Introduction

According to the World Health Organization (World Health Organization, 2019), approximately 26% of women worldwide have experienced severe violence from a male intimate partner at least once in their lifetime. To confront this challenge, research has dedicated significant efforts in recent decades to uncover the primary risk factors for intimate partner violence against women (IPVAW) perpetration (Capaldi et al., 2012). As a result, the affective domain has been recognized as a pivotal factor for IPVAW and its diverse manifestations (Birkley & Eckhardt, 2015; Expósito-Álvarez et al., 2023). Many IPVAW perpetrators report difficulties in identifying and expressing their own emotions, deficits defined as alexithymia (Romero-Martínez et al., 2021; Strickland et al., 2017). Furthermore, there is evidence of poor emotional regulation among this population (Marín-Morales et al., 2022), which has been associated with maladjusted internalization of their negative emotions (Maloney et al., 2022; Shorey et al., 2012). Accordingly, IPVAW perpetrators have shown an increased tendency to express anger externally and a reduced anger control during self-perceived distress situations (Birkley & Eckhardt, 2015; Sesar et al., 2018). These patterns of emotional processing have been associated with deficiencies in empathy, which involve the ability to understand and share another person's emotional experience (Capaldi et al., 2012; Ulloa & Hammett, 2016), two interrelated functions that facilitate social adjustment (Decety, 2015). Consequently, several authors have posited that empathic deficits could explain, at least partially, the difficulties observed in the behavioral regulation in this population (Godfrey et al., 2020; Leshem et al., 2019) and the increased risk of IPVAW (Romero-Martínez et al., 2022).

In this regard, sharing another person's emotional experience has been identified as a particularly important function for social adjustment (Stevens & Taber, 2021). An inadequate emotional response, whether absent or exacerbated, towards people experiencing negative emotions could evidence a poor engagement with others' feelings (Shamay-Tsoory, 2011). In IPVAW perpetrators, lower empathic concern (i.e., emotional response of compassion to others' suffering) have been described compared to controls (Romero-Martínez et al., 2021). Comes-Fayos et al. (2022) observed in IPVAW perpetrators an absent self-reported emotional response when witnessing others' suffering, unlike the non-violent group, which exhibited an increase in their negative mood state. Both studies suggest a diminished emotional response to others' suffering in IPVAW perpetrators. Additionally, Godfrey et al., (2020) found that a reduced self-reported emotional response to others' emotions was strongly linked to IPVAW perpetration. In fact, these authors observed a significant relationship between IPVAW and deficits in both empathy-related functions (i.e., to understand and share another person's emotional experience). However, their findings exposed that IPVAW inhibition might be more closely related to an adaptive emotional engagement than to the ability to cognitively adopt another person's perspective.

Nevertheless, studies focused on the emotional response of IPVAW perpetrators from a biopsychosocial perspective are scarce, especially the assessment of their emotional response to others' suffering. This fact is striking, given the high prevalence of biases in the affective domain among this population (e.g., low emotional self-perception, high social desirability or intentional manipulation; Romero-Martínez et al., 2021; Visschers et al., 2017). Therefore, there is an increasing emphasis on combining the psychosocial assessment of socioaffective functions in perpetrators with neuroscience techniques that provide data less susceptible to bias and manipulation (Pinto et al., 2010; Verdejo-Román et al., 2019). In this sense, the pattern of facial expressions in social situations has been recognized as a component of social behavior of special relevance, as it contains significant socio-affective information that influences emotional engagement and can foster empathic interactions (De Waal, 2008; Fridlund, 1996; Hess & Fischer, 2022; Weiß et al., 2021). Hence, analyzing the emotional facial response patterns of IPVAW perpetrators to others' emotions could offer valuable insights into understanding their socioaffective functioning, which is a recognized risk factor for IPVAW (Birkley & Eckhardt, 2015).

