Emotion Recognition Deficits in Children and Adolescents with Psychopathic Traits: A Systematic Review

Probably due to the prominence of the affective traits in the conceptualization and manifestation of psychopathic personality, affective impairments have been largely researched in relation to psychopathic traits at the neurobiological, cognitive, emotional and behavioral levels (Blair, 2013). Developmental models of psychopathy have commonly pointed to an amygdala dysfunction (e.g., Blair, 2003a), which is involved in emotion recognition (Phelps & LeDoux, 2005), a process that also seems to be affected in individuals high on psychopathic traits (Dawel et al., 2012). Emotion recognition refers to the ability to attribute emotional states in others, based on the identification of emotional cues relevant for socialization. Accurately processing emotional expressions, particularly facial expressions, is critical for everyday functioning as it facilitates appropriate interpersonal communication (Marsh & Blair, 2008), promotes shared affective experience (Hinnant & O’Brien, 2007), and motivates prosocial behavior (Marshall & Marshall, 2011). Conversely, deficits in emotion recognition may lead to dysfunctional interpersonal relationships and social adjustment (Kyranides et al., 2020), particularly when deficits in recognition affect distress emotions (e.g., fear, sadness). Hence, failing to recognize others’ distress may hinder the development of empathic concern through a process that would imply the absence of discomfort that typically follows wrong behaviors (e.g., guilt). This, in turn, would restrain the inhibition of those behaviors that may cause distress in others (e.g., Kochanska et al., 2010). Accordingly, deficits in emotion recognition have been suggested as a potential link between psychopathic traits and different forms of behavioral maladjustment (e.g., CP, antisocial behavior) commonly observed in high psychopathic individuals (Frick et al., 2014; Salekin, 2017).

Whether emotion recognition deficits are specific to distress emotions or reflect a more pervasive emotional impairment has been a question over debate. From one perspective, Blair (1995, 2006) suggested that psychopathy would be marked by specific deficits in the recognition of both fear and sadness. This would result in the no-experience of aversion after a behavior that may cause distress in others, allowing individuals with psychopathic traits to behave in a self-gratifying and goal-directed manner, without the negative consequence of feeling guilty and bad. Overall, these deficits would partially explain the callous, unremorseful and deceitful behavior in psychopathic personality. From an alternative approach, Dadds et al. (e.g., Dadds et al., 2006, 2008) have proposed a dysfunction on attentional mechanisms that would underlie emotion recognition deficits. More specifically, attentional deficits to socially relevant cues (i.e., the eyes, with some new evidence also for the mouth; Demetriou & Fanti, 2022), would serve as the basis for the emotion recognition deficits observed in high psychopathic individuals (Dadds et al., 2011, 2014). From this perspective, impairments in emotion recognition would be more pervasive rather than specific and would derive a more generalized deficit in socio-emotional functioning (Dawel et al., 2012).

Both theoretical approaches found support in previous meta-analytic studies, which mainly examined results from adult populations. The first documented meta-analysis on the topic provided support for a specific deficit in recognizing fear and, to a lesser extent, sadness from facial expressions in antisocial individuals, with no moderation of psychopathic traits (Marsh & Blair, 2008). Later studies reported a more pervasive deficit, with impairments in the recognition of multiple emotions (Wilson et al., 2011), also when more than facial cues (i.e., vocal, postural) were examined (Dawel et al., 2012). From these meta-analytic studies, only Dawel et al. (2012) distinguished between adult and young samples—including both middle-school children and adolescents, who showed deficits for all emotions, and particularly anger, fear and sadness, with greater effects for fear. Yet, most analyses were conducted for the total sample, with most adults from the forensic setting, and most children/adolescents from community or clinical settings. No additional distinction was made between child and adolescent samples, and no other relevant moderators (e.g., gender, age, sample type, the presence -or not- of concurrent disruptive behavior) were examined.

The multiple studies published so far in child and adolescent samples have yielded, to date, mixed results, probably due to a variety of designs, sampling procedures and methods that makes it difficult to extract firm conclusions (Northam & Dadds, 2020). Mirroring results from adult samples (e.g., Brislin & Patrick, 2019; Demetrioff et al., 2017; Kyranides et al., 2020), some studies provided additional evidence for the specific deficit in the recognition of fear (e.g., Fairchild et al., 2009), and other distress emotions, including anger or sadness (e.g., Muñoz, 2009; Powell et al., 2023). Others, in contrast, have found more pervasive deficits across different emotions including disgust or happiness (e.g., Kahn et al., 2017). There is also opposite evidence, with some studies showing increased recognition for distress emotions (e.g., Schwenck et al., 2014), or even no association between psychopathic traits and emotion recognition (e.g., Martin-Key et al., 2020). Understanding the diversity of results, and their potential causes and explanations, is of uttermost importance to clarify how psychopathic traits develop, and which mechanisms may boost or restrain their negative consequences. In this regard, applied implications from this knowledge may differ depending on whether emotion recognition deficits indeed exist, whether they are specific to one or two emotions (e.g., fear or sadness), or whether they reflect a more pervasive impairment in socioemotional functioning and the interpretation of others emotional states. Related with this last hypothesis, studies examining attention deficits related to emotion recognition also yielded mixed results, particularly when using eye-tracker methodologies (Demetriou & Fanti, 2022). Thus, even though there was evidence of reduced attention to other’s eyes in children and adolescents high on CU traits (e.g., Dadds et al., 2006, 2008; Martin-Key et al., 2018), some studies also revealed no association between CU traits and attention to the eyes (e.g., Bedford et al., 2017; Muñoz et al., 2021). In addition, the inclusion of other relevant face areas, such as the mouth, has made these results even more complex, with some studies suggesting that the deficits are more specific to the eye-region (e.g., Demetriou & Fanti, 2022) whilst others suggest that emotions linked to eye-region deficits (i.e., fear and anger) could differ from those observed in the mouth-region (i.e., sadness) (Hartmann & Schwenck, 2020).

