Abstract

Evidence that hormones and autonomic nervous system activity exert a joint influence on antisocial behavior suggests that they may work together to explain risk for criminal behavior (Bauer et al., 2002). Thus, the current study explored the interactive effects of hormones (testosterone, cortisol) and ANS activity (heart rate, skin conductance) on impulsive and violent criminal behavior in a large sample of University students (n = 495). Initial analyses found positive direct associations between impulsive and violent crime and both testosterone and cortisol, and a negative direct association between heart rate reactivity and impulsive and violent crime. Subsequent analyses revealed a statistically significant interaction between testosterone and heart rate in association with impulsive and violent criminal behavior. The positive association between testosterone and crime became statistically significant at -.29 standard deviations below the mean (HR = 76.44) and grew stronger as heart rate decreased from this point. The interaction between testosterone and heart rate remained statistically significant after the inclusion of controls for constructs that feature prominently in contemporary criminological theories. Other interactions between hormones and ANS activity were present but were not consistent across alternative model specifications.

Introduction

While research indicates that both hormones and autonomic nervous system (ANS) activity play a role in the etiology of crime (Ling et al., 2019; Moffitt et al., 2011), studies have yet to consider their joint influence on risk for criminal behavior. The potential moderation of associations between hormones such as testosterone and cortisol and risk for criminal behavior by ANS activity is directly indicated by evidence for interactions between ANS activity and cortisol in association with youth behavioral problems (Jones et al., 2020). In parallel, reciprocal associations between testosterone and stress system, of which the ANS is a major component, suggest that ANS activity may moderate the influence of testosterone on risk for criminal behavior (Handa and Weiser, 2014). Thus, our understanding of biological contributions to risk for criminal behavior may be further specified by a consideration of interactions between hormones and indicators of ANS activity including heart rate and skin conductance. The implications of such an approach for models of the etiology of criminal behavior would be maximized by attention to the known individual and socio-environmental correlates of crime.

Background

Hormones and Crime

A large body of research suggests that the hormones testosterone and cortisol are associated with antisocial behavior including aggression (Fairchild et al., 2018; Geniole et al., 2020; van Goozen et al., 2007). This work is complemented by a smaller group of studies showing that these associations extend to crime (Booth and Osgood, 1993; Dabbs et al., 1987; Johnson et al., 2015). Testosterone, a hormone product of the Hypothalamic Pituitary Gonadal (HPG) axis, is widely known for its role in reproductive physiology and behavior (Wingfield et al., 1990) and is thought to lead to increased antisocial and aggressive behaviors through associations with dominance and status seeking (Mazur and Booth, 1998). Cortisol is a steroid hormone released as part of the Hypothalamic Pituitary Adrenal (HPA) axis stress response (Chrousos and Gold, 1992). As part of the response to physiological or psychological stressors, cortisol is associated with an array of effects including the mobilization of energy, immune suppression, and cardiovascular changes (Sapolsky et al., 2000). The cumulative work on the relationship between stress system activity and antisocial, aggressive, and criminal behavior points to a bifurcated pattern of association with an increased risk through negative affect (e.g., anger and depression) when stress system activity is upregulated and cortisol levels are high (Kemeny and Shestyuk, 2008; Lykken, 1995; Patrick et al., 1993) and an increased risk through callousness and lack of affect when stress system activity is inhibited and cortisol levels are low (Lykken, 1995; Patrick et al., 1993).

ANS and Crime

A robust body of evidence finds that reduced ANS activity is associated with antisocial behavior including crime (De Looff et al., 2022; Lorber, 2004; Portnoy and Farrington, 2015). Studies relating ANS activity to antisocial behavior commonly measure ANS activity with heart rate and skin conductance. Skin conductance is influenced by the sympathetic nervous system (SNS) branch of the ANS, while heart rate is influence by both the SNS and the parasympathetic branch of the ANS. SNS activity is commonly characterized as potentiating the ‘fight or flight’ response to stress including cardiovascular changes, while parasympathetic nervous system (PNS) activity is largely associated with changes in support of a return to homeostasis after stress, sometimes referred to as the ‘rest and digest’ state.

The association between low resting heart rate and measures of criminal behavior including convictions for violence, convictions for serious offenses and officially recorded delinquency has been established by prospective longitudinal studies (Raine et al., 1990, 1995; Latvala et al., 2015; Murray et al., 2016). A lesser but still substantive research base points to an association between resting skin conductance and both antisocial and aggressive behavior (De Looff et al., 2022; Lorber, 2004). Work in this area also shows that the negative association between resting skin conductance and antisocial behavior extends to crime (Raine et al., 1990).

In addition to work demonstrating an association between ANS activity at rest (baseline ANS) and antisocial behavior, research has also assessed the association between ANS reactivity to stress and antisocial behavior. Early meta-analyses of work in this area identified negative associations between heart rate reactivity and antisocial behavior (Ortiz & Raine, 2004) and between skin conductance reactivity and psychopathy/sociopathy but not aggression (Lorber, 2004). However, a recent meta-analysis found heart rate reactivity and skin conductance reactivity were not associated with conduct disorder or physical aggression (de Loof, et al., 2021). In work specific to criminal behavior de Vries-Bouw (2011) found a negative association between heart rate reactivity and re-offending in juvenile males.

Reduced ANS activity may be associated with increased antisocial behavior through sensation seeking enacted to increase physiological arousal or through a lack of fear response to the aversive aspects of antisocial behavior (Lykken, 1957; Quay, 1965; Raine 2002). Studies considering the role of traits in the association between ANS activity and antisocial behavior have found evidence in support of the hypothesis that low heart rate is related to increased aggression through sensation seeking (Hammerton et al., 2018; Portnoy et al., 2014), but no support for the suggestion that fearlessness mediates the association between ANS activity and antisocial behavior (Portnoy et al., 2014).

Interactions between Biological Substrates

While work examining the effects of either hormones or ANS activity on criminal behavior is important, the accurate specification of the association between each of these biological substrates and criminal behavior depends on a consideration of the other. The potential utility of considering both hormones and ANS activity when attempting to parse the role of biology in criminal behavior was delineated in Bauer et al.’s (2002) seminal work.

Bauer et al. (2002) reviewed research showing that the SNS and HPA axis work together to enact physiological and behavioral responses to stress and the restoration of homeostasis. The authors then argued that a simultaneous consideration of both systems would “allow a more thorough understanding of the physiological corelates of behavior problems” (Bauer et al., 2002 p. 104). This simultaneous consideration of both SNS and HPA axis activity was thought to be important due to interconnectivity between the SNS and the HPA axis (Chrousos and Gold, 1992) and due to the interactive associations that may stem from this interconnectivity (Bauer et al., 2002). The two systems may have additive interactive effects when hyper-arousal across systems leads to risk for behavior problems through negative affect, or when coordinated hypo-arousal leads to problem behavior through under-arousal and a lack of fear (Kagan et al., 1988). Alternatively, negative interactions may occur as a result of dysregulation and asymmetrical activity in the SNS and HPA axis (Bauer et al., 2002). While Bauer’s arguments regarding the virtue of considering multiple biological substrates in the explanation of problem behavior have garnered attention with regard to the explanation of youth behavior problems they have yet to be applied to the explanation of criminal behavior. The potential utility of a consideration of such interactions for models of the role of biology in risk for criminal behavior is indicated by a growing body of work finding additive and interactive associations between SNS activity and cortisol in the explanation of youth behavior problems (Jones et al., 2020). For example, Chen et al. (2015) found increased externalizing and internalizing behaviors among boys with reduced basal cortisol and reduced basal SNS activity indexed with salivary alpha amylase (sAA).

