Externalizing and internalizing behaviors in children are significant risk factors for later mental health problems. The development and progression of these behaviors are heavily influenced by social, environmental, and biological factors. A better understanding of the elements that contribute to such behaviors in youth is critical given the economic and social costs to families and society at large. Increasingly, researchers are examining the contribution of both biological and psychosocial variables to better characterize the development of these externalizing and internalizing behaviors.

Both internalizing and externalizing behaviors are characterized by emotion dysregulation. Emotion regulation commonly operates in a series of steps: monitor and attend to a specific situation, evaluate or reappraise mentally, and update or modify a response in order to accomplish a specific goal (Gross, 2015). According to this model, emotion dysregulation is “a pattern of emotional experience and/or expression that interferes with appropriate goal directed behavior” (Beauchaine, 2012) by disrupting any one of the above steps. In this sense, some individuals with poor emotion regulation have difficulties in controlling emotional responses or inhibiting their behaviors, resulting in externalizing outbursts (Eisenberg et al., 2001). In the context of internalizing behaviors, emotion dysregulation occurs at the level of attention and mental reappraisal, with these individuals unable to inhibit negative thoughts resulting in excessive rumination and sadness (Eisenberg et al., 2001).

One of the accepted objective indicators of emotion regulation is respiratory sinus arrhythmia [RSA; (Beauchaine, 2015)]. RSA measures cardiac activity of the parasympathetic nervous system (PNS) and reflects the variations in inter-beat intervals of the heart rate across respiration (Porges, 1995, 2007). The parasympathetic branch of the autonomic nervous system (ANS) is critically involved in bodily rest and digest and overall homeostasis. Specifically, the PNS slows the heart rate through the activity of the vagus nerve. Increased parasympathetic activity at rest (e.g., high resting RSA) reflects the individual’s readiness to react efficiently to environmental stressors (Gyurak and Ayduk, 2008; Scarpa et al., 2010). Cortical structures, specifically regions within the central neural network (including the orbitofrontal and ventromedial prefrontal cortices, the amygdala, and the periaqueductal gray), feed forward information to preganglionic and parasympathetic nerve cells that in turn innervate the vagus nerve (Thayer et al., 2009). Neural activity in these regions affects downstream parasympathetic activity (Thayer and Lane, 2000; Lane et al., 2009; Thayer et al., 2009; Buchanan et al., 2010). Moreover, emotion dysregulation has been linked to deficiency in the prefrontal brain regions (Beauchaine, 2015). Taken together, RSA, a measure of parasympathetic functioning, is heavily influenced by prefrontal cortex activity and constitutes a good indicator of emotion regulation.

Given that a potential shared mechanism of internalizing and externalizing behaviors is emotion dysregulation, we would expect both behavioral problems to be associated with abnormal resting RSA. Many empirical studies have supported this proposition by showing that either externalizing, internalizing, or both are associated with low resting RSA (Forbes et al., 2006; Dietrich et al., 2007; Hinnant and El-Sheikh, 2009). For example, in their longitudinal study with children, Hinnant and El-Sheikh (2009) showed that lower resting RSA (in conjunction with either RSA suppression or RSA augmentation) predicted increased internalizing symptoms and externalizing symptoms 2 years later. This study highlights not only the importance of resting RSA as a biomarker for emotion (dys)regulation, but also shows the longitudinal predictive power of this biomarker. On the other hand, Dietrich et al. (2007) found externalizing behavior to be associated with elevated resting RSA while internalizing behavior was associated with lower resting RSA; however, the authors report that this physiological pattern may underlie externalizing behaviors that necessitate increased emotion regulation, such as proactive aggression versus reaction aggression.

