Introduction

Resting heart rate (RHR) is one of the most replicated, biological correlates of antisocial behavior (ASB; de Looff et al., 2022; Latvala et al., 2015), a set of behaviors that “bring harm to another person or involve the violation of the rights of others” (Tuvblad & Beaver, 2013). RHR has been examined in relation to ASB for decades, and the first study on this topic reported that refractory adolescent boys had lower RHR than controls (Davies & Maliphant, 1971). Since the study by Davies et al., (1971), the association between RHR and ASB has been explored in different samples; incarcerated individuals (Koegl et al., 2018), adult community samples (Jennings et al., 2013), total population (Latvala et al., 2015), child samples (Raine et al., 1997), across various cultures (Murray et al., 2016) and across different types of ASB; violent and non-violent behavior (Bertoldi et al., 2022), criminal convictions (Latvala et al., 2015), reoffending (Oskarsson et al., 2021), psychopathy (Kavish et al., 2017), and aggression (Raine et al., 2014). Several meta-analyses have concluded that RHR is a robust correlate of ASB, reporting effect sizes of Cohen’s d = − 0.38 (Lorber, 2004), d = − 0.44 (Ortiz & Raine, 2004), d = − 0.20 (Portnoy & Farrington, 2015), and Hedges g = − 0.17 (de Looff et al., 2022). The association between RHR and ASB remains after adjusting for potentially confounding factors, indicating that RHR has an independent effect on ASB (Latvala et al., 2015; Oskarsson et al., 2021).

Despite the robust findings supporting the link between RHR and ASB, recent evidence from a Mendelian randomization study suggests that confounding factors may account for the association between RHR and ASB (Kartwatowska et al., 2023). One proposed explanation by Kartwatowska et al. (2023) suggests that the association between RHR and ASB could be driven by sensation-seeking behavior. While the association between RHR and ASB is well-established, the precise mechanisms underlying this relationship remain largely unknown. Sensation-seeking emerges in the literature as a potential explanatory framework for the association between RHR and ASB (Beauchaine, 2012; Eyesenck, 1997). According to this theory, individuals with lower RHR exhibit reduced autonomic arousal which puts them at an unpleasant physiological state. Consequently, these individuals may engage in stimulating behaviors, such as ASB, to elevate their physiological arousal to a more optimal level. Sensation-seeking can be distinguished by autonomic arousal, with high sensation-seekers exhibiting lower autonomic arousal and low sensation-seekers exhibiting higher autonomic arousal (Zuckerman, 1994). The characteristic of sensation-seeking has been positively associated with ASB in prior work (Berg et al., 2015; Mann et al., 2017; Okuda et al., 2019; Wilson & Scarpa, 2011; Zuckerman, 1994) and sensation-seeking has been suggested to represent a personality endophenotype for ASB (Mann et al., 2017).

Although considered a well-established theory of ASB (Quay, 1965), few studies have undertaken the task to empirically validate the sensation-seeking theory (Portnoy et al., 2014; Wilson & Scarpa, 2014). A large prospective study utilized a sample of 1,752 individuals assessed RHR at age 11 and ASB at age 16 (Sijtsema et al., 2010). Results showed that among males, sensation-seeking at age 13.5 and 16 partially mediated the association between RHR at age 11 and ASB at age 16, lending support to sensation-seeking as a theoretical basis to explain the association between RHR and ASB. Another study used a subsample of 335 adolescents from the Pittsburg Youth Study (Loeber et al., 1998) to examine whether impulsive sensation-seeking mediated the association between RHR and aggression (Portnoy et al., 2014). Lower RHR was associated with a higher level of aggression, and this association was partly mediated by impulsive sensation-seeking. A third study, utilizing a sample of 4,046 individuals, showed that the association between lower RHR in childhood and higher levels of ASB in adolescence was mediated by sensation-seeking (Hammerton et al., 2018). Taken together, these findings suggest that sensation-seeking at least in part underlies the association between RHR and ASB. More recent findings suggest that disinhibition, a set of traits that include boredom proneness, difficulty delaying gratification, and a lack of behavioral and emotional restraint (Krueger et al., 2007), partially accounted for relations of RHR with ASB through the covariance between disinhibition and ASB (Bertoldi et al., 2022).

Despite some evidence of a mediating role of sensation-seeking in the association between RHR and ASB, other theoretical models suggest that sensation-seeking may instead have a moderating role. The biosocial model of ASB suggest that the biological risk factor RHR and the psychosocial risk factor sensation-seeking, interact to increase the risk of ASB (Barnes et al., 2022). In a similar fashion, the diathesis-stress model explains how negative psychosocial factors can amplify a biological risk, or “diathesis”, for a particular outcome (Barnes et al., 2022). The diathesis-stress model is supported if the likelihood of observing the outcome is highest among individuals who are exposed to both the psychosocial risk factor and the biological risk factor. Thus, this theoretical model suggests that a lower level of RHR in combination with a higher level of sensation-seeking behaviors may put individuals at risk of ASB.