Emotional facial expressions are conceived as biological primed automatic responses to external emotional cues involved in communication (Ekman, 1993), and they have been associated with neural processes important for emotional responsiveness to others' emotions (Hess & Fischer, 2022). It has been postulated that individuals emotionally mimic when engaged in emotional processing tasks, suggesting that facial expressions may reflect individual affective response to emotional cues in social interactions (Seibt et al., 2015). In a general population study, an aligned facial response pattern towards others' negative emotions was observed in individuals with higher levels of affective empathy (Rymarczyk et al., 2016). In a recent meta-analysis, Holland et al. (2021) found a significant, albeit weak, relationship between facial mimicry and empathy. Although these authors stress some methodological requirements to robust this relationship, they conclude that people who exhibited greater facial mimicry tended to have higher empathy dispositions, and reported sharing others' emotional states to a greater degree.

In violent populations, the research is more limited and focuses mainly on adolescents with conduct problems and adults with callous-unemotional traits, characterized by a deficient affect and a disregard for others' feelings. The literature suggests that individuals with a higher prevalence of violent behaviors may exhibit reduced emotional mimicry when exposed to others experiencing negative emotions (de Wied et al., 2012; Fanti et al., 2016). Specifically, studies carried out with facial electromyography (EMG), a muscle measurement methodology, pointed to a deficit in general unpleasantness facial expression in response to others’ negative emotions among adolescents with conduct problems (Frick et al., 2014). In this line, de Wied et al. (2012) reported that violent adolescents exhibited lower facial responsiveness during a negative-valenced empathic induction task, along with other psychological markers of reduced emotional responsiveness (i.e., lower self-reported empathy). In adults, another EMG study also found a reduced general unpleasantness facial expression in violent individuals with callous-unemotional traits in response to violent films (Fanti et al., 2016).

Adding to this work, Fanti et al. (2017) assessed the emotional facial response pattern among violent individuals with callous-unemotional traits employing a multi-method design that combined facial EMG and the FaceReader software, a facial coding software that categorizes individuals’ facial expressions into six basic emotions. As a result, the FaceReader analyses revealed that individuals with high callous-unemotional traits displayed reduced facial reactions of sadness and disgust in response to violent films. Notably, these findings were accompanied by a decrease in task-unpleasant expression recorded by the facial EMG methodology, suggesting a correspondence in the results derived from both methodologies. These authors connected the restricted display of negative facial responses with low concern towards the negative feelings of victims of violence. Taking this into account, it could be hypothesized that reduced emotional mimicry exists among violent individuals, which might reflect a restricted emotional response pattern. However, to the best of our knowledge, no study has addressed the emotional facial expression in IPVAW perpetrators in the context of witnessing others' suffering, which may be particularly relevant to further understanding of socioaffective functioning in this population.

Notably, the importance of accurately recognizing emotions, also known as emotional decoding, has been underscored as a preliminary step to adjusting emotional responses (Feldman et al., 1991; Hess & Kafetsios, 2022), holding particular significance in making inferences about others' mental states and helping to modulate our affective responses (Decety & Moriguchi, 2007). In the general population, the Reading the Mind in the Eyes Test (RMET; Baron-Cohen et al., 2001), a well-established instrument that assesses the ability to decode emotions expressed through human eyes, has been associated with various empathy-related constructs, such as mentalizing (Samur et al., 2018), empathic accuracy (Djikic et al., 2013), and interpersonal sensitivity (Fong et al., 2013). In IPVAW perpetrators, lower RMET scores and impairments in emotional decoding have been consistently reported, with particularly poor recognition of fear and sadness (Babcock et al., 2008; Nyline et al., 2018; Romero-Martínez et al., 2016). Indeed, the impairment in emotional decoding has been linked to greater IPVAW acceptance and recidivism (Romero-Martínez et al., 2019, 2022). It has been theorized that deficits in the emotional decoding of IPVAW perpetrators would impede an adequate understanding of others' contextual needs, hindering conflict resolution through conciliatory strategies (Romero-Martínez et al., 2022; Sun et al., 2015). A possible explanation is that deficits in emotional decoding processes might affect IPVAW perspective taking by diminishing how accurately they can understand others' emotional state (Holland et al., 2021), which may result in difficulties when responding to the feelings of others.