While Bauer et al.’s (2002) arguments are specific to the joint influence of the HPA axis and SNS, there is also reason to anticipate that ANS activity may moderate the association between testosterone and problem behaviors including crime. Androgens such as testosterone influence HPA axis function at multiple levels including the adrenal gland, pituitary gland, and the hypothalamus (Handa and Weiser, 2014). In turn, activity in these areas is associated with indicators of ANS activity including heart rate variability (Straus 2003). Thus, moderation of the association between testosterone and crime by ANS activity may occur as a result of reciprocal interconnections between HPG axis and stress system activity (Chichinadze and Chichinadze, 2008; Kutlikova et al., 2020). It is also possible that interactions occur as the result of the cumulative or sequential action of testosterone and ANS activity on reward seeking. Research has shown that sensation seeking mediates the association between heart rate and aggression (Hammerton et al., 2018; Portnoy et al., 2014; Sijtsema et al., 2010). Concurrently, studies indicate testosterone is associated with increased preference for risk as indicated by sensation seeking, impulsivity, and direct measures of risk preference (Kurath and Mata, 2018). Should associations between testosterone and risk preference be compounded by increased sensation seeking at lower levels heart rate, we may anticipate that any positive relationship between testosterone and criminal behavior would be strengthened when heart rate is low.

The Current Study

The current study explores interactions between hormone products of the endocrine system (i.e., testosterone and cortisol) and indicators of ANS activity (i.e., skin conductance and heart rate). Initial analyses focus on associations between criminal behavior and the direct and interactive effects of hormones and ANS measures. Subsequent models contextualize these associations within causal mechanisms central to contemporary criminological theory by including controls for individual differences in risk for criminal behavior and measures of the socio-environmental correlates of crime including parents, peers, and romantic partners.

While the current literature suggests that hormones and ANS activity exert a joint influence on criminal behavior, the analyses presented here are nonetheless exploratory. The results of studies on the role of neuroendocrine coordination in youth behavior problems provide evidence for both additive and interactive associations between cortisol and SNS activity as measured by sAA (Jones et al., 2020). Collectively, the studies reviewed by Jones et al. (2020) suggest that work attempting to specify the role of biological variation in risk for criminal behavior should consider both measures of ANS activity and cortisol. However, this work stops short of providing a research base that would lead to specific expectations regarding the nature of the joint influence of ANS measures and hormones on variation in crime. Similarly, the positive associations between testosterone and risk preference (Kurath and Mata, 2018) may be compounded by increased sensation seeking at lower levels of heart rate (Hammerton et al., 2018; Portnoy et al., 2014; Sijtsema et al., 2010) thus contributing to a stronger association between testosterone and criminal behavior when heart rate is low. However, direct evidence for increased risk preference with testosterone at lower heart is lacking. Therefore, the current analyses are framed as an initial exploration of the potential interactions between hormones and ANS activity in relation to risk for criminal behavior.

Methods

Data were gathered in 2016 as part of a larger project on the role of biology in the etiology of antisocial, aggressive, and criminal behavior conducted at a university in the southern United States. The current study builds directly on earlier work with these data (*citation redacted*). Analyses presented in this earlier work explored the joint effects of testosterone and cortisol on a measure of impulsive and violent criminal behavior and a measure of income generating crime. Results indicated that testosterone had a positive direct association with impulsive and violent crime. Hormones were not directly related to income generating crime but the association between testosterone and income generating crime was positive and statistically significant when cortisol was below -0.70 SD under the mean, and negative and statistically significant when cortisol is above 1.669 over the mean. We frame our initial exploratory analysis of the potential role of ANS activity in associations between hormones and crime around the direct association between testosterone and impulsive and violent crime while also considering potential interactions between cortisol and ANS activity as indicated by work reviewed earlier (Bauer et al., 2002; Jones et al., 2020). The analysis is framed in this manner to balance the complexity inherent in interaction models with disciplinary standards (and journal limitations) regarding manuscript length. To our knowledge, the current study is the first to consider interactions between hormones and ANS in association with risk for criminal behavior. Analyses begin by considering the effects of hormones and ANS activity on criminal behavior net of demographic controls after which models incorporate control measures for the socio-environmental correlates of crime and criminal propensity.

Sample

The study protocol was approved by the IRB of the institution where the study was administered. Participants were recruited by convenience from large Criminology and Criminal Justice classes. After obtaining informed consent, participants completed an in-class survey including measures of the socio-environmental correlates of crime, trait measures, and criminal behavior measures (n = 872). Following survey completion, participants were referred to a laboratory measurement protocol where ANS and hormone measures would be gathered. Subjects were asked to schedule a time for the laboratory protocol through signupgenius.com, an online scheduling tool. A total of 567 students attended the laboratory protocol. The analytic sample is comprised of 495 of these participants that provided complete measures of hormones and ANS, demographic information, and self-report criminal behavior. The self-identified race/ethnicity of these participants was 12.7% African American, 1.0% Asian, 37.2% Caucasian, 40.0% Hispanic, and 9.1% other. Participants in the analysis sample were 66.5% female and averaged 20.35 years of age (SD = 3.16).

Hormone and ANS Quantification

Analyses included measures of testosterone and cortisol taken at rest and measures of heart rate and skin conductance taken at rest and in response to a social stressor. Measures were gathered between the hours of 0800 and 1830. Participants were instructed to refrain from a variety of activities that may have affected testosterone and cortisol levels (e.g., smoking, eating, exercise) for at least one hour prior to reporting to the lab. Upon arrival at the lab participants were seated comfortably. Next, compliance with activity restrictions was confirmed and informed consent was verbally reaffirmed. Saliva samples for baseline hormone assays were gathered after a 30 second rest period with Salimetrics LLC Saliva Collection Aids. Each sample contained at least 1.5 mL of saliva. Prior to analysis, samples were stored in a freezer at -20 degrees Celsius. Saliva samples were analyzed using materials from and following established protocols for Salimetrics testosterone and cortisol enzyme immunoassay kits. All samples were tested in duplicate. Testosterone concentrations were recorded in pg/mL and cortisol concentrations in µg/dL. The mean intra-assay coefficient of variation for testosterone and cortisol were 5.98% and 11.01% respectively and the mean inter-assay coefficients of variation for testosterone and cortisol were 7.95% and 5.91% respectively.

