samples to calculate OT levels. Mean respiration rates for six participants fell outside the 0.12- to 0.40-Hz frequency band—these participants were excluded from all analyses, as respiratory frequencies outside this range have been observed to distort measurement of RSA (Berntson, Cacioppo, & Quigley, 1993). In addition, one participant did not yield usable RSA data; one participant did not provide information about their physical activity; and one participant did not complete the ERM survey. After these 11 participants were excluded, the final analysis sample was 73.

Initial descriptive statistics for the variables of interest are shown in Table 1. Because both OT and RSA showed pronounced positive skew, non-parametric descriptive statistics are reported, and the natural log form of each variable was used in statistical models. Values for the positive emotions variable were relatively normally distributed, whereas negative emotions clustered near zero with low variability. The low level of negative emotions across the various events is consistent with the positivity offset: In nonclinical samples, in the absence of specific stimuli, individuals tend to experience most events with mildly positive affect (Diener & Diener, 1996).

Table 1.

Person-level descriptive statistics

	Sociality	Positive emotions	Proportion of events with others present		OT (pg/mg creatinine)	RSA	Negative emotions
N	73	73	73	N	73	73	73
Mean	1.68	1.53	.61	Median	8.66	.04	.21
Std. Dev	.68	.70	.19	25th percentile	5.26	.03	.08
				75th percentile	19.61	.06	.42

Note. Mean values for sociality, positive emotions, negative emotions, and events with others present represent person-level estimates. Values for OT, RSA, and negative emotions were computed using non-parametric statistics due to skewness.

Tonic OT and RSA were uncorrelated (r = .04, 95% CI [−.19, .27], p = .74). The relationship remained non-significant after controlling for age and biological sex (r = .09, 95% CI [−.15, .31], p = .47). Individual event participation rates in the analysis sample ranged from 84% (drinking alcohol) to 100% (vigorous physical activity), with a mean of 95% (excluding smoking³). Exact rates, mean sociality, and mean emotion ratings for each event are provided in the online supplementary material.

Next, we computed the intraclass correlation (ICC) for positive emotions, measuring the distribution of variance among individual differences between participants and differences between events. The ICCs for both potential sources of variance were significant (person-level variance 38%, p < .0001; event-level variance 17%, p < .0001), suggesting that both individual differences and event-specific factors played a role in shaping participants’ positive affective responses to the typical life events assessed. In keeping with our hypotheses about individual differences, the analyses described below test whether some of the between-person variance in positive emotions can be predicted by individual differences in tonic OT and RSA.

Hypothesis Testing

All analyses reported below control for the following variables: age, biological sex (0 = female, 1 = male), BMI, frequency of physical activity, reproductive status (for females), steroid use and scores on the intimacy checklist. Biological sex, steroids, sexual activity/warm touch (assessed on intimacy checklist), and (for females only) reproductive status affect OT levels (Holt-Lunstad, Birmingham, & Light, 2008; Leake, Buster, Fisher, & Everett, 1984; Salonia et al., 2003; Share & Crofton, 1993), whereas age, biological sex, BMI and physical activity are correlated with RSA (Jensen-Urstad et al., 1997; Andrew et al., 2013; Melanson, 2000; Thayer, Yamamoto, & Brosschot, 2010; Bernston, et al., 1997). Analyses were also run without covariates. The pattern of significance for all hypothesis tests was largely consistent across analyses performed with and without covariates.⁴

Hypothesis 1: Individual differences in tonic levels of OT and RSA moderate the link between event sociality and positive emotions

To test this hypothesis, we fitted a multilevel model using the nlme package in R with REML and the “optim” optimizer. The model includes a person-level random effect for the intercept and allows error covariances to vary freely between events; more parsimonious structures were tested, but all failed to fit as well as an unconstrained structure. This unconstrained error covariance structure attributes between-event variance in the dependent variable to the random effects portion of the model; the fixed effects, which test our hypotheses, model only the common variance.