Hence, the objective of this study was threefold. First, we aimed to examine the emotional facial response pattern, as well as self-reported emotions, of IPVAW perpetrators in response to a violence-focused empathic induction task compared to a control group. Watching people suffering is expected to induce specific negative emotions; however, violent individuals have exhibited reduced emotional mimicry when exposed to others experiencing negative emotions (de Wied et al., 2012; Fanti et al., 2017). Also, a reduced self-reported emotional responsiveness to others' emotions has been suggested in IPVAW perpetrators (Comes-Fayos et al., 2022; Godfrey et al., 2020). Therefore, IPVAW perpetrators were expected to exhibit reduced facial expressions of negative emotions and fewer self-reported negative emotions when exposed to victims of violence. Second, we intended to explore whether participants' registered emotional facial expressions were related to self-reported emotions. We expect that a higher prevalence of negative emotional facial expressions will be related to greater self-reported negative emotions, based on the previously established relationship between facial mimicry and emotional empathy (Holland et al., 2021). Finally, we aimed to explore differences in emotional decoding between IPVAW perpetrators and controls and investigate if emotional decoding predicted the affective facial responding during the task. Based on previous conclusions, we would presume worse emotional decoding in IPVAW perpetrators (Nyline et al., 2018; Romero-Martínez et al., 2016). Additionally, we expect low emotional decoding to predict the buffered emotional mimicry (Sun et al., 2015). It is also worth noting that group belonging would play a significant role in this scenario.

Materials and methods

Participants

The sample consisted of 103 healthy males (55 IPVAW perpetrators and 48 controls). IPVAW perpetrators were recruited from the CONTEXTO Program at the University of Valencia (Spain), a community-based intervention program for men convicted for gender-based violence for up to 2 years (suspended under mandatory program attendance; Lila et al., 2018). Within the Spanish legislative framework, IPVAW perpetrated by a male partner is legally conceptualized within the concept of gender-based violence, (Government of Spain (2004), Organic Law 1/2004 on Comprehensive Protection Measures against Gender-Based Violence). Requirements for IPVAW perpetrators included: having being sentenced for gender-based violence, the absence of mental or neurological disorders; and having Spanish writing and speaking skills. The assessment of IPVAW criminal records was carried out through the data provided by the Spanish General Secretariat of Penitentiary Institutions. Recruitment was completed prior to the intervention program, with the assurance that their refusal to participate would not affect their legal disposition.

The control group was recruited in Valencia (Spain) through social media advertisements. The advertisement outlined the study's characteristics, detailed the primary participation requirements, and provided a contact number for interested individuals. Eligibility criteria for the control group included being male with comparable sociodemographic characteristics to the IPVAW perpetrators (age, educational level, annual income, marital status, and region of birth), with no criminal history, the absence of mental or neurological disorders, and having Spanish writing and speaking skills.

To ensure compliance with inclusion criteria for both groups, each interested individual underwent an ad-hoc semi-structured clinical interview conducted by two trained psychologists. The purpose of this interview was to explore the participants' criminal history, detect possible psychopathological symptoms (e.g., depressive, manic, psychotic, and/or neurological symptoms), and ensure similar sociodemographic characteristics in the control group comparable to those of IPVAW perpetrators. Furthermore, in order to eliminate potential IPVAW behaviors in the control group, the "Revised Conflict Tactics Scale" (Straus et al., 1996) was administered to both groups. As a result, the control group did not exhibit any moderate to severe IPVAW behavior, as determined by the standards set by Straus et al. (1996) (see supplementary Table 1).

The experiment was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of the University of Valencia (procedure number: H1538385543901). All participants were informed about the specific details of the study, voluntarily agreed to participate, gave written informed consent and received a financial retribution of 40 euros at the end of the study.

Procedure

The procedure was carried out in a two-hour session in the Psychobiology laboratories of the University of Valencia, with a constant temperature (21 ± 1 °C) and humidity. First, the informed consent was signed and a questionnaire on participants' baseline mood state was completed (i.e., the State-Trait Anxiety Inventory [STAI]). Then, the Noldus FaceReader 6.1 software was programmed for the online recording of participants' facial expressions. For this purpose, the participants were asked to set a neutral expression to calibrate the facial emotion recording software, correcting for person-specific biases towards a certain facial expression and/or lighting conditions. Once calibrated, the online facial recording of the participants was initiated.