After gathering the sample for hormone assays, sensors for the measurement of ANS activity were attached to the fingertips of the participant’s right hand. Heart rate was measured using plethysmograph based NeuLog infrared LED transmitter/receptor sensor. Skin conductance level was measured using NeuLog galvanic skin response sensors. Values for both were recorded on a single laptop using the NeuLog data acquisition and analysis application. Sampling rates were set at one measurement every two seconds. After sensors were attached, participants rested for three minutes and then were told that that they had 2 minutes to prepare a 2-minute speech on their biggest faults and weaknesses, and that the speech would be recorded and later evaluated. ANS recording continued for 30 seconds after the conclusion of the speech. ANS measures were estimated for four time periods during the measurement protocol. These include a baseline measure across the last 30 seconds of the rest period before the announcement of the task, a second measure across the last 30 seconds of speech preparation, a third measure comprised of the full 2 minutes of speech delivery, and a follow-up measure across the 30 seconds after the conclusion of the speech. For each of the measurement periods, ANS measures were estimated as the averaged sum of measures gathered every 2 seconds during the measurement period. ANS reactivity to stress was estimated as area under the curve with respect to increase (AUC_i) during speech preparation, delivery, and a short follow-up period (Pruessner et al., 2003). AUC_i represents the magnitude of change in the phenomenon of interest net of the starting value during the period under consideration. AUC_i has been used as a summary measure of change in ANS activity in prior work attempting to understand the joint contribution of hormones and ANS activity to human behavior (Rimmele et al., 2007; Romero-Martínez et al., 2013). We quantify change in ANS activity with AUC_i as it offers a single measure of change in ANS activity that incorporates all of the available ANS activity measurements. Descriptive statistics for study measures are presented Table 1 and correlations among ANS measures are presented in Supplementary Materials.

Survey Measures

The self-report survey included items measuring criminal behavior, measures of criminal propensity including self-control and psychopathy, and measures of the socio-environmental correlates of crime including delinquent peers, parental affection, and the presence of a romantic partner. A ten-item measure of impulsive and violent crime was derived from iterative principal components factor analyses of a 35-item self-report inventory of criminal behavior in the past year. Items in the self-report inventory were factor analyzed in light of evidence indicating that associations between the biological substrates under consideration here (i.e., ANS activity and hormones) vary across antisocial and aggressive behavior type (Armstrong et al., 2021; Popma et al., 2007; Ruttle et al., 2011; van Bokhoven et al., 2005). During the factor analysis, items were discarded unless the largest loading for the item was on the largest factor. This process was repeated until the largest loading for each of the items was on the only factor with an eigenvalue over one. In the first and second factor models, there were twelve and five factors with eigenvalues over one, respectively. On the third iteration, a set of ten items indicated a single factor with an eigenvalue over one. These items included indicators of vandalism, arson, assault, sexual assault, and petty theft and appeared to be reasonably well represented by the label impulsive and violent crime (Cronbach’s α = .71). Items were combined in a variety score, which increased by one for each different type of impulsive and violent crime that a participant indicated that they had engaged in during the past year. Variety scores are preferable when scales included both serious and less serious forms of criminal behavior, as they are not overly influenced by the increased frequency of less serious forms of crime (Sweeten 2012). After visual inspection of the distribution of impulsive and violent crime scores, a single non-continuous outlier was re-scored as the next highest continuous value.

Self-control was measured with the 24-item Grasmick et al. (1993) self-control scale (GSCS; Cronbach’s alpha = .84). Psychopathy was measured with the 26-item Levenson Self-Report Psychopathy Scale (LSRPS; Levenson et al., 1995; Cronbach’s alpha = .85). While there is evidence indicating the multidimensionality of the GSCS and LSRPS (Brinkley et al., 2008; Conner et al., 2009) here they are treated as indicators of general criminal propensity and analyses are based on aggregate scale scores.

Peer delinquency was assessed with 14 items asking respondents to specify the frequency with which their peers engaged in a variety of criminal activities ranging from substance use to property crime to violent crime (Cronbach’s alpha = .84). Responses were recorded on a 5-item Likert scale ranging from “none of them” to “all of them.” Quantified in this way each item in the peer delinquency scale indicates the proportion of friends that have any engagement in a particular delinquent behavior and the sum of these items indicates the average proportion of friends that engage in delinquent behaviors. Scores for the peer delinquency measure and the remaining scales were estimated as the averaged sum of item scores. Parental affection was measured with 4 items inquiring as to whether or not the respondent’s primary caregivers (e.g., mother/step-mother/foster-mother) loved and cared for the respondent before the age of 8 and from the age of 8 through age 16 (Cronbach’s alpha = .77). Responses were recorded on a four-point Likert scale ranging from “Not at all” to “Very much”. A dichotomous indicator of being in a romantic relationship was created from responses to a single item asking respondents to indicate their relationship status (married or not married but in a relationship = 1; divorced, separated, not in a relationship = 0).

Analyses also included controls for age in years, race and ethnicity as indicated with dummy variables for the two largest groups (Caucasian, 0 = no, 1 = yes; Hispanic, 0 = no, 1 = yes), and sex (0 = male, 1 = female).

Analytical Strategy

Prior to the estimation of inferential statistics, the distributions of hormone and ANS measures were considered. Outliers were rescored to 3 SD from the mean (Wilcox, 2010) and cortisol concentrations were log 10 transformed after the addition of a constant to address skewness. When measures exhibited sex differences, outliers were identified separately within sex. Details regarding this procedure are available in the Supplementary Materials.

Associations between hormones, ANS and impulsive and violent criminal behavior were tested with negative binomial regression models estimated with StataMP 15 (StataCorp, 2017). Negative binomial regression models were preferable to Poisson models due to overdispersion and are specifically designed to accommodate the non-normal distribution present in count variables (Hilbe, 2011). Zero inflated models were considered but the proposed associations failed to meet the theoretical assumptions of this model (e.g., there is a different causal process for zeros; Allison, 2012). To facilitate comparison of regression coefficients and the estimation of interaction terms, hormone and ANS measures were standardized (z-scores). When hormone or ANS measures exhibited sex differences, standardization was done within sex. In the first regression model the standardized hormone and ANS measures were regressed on impulsive and violent crime. Next, potential interactions between hormones and the ANS measures were tested by adding multiplicative interaction terms. Interaction terms modeled the interactive effects of each hormone measure with each type of ANS activity at both baseline and change. Finally, measures of criminal propensity (self-control and psychopathy) and controls for the effects of parents, peers, and romantic partners were added to the regression model. Potential multicollinearity was investigated with the estat vif function after rerunning each model with Ordinary Least Squares regression. Results indicated multicollinearity was not an issue (all vifs < 2.1). All regression models included control variables representing the two largest race/ethnic groups in the sample (1 = Caucasian, 0 = other; 1 = Hispanic, 0 = other), sex (0 = male, 1 = female), age in years, and time of data collection represented with a whole number for hours on the 24-hour clock and fraction of minutes within an hour to two decimal places. Statistically significant interactions were visualized with Johnson-Neyman region of significance plots (Bauer and Curran, 2005) generated with the tidyverse package (Wickham et al., 2019) in RStudio 4.2.1 (R Core Team, 2020).

Prior to the specification of the analytic sample, patterns of missing data were assessed among all those participating in the laboratory protocol. The frequency of missing data across hormone measures, ANS measures, control measures and criminal behavior ranged from 1.1% (crime measure) to 4.5% (heart rate responsivity). The frequency of missing data was similar for most of the theoretically motivated control measures ranging from 1.3% (romantic relationship) to 4.2% (self-control), with the exception of the psychopathy measure (11.2% missing data). Patterns of missing data were assessed with Little’s test of missing completely at random implemented with the mcartest command (Li, 2013; Little and Rubin, 2002). Results indicated data were not significantly different from missing completely at random for all analyses variables (χ²(541) = 547.27, p =.42).

