Research on callous-unemotional (CU) traits has made a significant contribution to our understanding of the development of conduct problems and delinquency in adolescence [1]. CU traits, which are characterized by a lack of empathy, remorse, and responsiveness to punishment, capture the affective component of psychopathy [2], and are associated with the most severe, persistent, and aggressive forms of antisocial behavior in youth [3, 4]. The inclusion of a Limited Prosocial Emotions specifier for the diagnosis of conduct disorder in the Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM-5; [5]) has further highlighted the clinical importance of CU traits for the development of delinquency. For example, children with high CU traits are more likely to display conduct problems early in their development [3] and are at the greatest risk for engaging in criminal activity in adulthood [6]. Furthermore, results from multiple large-scale meta-analyses suggest that the presence of CU or psychopathic traits in justice-involved youth are predictive of continued violent reoffending and recidivism [7, 8]. With one in seven incarcerated adolescents endorsing significant levels of CU traits [9], therefore, there is a pressing need to identify and intervene with these youth early in their developmental trajectories.

Current methods for assessing CU traits in justice-involved youth, however, are hindered by a number of limitations. For example, although the Psychopath Checklist-Youth Version (PCL-YV; [10]) is considered the “gold standard” clinical assessment of psychopathic and CU traits in children and adolescents, its utility in juvenile justice (JJ) settings is limited by its lengthy administration time and the need for extensively trained coders. In contrast, self-report measures, such as the Inventory of Callous-Unemotional traits (ICU; [11, 12]), are quick to administer and do not require any specialized training for scoring or interpretation. The ICU has also demonstrated good psychometric properties when used with justice-involved youth [13], including good internal consistency [12], strong convergent validity with self-reported aggression and delinquency [12, 14], and good concurrent and predictive validity with charge and arrest records [14, 15].

Nevertheless, self-report measures of CU traits are still limited by evidence that youth with CU traits may have difficulty recognizing their own as well as others’ emotions [16], and may be especially likely to underreport behavioral and emotional symptoms [15]. For example, in a study of 272 clinic-referred adolescents, youth scoring high on CU traits but low on anxiety reported significantly lower externalizing behaviors and cruelty relative to their parents’ reports [15]. Relatedly, a number of studies have found only modest correlations between youth and parent-reports of CU traits, with JJ youth consistently rating themselves as having significantly lower levels of CU traits than do their parents [14, 17]. This has led some researchers to suggest that parent-reports may be a more valid measure of CU traits than self-reports obtained from justice-involved youth (e.g., [14]). On the other hand, however, a recent psychometric study of 825 incarcerated adolescents concluded that parent-reports added little unique information about CU traits beyond what was already captured by youths’ self-report [18]. Therefore, although clinical interviews and questionnaire measures all have the potential to provide useful information about CU traits, the challenges discussed above demonstrate the need to find additional means of identifying youth with high levels of CU traits in JJ settings.

One promising direction for the assessment of CU traits involves utilizing psychophysiological indicators of emotional arousal. The rationale for using such methods draws on well-established evidence of the organizing effect of emotional arousal on the autonomic nervous system (ANS; [19]), and theoretical propositions suggesting that emotional under-arousal is a key feature in the development of CU traits (see [1] for a review). One relevant theoretical model suggests that, for emotionally under-aroused youth, the absence of negative affect following a transgression or punishment interferes with social learning and inhibits the normal development of prosocial emotions such as empathy, guilt, and remorse [20]. Similarly, Blair [21] has hypothesized that limited emotional and physiological responsivity to distress cues in others prevents chronically under-aroused youth from developing empathic concern through the typical process of associative learning. Furthermore, Frick and Morris [22] proposed that whereas youth with high emotional arousal and poor emotion regulation skills may develop conduct problems as a result of disinhibited aggression and/or impaired social relationships, children with temperaments characterized by fearlessness and emotional under-arousal may engage in antisocial behavior due to disrupted moral development and the adoption of a callous interpersonal style.

