Introduction

Humans are social animals by nature. They are born to form social relationships and interact with others. In fact, humans have a predisposition to prosocial behavior because connecting with and understanding others is important in order to navigate the social environment (Adolphs, 2003). Phylogenetic theories suggest that determined biological factors may predispose humans to be social (e.g., Porges, 2007). One such factor is resting vagally mediated heart rate variability (vmHRV), which derives from the synergistic actions of the sympathetic and parasympathetic autonomic nervous systems (ANS) that enable the cardiovascular system to adapt to the changing demands of the environment by regulating appropriate behavioral responses and modulating physiological states to support social behavior (Berntson et al., 1997; Smith et al., 2020). Findings suggest that vmHRV provides a non-invasive proximal measure of cardiac vagal modulation (e.g., Laborde et al., 2017), and its role in social engagement becomes evident when we consider that the vagus nerve (i.e., the 10th cranial nerve) richly innervates the entire body, including the face and the vocal apparatus.

Individuals with higher levels of vmHRV tend to exhibit higher social skills (reviewed in Colonnello et al., 2017) such as the ability to seize opportunities to reinforce positive emotions and social connections (Kok & Fredrickson, 2010), higher openness to experience (Costa & McCrae, 1985, 1992), and the tendency to affiliate with new groups (Sahdra et al., 2015). Moreover, vmHRV positively correlates with a greater ability to regulate emotional responses towards others (Eisenberg et al., 1996; Geisler et al., 2010, 2013; Lischke et al., 2019; Williams et al., 2015), with increased pro-sociality, altruistic behavior (Bornemann et al., 2016; Kogan et al., 2014; Kok & Fredrickson, 2010; Stellar & Keltner, 2017), compassion (Di Bello et al., 2020 for a meta-analysis), and a more accurate understanding of others’ mental states (Lischke et al., 2017). On the contrary, individuals with lower vmHRV appear to be characterized by inflexibility and a reduced ability to respond to situational demands (Thayer & Lane, 2000), poorer self-regulatory capacity (Thayer & Lane, 2009), a higher defensiveness personality trait (Lane et al., 1990; Weinberger, 1990) and diminished inhibition of the physiological stress responses to safe or ambiguous stimuli (e.g., Brosschot et al., 2017).

An important mechanism that affects various aspects of interpersonal interactions and social decisions is the tendency to make inferences about personality and intentions from facial appearance (Todorov et al., 2015). Based on the first impression of a face, these inferences can be made very quickly and long-lasting (Willis & Todorov, 2006). Regardless of an attribution’s accuracy, certain facial features are more likely to be linked to positive personality traits, more friendly behavior, and more favorable consequences in political elections, court cases, healthcare, and hiring (e.g., Bagnis et al., 2021; Bagnis et al., 2020; Schäfer et al., 2016; Todorov, 2005; Wilson & Rule, 2015; for a review, see Todorov et al., 2015). Face evaluation based on social traits involves an overgeneralization of adaptive mechanisms for inferring harmful or helpful intentions. This evaluation of faces is impacted by levels of trust, which best approximate the valence evaluation of faces (Oosterhof & Todorov, 2008, Todorov et al. 2008; Todorov et al., 2015) which affects approach/avoidance behavior (Todorov et al., 2015). Similarly, several studies have identified a relationship between vagal tone, social engagement, and approach/avoidance tendencies (Davidson, 1998; Porges et al., 1994; Porges, 1995; Shahrestani et al., 2014) - suggesting avenues to hypothesize about a relationship between vmHRV and the processing of social stimuli, such as faces. Recently, two independent studies demonstrated that vmHRV modulates facial memory, with individuals with higher levels of vmHRV gaining enhanced memory for faces from (i) the associated affectively relevant behavioral information (Mattarozzi et al., 2019) and (ii) facial appearance (Wendt et al., 2019). Given the existing evidence, it seems reasonable to assume that vmHRV may modulate not only memory formation of socially relevant stimuli but also face perception in terms of the automatic process of social attributions from facial appearance (Todorov & Oh, 2021). To our knowledge, the present study is the first to investigate the link between biologically based individual differences, such as vmHRV, and face evaluation based on social traits. Specifically, we first measured resting vmHRV in individuals and then asked them to rate faces based on their first impressions of various social traits, i.e., trustworthiness, dominance, typicality, familiarity, caring, and attractiveness. We chose these attributes due to their relevance in different social contexts. In particular, trustworthiness and dominance are the two most important trait dimensions that determine the social evaluation of faces, signaling whether a person should be approached or avoided and as an indication of people’s physical strength and weakness, respectively (Oosterhof & Todorov, 2008). Typicality and familiarity are important for social perception and cognition since people tend to prefer typical and familiar facial configurations, and judge familiar faces to be safer than unfamiliar ones (Bartlett et al., 1984; Zebrowitz et al., 2007). Notably, appearances perceived as more caring and attractive are associated with evolutionary advantages, as these traits respectively represent signals of safety (Mattarozzi et al., 2020) and health (Shackelford & Larsen, 1999). This study therefore tested the hypothesis that vmHRV could be associated to inferences based on facial appearance, such that higher levels of resting vmHRV would predict higher ratings of traits that favor approaching behaviors.