One previous study examined the moderating role of sensation-seeking in the association between RHR and ASB (Wilson & Scarpa, 2014). Results partially supported an interaction effect where a lower RHR was predictive of ASB among individuals with high sensation-seeking tendencies. However, this study utilized a small cross-sectional sample of 128 college students, which may not generalize to non-student samples.

With the current study, we aim to increase our understanding of the underlying mechanisms of the association between RHR and ASB. We sought to extend previous findings of sensation-seeking as a mediator in the association between RHR and ASB. We also sought to examine how sensation-seeking may moderate the association between RHR and ASB, while addressing some of the methodological limitations of previous research.

Our specific hypotheses were as follows:

  1. 1.

    RHR was expected to relate negatively to ASB, based on the extensive evidence for a relationship between RHR and ASB (see for example de Looff et al., 2022).

  2. 2.

    Given earlier findings of a positive association between sensation-seeking and ASB (see for example Okuda et al., 2019), we predicted a positive association between sensation-seeking and ASB.

  3. 3.

    Based on earlier findings supporting sensation-seeking as a mediator in the association between RHR and ASB (Portnoy et al., 2014; Sijtsema et al., 2010), we hypothesized that the association between RHR and ASB will be mediated by sensation-seeking. We further hypothesize that any indirect effects will remain significant after controlling for potential confounders.

  4. 4.

    Earlier findings of an interaction between RHR and sensation-seeking (Wilson & Scarpa, 2014) in combination with the biosocial model and the diathesis-stress model, suggest that sensation-seeking will interact with RHR to predict ASB. More specifically, we hypothesize that among individuals with high levels of sensation-seeking, a lower RHR will be associated with a higher risk of ASB.

Methods

Participants

We utilized data from the University of Southern California (USC) Risk Factors for Antisocial Behavior (RFAB) Project, which has been described in detail elsewhere (Baker et al., 2013). The base sample of the RFAB study included 1,673 participants, with 51.4% females. We included two subsamples in the current study: the first sample (W1 subsample) included all participants who had data on resting heart rate (RHR), body mass index (BMI), and race/ethnicity, all at W1, when participants were 9–10 years old, as well as data on sensation-seeking and ASB at the fifth wave (W5) when participants were 19–20 years old (n = 690, 56% female). The second subsample (W3 subsample) included all participants who had data on ethnicity at W1, data on RHR and BMI at the third wave (W3) when participants were 13–14 years old, as well as data on sensation-seeking and ASB at W5 (n = 391, 59% female). The two subsamples were not mutually exclusive.

Caregivers of all participants in the RFAB project provided informed consent during W1. During W5, the participants provided their own informed consent. The Institutional Review Board at USC approved of the project (Baker et al., 2013).

Measures

Resting Heart Rate

Resting heart rate (RHR) was measured as part of a larger psychophysiological measurement battery at W1 and W3. Participants had disposable electrodes attached to their ribs and a 38-channel Isolated Bioelectric Amplifier from the James Long Company (1999; Caroga Lake, New York) was used to acquire data. An Interbeat Interval Analysis Software program (James Long Company) was used to analyze heart rate data and RHR was defined as beats per minute (BPM) averaged across three minutes. There were 25 individuals who had less than one minute of usable RHR data, and they were therefore removed from further analyses (Bertoldi et al., 2022). The RHR between the two time-points was strongly positively associated (r = .46, p < .001) among participants who had their RHR measured at both waves.

Sensation-Seeking

Sensation-seeking was measured using the Urgency-Premeditation-Perseverance-Sensation Seeking-Positive Urgency (UPPS-P) Impulsive Behavior Scale, which is a self-report questionnaire that assesses impulsivity as a multidimensional construct (Lynam et al., 2006; Whiteside & Lynam, 2001). The UPPS-P consists of five subscales, each measuring a different facet of impulsivity. For the present study, the subscale sensation-seeking was used (W1 subsample α = 0.83; W3 subsample α = 0.83). Sensation-seeking in the UPPS-P is defined as the tendency to seek out exciting or novel experiences and take risks potentially associated with experiencing such sensations and thus corresponds to the theoretical ideas around the association between RHR and ASB (see Beauchaine, 2012; Eyesenck, 1997). Examples of items found in the sensation-seeking subscale are “I crave excitement and new sensations” and “I enjoy doing things that are a little frightening”. Participants are asked to indicate to what extent they agree or disagree with each statement: agree strongly (1), agree somewhat (2), disagree somewhat (3), disagree strongly (4). All items on the sensation-seeking subscale were reverse-coded and a higher score indicated higher levels of sensation-seeking.

Outcome

During the data collection at W5, participants were administered a questionnaire, the Self-Report Delinquency Interview (SR-DI; Wang et al., 2013) about antisocial behavior (ASB) that was specifically developed for the RFAB project. The SR-DI was inspired from several already existing measures of ASB, such as the Self-Report Delinquency in Adolescence (SRA) used in the Pittsburgh Youth Study (Loeber & Farrington, 1998). In this questionnaire, two standalone questions were provided to the participants to index life-time legal involvement: “Have you ever been in trouble with the police?” and “Have you ever been arrested”. For the present study, both variables were included as separate outcomes. The questions were dichotomous and coded 0 for no and 1 for yes.