Considerable empirical evidence for the association between CU traits and emotional under-arousal also can be drawn from studies utilizing psychophysiological indexes of emotional arousal. Emotional arousal is linked to physiological changes in the body through activation of the sympathetic and parasympathetic divisions of the ANS [19]. Therefore, psychophysiological measures that capture ANS activity can be used to infer changes in emotional arousal in response to a stimulus. Numerous such psychophysiological measures of the ANS exist; however, in the CU literature, the most common include skin conductance, an index of primarily sympathetic nervous system activity, and heart rate, which captures both sympathetic and parasympathetic activity [23]. After taking this variability in psychophysiological methods into account, two recent systematic review concluded that current empirical evidence supports the proposition that youth high in CU traits are under-aroused and display lower ANS reactivity to emotional stimuli relative to their low CU peers [23, 24]. Furthermore, Northam and Dadds [24] concluded that, relative to self-report questionnaires or behavioral assessments, psychophysiological indexes of ANS activity are one of the most robust predictors of reduced emotional responsiveness in youth with high CU traits. These findings are in line with previous research suggesting that blunted physiological reactivity to emotional stimuli is an important biomarker of CU traits (e.g., [25, 26]) and indicate that psychophysiological indexes of emotional arousal may present a promising addition to the current methods of assessing CU traits in JJ settings.

Despite the relevance of CU traits for delinquency, however, relatively few studies have used psychophysiological measures to explore the association between emotional arousal and CU traits in samples of JJ youth, and inconsistent results have emerged from those that have. For example, in line with theory, Kimonis and colleagues [12] and Muñoz and colleagues [27, 28] found that, in sample of 85–98 adolescent male offenders, boys with high CU traits displayed lower skin conductance reactivity when exposed to a simulated peer provocation task than did boys with low CU traits. More recently, MacDougall et al. [29] also found that higher levels of self-reported CU traits were associated with reduced skin conductance reactivity during a white noise countdown task in a sample of 56 detained boys. In a study of 238 incarcerated adolescent boys, Kimonis et al. [30] also reported that high CU traits were associated with lower physiological reactivity when assessed via reduced startle potentiation to aversive images; however, this was only true for those high-CU youth with low levels of maltreatment exposure. Finally, very few published studies have investigated the psychophysiological profiles of girls with high levels of CU traits (e.g., [31]), and the authors are not aware of any that have done so in a sample of justice-involved girls.

Heterogeneous results have also emerged from research using non-JJ samples of youth, with multiple studies reporting no differences in physiological responsivity to emotional stimuli depending on levels of CU traits (e.g., [32]). Indeed, both Fanti [23] and Northam and Dadds [24] noted that despite the overall results of their systematic reviews supporting an association between CU traits and decreased emotional arousal in youth, substantial heterogeneity still exists in individual study results due to variability in sample (e.g., child vs. adolescent), participant (e.g., presence of comorbid psychological disorders), and methodological (e.g., type of stimulus used) characteristics. Furthermore, both systematic reviews concluded that more research is needed to examine whether boys and girls with high CU traits differ in their physiological reactivity to emotional stimuli [23, 24]. It is clear that much remains to be understood about the emotional processes underlying CU traits in adolescents, and that further research into the psychophysiological profiles of youth with high CU traits is needed to determine whether psychophysiological measures have the potential improve CU trait assessment with JJ youth. A significant challenge to research in this area, however, is that many psychophysiological measures are limited in their practical utility in JJ settings due to the need for expensive equipment, specialized training, and time intensive and invasive procedures. Therefore, evidence supporting a more accessible and cost-effective psychophysiological index of emotional arousal would be of value for research involving JJ youth.

One such promising alternative to traditional psychophysiological measures draws on research demonstrating that emotional arousal may also be detected in human speech. In the first phase of human speech production, air passes over the vocal folds in the larynx as it moves from the lungs out the throat. The resulting vibrations in the air are consequently influenced by biological sex-based and anatomical differences (e.g., physical size) as well as conscious and unconscious processes affecting the force of the air and the tension of the vocal folds [33] Regarding unconscious processes, Scherer’s [34] Component Process Model of Emotion suggests that, in response to an individual’s cognitive and affective appraisal of an event, activation of the autonomic and somatic nervous systems leads to changes in respiration and muscle tension in the vocal tract [35]. In this way, emotional arousal is encoded in the acoustic properties of speech and can be extracted and interpreted objectively using features of the sound wave produced during vocalization.

Fundamental frequency (f0) is one of the most well-studied and emotionally salient features of vocally-encoded emotional arousal (see [33] and [36] for reviews). Perceived by the listener as pitch, f0 is the lowest harmonic frequency produced during speech, and is measured in cycles per second, or hertz (Hz) such that higher f0 corresponds with higher frequency vibrations and higher pitch of the voice [37]. Many f0-related parameters can be used as indicators of emotional arousal (e.g., mean f0, time to peak f0; [38]); however, current methodological best practices suggest that f0 range is the cleanest and most robust vocal correlate of emotional arousal [33, 36, 39]. By focusing on the difference between an individual’s maximum and minimum f0, f0 range is also advantageous in that it accounts for potential individual and sex-based differences in vocal expression and eliminates the need for a separate baseline measurement [40].