In the initial review of results, all coefficients at the traditional p < .05 threshold were considered substantive. Following this, the implications of a Bonferroni correction for multiple comparisons are discussed (Benjamini and Hochberg, 1995). This approach acknowledges both the importance of a consideration multiple comparisons and concern that corrections for multiple testing may stifle exploratory research when researchers deliberately limit the range of variables in data collection or analysis in order to increase the likelihood of meeting corrected thresholds for statistical significance (Nakagawa, 2004). We also present coefficients and confidence intervals when reporting results for regression models. Nakagawa (2004) argues this approach is preferable to correction for multiple testing.

Results

Direct Effects of Hormone and ANS Measures

Both testosterone and cortisol were positively associated with impulsive and violent crime (Table 2, Model 1; b = 0.17, SE = 0.09, p = 0.046, 95% CI [0.00, 0.34]; b = 0.21, SE = 0.09, p = 0.029, 95% CI [0.02, 0.40] respectively). Heart rate reactivity had a negative and statistically significant association with impulsive and violent crime (b = -0.20, SE = 0.09, p = 0.034, 95% CI [-0.38, -0.02]), but the other ANS measures were not. Time of data collection was positively associated with the impulsive and violent crime measure (b = 0.08, SE = 0.03, p = 0.022, 95% CI [0.01, 0.15]), while being female was associated with decreases in impulsive and violent crime (b = -0.72, SE = 0.17, p = 0.000, 95% CI [-1.06, -0.38]).

Hormone and ANS Measure Interaction Effects

The interaction between testosterone and baseline heart rate had a negative and statistically significant association with impulsive and violent crime (Table 2, Model 2; b = -0.29, SE = 0.10, p = 0.003, 95% CI [-0.48, -0.10]), as did the interaction between testosterone and baseline skin conductance (b = -0.20, SE = 0.09, p = 0.039, 95% CI [-0.38, -0.01]). Interactions between testosterone and ANS responsivity measures were not significantly associated with impulsive and violent crime. None of the interactions between cortisol and ANS measures were associated with the impulsive and violent crime measure. Region of significance plots show that the association between testosterone and impulsive and violent crime is positive and statistically significant as baseline heart rate decreases from -0.29 standard deviations below the mean (heart rate = 76.44) and negative and statistically significant as baseline heart rate is over 1.54 standard deviations above the mean (heart rate = 99.93; Figure 1 Panel A). In parallel the positive association between testosterone and impulsive and violent crime becomes significant when baseline skin conductance decreases from -.42 standard deviations under the mean (skin conductance = 2.76; Figure 1 Panel B).

Figure 1. Region of significance plots for interactions between hormones and ANS measures.

Panel A Conditional effect of testosterone on crime by heart rate

Panel B Conditional effect of testosterone on crime by skin conductance

Hormone and ANS Measure Interaction Effects with Trait and Socio-Environmental Controls

After the inclusion of measures quantifying traits and socio-environmental controls the interaction of testosterone with baseline heart rate continued to have a negative and statistically significant association with impulsive and violent crime (b = -0.22, SE = 0.10, p = 0.037, 95% CI [-0.42, -0.01]). The association between the interaction of testosterone with baseline skin conductance reactivity and impulsive and violent crime was no longer statistically significant, the interaction of cortisol with skin conductance reactivity emerged as statistically significant (b =0.22, SE = 0.10, p = 0.030, 95% CI [0.02, 0.41]). Full model results and region of significance plots of significant interactions are available as Supplemental Materials.

A Consideration of Multiple Comparisons and the Transformation of Cortisol Scores

A Bonferroni correction based on the number of direct and interactive associations between the measure of impulsive and violent crime and both hormone and ANS measures in each model resulted in a corrected p value of .00357 (.05/14). At this threshold, the interaction between testosterone and heart rate prior to the inclusion of theoretically motivated control variables remains statistically significant. All other associations fail to meet this threshold. While this highlights differences in the statistical significance of associations, we would suggest that given the novel nature of the analyses presented here all associations are strong candidates for replication. Variability in patterns of interaction between SNS and cortisol measures shown in research on childhood behavior problems further highlight the need for replication and necessary caution when making theoretical inference based on interactions between hormone and ANS measures (see Jones et al., 2020).

The Log 10 transformation of the cortisol measure undertaken prior to the estimation of inferential statistics is consistent with prominent studies on the role of hormones in the explanation of antisocial behaviors (e.g., Platje et al., 2015; Tackett et al., 2014). Nonetheless, we replicated all regression models with an untransformed baseline cortisol measure. In the direct effects model, the direction and statistical significance of all associations were the same. In the model with interactions and demographic controls, the interaction between testosterone and skin conductance was slightly attenuated (b =-0.17, SE = 0.09, p = 0.061, 95% CI [-0.36, -0.01]). In the model with interactions and both demographic and theoretical control measures, the direct effect of HRR emerged as statistically significant b = -.20, SE = .10, p = .048 (b =-0.20, SE = 0.10, p = 0.048, 95% CI [-0.39, <-0.01]).

Discussion

While strong conclusions await replication, the current results indicated that hormones and ANS activity interact to influence risk for impulsive and violent criminal behavior. Specifically, the positive association between testosterone and impulsive and violent crime became statistically significant as basal ANS activity decreased below its mean. This association was particularly robust for the interaction of testosterone with baseline heart rate (Table 2; Supplemental Table S2). In the broader literature informing associations between hormones, ANS activity, and antisocial behavior there is some suggestion that the interaction of testosterone with heart rate may be mediated by sensation seeking resulting from increased reward salience. An association between testosterone and crime through sensation seeking is indicated by studies showing that testosterone influences activity in the brain’s mesolimbic reward system (Laube et al., 2020; Op de Macks et al., 2011) and research showing a positive association between testosterone and risk preference (Kurath and Mata, 2018). In addition, a small but growing group of studies show that the direct association between resting heart rate and crime is mediated by sensation seeking (Hammerton et al., 2018; Portnoy et al., 2014; Sijtsema et al., 2010). Thus, a positive association between testosterone and risk for impulsive and violent criminal behavior through increased reward salience may be amplified by increased sensation seeking at lower levels of heart rate. This potential explanation of the association between the interaction of testosterone with impulsive and violent criminal behavior is of course highly speculative. The search for traits that moderate associations between hormones and human behavior often results in weak associations and null findings (Sundin et al., 2021), though perhaps the consideration of multiple biological substrates may help to clarify relationships between biological substrates and traits.

Results also indicated that the positive association between cortisol and impulsive and violent crime was stronger at higher levels of skin conductance reactivity after the inclusion of controls for the trait and socio-environmental correlates of crime (Supplemental Table S2, Supplemental Figure S1 Panel B). This result is consistent with the increased risk for problem behavior as a function of additive interactions between ANS and HPA axis activity as hypothesized by Bauer et al. (2002). At the trait level, elevated stress system activity indicated by the interaction of cortisol with skin conductance reactivity may result in increased risk for criminal behavior through emotional arousal including anger (Lobbestael et al., 2008; Moons et al, 2010). The weight given to this conjecture should be contextualized by the results of prior work exploring associations between youth problem behaviors and the interaction of cortisol with sAA, an alternative measure of SNS activity. A recent review of this research showed results varied with evidence of both positive and negative interactions between cortisol and SNS activity in association with youth problem behavior (Jones et al., 2020). This variability suggests inference regarding the interactive associations between HPA-axis, SNS activity and risk for criminal behavior should be regarded as tentative.