The full model without covariates is shown below using Raudenbush and Bryk (2002) notation, in which Social represents the reported sociality and PE represents mean positive emotions for an event:

$$\begin{array}{c}{PE}_{ti}={\gamma }_{00}+{\gamma }_{01}(OT)+{\gamma }_{02}(\mathit{RSA})+{\gamma }_{10}(\mathit{Social})+{\gamma }_{11}(\mathit{Social})(OT)+{\gamma }_{12}(\mathit{Social})(\mathit{RSA})+u_{0i}+e_{ti}\\ u_{0i}~N(0,{\sigma }_i^2),e_{ti}~N(0,{\sigma }^2),\mathit{cov}(P{E}_t,P{E}_{t^{\prime }})={\sigma }_{t,t^{\prime }}\end{array}$$

Results are shown in Table 2.

Table 2.

Multilevel model predicting positive emotions in daily life

Predictors	B(SE)	p	95% CI
γ01OT	−.19(.18)	.32	[−.55, .18]
γ02 RSA	−.01(.29)	.98	[−.59, .58]
γ10 Social	.20(.02)	<.0001	[.17, .23]
γ11 Social × OT	.08(.03)	.02	[.02, .15]
γ12 Social × RSA	.27(.05)	<.0001	[.17, .37]

Note. OT = tonic endogenous oxytocin; RSA = resting respiratory sinus arrhythmia. Covariates are age, sex, BMI, frequency of physical activity, reproductive status (for females), steroid use, and intimacy checklist. The “Social” independent variable is person-mean centered.

Consistent with Hypothesis 1, both OT and RSA moderated the association between sociality and positive emotions during events (see Table 2). As shown in Figure 1, higher levels of OT and RSA were independently associated with experiencing more positive emotions during situations of equivalent sociality. The positive emotions model was also fitted with negative emotions included as a covariate. The pattern of significance for OT, RSA, and their interaction with sociality did not change.

Figure 1.

(a) Positive emotions predicted by perceived sociality, as moderated by oxytocin (OT). (b) Positive emotions predicted by perceived sociality, as moderated by resting respiratory sinus arrhythmia (RSA). In both figures, regression lines are based on simple slopes derived from the models characterized in Table 2. Lower and higher levels of tonic OT and RSA are represented by the 25th percentile (dashed lines), and the 75^th percentile (solid lines), respectively.

Hypothesis 2: Individual differences in tonic RSA are linked to the probability of being with other people

For each event, participants indicated whether they were, or were not, in the presence of others. Although our hypothesis targets RSA, we include OT in the model to rule out any potential effects of OT. Therefore, we tested whether OT and RSA moderated the odds of being in the presence of others using a logistic multilevel model, run with the R package lme4 and optimizer bobyqa. The model included crossed random effects, with random intercepts for each person across events, and random intercepts for each event across persons. The model controlled for the same covariates as described previously. OT and RSA were included as the main predictors of interest. The full model without covariates is shown below using Raudenbush and Bryk (2002) notation, in which Others is a binary variable indicating the presence of other people during an event:

$$\begin{array}{c}{\mathit{Others}}_{ti}={\gamma }_{00}+{\gamma }_{01}(OT)+{\gamma }_{02}(\mathit{RSA})+u_{0i}+u_{0t}+e_{ti}\\ u_{0i}~N(0,{\sigma }_i^2),u_{0t}~N(0,{\sigma }_t^2),e_{ti}~N(0,{\sigma }^2)\end{array}$$

Results are shown in Table 3.

Table 3.

Logistic multilevel model predicting the presence of others in daily life

Predictors	B(SE)	OR (Others present)	p	95% CI
γ01 OT	−.02(.24)	0.98	.92	[.61, 1.55]
γ02 RSA	1.34(.39)	3.82	<.001	[1.78, 8.23]

Note. OT = tonic endogenous oxytocin; RSA = resting respiratory sinus arrhythmia. Beta coefficients represent the log-odds of being in the presence of others. OR represents the effect of a one-unit change in the predictor on the odds that someone else will be present. The 95% confidence interval represents the upper and lower bounds of OR. Covariates are age, sex, BMI, frequency of physical activity, reproductive status (for females), steroid use, and intimacy checklist.