Changes in RSA from baseline to task have also been used to assess parasympathetic reactivity in response to external threats. RSA reactivity can indicate either vagal/RSA withdrawal or vagal/RSA augmentation. During vagal withdrawal, the vagus nerve withdraws its control over the heart, resulting in increased PNS (RSA) reactivity. Appropriate levels of vagal withdrawal are adaptive and necessary for effectively reacting to external threats. In contrast, during vagal augmentation, the vagus nerve increases its control over the heart, resulting in decreased PNS (RSA) reactivity. In certain circumstances vagal augmentation is beneficial, particularly when completing cognitive tasks that require heightened attention and emotional control (DiPietro et al., 1992); however, in most cases vagal augmentation may lead to “increased control of emotionality” that can “attenuate painful experiences and insensitivity to socializing punishments” (Scarpa and Raine, 2004).

Findings linking RSA reactivity and externalizing and internalizing behaviors are mixed. Many studies previously found externalizing and internalizing to relate to blunted and increased RSA reactivity, respectively (Boyce et al., 2001; Calkins et al., 2007; Hinnant and El-Sheikh, 2009; Fortunato et al., 2013). For example, Hinnant and El-Sheikh (2009) found that children with high internalizing symptoms and low resting RSA had excessive RSA withdrawal during emotion tasks, indicating hyper-reactivity. They also saw that those high in externalizing symptoms and that had low resting RSA showed reduced RSA withdrawal, suggesting maladaptive emotional withdrawal during stressful situations (Hinnant and El-Sheikh, 2009). Similarly, children at risk for mixed internalizing and externalizing problems demonstrated consistently elevated levels of RSA withdrawal during various tasks while children at risk for high externalizing problems only displayed the smallest amount of vagal withdrawal compared to the mixed problem and low risk group (Calkins et al., 2007). Additionally, Fortunato et al. (2013) found that children with high RSA withdrawal in response to a negative emotion induction showed increased levels of internalizing, while children with reduced RSA withdrawal to a positive emotion induction showed increased levels of externalizing behavior. Taken together, these findings suggest that children with externalizing behavior alone exhibit diminished PNS reactivity, whereas children with internalizing (on its own or comorbid with externalizing) exhibit excessive PNS reactivity (Boyce et al., 2001; Calkins et al., 2007; Hinnant and El-Sheikh, 2009; Fortunato et al., 2013). However, another study reported that increased RSA withdrawal in children during independent free play was associated with externalizing behaviors (Hastings et al., 2008) and many studies have failed to establish a link between PNS reactivity and externalizing/internalizing problems all together (Gordis et al., 2010).

The absence of a consistent relationship between RSA and externalizing/internalizing problems potentially reflects a failure to consider the impact of environmental influences. Prior research has found that both psychosocial adversity and psychophysiological factors are important for understanding the etiologies of child behavioral problems. Some have demonstrated that psychophysiological abnormalities such as lower resting RSA and abnormal RSA reactivity increase susceptibility to psychosocial risk factors for behavioral deficits. For example, lower levels of resting RSA increased risks for externalizing problems whereas elevated levels of resting RSA buffered against externalizing problems for boys exposed to familial stressors such as inter-parental arguments (El-Sheikh et al., 2001). Similarly, a study on stress exposure and its effect on externalizing/internalizing behaviors showed that childhood adversity (including traumatic events, community violence exposure, and child abuse) and resting RSA interact such that at low levels of resting RSA and increased adversity, depression/withdrawal symptoms increase, but not for adolescents with high resting RSA (McLaughlin et al., 2015). With respect to RSA reactivity, children with high RSA reactivity to a cognitive task demonstrated increased externalizing symptoms under conditions of high marital conflict, and reduced externalizing symptoms under conditions of low marital conflict; externalizing symptoms of children with low RSA reactivity were unaffected by marital conflict (Obradović et al., 2011). Taken together, the combination of low resting RSA or high RSA reactivity and high psychosocial adversity seem to increase risks for both internalizing and externalizing behaviors.