We also included two subscales comprising of several questions from the SR-DI that assessed whether participants had ever engaged in different behaviors reflecting violent (nine items) and non-violent behavior (34 items). Violent behavior items focused on actions causing physical harm to others and the use of weapons whereas non-violent behavior encompassed behaviors such as theft, property damage, fraud, drug selling, disorderly conduct, and reckless driving. Questions were answered with yes (1) or no (0) and the items were averaged to form the scores for violent and non-violent behavior respectively. The combination of the two stand-alone questions together with the subscales of violent and non-violent behavior has been used several times in previous research to index ASB (see Bertoldi et al., 2022; Bertoldi et al., 2023). For more information about the creation of scales, please refer to Bertoldi et al. (2023).

Covariates

We adjusted for potential confounders to isolate the association between RHR and ASB. Firstly, we adjusted for BMI (measured at W1 for the W1 subsample and W3 for the W3 subsample), because it may be associated with both RHR and ASB (Beckley et al., 2014). We also adjusted for race/ethnicity, measured at W1, because evidence has been put forth about the overpolicing of Black and Hispanic individuals compared to White individuals in the United States (Gaston, 2019). Studies have also shown that the association between RHR and ASB is racially variant (Portnoy et al., 2020), supporting the need to include this as a covariateFootnote 1. Lastly, we included a measure the Hollingshead (1979) Index of socioeconomic status (SES; score range 8–66) during W1 as a covariate given its strong association with ASB (Bertoldi et al., 2022).

Statistical Analyses

All data management and analyses were performed using R Studio version 2022.12.0 (R Development Core Team, 2020). We started with examining zero-order associations between RHR, sensation-seeking, BMI, trouble with the police, arrests, violent and non-violent behavior.

Next, we used the PROCESS macro for R Studio (Hayes, 2022) to test the significance of indirect effects of sensation-seeking on ASB while controlling for covariates. We used a total of 5,000 bootstrapped samples drawn from the original data. For the binary outcomes, the a-path from the independent variable (RHR) to the mediator (sensation-seeking) was estimated with ordinary least squares regression, whereas the b-path from the mediator (sensation-seeking) to the dependent variable (trouble with the police/arrest) was estimated with logistic regression. In the results section, the a-path is reported using unstandardized regression coefficients whereas the b-path is reported in odds ratios (ORs) to ease interpretation. Log-odds metrics for the b-path can be found in the figures in the supporting information. The indirect effects of RHR on trouble with the police and arrests respectively, were calculated as the product of the ordinary least-squares coefficient for the relationship between RHR and sensation-seeking and the logistic regression coefficient for the relationship between sensation-seeking and trouble with the police/arrests. For the continuous outcomes, the paths were estimated with ordinary least squares regression. The indirect effects of RHR on violent and non-violent behavior were calculated as the product of the two ordinary least squares regression coefficients. All direct and indirect effects are reported as unstandardized coefficients.

Next, we used logistic regression to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for interactions between RHR and sensation-seeking in predicting trouble with the police and arrests. In the analyses, we included RHR and sensation-seeking as predictive factors, alongside an interaction term between the two predictors. Our focus was on predicting the two outcomes: trouble with the police and arrests. Throughout the analyses we accounted for the covariates listed above. This resulted in four separate models: two for each subsample (W1 and W3). All predictors were mean centered prior to inclusion in the moderation analyses. Unadjusted results are provided in Table S1 and S2 in the supporting information.

Lastly, linear regressions were used to estimate associations and 95% CIs for the continuous outcomes. In each linear regression model, RHR and sensation-seeking were entered as predictive factors along with an interaction term between the two predictors. Violent and non-violent behavior were entered as separate outcomes. All analyses adjusted for the covariates listed above. This resulted in a total of four separate models: two for each subsample (W1 and W3). Unadjusted results are provided in Table S3 and S4 in the supporting information.

It should be noted that associations between RHR, trouble with the police, and arrest have been published previously using slightly different subsamples from the RFAB (Bertoldi et al., 2022; Oskarsson et al., 2024). Given the interaction analyses in the present study, main effects could however not be omitted from results.

Sensitivity Analysis

We restricted our W1 subsample to include only those who in addition to RHR data during W1 also had RHR data during W3 (n = 376, 59% female; see Table S5 and S6 in the supporting information) to test the robustness of our results. The restriction of our sample enabled for examining whether the timing of the RHR measurement (i.e., childhood or adolescence) could explain the previously reported null-findings (Bertoldi et al., 2022; Oskarsson et al., 2024).

Results

Baseline characteristics of the two subsamples are shown in Table 1.