The pattern of mixed results obtained in previous research raises the need to systematically organize the current knowledge, providing compelling evidence about well-established findings, and identifying the gaps and inconsistencies that should be addressed in future research. This systematic review is devoted to accurately examining the association between psychopathic traits, addressing all its dimensions, and emotion recognition deficits in young samples, including early childhood, middle-childhood, and adolescence (mean age up to 18 years old). As it might be expected, CU traits have derived most of the research aimed at understanding the aforementioned impairments in emotion recognition and processing (see Northam & Dadds, 2020). However, as evidence has been collected on the multidimensionality of the construct (Salekin, 2017, 2022) studies addressing other psychopathy dimensions would be also considered. This may shed new light on the nature of emotion recognition deficits in psychopathic personality, leading to disentangle some of its mechanistic trends. It would also help to clarify whether previous findings on CU traits can be extrapolated to other psychopathy dimensions, whether there might be specific deficits for specific dimensions or whether a combination of high interpersonal, affective, and behavioral psychopathic traits may identify a distinctive etiological subgroup.

All studies published to date have been taken into account, with no time restrictions imposed. Even though there are some previous meta-analyses on this topic, they have been published more than 10 years ago, two of them considered together young and adult samples (Marsh & Blair, 2008; Wilson et al., 2011), and one distinguished between mid-child/adolescents and adult samples (Dawel et al., 2012). However, the study from Dawel et al. (2012) did not cover any study in preschool samples, did not distinguish between children and adolescents, and did not account for any other potential variables relevant to understand the association between emotion recognition and psychopathic traits in youngsters, such as the sample type or the co-occurrence of CP (Dawel et al., 2012). Therefore, all studies published up to 2022 were examined, and several methodological variables and sample characteristics of studies were compared in an attempt to account for the discrepancies observed in previous research. Due to the importance that attention deficits may have in the emotion recognition impairments observed in individuals high on psychopathic traits (Dadds et al., 2006), and with the aim to provide additional clarification in the aforementioned theoretical approaches, studies examining attention biases in the context of emotion recognition and psychopathic personality were also included.

More specifically, the present systematic review aims to respond to the following questions:

In accordance with previous literature, we anticipate (a) an association between emotion recognition and psychopathic traits; (b) the presence of more pervasive rather than specific deficits; (c) the influence of the different dimensions of psychopathy, (d) a relevant role of attention biases in emotion recognition deficits; and (e) the moderating role of socio-demographic variables, including sex and age, as well as the existence of clinical diagnoses (CD, ODD) or subclinical symptoms (CP).

In sum, this review is aimed to organize the available evidence in this field, by providing a comprehensive guide of studies and by addressing multiple sources of variability. Ultimately, this review is expected to identify shared conclusions across studies, and to delineate some guidelines to homogenize future protocols and favor additional replication.

The present review was performed according to the actualized Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Page et al., 2021). The PRISMA Checklist is provided in Table S1, available online. A protocol for this study was developed and registered on PROSPERO (Registration number: CRD42021276769).

A search strategy (available on Prospero and Supplement X) was developed to cover the key elements of this systematic review, including (1) psychopathic traits (callous* unemotion* or CU or psychopathy or psychopathic), (2) emotion recognition (emotion* recognition or emotion* process* or emotion* identification or eye gaze or eye track* or eye fix* or facial emotion* or emotion* attent*), and (3) the developmental period (child* or adolesc*). Note that emotion processing was also included in the search strategy since some studies used indistinctively emotion recognition and processing to examine specific deficits in the recognition of basic emotions. The search to identify relevant literature was conducted on September 22nd, 2021 on PsycInfo, Scopus, PubMed and Web of Science (WOS) databases. An update of the systematic search was conducted on December 23, 2022 in order to include all published studies up to 2022. The same search terms were applied across all databases with adjustments made to accommodate the specific requirements of the search sites (see Table 1). We did not impose a start data as most studies in this area have been published in the past 20 years, and it was our intention to provide a complete picture of the state of the art. Publications were restricted to peer review journal articles to ensure a minimum threshold for quality. In addition to the electronic search, the reference lists of the included studies were also checked to identify other potentially eligible studies.

This review aimed to include all published studies, in English or Spanish, that examined the relationship between psychopathic traits and emotion recognition in children and adolescents (mean age up to 18). Studies including participants aged ≥ 18 were included as long as they were based on a well-established adolescent sample (e.g., high-school students; forensic juvenile samples that in some contexts may involve participants up to age 21). The review included cross-sectional and longitudinal studies, with both correlational and/or between-groups designs. Specific inclusion criteria for studies in this review were:

Studies were excluded if they were based on adult samples (mean age < 18), or whether they explicitly included participants with known cognitive impairments that are likely to influence emotion recognition, including brain injuries, neurodevelopmental disorders (e.g., autism spectrum disorder; ASD) or current substance abuse. Studies including between-groups designs with participants meeting criteria for one or more of the aforementioned impairments were retained as long as it was possible to extract specific data for participants with psychopathic traits who do not meet criteria for those impairments. Studies exclusively based on populations with an attention deficit hyperactivity disorder (ADHD) diagnosis were also excluded. Yet, considering the high co-occurrence rates between CP and ADHD (Hudec & Mikami, 2017), studies were retained if they were based on populations with CP, who may also meet criteria for CD or ODD, even though they reported the inclusion of participants with comorbid ADHD (k = 17). To gain clarity about the potential influence of ADHD symptoms in emotion recognition deficits related to psychopathic traits, the percentage of participants with ADHD will be extracted, and information about control analysis (or the absence of) will be explicitly provided.