Method

Participants

We performed an a priori power analysis to determine the number of participants needed. G*Power indicated that we had to recruit 74 participants to have sufficient power (1 - β = 0.95, α = 0.05) to detect medium-sized effects. A total of 83 healthy participants volunteered for the study. We excluded data from four participants due to extreme outliers in vmHRV values (z-scores > ± 3 SD), yielding 79 participants (44 women, age 20.56 ± 1.41 years).

All participants were university students, receiving partial course credit for their participation. They gave written informed consent before participating in the study and were fully debriefed after they completed the study. The study was approved by the Ethics Committee of the University of Bologna and conducted in accordance with the 1964 Declaration of Helsinki and its later addenda.

Procedure

Participants were instructed not to engage in strenuous exercise on the day of the study and to abstain from alcohol, caffeine, and nicotine within twenty-four hours prior to their participation. Participants were asked to sit in a comfortable chair upon arrival at the laboratory. After the electrocardiography (ECG) electrodes were placed on their chest, participants were instructed to sit still, breathe spontaneously, and leaf through a neutral-valenced garden magazine during the HR recording (i.e., “vanilla baseline,“ Jennings et al., 1992). Immediately after receiving instructions, HR was recorded for a total duration of 8 min. Participants then completed the first impression rating task. At home, participants answered a series of lifestyle questions (e.g., nicotine, physical activity) via the online survey platform Qualtrics. After completing the questionnaire, they received a brief written description of the aims of the study.

HRV assessment

Heart rate was recorded as beat-to-beat interval in ms using BodyGuard 2 (FirstBeat Technologies Ltd., Jyväskylä, Finland; Parak & Korhonen, 2015). VmHRV was assessed by calculating the root mean square value of successive beat-to-beat interval differences (rMSSD), which reflects the vagal regulation of HR and is less susceptible to respiratory influences (Laborde et al., 2017; Task Force, 1996). Throughout the remainder of this manuscript, we will use the term vmHRV to specifically refer to rMSSD. VmHRV analyses were performed using Kubios HRV software (Tarvainen et al., 2014). This software uses an advanced detrending method based on the formulation of smoothing priorities, where the filtering effect is attenuated at the beginning and the end of the data, avoiding the distortion of the data endpoints. Furthermore, the frequency response of the method is fitted with a single smoothing parameter chosen so that the spectral components of interest are not significantly affected by detrending (Tarvainen et al., 2002). Kubios HRV includes two methods for correcting for artefacts and ectopic beats: (1) a threshold-based correction, where these are simply corrected by comparing each RR interval value to a local average interval, and (2) an automatic correction, where artefacts are detected from a time series consisting of the differences between successive RR intervals.