Consistent with Hypothesis 2, as reported in Table 3 and illustrated in Figure 2, the odds of being in the presence of others increased as resting RSA increased. OT had no effect on the odds of being in the presence of others.

Figure 2.

Model-estimated probabilities that other people will be present during an event. OT = oxytocin; RSA = resting respiratory sinus arrhythmia. OT and RSA values are log-transformed and mean centered. 95% confidence region shown in gray.

Additional models

Full results of all additional models are available in the online supplementary material. In ancillary analyses, we tested an alternative explanation: that social facilitation accounts for the finding that OT and RSA increase the positive emotions experienced in social events. In this explanation, positive emotions are higher not because people perceive a situation as social, but because physiological arousal increases due to the presence of others, and this arousal is interpreted positively (Uziel, 2007). We tested this explanation by re-fitting the model for Hypothesis 1, replacing “sociality” with the binary variable indicating whether another person was or was not physically present during each event. When presence of others replaced sociality in the model, the presence of others significantly interacted with neither OT (B = .03(.11), p = .78) nor RSA (B = .10(.17), p = .55). We thus conclude that it is perceived sociality, rather than the mere presence of others, that interacts with OT and RSA to predict positive emotions.

Because biological sex has been found to play a role in some effects of OT (Kubzansky, Mendes, Appleton, Block, & Adler, 2012), we fitted additional models with biological sex as a moderator of the OT and RSA effects. For the positive emotions model, there was neither a significant interaction of biological sex and OT nor a significant three-way interaction among sociality, biological sex, and OT. However, a significant and positive three-way interaction among sociality, biological sex, and RSA did emerge (B = .35(.11), p = .001). Contrasts revealed that, in our sample, sociality and RSA interacted to predict greater positive emotions in men (B = .37(.07), p < .0001) but not in women (B = .02(.08), p = .76). For the model predicting the presence of others, there was no significant interaction between biological sex and OT or between biological sex and RSA.

Finally, we tested whether OT and RSA might multiplicatively predict positive emotions or time spent with others, by fitting models for positive emotions and presence of others, respectively, which included an OT-by-RSA interaction term and (when relevant) an OT-by-RSA-by-sociality interaction term. For the positive emotions model, a marginally significant three-way interaction emerged for OT, RSA, and sociality (B = .16(.09), p = .098), which potentially suggests that individuals high in both OT and RSA may experience even greater pleasantness during social events than would be expected given the effects of each alone. No significant interaction effect emerged in the model predicting the presence of others.

Discussion

Given the myriad advantages of social relationships, it is reasonable to consider that human beings have evolved physiological systems to motivate pursuit of beneficial social interactions. Our evidence suggests that oxytocin (OT) and respiratory sinus arrhythmia (RSA) may index the activity of two overlapping, yet distinct, systems that promote the formation and maintenance of positive relationships. As we hypothesized, individual differences in tonic OT and RSA independently moderated the link between the perceived sociality of typical life events and experienced positive emotions. That is, for events rated as more social, both predictors accounted for significant, unique variance in positive emotions. Importantly, no direct effects of OT or RSA on positive emotions emerged independent of perceived sociality. Effects also held when negative emotions were included as a covariate (see online supplementary material). Ancillary analyses suggest that the effect of OT was consistent across males and females, whereas the effect of RSA moderating the link between perceived event sociality and positive emotions emerged for males only, an unanticipated result. OT was not related to the likelihood of being with others, but as hypothesized, participants higher in RSA were more likely to report being with other people during typical everyday events.