Title and abstract were screened by two researchers, and excluded if they did not meet the inclusion criteria. Both researchers were blinded to each other’s decisions and the first study selection was put in common once the study selection phase was finished. RefWorks was independently used by both researchers as a reference manager tool. Of the 3215 abstracts screened, 74 were included for full-text review, which was also performed by two researchers who worked independently. If studies met the inclusion criteria, they were finally included in the review. If they violated some of the eligibility assumptions, they were classified based on the reason of exclusion (e.g., adult samples, emotion processing instead of emotion recognition, no measures of psychopathic traits, no lab task for emotion recognition, participants with neurodevelopmental disabilities). At the full-text level, articles were excluded if they: Included participants with a mean age that exceed the 18 years of age (k = 3), measured emotion recognition through questionnaires or scenarios with additional contextual cues that require emotional inferences (k = 3), examined complex instead of basic emotions (k = 1), emotion recognition was computed as a function of a global measure (e.g., moral reasoning, emotional understanding) (k = 5), or it was restricted to inform about owns’ emotional state in response to emotional content (e.g., emotional responsiveness) (k = 2), emotion recognition deficits were not (exclusively) linked to psychopathic traits (k = 4), assessed emotional processing instead of emotion recognition (k = 4), and there was evidence of substance abuse (k = 2). If the studies met the inclusion criteria but sufficient data was not available to extract the main results, data were directly requested from the authors by email.

At all stages of the review process, disagreements were solved by discussion and consensus and, when needed, a third researcher settled the unresolved disagreements. The inter-rater agreement Cohen’s kappa was used to compare agreement between the researchers regarding the decision to include or exclude the eligible studies, and interpreted as ≤ 0, no agreement, 0.01–0.20, none to slight, 0.21–0.40 fair, 0.41–0.60 moderate, 0.61–0.80 substantial, 0.81—1.00 almost perfect agreement (McHugh, 2012).

Two independent reviewers assessed the quality of included studies using the Appraisal Tool for Cross-Sectional Studies (AXIS) (Downes et al., 2016), which contains 20 questions regarding introduction, methods, results and discussion of each study. Each question could be answered with “yes” (1 point) or “no”/“don’t know” (0 points). Longitudinal studies were assessed via the Critical Appraisal Skills Program (CASP), a 12-question checklist that addresses different aspects concerning the objectives, sample recruitment, measurement, attrition, results and implications. Disagreements on studies’ quality assessment were solved by consensus.

Relevant data for each included article were added to an extraction sheet developed for this review and refined when necessary. Data extraction included participants characteristics (N [females], sample type and age range [mean age]), psychopathic and/or CU traits measure and informant, emotion recognition task specificities (i.e., type of stimuli, emotions assessed and measurement outcome), and main results specifically related with the objectives of this review. For correlational studies we extracted results focused on the relationship between psychopathic traits and emotion recognition, whilst in between-group designs we focused on results obtained in comparisons between the high psychopathic and the comparison groups. Between-group designs where psychopathic traits were examined as potential covariates were also considered as long as it was possible to extract specific data on the effect of psychopathic traits on emotion recognition. For studies assessing attention biases, information about the Areas of interest (AOI) and the measurement outcome was also extracted. Studies were organized by their focus on specific CU traits (k = 26) (Table 2) or multidimensional psychopathic traits (Table 3), and the inclusion of attention biases measurement (Table 4). Additional information about main study characteristics was also collected, including: the main purpose of the study, the study design, sample definition, percentage (%) of males, location and ethnicity, psychopathy dimension, emotion recognition task specificities (i.e., stimuli, exposure duration, number of blocks and trials, and response format), and the inclusion (or not) of attention biases analyses including, when available, the type of measure and/or the measurement device (Table S3).

A narrative approach was used to synthesize the findings for each study, obtained by compiling the information from the extraction sheet form. A minimum of five studies reporting similar results was used as a criterion to extract firm conclusions. Nevertheless, inconsistent findings or descriptive results based on a smaller number of studies were also reported as they may provide relevant information to establish new ways of discussion and analysis in future research.

The systematic search was conducted in two different time points; the first one provided 1631 titles from the electronic database search, and two additional references located via reference-list searches. The second one (update) provided 3577 titles. After excluding duplicates, 4679 studies were screened based on titles and abstracts. A total of 76 articles were selected for full-text assessment of eligibility, and the remaining articles were excluded for being off-topic and because they failed to meet the minimum inclusion criteria for this systematic review. After the full text assessment, 50 articles were finally included in the review. The excluded articles and the reasons for their exclusion are available in Supplementary Material (Table S5). The entire selection process is represented in the flowchart of Fig. 1.

The inter-rater agreement Cohen’s Kappa revealed an almost perfect agreement (K = 0.89) between the researchers regarding the decisions to include or exclude the eligible studies.

Table S3, available online, provides a summary of the main characteristics of each included study. The 50 included studies were published between 2000 and 2022 and provide data from 12,139 children and adolescents (40.84% females [N = 4958]; M age = 12.27) belonging to clinical (k = 10), forensic (k = 6), community (k = 9), at-risk (k = 9) and combined (k = 16) samples. Depending on age, we can distinguish between samples of children (i.e., up to 12 years old; k = 13), samples of adolescents (i.e., from 12 years old; k = 16), and mixed samples of children and adolescents (k = 20). Within children’s samples, the majority of the analyzed articles have focused on middle childhood (k = 9). A few articles included combined samples from early and middle childhood (k = 3). Only one study examined emotion recognition deficits in preschoolers (Kimonis et al., 2016). Most of the population analyzed came from European samples (k = 26; 11 UK; 3 Germany; 3 Netherlands; 2 Cyprus; 2 Italy; 1 Spain; 1 Switzerland and five with no region specified). The next most frequent location was USA (k = 9), Australia (k = 4), Canada (k = 2) and Brazil (only 1 study). Ethnicity was reported in 27 studies being White/Caucasian (k = 19) and African American (k = 7) the most analyzed ethnic groups.

The most common diagnosis notified was CD (k = 16), followed by ODD (k = 7) and CP (k = 3). In some cases, the clinical sample was classified according to the level of CU traits (Bennet & Kerig, 2014; Demetriu & Fanti, 2022), the specifier with Limited Prosocial Emotions (LPE), under the label “offenders” (Bowen et al., 2014), based on levels of Low concern (LC) and punishment insensitivity (PI) (White et al., 2016), or based on the presence/absence of “psychopathic personality” (Blair et al., 2001, 2005; Lemos Vasconcellos et al., 2014).