Table 1 Descriptive information for the W1 and the W3 subsamples
Full size table

Zero-order associations for the W1 subsample are displayed in Table 2. In the W1 subsample, RHR was negatively associated with sensation-seeking, r = − .09, p < .05, getting in trouble with the police, r = − .15, p < .001, getting arrested, r = − .15, p < .001, violent behavior, r = − .16, p < .001, and non-violent behavior, r = − .14, p < .001. Sensation-seeking was positively related to getting in trouble with the police, r = .20, p < .001, getting arrested, r = .15, p < .001, violent behavior, r = .26, p < .001, and non-violent behavior, r = .31, p < .001.

Table 2 Zero-order associations between predictors, covariates, and outcomes for subsample W1 (n = 690)
Full size table

Zero-order associations for the W3 subsample are displayed in Table 3. In the W3 subsample, RHR was negatively associated with sensation-seeking, getting in trouble with the police, getting arrested, violent, and non-violent behavior, although only the association with sensation-seeking was significant, r = − .12, p < .05. Sensation-seeking was positively related to getting in trouble with the police, r = .16, p < .01, getting arrested, r = .10, p < .05, violent behavior, r = .23, p < .001, and non-violent behavior, r = .30, p < .001.

Table 3 Zero-order associations between predictors, covariates, and outcomes for subsample W3 (n = 391)
Full size table

Mediation Effects

W1 Subsample

In the W1 subsample, mediation models were constructed to assess the mediating role of sensation-seeking in the association between RHR with ASB, while accounting for race/ethnicity, BMI, and SES. In line with zero-order associations, results revealed that RHR was negatively associated with sensation-seeking (Est. = − 0.05, p = .022, 95% CI = − 0.10, − 0.01). An increased RHR was further associated with a decreased odds of getting in trouble with the police and being arrested (OR = 0.68, p < .001, 95% CI = 0.54, 0.83; OR = 0.55, p < .001, 95% CI = 0.41, 0.74). Lastly, RHR was negatively associated with violent and non-violent behavior (Est. = − 0.02, p < .001, 95% CI = − 0.03, − 0.01; Est. = − 0.01, p = .003, 95% CI = − 0.021, − 0.004).

An increased level of sensation-seeking was significantly associated with an increased odds of getting in trouble with the police and getting arrested (OR = 2.33, p < .001, 95% CI = 1.60, 3.38; OR = 2.36, p < .001, 95% CI = 1.43, 3.91). Sensation-seeking was further positively associated with violent and non-violent behavior (Est. = 0.05, p < .001, 95% CI = 0.03, 0.06; Est. = 0.06, p < .001, 95% CI = 0.04, 0.07). Results further revealed that sensation-seeking partly mediated the association between RHR and all of the four ASB outcomes, see Table 4. These results suggest that RHR is indirectly related to ASB through its association with sensation-seeking. Figures for the mediation analyses are shown in Supporting information.

Table 4 Summary of direct and indirect effects of RHR through sensation-seeking on antisocial behavior in the W1 subsample
Full size table

W3 Subsample

In the W3 subsample, mediation models were constructed to assess the mediating role of sensation-seeking in the association between RHR with ASB, while accounting for race/ethnicity, BMI, and SES. In line with zero-order associations, results revealed that RHR was negatively associated with sensation-seeking (Est. = − 0.07, p = .016, 95% CI = − 0.13, − 0.01). As expected, RHR was not significantly associated with getting in trouble with the police or being arrested (OR = 0.84, p = .248, 95% CI = 0.62, 1.13; OR = 0.80, p = .322, 95% CI = 0.52, 1.24). RHR was also not significantly associated with violent and non-violent behavior (Est. = − 0.003, p = .676, 95% CI = − 0.01, 0.01; Est. = − 0.001, p = .885, 95% CI = − 0.01, 0.01).

An increased level of sensation-seeking was significantly associated with an increased odds of getting in trouble with the police (OR = 1.84, p = .016, 95% CI = 1.12, 3.02) but not getting arrested (OR = 1.84, p = .097, 95% CI = 0.90, 3.79). Sensation-seeking was further positively associated with violent and non-violent behavior (Est. = 0.04, p < .001, 95% CI = 0.02, 0.06; Est. = 0.05, p < .001, 95% CI = 0.04, 0.07). Results further revealed that sensation-seeking partly mediated the association between RHR and three out of the four ASB outcomes, see Table 5. These results suggest that RHR is indirectly related to ASB through its association with sensation-seeking. Figures for the mediation analyses are shown in Supporting information.

Table 5 Summary of direct and indirect effects of RHR through sensation-seeking on antisocial behavior in the W3 subsample
Full size table

Interaction Effects

W1 Subsample

In the W1 subsample, we tested for interactive effects of RHR and sensation-seeking in predicting ASB, while accounting for race/ethnicity, BMI and SES. The first model using logistic regression with getting in trouble with the police as the outcome, showed no significant interaction (p > .05). Significant main effects were observed for both RHR and sensation-seeking (OR = 0.70, 95% CI = 0.56, 0.88 and OR = 1.59, 95% CI = 1.27, 1.99 respectively). Thus, a one unit increase in RHR was associated with 30% decreased odds of getting in trouble with the police, whereas a one unit increase in sensation-seeking was associated with 59% increased odds of getting in trouble with the police.