A number of studies have validated f0 range as an index of encoded emotional arousal in adult samples by demonstrating convergent validity with other psychophysiological measures of arousal (see [33, 38, 41] for reviews). For example, Weusthoff and colleagues [40] found that in a sample of 67 German couples, wider f0 range was associated with higher heart rate, higher blood pressure, and greater cortisol (the body’s main stress hormone) output during conflict-focused conversations. Few studies, however, have utilized f0 range as an index of emotional arousal in adolescents. In fact, to date, the authors are aware of only one relevant study in which f0 was extracted from audio recordings of problem-focussed discussions between 56 adolescents (ages 13–18) and their parents [39]. From this study, Baucom and colleagues [39] found that, for both boys and girls, wider mean f0 range was associated with greater cortisol output and higher levels of self-reported negative emotions during parent-child discussions. Boys, however, displayed significantly less variability in f0 range than girls did across the duration of the discussion, and this was especially true when their cortisol output was high. These results provide support for f0 range as an index of encoded emotional arousal in adolescence, while also highlighting the importance of examining sex differences in the association between f0 range and emotional arousal across development. In particular, because of the limited research on vocally-encoded emotional arousal in adolescence, little is known about how sex-based changes to the properties of f0 during puberty may differentially impact boys and girls vocal expression of emotional arousal. During adolescence, lengthening and thickening of the vocal folds results in decreased mean f0 for both sexes, as well as increased width and variability in f0 range, and greater differentiation between male and female voices [39, 42]. Therefore, more studies are needed that carefully examine such potential sex differences during adolescence.

The Current Study

Theoretical and empirical support for the association between CU traits and dampened emotional arousal suggests that psychophysiological indexes of emotional arousal have the potential to be useful in the assessment of CU traits in justice-involved youth. However, to date, little psychophysiological research into CU traits has involved JJ youth, and there is a particular dearth of studies including justice-involved girls with high CU traits. Emerging research on vocally-encoded emotional arousal suggests that f0 range may serve as an accessible and cost-effective psychophysiological marker of CU traits with the potential to be feasible within JJ settings. Currently, few studies have used f0 range as an indicator of emotional arousal in adolescents, and no studies have used this methodology in a sample of JJ youth. Therefore, an important first step towards validating f0 range as a marker of CU traits in JJ youth is to establish the nomological network between it and other theoretically related constructs in a relevant sample [43]. The primary aim of the present study is to make a critical contribution towards this goal by testing the convergent validity between f0 range, CU traits, and emotion dysregulation in a mixed-gender sample of justice-involved adolescents.

Based on the literature reviewed above, it was hypothesized that youth with higher levels of self-reported CU traits would experience lower emotional arousal during an emotionally evocative task and therefore exhibit narrower mean f0 range. Similarly, because we expected JJ youth with greater difficulty regulating their emotions to display higher levels of emotional arousal when exposed to an emotionally evocative stimulus, we hypothesized that higher self-reported emotion dysregulation would be associated with wider mean f0 range during our experimental task. Finally, although we expected boys to display narrower average f0 range than girls, the limited foregoing literature did not support specific hypotheses about gender differences in the associations between f0 range, CU traits, and emotion dysregulation.

Method

Participants

Participants included 170 youth (137 boys, 31 girls, 2 transgender) recruited from two juvenile detention centers in the Mountain West region of the United States. Transgender participants were excluded from analyses because of the limited representation in this sample and our interest in testing sex-based difference in mean f0 range. The remaining 168 participants were between 12.52 and 17.96 years old with a mean age for boys of 16.08 (SD = 1.27) and of 15.81 (SD = 1.36) for girls. The racial and ethnic composition of this sample was consistent with demographics of the juvenile justice population in this geographic region: 39% of the sample self-identified as Hispanic/Latinx, 36% as White/Caucasian, 6% as Black/African-American, 2% as Pacific Islander/Native Hawaiian, 2% as Asian/Asian-American, 1% as Native American/Alaskan Native, 9% as Multiracial, and 1% as an Other race. 4% of participants did not disclose their race.