First impression rating task

16 faces from the Chicago Face Database (Ma et al., 2015) were selected based on a standardized average (z-score) of their trustworthiness ratings, as in Oosterhof and Todorov (2008). Since it would have been impossible to select faces according to every trait examined in the study, we chose to select them according to trustworthiness. In fact, although people make a variety of inferences from faces, studies suggest that judgments about trustworthiness approximate the intrinsic valence of faces that underlies the other social judgments (Oosterhof & Todorov, 2008; Todorov et al., 2008). Specifically, we selected four male and four female faces that were rated as trustworthy (i.e., with a positive intrinsic valence, z = > 0.50) and four male and four female faces that were rated as untrustworthy (i.e., with a negative intrinsic valence, z = > − 0.50).

Each participant was instructed to report their first “gut impression” and was asked to rate the faces in terms of trustworthiness, dominance, typicality, familiarity, caring, and attractiveness. Participants rated the targets for all traits at once. Ratings were collected on a Likert-type scale ranging from 1 (not at all) to 9 (extremely). Using Qualtrics software, faces were presented individually and randomly in the center of a computer screen (18 cm high x 21 cm wide) at the 80 cm viewing distance from the center of a 19-in computer monitor, subtending a visual angle of approximately 12.83° − 14.95°. Each photo presentation was preceded by a central fixation cross (500 ms). There was no time limit or feedback during the task.

Statistical analysis

Data were first analyzed for skewness and kurtosis to inspect for non-normality. Data did not differ from normality beyond acceptable limits (Hair et al., 2010) and, therefore, parametric tests were used in all analyses. Preliminary analyses were conducted to examine the characteristics of the participants. For this purpose, correlations were calculated between the vmHRV and habits related to smoking, coffee consumption, alcohol consumption and physical activity. Bivariate correlations were performed to determine the relationship between vmHRV and traits ratings from facial appearance, while a multivariate simple linear regression analysis was computed with vmHRV as an independent variable and ratings of trustworthiness, dominance, typicality, familiarity, caring, and attractiveness as dependent variables to investigate whether inter-individual differences in vmHRV could predict trait ratings. Sex was used as a between-factor variable, while age and BMI were used as covariates.

All statistical analyses were conducted using SPSS® 28.0 software (SPSS, Chicago, IL, USA). The significance level was set at p < .05.

Results

Table 1 presents descriptive statistics and t-test results for sex differences in the main variables of the study. Habits in terms of smoking (r = − .10, p = .40), coffee consumption (r = − .04, p = .71), alcohol consumption (r = − .06, p = .59) and physical activity (r = .12, p = .29) were not significantly correlated with vmHRV. Positive correlations were found between vmHRV and ratings of trustworthiness (r = .26, p < .05), dominance (r = .23, p < .05) and caring (r = .39, p < .001).

Table 1 Descriptive statistics for female and male participants
Full size table

Multivariate simple linear regression showed a statistical significance in trait ratings scores based on vmHRV, F (6, 65) = 2.40, p < .05, Wilk’s Λ = 0.81, partial η2 = 0.18, while sex, age, and BMI were found not to be statistically significant. Univariate analyses revealed that vmHRV was able to significantly predict caring (F (1, 70) = 12.67; p < .001; partial η2 = 0.15, R2 adj = 0.18; B = 0.01, p < .001) and trustworthiness (F (1, 70) = 5.36; p < .05; partial η2 = 0.07, R2 adj = 0.07; B = 0.01, p < .05) rating scores. Figure 1 shows positive correlation between vmHRV and the significant dependent variables in the regression model (i.e., caring and trustworthiness rating scores).