Heart rate reactivity had a negative direct association with impulsive and violent crime that survived the inclusion of interaction terms. The association between heart rate reactivity and impulsive and violent crime is to a certain extent anticipated by Ortiz and Raine (2004) who found the association between heart rate during a stressor and antisocial behavior (d = -0.76) was much larger in magnitude than the effect size for the association between heart rate at rest and antisocial behavior (d = -0.44). In the context of the current results, lower heart rate reactivity to stress may result in a tolerance for the aversive aspects of impulsive and violent crime. However, the negative direct association between heart rate reactivity and impulsive and violent crime did not extend to skin conductance reactivity, and both baseline measures of ANS activity did not have direct associations with the impulsive and violent crime measure. The lack of association between baseline heart rate and impulsive violent crime is particularly surprising. Meta analyses have confirmed an association between resting heart rate and general antisocial behavior antisocial behavior (Portnoy & Farrington, 2015, deLoof et al., 2021). It is possible that the lack of association is specific to the content of the impulsive and violent crime measure. Estimates of the association between measures of baseline heart rate and antisocial behavior show considerable heterogeneity across findings (deLoof et al., 2021).

Broadly speaking, the results herein suggest that biological substrates interact to influence risk for criminal behavior though the mechanism of this influence remains at issue. Booth and Osgood (1993) found the association between testosterone and a measure of adult deviance largely defined by criminal behavior was mediated by the inclusion of measures of social integration and prior criminal behavior. It is possible that the interaction of testosterone and heart rate is associated with criminal behavior in a similar fashion. This interaction may contribute to a behavioral disposition whose consequences accrete over the life-course leading to increased risk for criminal behavior through state dependent effects and by eroding social integration. A meaningful test of this proposed set of associations would require longitudinal measures of hormones and social integration. Similarly, the data used here lack the structure and measures necessary to capture inter-relationships between aspects of biology and the proximal processes thought to directly influence patterns of development and behavior over the life-course. Nonetheless these data do include measures of constructs featuring prominently in criminological theory. The measures capture major classes of the correlates of criminal behavior including individual differences like self-control and psychopathy and indicators of key aspects of the socio-environment including parental affection and peer delinquency. Associations between criminal behavior and aspects of biology like hormones and autonomic nervous system activity in models that include controls for these measures indicate that comprehensive theoretical accounts of risk for crime need to address individual differences including aspects of biology in addition to constructs featuring prominently in traditional criminological theories like parental affection and peer delinquency. These results also indicate that indicators of biological risk for criminal behavior do not operate in silos but instead may interact to influence behavioral dispositions increasing risk for specific types of criminal behavior. The practical implications of the current work and for the broader literature on biological risk for criminal behavior remain remote as they require the development of a much more systematic knowledge base grounded in methodologically strong studies. However, it is possible that a better understanding of the role of biology in the explanation of risk for criminal behavior may facilitate better use of treatment and prevention resources through the matching of risk for specific types of criminal behavior to treatment and prevention modalities that are tailored towards ameliorating a given type of risk.

The implications of the current work should be contextualized by the characteristics of the sample under consideration and by other aspects of the research methodology. These concerns may be particularly salient with regard to the positive association between cortisol and impulsive and violent crime. As noted above, cortisol shows a bifurcated pattern of association with antisocial behavior including crime with negative associations often found in samples with high risk for criminal behavior and in studies employing case-control methodologies (c.f., van Bokhoven et al., 2005; van Goozen et al., 2007). Thus, the positive associations between cortisol and crime found in this study both directly and in interaction with SCR may not generalize to samples with increased risk for criminal behavior, to alternative methods of analysis, or to general measures of criminal behavior.

The weight given results should also be conditioned by the statistical power of the analysis and by the retrospective prediction of past year criminality by biological measures. To contextualize the statistical power of the current study, we conducted a series of post hoc power analysis with G*Power. The first indicated that a sample size of 759 would have been necessary to detect the very small effect sizes (d = .02) with .80 power at the p = .05 level of statistical significance (α = .05). While d = .02 is quite small, it is a reasonable benchmark for the magnitude of effect sizes found in moderation analyses. Required sample sizes for d = .15 and d = .35 were 109 and 53 respectively (.80 power at α = .05). Collectively, results from the power analyses indicate that even though our sample size is very large compared to the strong majority of work considering associations between hormones and either traits or behaviors, it falls short of the mark for the detection of very small effect sizes at conventional levels of power and statistical significance. This may reduce the stability of statistical estimates and increases the importance of replication.

Analyses in the current study focused on associations between biological measures and a retrospective measure of impulsive and violent crime. In support of this approach, we would note that studies testing the association between hormone measures and retrospective measures of antisocial behavior are not uncommon (e.g., Platje et al., 2015; Tackett et al., 2014), and that both hormone levels and antisocial behaviors including crime exhibit a fair amount of between individual stability over time (Nagin and Paternoster, 2000; Zhang et al., 2017; Zmuda et al., 1997). Heart rate and skin conductance measures taken both at rest and in response also show stability across time (Allen et al., 1987; El-Sheikh, et al., 2007; Hassellund et al., 2010). This stability appears to be stronger for heart rate than for skin conductance and there is some evidence that the stability of skin conductance reactivity is task specific (El-Sheikh, et al., 2007).

Nonetheless, stability in hormone and ANS measures is far from absolute and variation over time may introduce measurement error. A tendency to rely on retrospective behavioral measures is in part due to the costs of hormone analyses in both monetary terms and in time. While limited this approach may be an efficient use of resources that provides initial findings which can serve to frame well powered and prospective studies.

The results of the current study may also be influenced by aspects of the quantification of ANS function and hormone levels. The current study used plethysmography to measure heart rate. Plethysmography indexes arterial pulse to quantify heart rate. As an alternative to plethysmography an electrocardiogram can provide a more accurate measure of heart rate by directly quantifying cardiac activity (Berntson et al., 1997). Results may also be influenced by diurnal variation in testosterone and cortisol (Brownlee et al., 2005; Clow et al., 2010; Cumming, Quigley and Yen, 1983; Zhang et al., 2017). In order to account for diurnal variation studies often collect samples for hormone assays in either the morning or afternoon, though there are studies with wider data collection windows (e.g., Welker et al., 2014). To account for the potential influence of the time of day on associations between hormones and crime, the time that salvia samples for hormone assays were collected was included as a control variable in regression models. However, a better understanding of the impact of time of day on associations between hormones and crime may come from randomization into data collection windows. Associations between hormone measures and criminal behavior in the current work may also be influenced by the use of enzyme immunoassays rather than tandem mass spectrometry. Comparisons of hormone values gathered with these respective methods are limited but predominantly show statistically significant correlations. However, in some cases correlations between hormone values based on the respective methods lack statistical significance (Prasad et al., 2019; Welker et al., 2016).

Despite the concerns and caveats outlined above, the current work does show that risk for criminal behavior is influenced by the interaction between distinct biological substrates. Specifically, the interaction between hormones and ANS activity is associated with increased risk for impulsive and violent criminal behavior. While these results are preliminary and await replication, they demonstrate the potential utility of examining the interaction between biological substrates for attempts to specify the role of biology in the etiology of risk for criminal behavior. Future research in this direction should include the replication of the current results and the extension of the larger framework considering interactions between biological substrates in the explanation of risk for criminal behavior to other substrates such as brain structure and function. This research may also explore the extent to which the association between genetic variation and crime are conditioned by biological substrates known to influence risk for criminal behavior.