Additionally, evidence of gender differences with respect to psychosocial characteristics and physiological activity is becoming more apparent in the current literature. In one study, childhood stress exposure (i.e., abuse, community violence, trauma, etc.) interacted with resting RSA to predict anxiety/depression problems in male adolescents, but not in female adolescents (McLaughlin et al., 2015). Similarly, young boys at risk for externalizing showed lower resting RSA than low-risk boys but no such difference was observed between high- and low-risk girls (Calkins and Dedmon, 2000). Taken together, evidence suggests that RSA may be a stronger predictor of behavioral problems for males than for females, supporting the proposition that there are distinct mechanisms underlying the development of behavioral problems for the two gender groups (Beauchaine et al., 2008).

Prior studies have principally examined the concurrent associations between RSA and behavioral problems (Calkins and Dedmon, 2000; Calkins et al., 2007; Dietrich et al., 2007; Gordis et al., 2010; Fortunato et al., 2013; McLaughlin et al., 2015). There have been fewer longitudinal studies evaluating RSA and its interaction with environmental factors in predicting behavioral development over time (El-Sheikh and Whitson, 2006; Hinnant and El-Sheikh, 2009; El-Sheikh and Hinnant, 2011; Eisenberg et al., 2012). For example, a study on the longitudinal relationship between RSA, marital conflicts and child behaviors showed that lower initial resting RSA predicted externalizing behaviors from 8 to 11 years in boys (El-Sheikh and Hinnant, 2011). The aforementioned study of Hinnant and El-Sheikh (2009) found that the interaction between resting RSA and RSA reactivity predicted child externalizing/internalizing behaviors 2 years after. Developmental models might further clarify the stability and/or malleability of RSA activity, the process by which RSA may result in maladaptive behavioral development, and the potential of implications for preventive interventions.

In summary, atypical RSA, reflecting emotion dysregulation, is suggested to be a predisposing risk factor for both externalizing and internalizing disorders. In addition, the RSA-behavior problem association may be moderated by psychosocial factors and gender. The current study was conducted in order to examine the developmental trajectory of behavior problems in early childhood through the pre-teen years. Specifically, we aim to examine if RSA offers longitudinal prediction of externalizing or internalizing behaviors in male and female children from the community, and if social adversity moderates this relationship. We targeted children aged 7- to 10-years old in attempt to identify risk factors associated with the emergence and development of conduct disorder that is diagnosed at approximately age 10 [Diagnostic and statistical manual of mental disorders (DSM-5^®); American Psychiatric Association, 2013]. This age range is particularly relevant because previous studies have identified middle childhood as a period during which children are exposed to school-related demands and pathologies involving emotion dysregulation are increasingly diagnosed (Musser et al., 2011). Social adversity was assessed in Time 1 (T1) and psychophysiological measures and behavioral problems were assessed in both Time 1 (T1) and Time 2 (T2) with approximately 1 year apart. We hypothesized that (1) low resting RSA at T1 would predict high internalizing and externalizing behaviors at T2, (2) excessive RSA reactivity at T1 would predict high internalizing behavior at T2 whereas reduced RSA reactivity at T1 would predict high externalizing behavior at T2, and (3) RSA would interact with social adversity such that RSA abnormality in conjunction with high social adversity (but not low adversity) would predict greater externalizing or internalizing behaviors. Consistent with prior literature (El-Sheikh and Whitson, 2006; El-Sheikh et al., 2007; Hinnant and El-Sheikh, 2009), T1 behavioral measures were controlled for in our analyses to examine the influence of T1 RSA and social adversity over and beyond the effects of T1 behavioral measures on T2 measures. Since behavioral problems and pathologies involving emotion dysregulation emerge and increasingly develop during this age range, the effects of age were additionally controlled for in our analyses. In light of growing evidence that the mechanism underlying behavioral problems differs between males and females, we conducted analyses for boys and girls separately, though no formal gender-specific predictions were made given inconsistent findings in the literature.