Afterwards, a Spanish validated battery of emotion-inducing videos (Fernández-Megías et al., 2011) was employed for the empathic induction task. For the creation and subsequent validation in the Spanish population of this battery of emotion-inducing clips, the scenes described in previous studies in the English and French populations were employed (Rottenberg et al., 2007; Schaefer et al., 2010). From this battery, four violence-focused clips were selected based on their high negative affect (mean length of 1′ 21", see Supplementary Table 2). Also, due to recommendations from several studies that point to differential processing in IPVAW perpetrators when processing IPVAW-specific content, we selected two scenes of IPVAW and two scenes of general violence (Gracia et al., 2015; Romero-Martínez et al., 2019). Prior to visualization, participants were instructed to actively empathize with the victim in each scene, as indicated previously. Once the empathic induction task was initiated, participants were required to complete the self-report emotion questionnaire right after each scene. Additionally, after completion of the self-report questionnaire, participants had to relax for one minute before viewing the next scene (for more details, refer to Comes-Fayos et al., 2022).

After completing the empathic induction task, participants underwent a 15-min recovery phase. Once the recovery period was over, information on sociodemographic variables was collected (i.e., age, marital status, level of education, annual income, and geographical region of birth) and the emotional decoding task (i.e., the RMET) was administered. To conclude, the participants received their remuneration (see Fig. 1).

Fig. 1
figure 1

Experimental procedure. Note. The experimental protocol was divided into five blocks. Block I: A habituation phase aimed at acclimatizing participants to the experimental environment. Block II: The setup phase. The Noldus FaceReader 6.1 software was programmed for the online recording of participants' facial expressions. Additionally, the "State-Trait Anxiety Inventory" questionnaire was administered to register participants' baseline mood state. Block III: The administration of the empathic induction task, performed through violence-focused emotion-eliciting videos. Block IV: After a 15-min recovery period, sociodemographic data were collected. Block V: The “Reading the Mind in the Eyes Test” was employed to acquire scores for emotional decoding. IPVAW: Intimate partner violence against women; GV: General violence; DEQ: Discrete Emotions Questionnaire. Images adapted from FaceReader [Photograph], by Noldus Information Technology, 2012, Horizon (https://horizon.ac.uk)

Full size image

Measures

Facial expression analyses

To register participants’ emotional facial expressions, we employed the Noldus FaceReader 6.1 software (Noldus Information Technology, 2015). The FaceReader 6.1 is a program for facial analysis that can detect the emotional valence and identify Ekman’ six basic emotions: sadness, anger, disgust, fear, surprise and happiness (as well as a neutral state). FaceReader can recognize facial expressions with an accuracy of 90%, correcting for person-specific biases such as the person's gender, age, ethnicity, the amount of facial hair and whether the person is wearing glasses or not. FaceReader classifies faces in three consecutive steps: (1) "Face finding" (the position of the face is found using a specific algorithm developed for finding faces in images), (2) "Face modeling" (a model-based method is used to synthesize an artificial face model that will serve as a foundation for analyzing and integrating individuals' facial gestures) and (3) "Face classification" (the actual classification of the facial expressions, based on the six basic emotions described by Ekman). Emotional facial expressions were calculated for each scene of the empathic induction task based on the number of seconds a specific emotion was registered in the FaceReader 6.1 software.

Self-report emotions

The Discrete Emotions Questionnaire (DEQ) is a categorical self-report instrument designed to examine specific emotions. It is an adapted version of the "Post-Film Questionnaire" employed in the study conducted by Rottenberg et al. (2007). The questionnaire consists of 18 items, corresponding to different emotional labels, including: amusement, anger, anxiety, confusion, satisfaction, disgust, fear, guilt, happiness, interest, joy, tenderness, pride, sadness, embarrassment, surprise, unhappiness, and shame. In order to avoid an increase in type 1 error and to homogenize the self-reported emotions with those registered in the FaceReader software, Ekman's six basic emotions (sadness, anger, disgust, fear, surprise and happiness) will be taken into account (Ekman, 1993). Additionally, tenderness will also be considered due to its relevance to empathy (Kalawski, 2010). A Likert scale ranging from 1 (no emotion) to 7 (most intense emotion) scores the self-perceived emotion for each scene. Cronbach’s α was 0.94.