Procedure

As part of a larger study, all procedures for this project were approved by the institutional review boards of the University of Utah and the Utah Department of Human Services. After informed consent was obtained from legal guardians during visitation hours at the detention center, youth were invited to provide assent to participate. Study tasks were administered by trained research assistants in private visitation rooms at the detention center, and all data were anonymized through the use of unique identification codes. Following the completion of self-report questionnaires on a laptop, youth were invited to provide a two-minute verbal narrative describing a personally stressful experience. Youth were encouraged to focus on sensory details of the experience as well as the way they felt at the time and were given the option to describe their most recent arrest if they did not feel comfortable discussing another event. This task was used to evoke an emotional response in participants and has been found to effectively increase physiological arousal in other samples of adolescents [44] and adults [45]. With youth permission, narratives were recorded using a handheld audio recording device. No compensation was offered to youth for their participation.

Measures

Callous Unemotional Traits

Self-reported CU traits were assessed using the Inventory of Callous Unemotional traits (ICU; [11]). In this measure, youth are asked to rate how well each of 24 statements describe them using a 4-point Likert scale anchored by 0 (“not at all true”) and 3 (“definitely true”). All items were summed to create a composite score with higher values indicating greater CU traits. Sample items include “I do not show my emotions to others” and “I am concerned about the feelings of others” (reverse-scored). The ICU has demonstrated good psychometric properties when used with justice-involved youth including good internal consistency and strong construct validity [12, 14]. In the current study, Cronbach’s alphas were 0.81 and 0.85 for boys and girls, respectively.

Emotion Dysregulation

Total scores from the 18-item short form [46] of the Difficulties in Emotional Regulation Scale (DERS; [47]) were used as a measure of youths’ self-reported emotion dysregulation. Items inquire about awareness (e.g. “I pay attention to how I feel”, reverse coded), clarity (e.g. “I am confused about how I feel”), and acceptance (e.g. “when I am upset, I become embarrassed for feeling that way”) of emotions, as well as responses to negative emotional experiences (e.g. “when I am upset, I have difficulty concentrating”, “When I am upset, I lose control over my behaviors”). Items are rated on a five-point Likert scale (from 1 = “almost never” to 5 = “almost always”) and summed to create a composite such that higher scores indicate greater dysregulation. Validation of this measure in a sample of adolescents indicated that it possesses excellent psychometric properties equivalent to those of the full DERS [46], and its reliability and predictive validity have been confirmed in samples of JJ youth [16, 48]. Cronbach’s alphas for the current sample were 0.81 and 0.89 for boys and girls, respectively.

Fundamental Frequency

Emotional arousal was operationalized as youths’ mean f0 range during their verbal recounting of a personally stressful experience. Following the methods outlined by Baucom et al. [39] and Owren and Bachorowski [49], vocal data were analyzed in 0.25 s intervals and f0 was extracted from the audio recordings using Praat software, version 6.0.43 [50]. F0 range was calculated by subtracting the minimum f0 value detected during their entire narrative, from the maximum f0 value measured at each 0.25 s interval. These values were then used to compute an average f0 range for each individual. Data cleaning involved first manually silencing any non-verbal vocalizations (e.g., coughing, laughing) or speech uttered by someone other than the participant. Time series and regression plots of f0 range were then visually inspected for impossible or extreme values (i.e., errors in pitch extraction), and problematic recordings were re-inspected/re-analyzed. Raw audio files in the current study ranged between 28 s and 12 min 41 s in length (M = 2 min 16 s, SD = 1 min 19 s) and, after cleaning, participants’ total usable talk time was between 5 s and 10 min 10 s (M = 1 min 14 s, SD = 1 min 2 s). After the first round of manual cleaning, all f0 values fell within the acceptable range for adolescents [49], and re-inspection of audio files flagged for potentially extreme values revealed no obvious recording errors. Given that the computation of f0 range is reliant on having sufficient data to capture how individuals’ f0 varies over time, participants with less than 30 s of usable audio data after audio cleaning (nboys = 20, ngirls = 2) were flagged and used in sensitivity analyses to determine whether their data were biasing the results.