Fig. 1
figure 1

Scatter plots illustrating the positive correlations between vmHRV and rating scores

Note. Scatter plots with marginal histograms and density lines showing relationship between vmHRV (rMSSD) and rating scores. Panel A shows the correlation between vmHRV and caring scores, while panel B shows the correlation between vmHRV and trustworthiness scores. The small circles represent the plotted values obtained for each of the variables while the blue lines represent the fitted regression lines. The bands around the regression lines represent the 95% confidence interval. rMSSD = root mean square value of successive beat-to-beat interval differences

Full size image

Discussion

Autonomic responses modulate the individual’s affective experience, influencing engagement in the social environment (Berntson et al., 1997). We hypothesized that resting vmHRV would predict social attributions from facial appearance. Results showed that higher levels of vmHRV were associated with higher scores on caring and trustworthiness ratings, suggesting that strangers’ faces were more likely to be perceived as safer and less threatening by these individuals. This is consistent with phylogenetic theories linking vagal modulation to social bonding, social engagement, and prosocial behavior (Porges, 2007; Thayer & Lane, 2009). Higher vmHRV has been associated with better adaptability and flexibility in coping with various situational demands, as well as with prosocial motives and emotions that help alleviate and prevent psychological distress (Petrocchi & Cheli, 2019). On the other hand, lower vmHRV has been associated with the release of the default physiological defensive response in a safe environment (e.g., Brosschot et al., 2017). Current data suggest that resting vmHRV contributes to shape social attributions about strangers such that individuals with higher vmHRV tend to perceive others as more caring and trustworthy, and therefore being more willing to approach and bond with others.

Caring and bonding motivations are ingrained in human nature and rooted in evolutionarily developed brain systems that we share with other mammals (Panksepp, 2004). From birth, survival depends on others, especially on caretakers. Humans learn to recognize them as important, to see them as social buffers, and to perceive the environment as safer in their presence (Bowlby, 1958). Perceiving strangers as more caring may contribute to fostering social behaviors that have an evolutionary advantage. As such, we found that higher levels of vmHRV were associated with higher ratings on caring, i.e., with a greater tendency to maximize the perception of facial cues that suggest more caring behavior and a feeling of safety. Consistently, higher vmHRV has predicted a higher feeling of safeness during stressful circumstances (Makovac et al., 2022), while a reduction in vmHRV has been associated with social inhibition, avoidance, and anxiety (Alvares et al., 2013).

Facial appearance-based ratings of trustworthiness have been associated with the functional role of inferring behavioral intention and preparing the decision to approach or avoid a stranger during social interactions (Todorov, 2008). Perceiving strangers as more trustworthy (i.e., as less threatening) may increase the chance of cooperation and affiliation and serve to facilitate social interactions. Not surprisingly, we found that higher levels of vmHRV are associated with higher ratings on trustworthiness, maybe because individuals with higher vmHRV are more prone to detect potential relevant positive cues in the absence of other information. This is consistent with studies showing that higher vmHRV is associated with lower negativity bias, i.e., the tendency to favor negative information over positive (Shook et al., 2007). In general, it seems that individuals with higher vmHRV are adaptively engaged in the social world. For example, compared to individuals with lower vmHRV, they are more inclined to perceive neutral stimuli as potential cooperators and to remember negative stimuli to avoid potential danger (Lischke et al., 2018; Mattarozzi et al., 2019; Wendt et al., 2019).

In this study, we found no evidence that vmHRV may be a predictor of judgments of typicality, familiarity, and attractiveness. One possible reason for this is that these social traits differ from trustworthiness and caring in their ability to predict the behavior or decision of others in an unknown situation. As Todorov and Ho’s (2021) model suggests, judgments of trustworthiness and caring are related to perceived valence and represent an overgeneralization of adaptive mechanisms to infer behavioral intentions and signal whether or not to approach a person. Instead, judgments of typicality, familiarity and attractiveness are more rooted in individual experience and are associated with less stable dispositions (Verosky & Todorov, 2013). People who interact with someone whose facial features are typical or familiar tend to generalize their evaluation and transfer the characteristics of familiar people to strangers. However, because our experiences with familiar people are mixed and heterogeneous (i.e., both positive and negative), predicting the intentions of strangers with typical or familiar facial features in terms of valence would be more difficult and less accurate than predicting the intentions of strangers perceived as more trustworthy or caring. Similarly, people cannot infer the intentions of others from perceived attractiveness, as research suggests that cues of attractive facial features alone do not signal a potentially effective interaction but require more powerful cues (e.g., a smile) (Okubo et al., 2015).