The second model with arrest as the outcome revealed a moderating effect of sensation-seeking (p = .003), and to characterize the nature of the effect, we probed the association between RHR and arrests at low (-1 SD) and high (+ 1 SD) sensation-seeking. Whereas no significant effect of RHR was observed at 1 SD below the mean of sensation-seeking (OR = 0.95, 95% CI = 0.58, 1.45), a significant effect of RHR was evident at 1 SD above the mean of sensation-seeking (OR = 0.40, 95% CI = 0.26, 0.59). Thus, at high levels of sensation-seeking, a one SD increase in RHR was associated with 60% decreased odds of getting arrested (see Fig. 1). A main effect was evident for RHR (OR = 0.55, 95% CI = 0.40, 0.73), indicating that a one unit increase in RHR, decreased the odds of getting arrested with 45%. A main effect was evident also for sensation-seeking (OR = 1.66, 95% CI = 1.24, 2.26), indicating that a one unit increase in sensation-seeking increased the odds of getting arrested with 66%.

Fig. 1
figure 1

Note. Associations between RHR and arrests with sensation-seeking in the W1 subsample. The figure displays regression lines illustrating levels of sensation-seeking: 1 SD below the sample mean and 1 SD above the sample mean. Shaded areas represent confidence intervals

Full size image

The third model with violent behavior as the outcome revealed a moderating effect of sensation-seeking (p = .016), and to characterize the nature of the effect, we probed the association between RHR and violent behavior at low (-1 SD) and high (+ 1 SD) sensation-seeking. Whereas no significant effect of RHR was observed at 1 SD below the mean of sensation-seeking (Est. = − 0.01, 95% CI = − 0.03, 0.01, p > .05), a significant effect of RHR was evident at 1 SD above the mean of sensation-seeking (Est. = − 0.05, 95% CI = − 0.07, − 0.03, p < .001). Thus, at high levels of sensation-seeking, a 1 SD increase in RHR predicted a 0.05 unit decrease in violent behavior, see Fig. 2.

Fig. 2
figure 2

Note. Associations between RHR and violent behavior with sensation-seeking in the W1 subsample. The figure displays regression lines illustrating levels of sensation-seeking: 1 SD below the sample mean and 1 SD above the sample mean. Shaded areas represent confidence intervals

Full size image

The fourth model with non-violent behavior as the outcome, showed no significant interaction (p > .05). Significant main effects were observed for both RHR and sensation-seeking (Est. = − 0.01, 95% CI = − 0.02, − 0.01, p = .005; Est. = − 0.04, 95% CI = 0.03, 0.05, p < .001). Thus, higher levels of RHR predicted lower levels of non-violent behavior whereas higher levels of sensation-seeking predicted higher levels of non-violent behavior. For a detailed presentation of all results from the W1 subsample, please refer to Tables 6 and 7.

Table 6 Logistic regression results for W1 RHR with W5 ASB outcomes and interaction effects with W5 sensation-seeking
Full size table
Table 7 Linear regression results for W1 RHR with W5 violent and non-violent behavior and interaction effects with W5 sensation-seeking
Full size table

W3 Subsample

In the W3 subsample, we tested for interactive effects of RHR and sensation-seeking in predicting ASB, while accounting for race/ethnicity, BMI and SES. The first model using logistic regression with getting in trouble with the police as the outcome, showed a significant interaction (p = .002), and to characterize the nature of the effect, we probed the association between RHR and trouble with the police at low (-1 SD) and high (+ 1 SD) sensation-seeking. Whereas no significant effect of RHR was observed at 1 SD below the mean of sensation-seeking (OR = 1.36, 95% CI = 0.90, 2.00), a significant effect of RHR was evident at 1 SD above the mean of sensation-seeking (OR = 0.55, 95% CI = 0.36, 0.82). Thus, at high levels of sensation-seeking, a one SD increase in RHR was associated with 45% decreased odds of getting in trouble with the police (see Fig. 3). A main effect was evident for sensation-seeking (OR = 1.44, 95% CI = 1.08, 1.95), indicating that a one unit increase in sensation-seeking increased the odds of getting in trouble with the police with 44%. No main effect was found for RHR (p > .05).

Fig. 3
figure 3

Note. Associations between RHR and trouble with the police with sensation-seeking in the W3 subsample. The figure displays regression lines illustrating levels of sensation-seeking: 1 SD below the sample mean and 1 SD above the sample mean. Shaded areas represent confidence intervals

Full size image

A second model with arrest as the outcome showed a significant interaction (p < .001), and to characterize the nature of the interaction effect, we probed the association between RHR and arrests at low (-1 SD) and high (+ 1 SD) sensation-seeking. Whereas no significant effect of RHR was observed at 1 SD below the mean of sensation-seeking (OR = 1.66, 95% CI = 0.93, 2.80), a significant effect of RHR was evident at 1 SD above the mean of sensation-seeking (OR = 0.46, 95% CI = 0.26, 0.79). Thus, at high levels of sensation-seeking, a one SD increase in RHR was associated with 54% decreased odds of getting arrested (see Fig. 4). There were no main effects for RHR or sensation-seeking (p’s > 0.05).