These results provide some of the first evidence to differentiate the social correlates of OT and RSA in humans. Whereas both biological traits are associated with greater natural rewards in everyday social interactions, only RSA appears to be linked with social approach behavior, as assessed by a higher likelihood of being in the presence of others. These findings contribute to the small but growing body of literature that has linked the OT system in humans to positive affectivity in benign social contexts (Algoe & Way, 2014; Isgett, Algoe, Boulton, Way, & Fredrickson, 2016). Although our data do not address neurocognitive mechanisms, our results are nonetheless consistent with the emergent body of evidence suggesting that OT amplifies the salience of positive or negative social information, which in turn modulates social emotions and behaviors in response to the social milieu (Shamay-Tsoory & Abu-Akel, 2016). In our data, OT did not appear to promote the pursuit of social interactions. Rather, our findings suggest that OT may influence affective processes within existing relationships, and not global social approach behavior (e.g., Kosfeld, Heinrichs, Zak, Fischbacher, & Fehr, 2005). In contrast, RSA, as a measure of activity of the vagus nerve, may represent a physiological system tuned to seek out social engagement, by fostering eye contact with others or picking out the sound of a human voice within a noisy environment (Porges, 2011). Thus, we speculate that higher vagal activity may help to form new relationships, as individuals seek out interaction opportunities and express positive emotions in ways that make them increasingly attractive interaction partners (Berry & Hansen, 1996). Once a positive relationship is established, OT may play a larger role in maintaining the relationship, by reinforcing the positive emotional associations that are the foundation of enduring social bonds (Algoe & Way, 2014; Brown & Brown, 2006). In concert, these systems would theoretically result in more frequent high-quality social interactions, building social bonds that have been essential for optimal human functioning from the evolutionary past to the present day.

Our ancillary analyses also tentatively suggest that OT and RSA may, in combination, predict even more positive emotions within everyday social encounters. Although only marginally significant, we speculate that this correlational three-way interaction among OT, RSA, and sociality may reflect the consequences of greater social context sensitivity for people with higher tonic OT, which may in turn amplify the emotional benefits of an RSA-linked social approach orientation. Furthermore, OT may strengthen this social approach orientation by acting on the vagus nerve in a positive feedback loop that forms the physiological basis for upward spirals that promote health (Fredrickson, 2013; Fredrickson & Joiner, in press). This interpretation is congruent with research from both pharmacological and genetic perspectives that suggests a context-specific relationship between oxytocinergic and parasympathetic systems (Gamer & Büchel, 2012; Kanthak et al., 2016). These possible explanations are worthy of future investigation with larger sample sizes to understand the stability of the three-way interaction observed here.

Future Directions

The event reconstruction method (ERM) survey asked participants to recall particular events, report on the perceived sociality of these events, and indicate their state of mind and emotions during each. Although previous work has supported the ERM’s ecological validity and resistance to recollection bias (Grube, et al., 2008), because we asked participants to reconstruct events from their past, we acknowledge that their reports may have been shaded by emotional states experienced at the time of reporting, or by their schemas concerning specific events. However, mitigating this concern, the ERM survey and the physiological measures were collected on different days and in different environments. Furthermore, both OT and RSA were intentionally assessed in ways designed to be resistant to the influence of momentary, state-driven fluctuations in an effort to capture tonic person-level variance. This increases our confidence that the links between these biological variables and data gathered via the ERM survey cannot be attributed solely to participants’ physiological state when completing the self-reports. Nevertheless, it is possible that participants’ recall may have been biased toward more memorable or interesting events, and they may have forgotten more typical instances of a given event. Other methods of assessing daily experience, such as ecological momentary assessment or the Day Reconstruction Method (Kahneman et al., 2004), may provide improved capture of typical daily experiences.