Regarding the clinical samples (i.e., children and adolescents with (sub)clinical levels of CP, ODD and CD) (n = 2724; 22.44% of total sample), some articles reported the number of comorbid ADHD (n = 2724; 24.63% of clinical samples). Also, despite not being considered in this review, in three cases ASD was notified (Bours et al., 2018; Klapwijk et al., 2016; Schwenck et al., 2011). All this information was added to the descriptive sample section, in all results tables (see Tables 2, 3, 4).

The presence and intensity of psychopathic traits have been assessed with various instruments, including the Youth Psychopathic Traits Inventory (YPI; Andershed et al., 2002) (k = 9), the Inventory of Callous Unemotional Traits (ICU; Frick, 2004) (k = 24; 48%), the Antisocial Process Screening Device (APSD; Frick y Hare, 2002) (k = 12), the Psychopathy Screening Device (PSD; Frick et al., 1994) (k = 3), the Strengths and Difficulties Questionnaire (SDQ; Goodman, 1997) (k = 4), the Child Problematic Traits Inventory (CPTI; Colins et al., 2014) (k = 1), the Psychopathy Checklist: Youth Version (PCL: YV; Forth et al., 2003) (k = 3), Items ad hoc (k = 1) and the Multidimensional Assessment Profile of Disruptive Behaviour (MAP-DB; Wakschlag et al., 2012) (k = 1). Sometimes cut-off points were used to determine the presence or absence of psychopathic traits (especially CU traits), which have been arbitrarily chosen according to different studies. Regarding CU traits, sometimes the subdivision in three dimensions (i.e., callouness, uncaring, and unemotional) was also considered.

The affective factor of psychopathy (i.e., CU traits) has been the most evaluated psychopathy dimension (k = 33; 66%). Studies that just focused on CU traits are described in Table 2. These traits can be found in different degrees of intensity (high/low) or can be conceptualized dichotomously (presence/absence). Also, 17 studies considered psychopathic traits as a multidimensional construct. Most of them (k = 9) considered psychopathic personality as a composite total score, with 3 also providing data on specific CU traits (Bours et al., 2018; Bowen et al., 2014; Fairchild et al., 2010). The remaining studies focused on different subdimensions (i.e., GM, CU, INS) (k = 5), or used alternative combinations (e.g., GM-INS, CU) (Dadds et al., 2006, 2008, 2011). Articles assessing psychopathic traits from a multidimensional perspective are reported in Table 3.

Considering design, 24 studies compared performance in emotion recognition across the whole sample using a correlational design. In the remaining (k = 26), a between-group design was used. To determine membership of the comparison group, one of the following criteria was considered: CU levels (high vs. low), cut-off points based on psychopathy assessment instruments, no clinical disorders and not being an offender. Most studies were cross-sectional (k = 45; 90%), whilst just five studies provided longitudinal results (Bedford et al., 2017; De Ridder et al., 2016; Peticlerc et al., 2019; Rehder et al., 2017).

All studies included in the present review employed an emotional recognition task. Most of the articles examined the recognition of basic emotions (Ekman, 1992). However, the most repeated emotions were distress emotions (fear and sadness). Fear was analyzed in all articles except 1 (98%) (Milone et al., 2019), whilst sadness was considered in 44 articles (88%). Some articles assessed more emotions than the basic ones (i.e., excitement, shame, pain, mad, scared) (Bedford et al., 2017; Bennett & Kerig, 2014; Kimonis et al., 2016), but these results were not considered for the purpose of the current review.

The most frequent emotion recognition tasks used were UNSW Facial Emotion Task (k = 6), The Emotion Hexagon Task (k = 5), NimStim Set of Facial Expressions (k = 4), Facial Emotion Recognition Task (k = 3), RaFD (k = 3), DANVA-II (k = 2) y CET (k = 2). Both accuracy and reaction time were the most used measures for emotion recognition. The emotion recognition task was sometimes complemented by another task to assess attentional bias (k = 11). The characteristics of the attentional tasks, with their specific results, are presented in Table 4. Only two of these articles examined attentional bias by assessing gaze direction between mothers and children (Bedford et al., 2017; Dadds et al., 2011). In the others, eye tracker devices were used. To assess the attentional pattern, AOI were established: eyes and mouth (k = 8), only the face (Bedford et al., 2017), eyes (Dadds et al., 2011) or arms (Martin-Key et al., 2021). The analysis of the response considered accuracy, reaction time (RT), number and duration of fixations.

The most analyzed stimulus type was facial and human (k = 43; 86%). Bodily stimuli (Martin-Key et al., 2021; Muñoz, 2009; Wolf & Muñoz, 2014) and vocal tones were also examined (Blair et al., 2005; Gillen et al., 2018; Stevens et al., 2001), as well as non-human stimuli, including doll faces (O’Kearney et al., 2017, 2020) and emoticons (Ezpeleta et al., 2017).

Stimuli presentation time ranged from 200 ms (Lemos Vasconcellos et al., 2014) to 6 s for static images (Bours et al., 2018), and from 1 s (Kimonis et al., 2016; Martin-Key et al., 2018) to 9 s (Schwenck et al., 2011, 2014) for dynamic emotional stimuli. In the case of video clips, the presentation ranged between 3 s (Martin-Key et al., 2021) and 144 s (Martin-Key et al., 2017, 2020).

The evaluation made from AXIS and CASP suggests overall a low to moderate level of bias among the eligible studies (see Tables S4, S5, available online). Most of the included studies provided good indicators of quality, suggesting a low risk of bias. Exceptionally, some articles that did not provide sufficient information for replicability (Milone et al., 2019) or used subjective measures that may introduce considerable bias into the study (De Ridder et al., 2016).

The main results obtained in the articles included in this systematic review are reported in Tables 2, 3 and 4. The first two analyze those studies that consider CU traits (Table 2) and all psychopathic traits (Table 3) respectively. The third one (Table 4), groups together all those articles that evaluated attentional bias in addition to emotion recognition.