Fig. 4
figure 4

Note. Associations between RHR and arrests with sensation-seeking in the W3 subsample. The figure displays regression lines illustrating levels of sensation-seeking: 1 SD below the sample mean and 1 SD above the sample mean. Shaded areas represent confidence intervals

Full size image

The third model with violent behavior as the outcome, showed no significant interaction (p > .05). There was no main effect for RHR (p > .05). There was a main effect of sensation-seeking (Est. = 0.04, 95% CI = 0.02, 0.07, p < .001) such that higher levels of sensation-seeking predicted higher levels of violent behavior.

The fourth model with non-violent behavior as the outcome, showed no significant interaction (p > .05). There was no main effect for RHR (p > .05). There was a main effect of sensation-seeking (Est. = 0.04, 95% CI = 0.03, 0.06, p < .001) such that higher levels of sensation-seeking predicted higher levels of non-violent behavior. For a detailed presentation of all results from the W3 subsample, please refer to Tables 8 and 9.

Table 8 Logistic regression results for W3 RHR with W5 ASB outcomes and interaction effects with W5 sensation-seeking
Full size table
Table 9 Linear regression results for W3 RHR with W5 violent and non-violent behavior and interaction effects with W5 sensation-seeking
Full size table

Sensitivity Analysis

The estimates were not affected by restricting the W1 subsample to include only individuals who had data on RHR at both waves (i.e., W1 and W3; Table S5-S6).

Discussion

Although the association between resting heart rate (RHR) and antisocial behavior (ASB) has been extensively studied, the underlying mechanisms remain insufficiently understood. Our most important findings are that the overall pattern observed within the data of the present study suggests that sensation-seeking partially mediates the association between RHR and ASB as well as interacts with RHR in predicting ASB. Neither of these findings have previously been demonstrated using data from the RFAB study. Thus, our findings largely support sensation-seeking as a theoretical explanation to why individuals with lower RHR have an increased risk of engaging in ASB. Our findings further support alternative theoretical models such as the biosocial model and the diathesis-stress model, which suggests that sensation-seeking has a moderating influence on the association between RHR and ASB (Barnes et al., 2022; Raine, 2002). These findings highlight the complexity of the association between RHR and ASB, suggesting that the relationship varies depending on the level of sensation-seeking.

Sensation-seeking theory has been used as a way to explain why individuals with lower RHR have a higher risk of engaging in ASB compared to individuals with higher RHR (Beauchaine, 2012; Eyesenck, 1997). Sensation-seeking theory rests on the premise that individuals with lower RHR have a lower level of arousal that is perceived as unpleasant. The theory further suggests that individuals with lower RHR may engage in stimulating behaviors, such as ASB, to increase their arousal to a more optimal level (Beauchaine, 2012; Eyesenck, 1997; Portnoy et al., 2014). These theoretical ideas have been supported by a comprehensive meta-analysis encompassing 40 studies, illustrating a robust positive association between sensation-seeking and aggression (Wilson & Scarpa, 2011). Moreover, scholars have emphasized a notable link between impulsivity and sensation-seeking (Zuckerman, 1994), commonly depicting these as distinct yet intertwined concepts (Ravert & Donnellan, 2021). Consequently, sensation-seeking manifests through behavioral tendencies that actively pursue novel, intense, and thrilling experiences, typically entailing psychological features such as risk-taking and impulsivity – traits intricately associated with ASB (e.g., Portnoy et al., 2014). However, there has been little empirical evidence to support sensation-seeking as a mediator in the association between RHR and ASB (Portnoy et al., 2014; Sijtsema et al., 2010). Our results demonstrate that sensation-seeking partially mediated the association between RHR and ASB, except for the arrest outcome in the W3 subsample. It is likely that the non-significant indirect effect observed was due to limited statistical power given the few individuals who reported getting arrested in the W3 subsample. This potential statistical power issue is also evident from the estimate and the wide confidence intervals in the interaction analyses in the W3 subsample, where results revealed a non-significant main effect for sensation-seeking in predicting getting arrested. Overall, our results are in line with sensation-seeking theory which suggests that sensation-seeking serves as a potential pathway though which low RHR influences ASB, which adds to the body of knowledge trying to elucidate the association between RHR and ASB.

In addition to sensation-seeking mediating the association between RHR and ASB, we also demonstrate that sensation-seeking interacted with RHR in predicting ASB. These findings are in line with results by Wilson and Scarpa (2014), who demonstrated sensation-seeking to partly moderate the association between RHR and aggression. More specifically, our results showed that a lower RHR increased the odds of getting in trouble with the police in the W3 subsample and getting arrested in both the W1 and the W3 subsample, as well as the violent behavior outcome in the W1 subsample, among individuals with higher but not lower levels of sensation-seeking. Furthermore, we identified main effects for both RHR and sensation-seeking in predicting ASB, including all four outcomes, in the W1 subsample. Lastly, RHR did not have a main effect on ASB in the W3 subsample. Taken together, this study contributes to the field by demonstrating that sensation-seeking serves not only as a mediator as reported by previous studies (Portnoy et al., 2014; Sijtsema et al., 2010), but also as a moderator in the RHR-ASB relationship.