Data Analysis

All analyses were conducted using R, version 3.6.3 [51]. We used multiple regression to test our hypotheses. Post-hoc diagnostics conducted using only complete cases (nboys = 121, ngirls = 29) were consistent with the assumptions of linear regression. Separate models were computed for boys and girls to account for sex-based differences in mean f0 range. Age, measured in years, was included to control for differences in adolescents’ physical maturation; while ethnicity, coded as White = 0, Hispanic/Latinx = 1, Other Race = 2, was added to the models in light of some evidence suggesting ethnic differences in physiological responsivity (e.g., [52]). All continuous variables were mean-centered prior to analysis. Missing data ranged from 3.2 to 6.5% for items assessing CU traits and emotion dysregulation. Data were missing because youth were given the option of “choose to skip” for every question on the ICU and DERS-SF. Based on current best practices for handling missing data [53], model-based multiple imputation with chained equations was performed using the mice package [54] in R. This approach uses a series of univariate regression models to generate the imputations [53] and is recommended for models including both continuous and categorical variables [55]. Imputations were conducted using the combined boys’ and girls’ data, and participant self-reported biological sex and average anxiety scores (computed from the short form of the Revised Children’s Manifest Anxiety Scale-2; [56, 57]) were included as auxiliary variables in the imputation models. All reported model results were computed by separating the boys’ and girls’ data and averaging across fifty imputed data sets to create pooled estimates. Sensitivity analyses comparing model results using the imputed datasets to those obtained with listwise deletion suggested that multiple imputation improved the accuracy of the estimates (see Table S1 in the supplemental materials). Sensitivity analyses also indicated that data from the 22 participants with less than 30 s of talk time were not impacting the overall pattern of results for boys or girls (see Table S2 in the supplemental materials); therefore, analyses were conducted using data from all 137 boys and 31 girls.

Results

Descriptive Statistics

To characterize the samples of boys and girls, we conducted exploratory analyses including descriptive statistics (see Table 1) and zero-order correlations (see Table 2). These results are summarized below to provide additional context and understanding of our inferential analyses. As expected, boys’ mean f0 range was narrower than girls’; however, in this sample, boys’ and girls’ scores on measures of CU traits and emotion dysregulation scores were relatively similar. Also as expected, mean f0 range was positively associated with emotion dysregulation for boys ( r = .23; 95% CI: 0.06, 0.39). Although a correlation of similar magnitude emerged for girls (r = .20), the associated 95% confidence interval contained zero (-0.17, 0.52). Confidence intervals for the association between CU traits and mean f0 range also contained zero for boys ( r = − .12; 95% CI: − 0.29, 0.06) and girls (r = .02; 95% CI: − 0.35, 0.39). Finally, a medium sized positive correlation between CU traits and emotion dysregulation was observed for girls (r = .56; 95% CI: 0.24, 0.77), but not boys (r = .17; 95% CI: − 0.01, 0.33).

Table 1 Descriptive Statistics for Boys and Girls on all Study Variables
Full size table
Table 2 Zero-Order Correlations and 95% Confidence Intervals Between Variables of Interest for Boys and Girls
Full size table

Associations Among Mean f0 Range, Emotion Dysregulation, and CU Traits

Study hypotheses were tested by regressing mean f0 range onto CU traits, emotion dysregulation, age, and ethnicity. As a group, the predictors accounted for a small but significant proportion of variance in mean f0 range for boys [Adj R2 = 0.10, F(5, 131) = 4.18, p = .003], but not girls [Adj R2 < 0.00, F(5, 25) = 0.45, p = .811]. Full model results are presented in Table 3. Pooled results indicated that although zero-order correlations between CU traits and mean f0 range were non-significant for both sexes, after controlling for individual differences in emotion dysregulation, age, and ethnicity, the negative association between CU traits and mean f0 range was significant for boys, B = –0.55, p = .038. This pattern of results is indicative of a suppression effect which is an effect that emerges when two variables are positively correlated with each other and one of these variables is positively correlated with a third while the second is negatively correlated with a third variable [58]. Including all three variables in the same model allows for correct partitioning of this complex pattern of associations and for suppressed effects to emerge. Post hoc analysesFootnote 1 determined that the association between mean f0 range and CU traits was significant only when all three additional variables were included in the model, which suggests that age, ethnicity, and emotion dysregulation all account for unique aspects of f0 range that are unrelated to CU traits, and vice versa. Furthermore, after controlling for age, ethnicity, and CU traits, emotion dysregulation remained significantly positively associated with mean f0 range for boys, B = 0.52, p = .011. Finally, after controlling for the other variables in the model, two significant effects emerged with respect to our covariates. Boys who were older displayed narrower mean f0 range during the experimental task (B = − 3.43, p = .047), and boys or who identified as neither White nor Latinx had significantly narrower mean f0 range than those who identified as White (B = − 12.31, p = .040). No significant results emerged for girls.