Findings pointed out that variability is the main relevant result attained. Yet, we found a recurrent trend for pervasive deficits in emotion recognition considering both overall psychopathic traits and specific CU, particularly in non-clinical samples. This result would be in line with past metanalytical results (Dawel et al., 2012), and would provide additional support for a general impairment in emotion recognition (e.g., Dadds et al., 2018). However, in line with preliminary Blair’s hypothesis (1995, 2006), specific deficits in recognizing distress emotions (i.e., particularly fear and also sadness) were also found, a result that also emerges in a previous meta-analysis conducted in adult and young populations (Wilson et al., 2011), and which was repeatedly observed in most of the studies included in the present review. These deficits affected both accuracy and processing, measured as the reaction time, and were more notable for complex than simple emotions (e.g., Bowen et al., 2014; Sharp et al., 2014). Importantly, the observed deficits emerged across all modalities of emotion presented (facial, vocal, bodily, non-human) and all emotional stimuli used, a result that has been recently replicated in a study specifically aimed at examining the role of stimulus characteristics in emotion recognition related to CU traits (Powell et al., 2023).

Notwithstanding these findings, some patterns of mixed results were also found. Hence, some studies showed no relationship between emotion recognition and CU traits (e.g., Aghajani et al., 2021; Hartmann & Schwenck, 2020; Martin-Key et al., 2017, 2020), whilst others found an increased ability for emotion recognition in relation to CU traits (e.g., Bowen et al., 2014), a result that might be dependent of the assessment measurement used (Hartmann & Schwenck, 2020).

Accounting for CU subdimensions, or even CU variants, could help to clarify these results. However, in the present review only two studies considered the subdimensions of CU traits (Kimonis et al., 2016; Moore et al., 2019), with no conclusive results probably due to the lack of homogeneity in clustering the subdimensions of CU traits. It could be suggested that emotion recognition deficits are more related with both callousness and uncaring dimensions, whilst unemotional traits would show a distinctive pattern of results, with improved emotion recognition for disgust and surprise (Moore et al., 2019). These results would be in line with previous studies raising the usefulness of the callousness-uncaring combination in the CU conceptualization (Cardinale & Marsh, 2020), with some suggesting reconsiderations to the inclusion of unemotional traits as a facet of CU traits (Hawes et al., 2014). However, much more research is needed to clarify the role of each subdimension, if some suppression effect might be expected when using a total score, and whether measurement biases could be influencing these results. More research is also needed regarding the CU variants, with only one study in this review specifically focused on CU variants (Bennett & Kerig, 2014), and two more addressing the combination between CU traits and anxiety (Kahn et al., 2017) or maltreatment (Dadds et al., 2018) as a proxy of this CU distinction. Results again are mixed, probably due to methodological variability, as observed in a previous review on the topic (Craig et al., 2021). Thereby, some results suggest that primarily CU individuals, characterized by low levels of anxiety, could show an enhanced emotion recognition for some specific emotions like anger (Bennett & Kerig, 2014) or fear (Kahn et al., 2016), whilst secondary CU would be related with improved recognition of disgust (Bennett & Kerig, 2014). Others, in contrast, suggest that at low levels of anxiety (i.e., primary CU) CU traits would be related with pervasive emotion recognition deficits, affecting all emotions except happiness (Dadds et al., 2018). Considering the important distinction between primary and secondary variants, and the role that anxiety may play as moderator in emotion recognition, interesting questions arise about the precise nature of emotional processing among youths within each variant. As suggested by Craig et al. (2021) it is possible, for example, that differences may be apparent in processing emotional information about others, but not necessarily about the self. Additional research, aimed at clarifying differences in emotional processing and recognition between the two variants is particularly encourage. It might help to elucidate how anxiety, and other early adversity variables, may influence in the development of emotion recognition and awareness.

Notably, the pattern of mixed results observed in the current review was more frequently observed for CU traits, but not when other psychopathy dimensions were examined, with one exception (i.e., Gillen et al., 2018). One potential explanation of these results, beyond potential measurement biases, could be the presence of different forms of disruptive behavior, including CP, ODD and CD. Results showed that it was the combination between disruptive behavior and multidimensional psychopathic traits which was related to a dampened ability for emotion recognition (Fairchild et al., 2009, 2010; Levantini et al., 2022a), with some specific deficits for CU traits, related with worse recognition of sadness, and for GM, related with worse recognition of disgust (Levantini et al., 2022a). Further, within this (sub)clinical samples, some studies revealed that it might be the level of disruptive behavior instead of CU traits which mostly determined the impaired emotion recognition (Woodworth & Waschbusch, 2007). In this regard, several studies showed no relationship with CU traits in different (sub)clinical groups (Aghajani et al., 2021; Khols, Baumann et al., 2020; Ezpeleta et al., 2017; Kohls et al., 2020b; Klapwijk et al., 2016; Martin-Key et al., 2017, 2020, 2021; Milone et al., 2019; Schwenck et al., 2011). Less consistent but also relevant was the presence of an enhanced emotion recognition in high CU participants within (sub)clinical samples, particularly in the recognition of fear (e.g., Ezpeleta et al., 2017; Schwenck et al., 2014; Woodworth & Waschbusch, 2007). These results were also replicated for reaction time, with less time needed to recognize emotions in general (Klapwijk et al., 2016) and fear in particular (Martin-Key et al., 2018).

These results raise two important questions to be clarified in future research. First, the importance of testing the potential moderator role of disruptive behavior, particularly in relation to CU traits. The extracted results suggest that CU traits are indeed associated with deficits in emotion recognition, with a more pronounced effect in the absence or at low levels of disruptive behavior (Woodworth & Waschbusch, 2007). Second, the need to account for all psychopathy dimensions. Certainly, CU traits has been the most analyzed dimension, with great advances on delineating a potential developmental precursors of psychopathy (Frick, 2022; Frick et al., 2014). However, some voices have claimed that all psychopathy dimensions should be better conceptualized within the construct, which may help to better understand how psychopathic traits identify a distinctive group of problematic children and adolescents (Salekin, 2017, 2022). It implies that developmental models of psychopathic personality, and their related research, also include additional psychopathy dimensions to clearly disentangle how they develop, and which mechanistic processes might be underlying (i.e., similar or distinctive than those traditionally observed for CU traits).