The principle aim of the current study was to investigate if resting RSA or RSA reactivity, an indicator of emotion regulation, interacts with social adversity in prospectively predicting externalizing and internalizing behavior 1 year later in male and female children from the community. We found that in boys, lower resting RSA at T1 was associated with elevated levels of externalizing and internalizing behaviors at T2, over and above the effects of T1 behavioral problems and T2 age. Moreover, significant biosocial interaction effects were observed such that low T1 resting RSA predicted high externalizing and internalizing behaviors in the condition of high but not low social adversity. Follow-up analyses indicated that these predictive effects were stronger for externalizing behavior than for internalizing behavior. In contrast, neither T1 RSA reactivity nor its interaction with social adversity significantly predicted externalizing or internalizing behavior. Finally, none of these variables were significant in predicting behavioral problems in girls.

Consistent with our prediction, low T1 resting RSA was associated with high externalizing and internalizing behaviors T2, although these effects were found for boys only. These results support the theory that low resting RSA is an indicator of poor emotion and attentional regulation due to inefficient prefrontal functioning (Beauchaine, 2001, 2015; Crowell et al., 2006), and demonstrate that resting RSA offers independent longitudinal prediction of behavior problems, over and above the effects of concurrent behavioral problems. Taken together, RSA provides a non-invasive, non-intrusive, and reliable biological tool to assess the PNS activity that reflects emotion regulation, and low resting RSA may serve as a biomarker that specifically predisposes male children to externalizing behavior.

More importantly, we found that the predictive effect of resting RSA was only significant among those with adverse home backgrounds. In fact, low resting RSA and high social adversity in combination predicted the highest levels of both externalizing and internalizing behavioral problems. In contrast, high resting RSA served to protect children from developing either externalizing or internalizing behavioral problems in the face of adverse social environments (Figure 1). This is consistent with prior studies that have suggested high resting RSA may buffer against familial and environmental adversity (Eisenberg et al., 2012; McLaughlin et al., 2015), and offer protection under unpredictable or challenging circumstances (Appelhans and Luecken, 2006; Gyurak and Ayduk, 2008). High resting RSA has been associated with adaptive characteristics in children, including better physiological recovery from stress (Lane et al., 1992; Fabes et al., 1993; McLaughlin et al., 2014). Mechanistically, children with higher resting RSA (reflecting increased baseline parasympathetic physiological activity) may be better able to flexibly regulate emotions while experiencing adversity and recover sooner after exposure to high-stress events.

This finding is also in line with the biological sensitivity to context theory that has argued that certain biological phenotypes are differentially affected by environmental stressors. For example, children high in physiological arousal are capable of both flourishing in protective environments and suffering negative behavioral outcomes under hardship (Boyce and Ellis, 2005). One such biological phenotypic marker is low resting RSA, which indicates sympathetic nervous system dominance at rest, resulting in increased vigilance and limited physiological flexibility to meet environmental demands (Thayer and Lane, 2000; Hinnant and El-Sheikh, 2009; Viana et al., 2017). This in turn may predispose individuals to more negative behavioral outcomes under adversity (Boyce and Ellis, 2005). This argument is also partially consistent with the temperament-based theory of antisocial behavior which argues that temperamental constructs such as impulsivity, emotion regulation, and negative affect can contribute to the development of externalizing behaviors (DeLisi and Vaughn, 2014). These issues in contexts of increased adversity are exacerbated, while in positive or nourishing environments may not be as detrimental; alternatively, positive emotionality and good emotion regulation (such as high resting RSA) can protect a child in a highly adverse environment from manifesting externalizing behaviors (DeLisi and Vaughn, 2014).

Interestingly, low resting RSA and its interaction with social adversity were robustly associated with externalizing behavior even after controlling for the effect of concurrent internalizing behavior. However, the effects were somehow attenuated when predicting internalizing behavior after controlling for concurrent externalizing behavior. This finding may indicate some specificity of physiological risk factor, i.e., low resting RSA, for externalizing behavior and further suggest that externalizing and internalizing are two overlapping but distinct constructs, at least in middle childhood. Alternatively, the weaker predictive effects for internalizing behavior may be a function of informant under-reporting of internalizing symptoms that are more difficult to gauge from an outside perspective (Dietrich et al., 2007). Future studies utilizing multi-informant approaches are needed to replicate these findings.