These findings align with the principles of biopsychosocial criminology, specifically highlighting the relevance of the biosocial model and the diathesis-stress model (Barnes et al., 2022) for understanding the development of ASB. Our findings support the biosocial model by revealing that ASB is predicted by the interaction between a biological risk factor (RHR) and a psychosocial risk factor (sensation-seeking; Barnes et al., 2022). Our findings also support the diathesis-stress model, which suggests that ASB arises when a “diathesis” (such as a biological vulnerability, i.e., RHR) combines with a psychosocial risk factor (i.e., sensation-seeking; Barnes et al., 2022). Our interpretation underscores that a biological vulnerability alone, as represented by a low RHR, is insufficient to fully explain why individuals engage in ASB. Instead, the risk of ASB escalates when this biological vulnerability interacts with a psychosocial risk factor, such as high sensation-seeking.

The interaction between biological factors, such as a low RHR, and psychosocial factors, such as sensation-seeking, may create a synergistic effect that significantly amplifies the risk of engaging in ASB. Sensation-seeking behavior, characterized by a desire for novel and intense experiences may drive individuals to seek out risky and thrilling activities. When combined with a low RHR, which is indicative of reduced physiological arousal, this propensity for sensation-seeking can exacerbate the effects of low arousal. Individuals with low RHR may have difficulty experiencing arousal or excitement from everyday stimuli. As a result, they may seek out more extreme or dangerous activities to achieve the heightened arousal they crave. Sensation-seeking behavior serves to compensate for this deficit in arousal, leading individuals to engage in impulsive and antisocial actions in pursuit of stimulation. Moreover, the combination of sensation-seeking and RHR may impair inhibitory control and decision-making processes. Sensation-seeking individuals with low physiological arousal may be more prone to act on impulse, without fully considering the consequences of their actions (Beauchaine, 2012).

While our analyses revealed a significant main effect of RHR in predicting ASB within the W1 subsample, the interaction analysis in the W3 subsample unveils a more nuanced picture. It shows that the association between RHR and ASB is particularly evident among individuals who, in combination with a low RHR, also exhibit high levels of sensation-seeking. This suggests that a lower level of sensation-seeking may serve as a protective buffer against the biological risk factor, thereby reducing the likelihood of engaging in ASB. In a similar fashion, a higher level of sensation-seeking amplifies the influence of lower RHR as a biological risk factor to predict an increased risk of ASB (Barnes et al., 2022). These findings may have implications for the field of biopsychosocial criminology, indicating the necessity of studying the association between RHR and ASB within a broader context. Considering multiple risk factors and their interactions becomes crucial for understanding the development of ASB.

Our study findings are based on a young sample, in which individuals are assessed in childhood, adolescence, and young adulthood. Adolescence is recognized as a developmental period characterized by heightened risk-taking behaviors (Farrington et al., 2006). This developmental context may further magnify the significance of our results. During adolescence, individuals are more inclined to seek out novel and thrilling experiences, and this propensity for sensation-seeking may interact with biological factors, such as RHR, to escalate risk-taking behaviors. Adolescence is also a time when peer interactions play a significant role in shaping behavior (Moffitt, 1993). The findings regarding the interaction between RHR and sensation-seeking, as well as the mediating effect of sensation-seeking on the association between RHR and ASB, may have implications for understanding how peer dynamics contribute to the development of ASB. Adolescents who are sensation-seeking and have low RHR may be more susceptible to peer influences that promote ASB, such as peer pressure or social norms endorsing risky conduct.

Our findings support lower RHR as a risk factor for ASB. Given the extensive support for this association, future research should aim to include RHR in risk prediction models to evaluate the potential predictive value in risk assessment protocols. The mediation of sensation-seeking in the RHR-ASB relationship, as well as the interaction between RHR and sensation-seeking in predicting ASB holds potential implications for intervention and prevention strategies. It suggests that targeting both low RHR and high sensation-seeking tendencies may be particularly effective in reducing the risk of ASB. In practical terms, this implies that individuals with both low RHR and high sensation-seeking tendencies could benefit from interventions that offer alternative, positive outlets for their desire for stimulation. For example, engaging in sports or other activities that provide an adrenaline rush could serve as constructive alternatives to more antisocial behaviors. Earlier work has suggested that low RHR is not only associated with ASB (Latvala et al., 2015) but also with extreme sports (Regoli & Hewitt, 2008) and risky jobs (Raine, 2013), which may indicate that low RHR can be associated with both prosocial and antisocial behaviors. Thus, individuals with a physiological risk profile may benefit from engaging in more prosocial activities.