Even though within this review a lower number of studies focused on psychopathic personality as a multidimensional construct, the results extracted were more robust, clearly suggesting a close relationship with emotion recognition deficits (e.g., Blair & Coles, 2000; Blair et al., 2001; Bowen et al., 2014; Gillen et al., 2018). This result was even more remarkable within problematic or (sub)clinical samples (e.g., Fairchild et al., 2009, 2010; Levantini et al., 2022a). It should be noted that most of these studies relied on a global psychopathy score, which limits the possibility of examining potential distinctive results across dimensions. Even in some studies assessing all psychopathic traits, there was a specific focus on CU traits (e.g., Dadds et al., 2006, 2008, 2011) with GM and INS dimensions being used as potential covariates. When different dimensions were specifically examined within the psychopathy construct, CU traits seem to be more related with impairments in emotion recognition. GM traits, on the other side, would be to some extent also related with better accuracy—and higher fixation to the mouth—for angry faces, when controlling for the other two dimensions (Gillen et al., 2018; Levantini et al., 2022a). As observed in previous studies, it seems that the combination between all psychopathy dimensions is what better identifies a more serious group of problematic children and youths (e.g., Bergstrøm & Farrington, 2022; Burke et al., 2022; Colins et al., 2022; Fanti et al., 2018; López-Romero et al., 2021, 2022). According to the current results, it could be the case that these individuals, with overall high levels on psychopathic traits, would also be the most impaired in terms of emotion recognition. Interestingly, these results held in both correlational and between-groups designs. In this regard, even though correlational designs allow accounting for all the dimensional spectrum (e.g., Kahn et al., 2016; Sharp et al., 2014), when high versus low psychopathic groups are considered, deficits in emotion recognition became more evident (e.g., Blair et al., 2005; Fairchild et al., 2010).

As was previously mentioned, the noticeable variability in the assessment of psychopathic traits could have impacted the aforementioned mixed results. For instance, there were variation across studies in terms of the version/informant (i.e., parent-, teacher-, and self-reports), the dimension used to measure psychopathic traits (e.g., Total score versus CU), the dimensional versus categorical conceptualization of the psychopathy dimension or the cut-offs used to classify individuals based on the psychopathy score. Interestingly, results seem to suggest that mixed results might not be directly related with the instrument. Instead, variation in results could be to some extent related with the informant, a result that clearly emerged with the ICU, with emotion recognition deficits being more evident with parent- and teacher-reported version of the ICU. It could be the case that response biases (e.g., social desirability) could affect more the self-report of personality aspects that entails negative connotations, or that a lack of insight or even the unwillingness to respond (e.g., tendency to lie) could have an impact on self-reports (Ray et al., 2013). This hypothesis that was not well-supported on previous research (Kelley et al., 2018; Miller et al., 2011), not either in the current review, with other self-reported measures, assessing different psychopathy dimensions (i.e., APSD, YPI) showing significant results in relation to emotion recognition. Similarly, in a recent meta-analytic review, results showed a greater relationship between psychopathic traits and negative outcomes (e.g., externalizing problems) for other- than self-reported psychopathy, although differences were not to a substantial extent (Mendez et al., 2023). The dimension addressed with the instrument (e.g., CU versus the multidimensional construct, with firmer results for the later), and other moderator variables (e.g., sex, ethnicity, anxiety, sample type) seem to also affect the results more than the instrument itself. Additional studies specifically aimed at comparing different measures would be of great interest. Future research involving all psychopathy dimensions across different samples and settings should be particularly encouraged, as well as additional comparisons between dimensional versus categorical approaches, considering similar cut-off points. Clarifying the best way to conceptualize and assess psychopathic traits will help to further examine those underlying mechanisms that, as in the case of emotion recognition, may affect socioemotional functioning, having an impact on later behavioral, emotional and psychosocial adjustment.

When categorizing facial emotions, attention is typically directed towards critical facial features, most notably the eyes and the mouth (Eisenbarth & Alpers, 2011; Wells et al., 2016). In fact, a failure to attend to these regions may lead to difficulties in judging the expressed emotions. It has been suggested that one mechanism whereby amygdala dysfunction, largely linked to psychopathic personality (Yang & Raine, 2018), may contribute to pervasive emotion recognition deficits is the abnormal attention patters to the face regions, particularly the eyes (Dawel et al., 2012), a result that has been replicated across laboratory tasks and real-life interactions.

The results of the current review support Dadds’s theory of pervasive deficit in emotion recognition (Dadds et al., 2006, 2008), which would be derived by a dysfunction on attentional mechanisms that underlie emotion recognition. Psychopathic traits were associated with atypical eye scan paths for emotional faces in children and adolescents (Billeci et al., 2019; Dadds et al., 2008; Martin-Key et al., 2018) and with different patterns of attention regarding an alternative area of interest; i.e., the mouth (Demetriou & Fanti, 2022; Hartmann & Schwenck, 2020). These results also corroborate previous findings observed in adult samples (e.g., Dargis et al., 2018; Gillespie et al., 2015). The small number of articles in the current review dealing with traits other than CU (only Levantini et al., 2022a account for each psychopathy dimension) makes it difficult to establish a robust connection with different psychopathy dimensions in samples of children and adolescents. Moreover, less is known about the role of attention in clinical or forensic samples, but preliminary results seem to go in the same direction, i.e., the presence of clinical disorders could be moderating the associations (Hartmann & Schwenck, 2020). The same occurs when the area of interest is considered. Hence, in clinical and forensic groups, atypical fixation patterns are observed not only to the priority face areas of interest, i.e., eyes (Billeci et al., 2019) and mouth (Levantini et al., 2022a), but also deficits were seen according to CD status when the area of interest was bodily posture (Martin-Key et al., 2021).