Nonetheless, we found non-significant concurrent relationships between resting RSA and externalizing/internalizing behaviors in both genders at both time points (Table 2). While our results replicate a number of null findings of prior research (Hinnant and El-Sheikh, 2009; McLaughlin et al., 2015), they are inconsistent with several previous studies that documented significant associations between resting RSA and behavioral problems at the same time period. For instance, these studies found RSA was associated with concurrent behavior problems (i.e., externalizing or internalizing problems) in 2- to 3-year-old boys with clinical levels of externalizing behaviors (Calkins and Dedmon, 2000), 3- to 9-year-old children born to mothers with childhood-onset depression (Forbes et al., 2006), 8- to 12-year-old externalizing boys with severe conduct problems (Beauchaine et al., 2008), and 10- to 13-year-old externalizers with early-onset behavior problems (at ages 4–5) (Dietrich et al., 2007). Behavior problems assessed in the current community sample may represent a mild form of psychopathology, thus the inconsistency between the present results may be partially attributed to sample differences. Moreover, caution must be taken in interpreting the correlational findings, as we cannot draw any possible conclusion about cause and effect. Using a short-term longitudinal model, we were able to find that children with high resting RSA early in development are less likely to develop behavior problems in the context of high adversity than children with low resting RSA. This pattern further suggests that lower resting RSA represents a diathesis for behavioral problems upon exposure to environmental stressors.

Inconsistent with prior findings, we failed to find significant main or interaction effects involving RSA reactivity in predicting problematic behaviors. Differences in sample characteristics and types of tasks may partly contribute to this inconsistency. For example, significant associations between RSA reactivity and behavioral problems have been reported in a targeted sample of high externalizers from a typical population (Boyce et al., 2001), and preschoolers (Mage = 5) (Obradović et al., 2011). In addition, these studies have used passive emotional tasks or cognitive tasks (Boyce et al., 2001; Hinnant and El-Sheikh, 2009; Obradović et al., 2011) whereas ours is an effortful task involving explicit emotion regulation. Obradović et al. (2011) argued that elevated RSA activity in a cognitive task could indicate “adaptive engagement” for children in low conflict environments while conferring limitations under cognitive stress in high conflict environments (p. 110). Meanwhile, children with elevated RSA reactivity in an interpersonal task had consistent(ly) higher levels of externalizing behaviors across high and low conflict environments compared to children with low RSA reactivity to the interpersonal task. The mixed emotional and cognitive nature of our task (which is specific to emotion regulation rather than a stress response) makes it difficult to compare with prior studies. Finally, the moderate change of RSA in our sample may indicate that the emotion regulation task may not be strong enough to induce stable RSA reactivity. Future investigations using a consistent category of emotional task and a typical population of children in middle childhood are needed to clarify if RSA reactivity during these tasks can longitudinally predict behavioral problems.

Alternatively, this null finding for RSA reactivity may in part reflect the developmental characteristics of RSA measures in childhood. Prior research has suggested that resting RSA has better longitudinal stability than RSA reactivity (Bornstein and Suess, 2000; Hinnant and El-Sheikh, 2009; Pang and Beauchaine, 2013). For instance, a study in 8- and 12-year-old children with conduct problems and/or depression showed that resting RSA was relatively stable (inter-correlation from 1 year to the next ranged from 0.35 to 0.54), whereas RSA reactivity increased across time (not significantly correlated) (Pang and Beauchaine, 2013). Our sample additionally demonstrated this pattern: while resting RSA values at two times were significantly correlated (r ranged from 0.55 to 0.57, p < 0.001), RSA reactivity at T1 and T2 were not significantly correlated (r ranged from 0.07 to 0.22, p > 0.05). However, the degree to which this higher developmental malleability in RSA reactivity (as opposed to baseline measures) influences our findings is unknown.