Participants higher in RSA were more likely to report being in the presence of others, which might suggest that RSA is associated with social approach behaviors such as seeking out others. However, the correlational nature of the data leaves open other potential explanations. Perhaps the enhanced vagal activity reflected by high RSA increases awareness of, and memory for, the presence of others, creating the appearance of spending more time with others. It is also possible that frequently being in the presence of others causes RSA to increase over time. For example, Schwerdtfeger and Friedrich-Mai (2009) found that when depressive participants were alone they evidenced lower heart rate variability than when they were engaged in social interactions with a partner, family members, or friends. Furthermore, we found that both OT and RSA moderated the association between the perceived sociality of commonly-experienced events and positive emotions. This pattern may occur because OT and RSA each amplify positive affective responses to benign social interactions, or it may be that people who extract the greatest natural reward from social interactions develop higher tonic endogenous OT and RSA over time. Any or all of these explanations may be true; additional research that experimentally manipulates OT, RSA, and sociality—either singly or in conjunction—is necessary to determine the directions and nature of causality within the observed relationships.

There has been disagreement in the oxytocin literature concerning the relationship between central OT and peripheral OT as measured in saliva, blood, or urine (McCullough, Churchland, & Mendez, 2013; Leng & Ludwig, 2016). We acknowledge that this is a literature in development and that further work is needed to assess the magnitude of the link between peripheral and central measures of OT. There is also controversy regarding the specificity of various OT assays that have been used in prior studies (Carter, 2014; Szeto et al., 2011). We note, however, that the EIA with extraction method of oxytocin assay used in the present study has been identified as among the best laboratory practices for obtaining accurate endogenous OT levels (McCullough et al., 2013).

Similarly, the physiological measure RSA has been of interest to psychologists because it is believed to reflect the activity of the vagus nerve and the parasympathetic nervous system (PNS) more generally. However, RSA levels can at times be confounded with other physiological factors, including respiratory frequency and amplitude, that are not necessarily linked to either the vagus or the PNS (Grossman & Taylor, 2007). These other factors may not be fully accounted for with the free-respiration resting period assessment employed here, and as such, inferences about vagal functioning cannot be made with certainty. More generally, respiratory confounds have been a concern within the literature on RSA, and it is worth noting that some work suggests the relationship between vagal activity and prosocial outcomes is quadratic (Kogan, et al., 2014). Future studies may benefit from using multiple, paced breathing exercises (Grossman & Taylor, 2007) to capture a “purified” RSA measure that may serve as a more accurate index of cardiac vagal tone.

Finally, whereas this study focused on OT and RSA because they represent the physiological markers most widely linked to social relationships, there are other hormones that also appear to influence social behavior, including testosterone (Mehta & Josephs, 2011), progesterone (Brown et al., 2009), and cortisol (Mehta & Josephs, 2011). Furthermore, social interactions involve many brain areas that form dynamic networks of subcortical (amygdala, hippocampus, striatum) and cortico-limbic (insula, cingulate) structures, as well as the medial prefrontal cortex (mPFC), superior temporal sulcus (STS), temporo-parietal junction (TPJ), and precuneus, which are specifically related to mental state attribution for self and others (Van Overwalle & Baetens, 2009). Apart from OT, the release of dopamine, vasopressin, testosterone, and other neurotransmitters appear to influence social behavior by targeting the aforementioned structures (Depue & Morrone-Strupinsky, 2005; McCall & Singer, 2012). Many of these systems are hypothesized to work together with OT to create positive social behavior (e.g. Gordon, Martin, Feldman, Leckman, 2011). To achieve an accurate picture of how physiology influences social behavior, further interdisciplinary work involving these additional variables will be essential.

Conclusion

Although OT and RSA have been linked to positive social experiences, it is important to understand how these interactions emerge. Using an ecologically valid measure of everyday experiences, we found that tonic levels of endogenous OT and RSA, which were themselves uncorrelated, each independently predicted the reward value of sociality, whereas RSA alone was associated with a greater likelihood of being in the presence of others. This pattern of results is consistent with the logic that OT may be linked to positive relationship outcomes by increasing the salience of social contexts, which are often benign in daily life, whereas RSA may be linked to positive relationship outcomes through different processes, such as seeking out social contexts and augmenting their affective rewards. Our findings of the separable effects for OT and RSA argue for the importance of well-defined and specific psychophysiological theories that model social behavior as a multi-part system with many pathways to the same beneficial social outcomes.