The identification of emotional expressions can be affected by several factors that have been researched and considered in the field of emotion recognition. The most widely studied have been age, sex and ethnic origin. The results, in line with what has been discussed throughout the study, have been mixed.

According to the biopsychosocial perspective, the development of various aspects of emotional and behavioral functioning occurs through interactions and coactions at multiple levels of analysis, including genetic aspects, neural and cognitive activity, behavior and social and cultural environments (Gottlieb, 2007). In line with this last point, patterns of socialization or upbringing may influence children's emotional competencies through verbal and non-verbal information communicated by parents to children (Dunbar et al., 2015). In an attempt to clarify the matter, different ethnic stimuli have been used as emotional tasks and ethnicity has been considered in the sample description of the studies; yet, the role of ethnicity has been barely considered when interpreting the results due to a lack of studies specifically accounting for this variable. Recently, Laukka and Elfenbein (2021) pointed out evidence of advantages related to belonging to the same ethnic group, with vocal emotion being more accurately recognized when expressers and perceivers were from the same cultural group. These differences in terms of ethnicity were preliminary supported in this review, with one study showing that the observed results were moderated by ethnicity, with European-American participants showing better accuracy in emotion recognition when CP were not present (Rehder et al., 2017).

As for age, the trend towards greater emotional recognition with increasing age has become apparent in recent years in samples of children (Schaan et al., 2019) and adults (Hayes et al., 2020). These results held in the current review for groups of children and adolescents with psychopathic traits in several modalities of presentation such as vocal (Blair et al., 2005) and image (Kimonis et al., 2016). However, it seems that magnitude and direction of age effects may be influenced by elements of task design in emotional recognition tasks, with larger age effects in combined tasks (i.e., static and dynamic; Hayes et al., 2020; Kimonis et al., 2016), an interaction that should be considered. Of note, most of the studies conducted in childhood covered school-aged children, so caution is needed when extending these findings to preschool samples.

Undoubtedly, the sex variable has been the most considered and the one that has generated most debate about its implication in emotional recognition. Many studies underline the advantage of woman in decoding emotions (Olderbak et al., 2018) and better outperformance of females in clinical (Kohls et al., 2019) and normative youth samples (Bek et al., 2022; Wells et al., 2016) and adult samples (Abbruzzese et al., 2019). Further, some investigations about sex differences were conducted to understand the neural basis of emotional processing and recognition (Derntl et al., 2010; Li et al., 2020). However, in the present review, the results obtained in terms of sex, especially when analyzing studies focused on female samples, show the same trends and deficits as their male counterparts (Fairchild et al., 2010; Martin-Key et al., 2020; Schwenck et al., 2014). Similar results were obtained in mixed samples that controlled for sex (Blair & Coles, 2000; Ezpeleta et al., 2017; White et al., 2016; Woodworth & Waschbusch, 2007). Additional research conducted by sex, or specifically in female samples, would be extremely needed to corroborate whether similar deficits are present, or an alternative pattern of dysfunction might be identified. These studies should be in line with future research aimed at examining the structure of psychopathic personality in girls, clarifying whether the current conceptualization of psychopathy, largely based on males’ representation, accurately replicates for different female samples (Haneveld et al., 2022).

In sum, while is crucial to consider the moderating variables mentioned above, it is equally important not to overlook the influence of task characteristics. The interaction among all these factors seems to provide a better explanation of the differences observed in the results (Pauli et al., 2021; Martin-Key et al., 2018; Wells et al., 2016). This fact would be in line with the premise that there exist multiple contributors to emotional recognition, including genetic, biochemical and environmental factors (Christov-Moore et al., 2014). Accounting for most of them, including other relevant moderators such as IQ family SES, could help to clarify the marked pattern of mixed results.

This is the first review that examines the entire literature up to 2022 on the relation of emotion recognition and psychopathic traits in childhood and adolescence. The obtained results reflect relatively strong trends in the data that at least deserve to be the subject of future research. Some of the most robust conclusions indicate the presence of deficits in emotional recognition—especially distress emotions—and raise the need to explore the role of attention and sociodemographic variables as moderators of emotional recognition, as well as to consider all psychopathy dimensions and the role of disruptive behavioral disorders in understanding emotion recognition in relation to psychopathic traits.

Nevertheless, the results attained in this systematic review and the conclusions that have been reached should be assessed in the light of some important limitations. The most remarkable concerns the high heterogeneity found in the different studies. It is hard to identify a unique source for this variability, but there are some possible factors that might be of influence. Multiple emotions were presented, but not the same across all studies. Most of them considered the basic emotions (Ekman, 1992), but at times, the emotions chosen were in response to the researcher's objectives (e.g., distress emotions, complex emotions). Intensity at presentation, complexity and exposure times were variable too, both being characteristics relevant for emotion recognition. Also, occasionally, emotional stimuli were presented several times, so the learning effect cannot be dismissed. The environmental conditions (e.g., light, noise, spaces) of the experimental setup and the way information is collected (e.g., self-registration, third part collection) can lead to differences in the results. The wide range of different instruments used to assess psychopathic traits, with different informants (i.e., teachers, parents, self-report), the consideration of psychopathic traits from a dimensional versus a categorical perspective, or the arbitrary cut-off point used to determine the presence/absence of psychopathic traits, may also lead to inconsistencies in the results. As a result of this variability, at times, the shortage of studies made it challenging to extract robust results and draw generalizable conclusions. In addition, some studies did not provide enough information for replication. In this regard, while certain recognition deficits were replicated across stimuli and presentation modalities, which is relevant to generalization, future research would benefit from homogenizing experimental conditions and promoting replication studies across diverse samples, contexts, and settings.