In both gender groups resting RSA was positively associated with social adversity, supporting the proposition that early-life social adversity may directly alter the HPA axis in order to bring the stress response to more adaptive levels. Early adversity experiences have been related to changes in physiological response to stress, although the impacts on vagal activity have not been clearly demonstrated (Gunnar and Quevedo, 2007). In addition, boys had higher levels of externalizing behavior compared to girls at both time points, whereas girls had higher levels of internalizing behaviors than boys, although the latter differences were non-significant. This is in line with literature showing that the increased levels of internalizing and depressive symptoms in girls emerge during early adolescence (Ge et al., 2001). While it is also possible that boys simply exhibit more externalizing behaviors than girls, a likely alternative is that measures of externalizing behaviors are biased toward the kind of physical aggression more often seen in males (Zoccolillo, 1993; Crick and Grotpeter, 1995). Importantly, the predictive effects of RSA and social adversity were found for boys only. This compliments a growing body of literature showing that boys and girls do not uniformly express the same patterns of association between behavior problems and autonomic physiological activity. For example, Beauchaine et al. (2008) found that lower resting RSA was associated with increased aggression in males but no such relationship was observed in females. El-Sheikh and Hinnant (2011) found the initial resting RSA was negatively related to externalizing behaviors overtime for boys but not for girls. This again could be a function of a poor translation of current aggression measures to female aggressive behavior. Regardless, additional studies looking at measures of autonomic activity such as heart rate and skin conductance level similarly saw no relationship between behavior problems and physiological responding in girls (Isen et al., 2010; Fagan et al., 2017). Some posit that social and environmental influences are stronger, if not dominant, in predicting female externalizing behaviors, and that there could be sex-specific genetic effects driving the development of these behaviors (Beauchaine et al., 2008). Our correlational results support this idea as they show a direct relationship between social adversity and externalizing and internalizing behavior in females but not in males (Table 2).

One limitation of the current study is that we only offer two data points, 1 year apart, due to funding limitations. Although this is not sufficient to gauge a stable developmental trajectory, it does offer some preliminary examination of a developmental critical period (early childhood to pre-adolescence) and the (relative) long-term influence of physiological activity on behavior. In addition, our reports of both internalizing and externalizing behaviors were from caregivers only; having multiple informants including teacher/self-report and behavioral observation would provide a more holistic and reliable assessment. Lastly, we assessed youths from a community setting where the base rate of high scorers is generally low. Future studies integrating prospective longitudinal assessment of RSA, laboratory-based research methods, and multi-informant measures of clinical and non-clinical samples are needed to replicate our findings.

Overall, our results provide further longitudinal evidence (El-Sheikh and Whitson, 2006; Hinnant and El-Sheikh, 2009; El-Sheikh and Hinnant, 2011; Eisenberg et al., 2012) that low resting RSA may be a transdiagnostic biomarker of emotion dysregulation and that it is a predisposing risk factor for both externalizing and internalizing behavior problems during development. Further understanding of the role of RSA is consistent with the aim of the Research Domain Criteria (RDoC) initiative of the National Institute for Mental Health (Insel et al., 2010) which focuses on “new ways of classifying psychopathology based on dimensions of observable behavior and neurobiological measures” with the objective of defining “basic dimensions of functioning … cutting across disorders as traditionally defined” (Cuthbert and Insel, 2013). Our findings also highlight the importance of examining biosocial interaction effects on childhood psychopathology for males and females separately.

WZ and SF carried out data collection, data analysis and interpretation, drafting of the article, and critical revision of the article. YG carried out conception and design of the study, coordination and supervision of data collection, drafting the article, and critical revision of the article. All authors have approved the final article.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.