Emotional decoding

The RMET is a facial emotion recognition task that assesses the ability to decode emotions based on the eye region of the face photographs (Baron-Cohen et al., 2001). Participants are presented with 36 black-and-white photographs of both men and women and are asked to attribute their mental state by selecting one of four words that best describe their emotional state. Each photograph focused exclusively on the individuals' eye region and was surrounded by four single-word emotional state descriptors (e.g., nervous, guilty), one at each corner. One of these descriptors was aligned with the mental state depicted in the photograph, while the remaining descriptors served as distractors. Participants had a single attempt, and could take as much time as required for each item. Scores are calculated based on the total number of correct choices across all 36 photographs. Cronbach’s α was 0.62 for the total score.

State anxiety

The STAI is a validated questionnaire used to measure both trait and state anxiety (Spielberger, 1983). In the present study, we employed the questionnaire form focused on "state anxiety" (hereafter referred to as STAI-S). The STAI-S consists of 20 items, with instructions that require participants to indicate their anxiety-related at a specific moment. Each item is rated on a 4-point scale. Scores on the STAI-S range from a minimum of 20 to a maximum of 80, and scores are typically categorized as follows: “no or low anxiety” (20–37), “moderate anxiety” (38–44), and “high anxiety” (45–80). Cronbach’s α was 0.89 for the STAI-S total score.

Data analysis

After testing the normality of the data with the Shapiro–Wilk test (p < 0.05), the data for recorded emotional facial expressions and self-reported emotions were transformed using a square root transformation (sqrt). Afterwards, t-tests were performed for assessing group (IPVAW perpetrators and controls) differences for age, state anxiety, registered emotional facial expressions, self-reported emotions and emotional decoding scores. To further explore participants' emotional response to the task, analyses of recorded emotional facial expressions and self-reported emotions were conducted for the entire empathic induction task and splitting the task into IPVAW versus general violence scenes. A chi-squared analysis was also performed for marital status, education and income of participants. Cohen's d and Cramer’s V were used as the effect size measure for t-test and chi-square analyses (values of 0.10, 0.30, and 0.50 were considered small, moderate, and large effect sizes, respectively; Cohen, 1988). Additionally, Pearson’s correlation coefficients were calculated to assess the relationship between registered emotional facial expressions and self-reported emotions for all participants. This is intended to provide insight into the relationship between the two measures of emotional response. Finally, a linear regression model was constructed by jointly using the total scores of RMET and group (dummy coded as 0 for IPVAW perpetrators and 1 for controls) as the independent variables and the registered facial expressions as the dependent variable. Data analyses were conducted using IBM SPSS Statistics for Windows, version 28.0 (Armonk, NY). Values of p < 0.05 were considered statistically significant. The average values are reported in the tables as mean ± Standard deviation (SD).

Results

Sociodemographic characteristics and baseline mood state of the participants

There were no significant differences between groups in age [t (101) = 1.680, p = 0.096; d = 0.33], marital status (χ2 = 3.547, p = 0.060; V = 0.19), level of education (χ2 = 5.856, p = 0.119; V = 0.24), annual income (χ2 = 1.708, p = 0.426; V = 0.13), or geographical region of birth (χ2 = 2.045, p = 0.563; V = 0.14; see Table 1). No significant differences were found in STAI-S scores between groups [t (101) = 0.593, p = 0.555; d = 0.33; IPVAW perpetrators, M = 34.58, SD = 9.75; control group, M = 33.56, SD = 7.33].

Table 1 Means, standard deviations, percentages, and means comparisons for sociodemographic data for all groups
Full size table

Registered emotional facial expression

The analysis performed for emotional facial expressions revealed that IPVAW perpetrators reported lower registration of sad facial expression (p < 0.001) and higher registration of happy facial expression (p < 0.001), compared to controls. Furthermore, there was a tendency towards a significance to higher prevalence of angry facial expressions in IPVAW perpetrators (p = 0.062; see Fig. 2 and Table 2).