References

Allen, M. T., Sherwood, A., Obrist, P. A., Crowell, M. D., & Grange, L. A. (1987). Stability of cardiovascular reactivity to laboratory stressors: a 2 1/2 yr follow-up. Journal of Psychosomatic Research, 31(5), 639-645.

Allison, P. (2012). Do We Really Need Zero Inflated Models? Statistical Horizons. Retrieved from http:// statisticalhorizons.com/zero-inflated-models.

Armstrong, T., Boisvert, D., Wells, J., Lewis, R, Cooke, E., & Woeckner, M. (2020). Assessing Overlap between Self-Control and Psychopathy: A Consideration of the Grasmick Self-Control Scale and Levenson Self-Report Psychopathy Scale. Journal of Criminal Justice, 70, 101725. https://doi.org/10.1016/j.jcrimjus.2020.101725. Online ahead of print.

Armstrong, T., Wells, J., Boisvert, D., Lewis, R., Cooke, E., Woeckener, M., & Kavish, N. (2021). An exploratory analysis of testosterone, cortisol, and aggressive behavior type in men and women. Biological Psychology, 161, 108073, https://doi.org/10.1016/j.biopsycho.2021.108073.

Bauer, A. M., Quas, J. A., & Boyce, W. T. (2002). Associations between physiological reactivity and children’s behavior: Advantages of a multisystem approach. Journal of Developmental & Behavioral Pediatrics, 23(2), 102−113.

Bauer, D. J., & Curran, P. J. (2005). Probing interactions in fixed and multilevel regression: Inferential and graphical techniques. Multivariate Behavioral Research, 40, 373–400. 

Benjamini, Y., and Y. Hochberg. (1995). Controlling for the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society: Series B (Methodological), 57(1), 289-300.

Berntson, G. G., J. T. Bigger Jr, D. L. Eckberg, P. Grossman, P. G. Kaufmann, M. Malik, H. N. Nagaraja et al. (1997). Heart rate variability: Origins, methods, and interpretive caveats. Psychophysiology, 34(6), 623-648.

Booth, A., & Osgood, D. W. (1993). The influence of testosterone on deviance in adulthood: Assessing and explaining the relationship. Criminology, 31(1), 93−117.

Brennan, P. A., Raine, A., Schulsinger, F., Kirkegaard-Sorensen, L., Knop, J., Hutchings, B., ... & Mednick, S. A. (1997). Psychophysiological protective factors for male subjects at high risk for criminal behavior. American Journal of Psychiatry, 154(6), 853−855.

Brinkley, C. A., Diamond, P. M., Magaletta, P. R., & Heigel, C. P. (2008). Cross-validation of Levenson’s psychopathy scale in a sample of federal female inmates. Assessment, 15(4), 464−482.

Brinkley, C. A., Schmitt, W. A., Smith, S. S., & Newman, J. P. (2001). Construct validation of a self-report psychopathy scale: Does Levenson’s self-report psychopathy scale measure the same constructs as Hare’s psychopathy checklist-revised? Personality and Individual Differences, 31, 1021−1038.

Brownlee, K. K., Moore, A. W., & Hackney, A. C. (2005). Relationship between circulating cortisol and testosterone: influence of physical exercise. Journal of Sports Science & Medicine, 4(1), 76–83.

Chen, F. R., A. Raine, A. L. Glenn, and D. A. Granger. (2016). Hypothalamic pituitary adrenal activity and autonomic nervous system arousal predict developmental trajectories of children’s comorbid behavior problems. Developmental Psychobiology, 58(3), 393-405.

Chen, F. R., A. Raine, and D. A. Granger. (2015). Tactics for modeling multiple salivary analyte data in relation to behavior problems: Additive, ration, and interaction effects. Psychoneuroendocrinology, 51, 188-200.

Chichinadze, K., & Chichinadze, N. (2008). Stress-induced increase of testosterone: Contributions of social status and sympathetic reactivity. Physiology & Behavior, 94(4), 595–603.

Chrousos, G. P., & Gold, P. W. (1992). The concepts of stress and stress system disorders: Overview of physical and behavioral homeostasis. JAMA, 267(9), 1244–1252.

Clow, A., Hucklebridge, F., Stalder, T., Evans, P., & Thorn, L. (2010). The cortisol awakening response: More than a measure of HPA axis function. Neuroscience & Biobehavioral Reviews, 35(1), 97–103.

Conner, B. T., Stein, J. A., & Longshore, D. (2009). Examining self-control as a multidimensional predictor of crime and drug use in adolescents with criminal histories. The Journal of Behavioral Health Services & Research, 36, 137–149.

Cumming, D. C., Quigley, M. E., & Yen, S. S. C. (1983). Acute suppression of circulating testosterone levels by cortisol in men. The Journal of Clinical Endocrinology & Metabolism, 57(3), 671–673.

Dabbs Jr., J.M., Frady, R.L., Carr, T.S., Besch, N.F. (1987). Saliva testosterone and criminal violence in young adult prison inmates. Psychosomatic Medicine, 49(2), 174–182.

De Looff, P. C., Cornet, L. J., De Kogel, C. H., Fernández-Castilla, B., Embregts, P. J., Didden, R., & Nijman, H. L. (2022). Heart rate and skin conductance associations with physical aggression, psychopathy, antisocial personality disorder and conduct disorder: An updated meta-analysis. Neuroscience & Biobehavioral Reviews, 132, 553–582 https://doi.org/10.1016/j.neubiorev.2021.11.003.

De Vries‐Bouw, M., Popma, A., Vermeiren, R., Doreleijers, T. A., Van De Ven, P. M., & Jansen, L. M. (2011). The predictive value of low heart rate and heart rate variability during stress for reoffending in delinquent male adolescents. Psychophysiology, 48(11), 1597-1604.

El‐Sheikh, M. (2007). Children's skin conductance level and reactivity: Are these measures stable over time and across tasks?. Developmental Psychobiology, 49(2), 180-186.

Fairchild, G., Baker, E., & Eaton, S. (2018). Hypothalamic-pituitary-adrenal Axis function in children and adults with severe antisocial behavior and the impact of early adversity. Current Psychiatry Reports, 20(10), 1–9.

Farrington, D. (1997). The relationship between low resting heart rate and violence. In A. Raine, P. Brennan, D. P. Farrington, & S. Mednick (Eds.), Biosocial bases of violence (pp. 89–106). New York: Plenum.

Geniole, S. N., Bird, B. M., McVittie, J. S., Purcell, R. B., Archer, J., & Carré, J. M. (2019). Is testosterone linked to human aggression? A meta-analytic examination of the relationship between baseline, dynamic, and manipulated testosterone on human aggression. Hormones and Behavior, 123, 104644. https://doi.org/10.1016/j.yhbeh.2019.104644.

Grasmick, H. G., Tittle, C. R., Bursik Jr, R. J., & Arneklev, B. J. (1993). Testing the core empirical implications of Gottfredson and Hirschi's general theory of crime. Journal of Research in Crime and Delinquency, 30(1), 5-29.

Hammerton, G., Heron, J., Mahedy, L., Maughan, B., Hickman, M., & Murray, J. (2018). Low resting heart rate, sensation seeking and the course of antisocial behaviour across adolescence and young adulthood. Psychological Medicine, 48(13), 2194–2201.