Other factors that affect the variability in results include the limited sample size in some of the included studies and the non-differentiation of the sample with respect to sex. A better understanding of this phenomenon inevitably requires the establishment of differentiated samples between boys and girls and the recruitment of a sufficient sample size to be able to draw reliable conclusions, particularly when between-groups designs are considered. Also important is to continue examining the longitudinal association between emotion recognition and psychopathic traits, as most of the studies included in this review were cross-sectional and hindered the possibility to interpret the directionality of the effects. New longitudinal research will help to disentangle the potential causal mechanisms, addressing linked deficits in other relevant brain structures and functional areas. Hence, complementary measures of emotional processing, including psychophysiological recordings (i.e., heart rate, skin conductance), are particularly needed to provide further evidence on underlying mechanisms that might be influencing the relationship between psychopathic traits and emotion recognition. Future research would also benefit from including preschool samples, as the available results are scarce in this developmental period. This would help to elucidate whether deficits in emotion recognition in early childhood could be somehow explained by the presence of psychopathic traits beyond age-related reasons (i.e., increased ability to recognize different emotions as children grow up). Finally, additional suggestions for future research would include stimuli with different emotional intensities. Limiting the recognition methodology only to images of high emotional intensity may not provide a clear picture of emotional recognition difficulties, whilst contemplation of various emotional intensities may be best suited to reveal sensitivity and more subtle differences in recognition (Adolphs & Tranel, 2004). More attention should be also paid to other modalities such as vocal affect, as they represent a great complement to facial and bodily stimuli in socialization processes. Lastly, promotion of cross-national and cross-cultural studies with the aim of outlining cultural differences in emotional recognition should be also encouraged.

Some other limitations concerning this study should also be outlined. It covers a wide period of time (i.e., more than 20 years), and two distinctive developmental periods (i.e., childhood and adolescence), which results in an appreciable number of studies, with multiple results to be extracted. This fact, along with the inclusion of mixed samples, makes complex to provide a finer-grained extraction, with studies specifically examined by age, sex or sample type. Relatedly, to address all the intended objectives, an extensive search equation was used, providing a great number of results to be screened. The great number of eligible studies required an organized extraction. For the current study, and based on our objectives, results were organized based on the psychopathy dimension analyzed, and the assessment of attention deficits, but other forms or organization could be possible, including the sample type, or the developmental period. However, because some studies included mixed samples, these alternative forms of organization were finally discarded. The inclusion of adolescent samples led to include some participants older than 18, resulting in a wide age range. Yet, their inclusion was justifiable as they were part of well-defined adolescent samples (e.g., high school, samples within juvenile forensic systems that include participants up to 21 years old). Finally, restricting the search to published studies (i.e., excluding grey literature) may raise the likelihood of publication biases due to the file drawer effect. The inclusion of multiple results, in different directions, could have attenuated this effect but the potential influence of this kind of biases cannot be diminished when interpreting the results.

Emotion recognition of facial expressions represents a crucial component of human social interaction, allowing the observer to infer another's emotional state and adjust their behavior (Blair, 2003b). This review provides important clinical, social and scientific implications. On a clinical and scientific level, it raises the need to attend to this problem at an early age, as we already know that psychopathic traits can be reliable identified early in development (Colins et al., 2014; López-Romero et al., 2019) and, what is even more important, that these traits could be more malleable at childhood when behavioral patterns are not firmly established. Disentangling the mechanisms underlying the development of psychopathic traits will provide additional insight about a construct that has proved its value for child and youth CP. This, in turn, will help to improve the development of more tailored preventive and intervention approaches, aimed at restraining high-risk patterns of behavioral maladjustment, with related benefits in the policy and social fields. However, more research is needed to clarify how preventive interventions could be improved from research in emotion recognition, particularly in early childhood.

In (sub)clinical samples, parent training seems to be the option that works best to reduce conduct problems (Romero et al., 2023). Recently, attempts have been made to improve behavioral parent training interventions to increase efficacy for children with CP + CU (Dadds et al., 2019; Kimonis et al., 2019; Waschbusch et al., 2019) with some promising results for high CU children evidenced in a recent meta-analysis (Perlstein et al., 2023). Based on previous research on the association between parenting practices and psychopathic traits (e.g., Waller et al., 2018), these tailored programs emphasized the importance of improving the quality of parent–child relationships by increasing sensitive and receptive parenting interactions (Kimonis et al., 2019). Aligning with prior recommendations of translating emotion-related evidence to prevention (e.g., Izard, 2002), some recent interventions have also incorporated an element of distress cue and emotion recognition training, with the goal of increasing socioemotional skills (e.g., empathy) in children with high CU traits. Interestingly, promising results have been reported in this regard (Fleming et al., 2022), even when considering attention deficits as the target of the intervention (Muñoz et al., 2021). Because previous research has suggested that parenting practices could be moderating the association between psychopathic traits and emotion recognition deficits (Levantini et al., 2022b), additional knowledge on these interactions would serve to continue refining parenting programs with socioemotional components specifically tailored to children and adolescents with psychopathic traits.

Yet, these promising implications at the practical level should also be interpreted with caution. Hence, some studies included in this review showed no deficits (e.g., Martin-Key et al., 2020), or even better recognition (e.g., Ezpeleta et al., 2017) in relation to psychopathic traits. This is particularly important within the (sub)clinical groups, with psychopathic traits, and not just CU, being linked with impairments in emotion recognition. However, high CU individuals tend to be the target of the current interventions intended to reduce child CP (Perlstein et al., 2023), whilst other psychopathic dimensions have been overlooked in intervention. Also, if psychopathic individuals show enhanced abilities for emotion recognition, as revealed but some studies in this review (e.g., Ezpeleta et al., 2017; Klapwijk et al., 2016; Martin-Key et al., 2018), this kind of interventions may come with some iatrogenic results. Improving the current knowledge on emotion recognition in relation to psychopathic personality, accounting for all its dimensions and potential moderators, would be decisive to keep moving forward at the practical level, with refinements in evidence-based programs that really account for deficits in this population.

Homogenizing research conditions and favoring replication studies could help in this purpose, allowing the establishment of more robust and clearer conclusions on the association between emotion recognition and psychopathic traits and, in turn, on the transfer of those results to the applied context.