Table 3 Regression Model Results for Mean Fundamental Frequency Range Regressed onto Emotion Dysregulation, CU Traits, and Covariates
Full size table

Discussion

This study is the first to examine the associations between f0 range measured during an emotionally evocative task, self-reported CU traits, and self-reported emotion dysregulation in a mixed-gender sample of justice-involved adolescents. In line with hypotheses, results demonstrated that boys who reported higher levels of CU traits displayed narrower f0 range while recounting a personally stressful experience. This finding indicates that justice-involved boys with higher CU traits displayed lower emotional arousal during the experimental task, which is supported by theoretical models suggesting that emotional under-arousal and limited physiological reactivity to emotional stimuli are key biomarkers of CU traits in youth [1, 25, 26]. This finding is also concordant with previous empirical evidence demonstrating that youth with higher CU or psychopathic traits display lower reactivity on a range of psychophysiological indexes of ANS activity [23, 24, 56]. Furthermore, the results of this study add to the limited existing psychophysiological research investigating CU traits in justice-involved youth. In line with the work of Kimonis and colleagues [12], Munoz and colleagues [27, 28], and MacDougall and colleagues [29], the current study confirms that for JJ-involved boys, higher levels of CU traits are associated with lower physiological responsivity during an emotionally evocative task; however, similar to Kimonis and colleagues’ [30] more recent research, this pattern of results was only evident after accounting for other individual differences.

The fact that in the current sample of boys, a significant association between CU traits and mean f0 range only emerged when age, ethnicity, and emotion dysregulation were included in the model highlights that these variables all account for meaningful variation in f0 range and CU traits. Recent systematic reviews have concluded that sample and methodological variability largely account for inconsistent findings in the association between CU traits and physiological responsivity in youth [23, 24]. Therefore, future research using f0 range to investigate CU traits in youth may benefit from establishing a core set of theoretically supported covariates. Furthermore, given the results of the current study, future research should investigate how differences in ethnicity and emotion dysregulation may impact the development and identification of CU traits in adolescents. For example, a growing body of literature suggests that high levels of emotion dysregulation may characterize a subgroup of high CU youth (e.g., [60]).

The finding that higher self-reported emotion dysregulation was associated with wider mean f0 range during the experimental task also provides further support for f0 range as a valid indicator of emotional arousal in adolescent boys, including those involved with the JJ system. This corresponds with the theoretical proposition that for some youth, difficulties with emotion regulation may contribute to the development of conduct problems due to high emotional reactivity and disinhibited aggression in response to provocation [22]. Furthermore, the observed association between f0 range and emotion dysregulation is in line with previous research demonstrating that adolescents who display wider f0 range during an emotionally evocative task also endorse experiencing higher levels of subjective negative emotions [39]. The results of this study extend Baucom and colleagues’ [39] findings by providing support for f0 range as an index of encoded emotional arousal in an ethnically diverse and high-risk sample of boys.

Taken together, the pattern of associations that was observed between f0 range, CU traits, and emotion dysregulation in the current sample suggests that narrower mean f0 range during an emotionally evocative task is an indicator of emotional under-arousal in justice-involved boys and may serve as a marker of higher CU traits. We were not able to draw similar conclusions for girls, however, due to the limited number of significant findings that emerged and the small number of girls available for analysis in the current study. Indeed, given our sample size of girls, our analyses were only sufficiently powered to detect effects of at least r ≥ .50 at a significance of 0.05. It is therefore especially notable that a moderate and significant association between emotion dysregulation and CU traits emerged for girls but not boys. This suggests that in the current sample, girls who rated themselves as higher in CU traits also endorsed having greater difficulty engaging in self-regulatory processes when experiencing strong emotions. This finding may initially seem counter-intuitive given the previously cited literature suggesting that poorer emotion dysregulation is a feature of youth with conduct problems and low CU traits [22], as well as previous studies indicating that physiological responsivity to emotional stimuli is associated with lower CU traits [23, 24] and higher levels of emotion dysregulation (e.g., [61]). It is important to note, however, that emotion dysregulation is context dependent—one must first experience an emotional response in order to have difficulty modulating it [62]. Therefore, although youth with high CU traits may display little physiological or emotional arousal in response to some emotional stimuli (e.g., fear, distress in others; [63]), they have also been found to display intense emotional and behavioral reactions to other stimuli, such as negative social interactions (e.g., [64]). Nevertheless, the fact that the association between high CU traits and emotion dysregulation emerged only for girls suggests that, in the current study, boys and girls with high CU traits differed in their reported levels of emotion dysregulation. This is particularly interesting given that no mean differences were observed between boys and girls on self-reported emotion dysregulation or CU traits. Combined with the finding that CU traits and emotion dysregulation were associated with opposing levels of emotional arousal for boys, this pattern of results suggests that a direction for future research may be to examine whether there are meaningful subgroups of youth characterized by different profiles of CU traits, emotion dysregulation, and f0 range or emotional arousal.