Handa, R. J., & Weiser, M. J. (2014). Gonadal steroid hormones and the hypothalamo–pituitary–adrenal axis. Frontiers in Neuroendocrinology, 35(2), 197–220.

Hassellund, S. S., Flaa, A., Sandvik, L., Kjeldsen, S. E., & Rostrup, M. (2010). Long-term stability of cardiovascular and catecholamine responses to stress tests: an 18-year follow-up study. Hypertension, 55(1), 131-136.

Hilbe, J. M. (2011). Negative binomial regression. Cambridge, UK: Cambridge University Press.

Johnson, M. M., Mikolajewski, A., Shirtcliff, E. A., Eckel, L. A., & Taylor, J. (2015). The association between affective psychopathic traits, time incarcerated, and cortisol response to psychosocial stress. Hormones and Behavior, 72, 20–27.

Jones, E. J., Rohleder, N., & Schreier, H. M. (2020). Neuroendocrine coordination and youth behavior problems: A review of studies assessing sympathetic nervous system and hypothalamic-pituitary adrenal axis activity using salivary alpha amylase and salivary cortisol. Hormones and Behavior, 122, 104750.

Kagan, J., J. S. Reznick, and N. Snidman. (1988). Biological bases of childhood shyness. Science, 240(4849), 167-171.

Kemeny, M. E., & Shestyuk, A. (2008). Emotions, the neuroendocrine and immune systems, and health. In M. Lewis, J. M. Haviland-Jones, & L. F. Barrett (Eds.), Handbook of emotions (pp. 661–675). New York: Guilford Press.

Kutlikova, H. H., Durdiakova, J. B., Wagner, B., Vlček, M., Eisenegger, C., Lamm, C., & Riečanský, I. (2020). The effects of testosterone on the physiological response to social and somatic stressors. Psychoneuroendocrinology, 117, 104693.

Kurath, J., & Mata, R. (2018). Individual differences in risk taking and endogeneous levels of testosterone, estradiol, and cortisol: A systematic literature search and three independent meta-analyses. Neuroscience & Biobehavioral Reviews, 90, 428–446.

Latvala, A., Kuja-Halkola, R., Almqvist, C., Larsson, H., & Lichtenstein, P. (2015). A longitudinal study of resting heart rate and violent criminality in more than 700 000 men. JAMA Psychiatry, 72(10), 971–978.

Laube, C., Lorenz, R., & van den Bos, W. (2020). Pubertal testosterone correlates with adolescent impatience and dorsal striatal activity. Developmental Cognitive Neuroscience, 42, 100749.

Levenson, M. R., Kiehl, K. A., & Fitzpatrick, C. M. (1995). Assessing psychopathic attributes in a noninstitutional population. Journal of Personality and Social Psychology, 68, 151–158.

Li, C. (2013). Little's test of missing completely at random. The Stata Journal, 13(4), 795–809.

Ling, S., Umbach, R., & Raine, A. (2019). Biological explanations of criminal behavior. Psychology, Crime & Law, 25(6), 626–640.

Little, R. J., & Rubin, D. B. (2002). Statistical analysis with missing data (2^nd ed.). New York: Wiley.

Lobbestael, J., Arntz, A., & Wiers, R. W. (2008). How to push someone's buttons: A comparison of four anger-induction methods. Cognition & Emotion, 22(2), 353–373.

Lorber, M. F. (2004). Psychophysiology of aggression, psychopathy, and conduct problems: A meta-analysis. Psychological Bulletin, 130(4), 531–552. http://dx.doi.org/ 10.1037/0033-2909.130.4.531.

Lykken, D. T. (1957). A study of anxiety in the sociopathic personality. Journal of Abnormal and Social Psychology, 55, 6-10.

Lykken, D.T., (1995). The Antisocial Personalities. Lawrence Erlbaum Associates, Hillsdale, NJ.

Mazur, A., & Booth, A. (1998). Testosterone and dominance in men. Behavioral and Brain Sciences, 21(3), 353–363.

Moffitt, T. E., Ross, S., & Raine, A. (2011). Crime and biology. Crime and Public Policy, 53–87.

Moons, W. G., Eisenberger, N. I., & Taylor, S. E. (2010). Anger and fear responses to stress have different biological profiles. Brain, Behavior, and Immunity, 24(2), 215–219.

Murray, J., Hallal, P. C., Mielke, G. I., Raine, A., Wehrmeister, F. C., Anselmi, L., & Barros, F. C. (2016). Low resting heart rate is associated with violence in late adolescence: A prospective birth cohort study in Brazil. International Journal of Epidemiology, 45(2), 491–500.

Nagin, D., & Paternoster, R. (2000). Population heterogeneity and state dependence: State of the evidence and directions for future research. Journal of Quantitative Criminology, 16(2), 117–144.

Nakagawa, S. (2004). A farewell to Bonferroni: The problems of low statistical power and publication bias. Behavioral Ecology 15(6): 1044-1045.

Op de Macks, Z. A., Moor, B. G., Overgaauw, S., Güroğlu, B., Dahl, R. E., & Crone, E. A. (2011). Testosterone levels correspond with increased ventral striatum activation in response to monetary rewards in adolescents. Developmental Cognitive Neuroscience, 1(4), 506–516.

Ortiz, J., & Raine, A. (2004). Heart rate level and antisocial behavior in children and adolescents: A meta-analysis. Journal of the American Academy of Child & Adolescent Psychiatry, 43(2), 154-162.

Patrick, C. J., Bradley, M. M., & Lang, P. J. (1993). Emotion in the criminal psychopath: Startle reflex modulation. Journal of Abnormal Psychology, 102(1), 82–92.

Platje, E., Popma, A., Vermeiren, R. R., Doreleijers, T. A., Meeus, W. H., van Lier, P. A., ... & Jansen, L. M. (2015). Testosterone and cortisol in relation to aggression in a non‐clinical sample of boys and girls. Aggressive Behavior, 41(5), 478–487.

Popma, A., Vermeiren, R., Geluk, C. A., Rinne, T., van den Brink, W., Knol, D. L., ... & Doreleijers, T. A. (2007). Cortisol moderates the relationship between testosterone and aggression in delinquent male adolescents. Biological Psychiatry, 61(3), 405–411.

Portnoy, J., & Farrington, D. P. (2015). Resting heart rate and antisocial behavior: An updated systematic review and meta-analysis. Aggression and Violent Behavior, 22, 33–45.

Portnoy, J., Raine, A., Chen, F. R., Pardini, D., Loeber, R., & Jennings, J. R. (2014). Heart rate and antisocial behavior: The mediating role of impulsive sensation seeking. Criminology, 52(2), 292–311.

Prasad, S., Lassetter, B., Welker, K. M., & Mehta, P. H. (2019). Unstable correspondence between salivary testosterone measured with enzyme immunoassays and tandem mass spectrometry. Psychoneuroendocrinology, 109, 104373.

Pruessner, J. C., Kirschbaum, C., Meinlschmid, G., & Hellhammer, D. H. (2003). Two formulas for computation of the area under the curve represent measures of total hormone concentration versus time-dependent change. Psychoneuroendocrinology, 28(7), 916–931.

Quay, H. C. (1965). Psychopathic personality as pathological stimulation-seeking. American Journal of Psychiatry, 122(2), 180-183.

R Core Team. (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Retrieved from https://www.R-project.org/.