Our results revealed no main effects of RHR on ASB in the W3 subsample, whereas RHR did have main effects on AS in the W1 subsample. These findings have been documented in prior studies (e.g., Bertoldi et al., 2022) and are not groundbreaking in themselves. However, due to variations in inclusion and exclusion criteria, our study employed a somewhat distinct sample than earlier studies reporting this association. Alongside the necessity to include main effects in our analyses, we deemed it justified to incorporate these results in the present study. A sensitivity analysis in which we restricted our W1 subsample to remove individuals with missing data on RHR at W3 showed that RHR still had a main effect on ASB. Thus, RHR may be a more salient risk factor for ASB when measured in childhood as opposed to early adolescence. RHR is primarily indexing parasympathetic nervous system activity, which has been found to have a different maturational pattern compared to the sympathetic nervous system (Harteveld et al., 2021). The hormonal changes during adolescence and puberty may give rise to these differences in maturational patterns, which may also give rise to complex changes in RHR. Thus, our results for the association between RHR and ASB at W3 may be due to biological factors.

The results should be considered in the light of some limitations. While the current study sought to address the methodological limitations of a previous study (Wilson & Scarpa, 2014), specifically a small sample, we still may suffer from insufficient statistical power even though our sample should be considered fairly large for a longitudinal, psychophysiological study. The uncertainty surrounding the indirect effects from the mediation analyses as well as the interaction effects from the moderation analyses are evident from the wide confidence intervals which makes it important to interpret findings with caution, along with the relatively small effects in the current study in general. Future studies should aim to replicate these results in other, larger samples. It should also be noted that the discrepancy in significance levels may be influenced by the variance of RHR. Variability in RHR across the sample may impact the precision of estimates and the strengths of associations observed.

It also remains important to recognize that this study relies on self-report measures of both sensation-seeking and ASB. It is therefore crucial for future research to utilize diverse measurement methods, such as behavioral observations or physiological assessments but also data from multiple informants, to complement self-report measures. Future research should also consider implementing other statistical techniques such as structural equation modeling to control for shared method variance when analyzing data.

In the current study, we found that the association between RHR and ASB in the W3 subsample was non-significant. Despite this, there are compelling reasons to delve into mediation and moderation analyses involving RHR, sensation-seeking, and ASB for theoretical insights. According to sensation-seeking theory, individuals with a low RHR actively seek out varied, novel, and intense sensations and experiences, which have been associated with an increased risk of ASB. This framework suggests that sensation-seeking may serve as a mediator in the relationship between RHR and ASB, offering a plausible explanation for their interconnection. Similarly, other theoretical models propose that the biological risk factor of RHR and the psychosocial risk factor of sensation-seeking interact to heighten the risk of ASB. The diathesis-stress model extends this idea, suggesting that negative psychosocial factors can amplify a biological risk, or “diathesis”, for a particular outcome. These theories provide a strong rationale for exploring both mediation and moderation effects between RHR, sensation-seeking, and ASB despite the lack of significant association between RHR and the outcome variables in the W3 subsample. Furthermore, even if the direct effect of RHR on ASB is not statistically significant, sensation-seeking may still play a role in this relationship, implying partial mediation. This scenario aligns with our findings in the current study and underscores the importance of considering nuanced pathways in understanding the complex associations between physiological and psychological factors and ASB.

Additionally, as with all observational studies there may be residual confounding not captured by the present study. Future studies investigating the role of sensation-seeking in the association between RHR and ASB should therefore aim to include more confounders, such as physical activity. It is important to acknowledge that the two subsamples might exhibit variations in participant characteristics compared to the full sample. Due to the non-participation of some individuals in specific waves of data collection, direct comparisons of key variables are unfeasible. Interpretation of results should therefore be done cautiously. The two subsamples have been analyzed in previous studies (see for example Bertoldi et al., 2023) and as previously discussed, the differences in results may be due to developmental changes. However, it is important to recognize that results need to be interpreted with caution because changes between W1 and W3 may not necessarily reflect true developmental changes but could be influenced by various factors such as measurement error or contextual influences. Future studies should aim to investigate this further. Lastly, we included two variables from the SR-DI to index ASB; getting in trouble with the police and getting arrested, together with two subscales of the SR-DI measuring violent and non-violent behavior (Bertoldi et al., 2023). Although this complements earlier studies utilizing measures of aggression, psychopathy, violent and nonviolent delinquency to study the mediating effect of sensation-seeking (Portnoy et al., 2014; Sijtsema et al., 2010) and aggression to study the moderating effect of sensation-seeking (Wilson & Scarpa, 2014), future research should aim to test sensation-seeking as both a mediator and a moderator in the association between RHR and other measures of ASB.

In conclusion, our study expands on previous research by demonstrating that sensation-seeking tendencies during young adulthood play a mediating as well as a moderating role in the association between RHR in adolescence and ASB in young adulthood. Specifically, we found that among individuals with high levels of sensation-seeking, a low RHR was associated with higher odds of getting in trouble with the police. These findings underscore the importance of considering multiple risk factors when predicting ASB and highlight a potential need for tailored interventions targeting individuals with a psychophysiological risk profile. Encouraging prosocial behaviors as a means of fulfilling the need for stimulation in individuals with a psychophysiological risk profile could serve as an effective approach in redirecting their behaviors towards positive outcomes.