Finally, it is worth noting that, although vmHRV does not appear to be a specific predictor of dominance judgements, we found a positive correlation between the level of vmHRV and dominance judgements. This result is consistent with the findings on the traits of trustworthiness and caring, as dominance (i.e., social cues that signal power, strength, and potential threat) represents the other component of Todorov and Ho’s (2021) model that leads to inferences about intent and ability to harm. Overall our significant and non-significant results suggest that differences in vmHRV are more likely to be associated with differences in the evaluation of social traits (i.e., trustworthiness, caring, and dominance) that signal the behavioral intentions of others.

There are limitations to the interpretation of the present results. Firstly, although the relationship between vmHRV and social attributions and behavior is well recognized and supported by existing meta-analytic studies on the topic (e.g., Di Bello et al., 2020; Shahrestani et al., 2014; Zammuto et al., 2021), effect sizes are rather small. This is not surprising if we consider that the current study focused on granular constructs (e.g., trustworthiness) and attempted for the first time to disentangle fine effects (e.g., trustworthiness versus familiarity or attractiveness). In contrast, previous investigations mostly assessed the association of vmHRV with broader social constructs, such as empathic behavior (e.g., Sassenrath et al., 2021), with effects of larger size. Secondly, findings suggest that both vmHRV (Schmalenberger et al., 2020) and face perception (Macrae et al., 2002) are influenced by menstrual cycle and this potential confounder was not taken into account in the current investigation. Similarly, the emotional state of participants at baseline was not assessed and this confounding variable may have potentially biased the results; however, Vanilla baseline has been used as a way to minimize individual differences in worrisome thoughts and anticipatory state anxiety. Moreover, vmHRV measures are known to decrease with age (Agelink et al., 2001), and the sample of the present study consisted of healthy students; it is mandatory that the current findings are replicated in different age groups from the general population.

Conclusion

Overall, these data suggest that tonic vmHRV may be an important individual difference that biases trait inferences from faces. Specifically, we have shown that higher levels of vmHRV are associated with a greater tendency to perceive a stranger’s face as less negative (i.e., more trustworthy, and more caring), thereby minimizing social evaluative threat and maximizing affiliative social cues. As a result, individuals with higher resting vmHRV might be more likely to form positive relationships, which could help them cooperate and better manage stressful experiences in social situations where they are surrounded by strangers. On the other hand, individuals with lower vmHRV (compared with individuals with higher vmHRV) might be inclined to judge others more negatively (i.e., to “err on the side of caution”) and thus, to carry out avoidance behaviors, which can hamper adaptive social strategies, such as cooperation, with less evolutionary advantages.

The current study supports phylogenetic theories underpinning the view that vmHRV may function as a biomarker of approach-related behavior and social engagement (Porges, 2007), and extend previous findings further pointing that tonic vagal modulation of the heart may contribute to the rapid and unconscious processing streams underlying facial appearance-based judgments (e.g., Frisanco et al., 2021). This is not unexpected, since that high resting vmHRV is considered a proxy for functional ventromedial prefrontal cortex inhibitory control (Thayer et al., 2012 for a meta-analysis), and the latter is required for predicting others’ intentions (Leopold et al., 2012).

Given that the tendency to adopt a “better-safe-than-sorry” processing strategy has been proposed as a risk factor for mental and somatic health (Van den Bergh et al., 2021), the present results are also of clinical relevance. Several strategies (e.g., physical exercise, slow breathing, psychotherapeutic interventions) have been identified that are capable of increasing resting vmHRV (Vanderhasselt & Ottaviani, 2022), and Bornemann and colleagues (2016) have already reported that the ability to voluntarily upregulate vmHRV by the use of biofeedback predicts individual differences in altruistic prosocial behavior. In conclusion, future studies are warranted to assess whether the experimental manipulation of vmHRV has the potential to change individuals’ first impressions of strangers and its consequences.