Due to our sample size, we were not able to directly investigate how different profiles of CU traits, emotion dysregulation, and emotional arousal may have influenced the current study’s findings. Nevertheless, emerging research suggests that there may be variants of youth with high CU traits who display different patterns of psychophysiology in response to emotional stimuli (see [65] for a review), and future research using f0 range with JJ youth should consider these distinctions. Moreover, results from the present study suggest that JJ boys and girls who endorse high levels of CU traits may also experience different levels of emotion dysregulation. This has valuable implications for clinicians working with high-CU youth in JJ settings. For example, although youth with high CU traits are commonly thought of as being unemotional, for high-CU girls in particular, interventions targeting emotion regulation skills may be beneficial. Future research is needed to investigate these claims and to re-examine whether f0 range may serve as an indicator of emotional arousal and CU traits in justice-involved girls.

Results from the current study also have the potential to make an important contribution to the assessment of CU traits in JJ settings. In particular, vocal data analysis offers a number of advantages over clinical interviews and self-report measures of CU traits. In contrast to clinical interviews, vocal data collection is simple to perform and requires no specialized training or equipment. Furthermore, f0 range can be extracted easily using the free software package Praat [50], and, as demonstrated by the current study, even very short (i.e., less than two minute) audio recordings can provide meaningful information. Results from the current study also suggest that the task used to evoke an emotional response in youth can have relatively low demand characteristics in comparison to interview or questionnaire items designed to assess for CU traits. This is an important feature given that youth with high levels of CU traits may be highly motivated to engage in impression management. Therefore, although f0 range is subject to some degree of conscious control [33], the fact that it is primarily indexes ANS activity in response to unconscious cognitive and affective processes [34] suggests that it may also serve as a relatively objective indicator of callousness and emotional under-arousal in JJ boys.

Further research is needed to build upon the current study’s findings and to establish the validity of f0 range as an indicator of CU traits in justice-involved boys. This includes replication of the present results in other samples of JJ boys to determine their robustness to sample characteristics (e.g., ethnic composition). Furthermore, given the relevance of CU traits for continued antisocial behavior and recidivism, it will be important to examine the sensitivity and specificity of f0 range for classifying justice-involved boys on relevant external criteria (e.g., violent offending), as well as its predictive validity with respect to future recidivism, and its incremental validity over clinical interviews and self-report measures of CU traits [66]. Future studies involving the PCL-YV [10] and f0 range may be particularly valuable for advancing the state of CU trait assessment with JJ boys in light of discrepancies between the strong support for the PCL-YV’s ability to predict future recidivism and more recent research questioning its efficacy for discriminating between JJ youth with and without high psychopathic traits [67].

The results of the current study also suggest that vocal data have the potential to significantly increase the accessibility of psychophysiological research on CU traits in JJ settings. Traditional methods of collecting psychophysiological data, such as skin conductance and heart rate, present a number of challenges to researchers interested in studying CU traits in justice-involved youth. These challenges include the need for expensive equipment and the investment of significant time and financial resources to ensure that research assistants are thoroughly trained in the complex procedures necessary for accurate collection, scoring, and interpretation of psychophysiological data. Many traditional psychophysiological measures also require electrodes to be attached to participants’ skin (e.g., hand, torso). This can be a challenge regardless of the setting or sample, given that individual differences in skin type (e.g., oiliness, dampness), as well as participant movement during the procedure can interfere with measurement accuracy [68]. Furthermore, youth in general, and JJ youth in specific, may feel apprehensive about such physically invasive procedures, which can in turn interfere with their ability to remain still throughout study tasks. Finally, data collection in JJ settings is prone to interruptions due to the scheduling demands of multiple overlapping systems (e.g., court appointments, facility agendas, visitation schedules) and uncontrollable stimuli (e.g., overhead pages and alarms), which can interfere with the accuracy of traditional psychophysiological methods that require continuous and consecutive baseline and task data collection intervals in order to establish valid measurements of participants’ physiological reactions to the experimental stimulus.