Raine, A. (2002). Annotation: The role of prefrontal deficits, low autonomic arousal, and early health factors in the development of antisocial and aggressive behavior in children. Journal of Child Psychology and Psychiatry, 43(4), 417-434.

Raine, A., Venables, P. H., & Williams, M. (1990). Relationships between central and autonomic measures of arousal at age 15 years and criminality at age 24 years. Archives of General Psychiatry, 47(11), 1003–1007.

Raine, A., Venables, P. H., & Williams, M. (1995). High autonomic arousal and electrodermal orienting at age 15 years as protective factors against criminal behavior at age 29 years. American Journal of Psychiatry, 152(11), 1595–1600.

Rimmele, U., Zellweger, B. C., Marti, B., Seiler, R., Mohiyeddini, C., Ehlert, U., & Heinrichs, M. (2007). Trained men show lower cortisol, heart rate and psychological responses to psychosocial stress compared with untrained men. Psychoneuroendocrinology, 32(6), 627-635.

Romero-Martínez, A., Lila, M., Williams, R. K., González-Bono, E., & Moya-Albiol, L. (2013). Skin conductance rises in preparation and recovery to psychosocial stress and its relationship with impulsivity and testosterone in intimate partner violence perpetrators. International Journal of Psychophysiology, 90(3), 329-333.

Ruttle, P. L., Shirtcliff, E. A., Serbin, L. A., Fisher, D. B. D., Stack, D. M., & Schwartzman, A. E. (2011). Disentangling psychobiological mechanisms underlying internalizing and externalizing behaviors in youth: Longitudinal and concurrent associations with cortisol. Hormones and Behavior, 59(1), 123–132.

Sapolsky, R. M., Romero, L. M., & Munck, A. U. (2000). How do glucocorticoids influence stress responses? Integrating permissive, suppressive, stimulatory, and preparative actions. Endocrine Reviews, 21(1), 55–89.

StataCorp. 2017. Stata Statistical Software: Release 15. College Station, TX: StataCorp LLC.

Stauss, H. M. (2003). Heart rate variability. American Journal of Physiology-Regulatory, Integrative and Comparative Physiology, 285(5), R927-R931.

Sundin, Z. W., Chopik, W. J., Welker, K. M., Ascigil, E., Brandes, C. M., Chin, K., ... & Tackett, J. L. (2021). Estimating the associations between big five personality traits, testosterone, and cortisol. Adaptive Human Behavior and Physiology, 7(3), 307–340.

Sweeten, G. (2012). Scaling criminal offending. Journal of Quantitative Criminology, 28(3), 533–557.

Tackett, J. L., Herzhoff, K., Harden, K. P., Page-Gould, E., & Josephs, R. A. (2014). Personality× hormone interactions in adolescent externalizing psychopathology. Personality Disorders: Theory, Research, and Treatment, 5(3), 235–246.

van Bokhoven, I., Van Goozen, S. H. M., Van Engeland, H., Schaal, B., Arseneault, L., Séguin, J. R., ... & Tremblay, R. E. (2005). Salivary cortisol and aggression in a population-based longitudinal study of adolescent males. Journal of Neural Transmission, 112(8), 1083–1096.

van Goozen, S. H., Fairchild, G., Snoek, H., & Harold, G. T. (2007). The evidence for a neurobiological model of childhood antisocial behavior. Psychological Bulletin, 133(1), 149–182. https://psycnet.apa.org/doi/10.1037/0033-2909.133.1.149

Virkkunen, M. (1985). Urinary free cortisol secretion in habitually violent offenders. Acta Psychiatrica Scandinavica, 72(1), 40–44.

Welker, K. M., Lassetter, B., Brandes, C. M., Prasad, S., Koop, D. R., & Mehta, P. H. (2016). A comparison of salivary testosterone measurement using immunoassays and tandem mass spectrometry. Psychoneuroendocrinology, 71, 180–188.

Welker, K. M., Lozoya, E., Campbell, J. A., Neumann, C. S., & Carré, J. M. (2014). Testosterone, cortisol, and psychopathic traits in men and women. Physiology & Behavior, 129, 230–236.

Wickham, H., Averick, M., Bryan, J., Chang, W., McGowan, L. D., François, R., Grolemund, G., Hayes, A., Henry, L., Hester, J., Kuhn, M., Pedersen, T.L., Miller, E., Bache, S. M., Müller, K., Ooms, J., Robinson, D., Seidel, D. P., Spinu, V., Takahashi, K., Vaughan, D., Wilke, C., Woo, K., Yutani, H. (2019). Welcome to the tidyverse. Journal of Open Source Software, 4(43), 1686.

Wilcox, R. R. (2010). Fundamentals of modern statistical methods: Substantially improving power and accuracy. New York: Springer.

Wingfield, J. C., Hegner, R. E., Dufty Jr, A. M., & Ball, G. F. (1990). The "challenge hypothesis": Theoretical implications for patterns of testosterone secretion, mating systems, and breeding strategies. The American Naturalist, 136(6), 829–846.

Zhang, Q., Chen, Z., Chen, S., Xu, Y., & Deng, H. (2017). Intraindividual stability of cortisol and cortisone and the ratio of cortisol to cortisone in saliva, urine and hair. Steroids, 118, 61–67.

Zmuda, J. M., Cauley, J. A., Kriska, A., Glynn, N. W., Gutai, J. P., & Kuller, L. H. (1997). Longitudinal relation between endogenous testosterone and cardiovascular disease risk factors in middle-aged men: a 13-year follow-up of former multiple risk factor intervention trial participants. American Journal of Epidemiology, 146(8), 609–617.

Supplemental Material

Addressing Outliers and Non-normality in Hormone and ANS Measures

The distributions of testosterone and cortisol scores were examined and transformed where appropriate prior to the estimation of inferential statistics. Due to large sex differences in testosterone the distributions of women and men were considered separately (Women: M = 47.07, SD = 21.41; Men: M = 124.31, SD = 61.16; t(519) = 16.17, p < 0.001). Testosterone concentrations were not skewed in women or men. Outliers in testosterone concentrations for women and men were rescored to 3 SD from the mean (Wilcox 2010). There were no sex differences in cortisol (Women: M = .22, SD = .17; Men: M = .22, SD = .16; t(519) = .42, p = .83). Cortisol concentrations were skewed and therefore log 10 transformed after the addition of a constant. As with testosterone scores, outliers in the cortisol measure were rescored to 3 SD from the mean. There were sex differences in baseline heart rate (Women: M = 81.80, SD = 12.70; Men: M = 76.91, SD = 12.57; t(519) = -4.16, p < 0.001) and skin conductance (Women: M = 3.16, SD = 1.79; Men: M = 4.63, SD = 2.37; t(275) = 7.24, p < 0.001). Therefore, baseline ANS measures were evaluated within sex and ANS outliers identified within sex were rescored to 3 SD from the mean. There were no sex differences in AUC_i for heart rate (Women: M = 1265.11, SD = 1883.62; Men: M = 969.90, SD = 61.16; t(519) = -1.72, p = 09) or skin conductance (Women: M = 229.90, SD = 208.91; Men: M = 228.63, SD = 240.72; t(307) = -.06, p = .953).

Supplemental Figure S1. Region of significance plots for interactions between hormones and ANS measures in Supplemental Table S2.

Panel A Conditional effect of testosterone on crime by heart rate

Panel B Conditional effect of cortisol on crime by skin conductance reactivity