As demonstrated by the results of the current study, vocal data offer researchers an alternative indicator of encoded emotional arousal in JJ boys that is inexpensive, non-invasive, fast, and easy to administer. Notably, because vocal data collection does not require specialized training, it also invites the possibility of including detention staff in this stage of the research process. For example, with proper consent and assent, detention staff could potentially collect audio samples from boys describing how they are feeling during intake or before/after certain events (e.g., a court appearance, family visitation). Not only could these accessible procedures significantly increase researchers’ access to repeated measures of emotional arousal in larger samples of JJ boys, but they could also provide valuable insights into the emotional experiences of boys with and without high CU traits as they move through the JJ system. Furthermore, because vocal data collection does not require participants to be attached to any equipment or to remain still, it also has the potential to expand the range of study stimuli used in CU research both in and out of JJ settings. To date, most of psychophysiological research on CU traits has used task-based paradigms designed to elicit a particular emotional state in participants (e.g., fear, anger, empathy; see [24] for a review). These studies have been critical for guiding the field’s understanding of how youth high in CU traits recognize, process, and respond to specific emotional stimuli; however, their generalizability is limited, particularly to scenarios involving more complex emotional dynamics such as might be encountered outside the laboratory setting. Given that interpersonal deficits are a key feature of CU traits [1], psychophysiological studies focusing on more naturalistic interpersonal interactions could therefore make a valuable contribution to the field. Additionally, accruing evidence from dyadic and intervention research suggests that greater f0 synchrony during therapist-client interactions is correlated with higher ratings of empathy and emotional bond, and may have a regulating effect on the client’s emotional arousal [69, 70]. Therefore, future research investigating f0 synchrony between clinicians and justice-involved youth may provide further information about the deficits in empathy associated with high CU traits as well as the reciprocal effect that youth with high CU traits might have on the emotional arousal of others during interpersonal interactions.

Limitations

A number of limitations are important to note when considering the results of the current study. First, only one indicator of physiological arousal, f0 range, was assessed. Given previous research demonstrating strong and consistent correlations between f0 range and other psychophysiological measures (e.g., skin conductance, cortisol output; [39, 40]), the current pattern of results does support f0 range as an indicator of encoded emotional arousal in JJ boys. Nevertheless, in studies of CU traits as well as in the broader psychophysiological literature, patterns of physiological responsivity have been found to vary across indexes of ANS activity [23, 24, 65, 71]. Therefore, future replication of these findings using additional psychophysiological measures (e.g., heart rate, skin conductance) is warranted. Similarly, although f0 range is one of the most well-studied and robust f0-related indicators of emotional arousal [33], there are other relevant vocal features (e.g., speech rate, voice intensity, jitter, shimmer; [72]) that were not included in the current study. Examination of a wider range of vocal parameters may provide additional information about how emotional arousal is encoded in the voices of youth with high CU traits. Finally, we note that although we ran only a single regression model on each of the boys’ and girls’ datasets, under a Bonferroni corrected alpha to account for the number of hypotheses we tested on each sample (i.e., \(\alpha =\) 0.05/2 = 0.025), our effect for CU traits in the boys’ sample would no longer meet this more conservative threshold for significance. The sample of justice-involved youth included in this study are also representative of only one specific geographic region and therefore the present results may not generalize to the broader JJ population. Consequently, further research using larger and more diverse samples of JJ-involved boys are needed to explore the replicability of the current results.

Summary

The current study aimed to examine the associations between f0 range measured during an emotionally evocative task, self-reported CU traits, and self-reported emotion dysregulation in a mixed-gender sample of justice-involved adolescents. This study represents the first known application of vocal data analysis to a sample of justice-involved youth and provides preliminary evidence that f0 range may serve as a marker of CU traits in JJ-involved boys. Thus, vocal data may offer an inexpensive, non-invasive, and accessible alternative to traditional psychophysiological measures of emotional arousal, which has promising implications for future clinical assessment and research on CU traits in JJ settings. Although no significant associations with f0 range emerged for girls, this was likely due to the small number of girls available for analysis. Furthermore, higher levels of CU traits were found to be associated with higher levels of emotion dysregulation for girls only, which suggests that JJ-involved girls and boys are characterized by different profiles of CU traits and emotion dysregulation. These findings have valuable implications for our understanding of the etiology of conduct problems in JJ boys and girls as well as future intervention research targeting youth with high CU traits.