Fig. 2
figure 2

Means comparisons for registered emotional facial expression for both groups. Note: IPV: Intimate Partner Violence Against Women. Statistical significance, ***p < 0.001

Full size image
Table 2 Mean comparisons for registered emotional facial expression for both groups
Full size table

When facing IPVAW scenes, IPVAW perpetrators also differed in their sad facial expression (p < 0.001) and happy facial expression (p < 0.001), as well as a tendency in angry facial expression (p = 0.066), with IPVAW perpetrators reporting a lower prevalence of sad expressions and higher prevalence of happy and angry expressions than controls. As for general violence scenes, analysis revealed similar results with lower prevalence of sad expressions (p < 0.001) and higher prevalence of happy expressions (p < 0.001) in IPVAW perpetrators compared to controls (see Table 2).

Self-reported emotional state

As for self-reported emotions, the analysis informed that IPVAW perpetrators differed in their self-report level of sadness (p = 0.047), anger (p = 0.004), disgust (p = 0.024), and tenderness (p = 0.045), revealing lower self-report emotional levels after the empathic induction task, compared to the control group (see Table 3).

Table 3 Means comparisons for self-reported emotional state for both groups
Full size table

When dividing the scenes by types of violence, the analysis exhibited lower levels of sadness (p = 0.045), anger (p = 0.016), and tenderness (p = 0.035) in IPVAW perpetrators compared to the control group in response to IPVAW scenes. With respect to general violence, lower levels of self-reported anger (p = 0.002) and disgust (p = 0.019), as well as a trend toward significance in self-reported sadness (p = 0.061), were also found in IPVAW perpetrators in comparison to controls (see Table 3).

Relationships between self-reported emotions and registered facial emotions

With regard to the relationship between the self-reported emotional state and the registered emotional facial expression during the task, after controlling the group effect, a partial correlation analysis showed that there was a significant correlation between sad facial expressions and self-reported sadness (r = 0.228, p = 0.022), self-reported fear (r = 0.228, p = 0.022), self-reported tenderness (r = 0.260, p = 0.009). There was no other significant correlation.

Differences in emotional decoding and predictive capacity on facial expressions

IPVAW perpetrators differed from controls in their RMET total scores [t (101) =—3.581, p = 0.001, d = 0.71], with IPVAW perpetrators exhibiting lower scores than controls (IPVAW perpetrators, M = 19.58, SD = 3.95; control group, M = 22.38, SD = 3.95).

Regarding the regression analysis, the RMET total score and the group variable predicted 11.8% of the variance for sad facial expression during the empathic induction task (adj R2 = 0.118, F (2, 102) = 7.801, p = 0.001). The association was no significant for the RMET total score (β = 0.061, p = 0.539). However, a significant association was found for the group variable (β = 0.342, p = 0.001), indicating that belonging to the control group predicted the appearance of sad facial expressions.

Additionally, the model predicted 14.5% of variance for happy facial expressions during the task (adj R2 = 0.145, F [2, 102] = 9.627, p = 0.000). The association was significant for the RMET total score (β =—0.262, p = 0.008) and the variable group (β =—0.230, p = 0.020), suggesting that both lower RMET total score and belonging to the IPVAW group and predicted happy facial expressions in the task.

Discussion

Research has indicated that difficulties in emotional engagement with others' negative emotions emerge as an important risk factor for IPVAW, as it can hinder empathy-based social interactions. Our current findings reveal that IPVAW perpetrators exhibited a distinctive emotional response pattern when observing others' suffering, characterized by lower display of sadness and a higher display of happiness in their facial expressions, along with lower self-reported emotional responses. Furthermore, the lower prevalence of sadness facial expression during the task was associated with lower self-reported sadness, fear and tenderness, regardless of group. Notably, both poorer emotional decoding and IPVAW group belonging predicted lower prevalence of sadness facial expression and higher prevalence of happiness expression during the task.

Supporting our first hypothesis, we observed a reduced tendency to display sad facial expressions and an increased inclination toward happy facial expressions among IPVAW perpetrators when viewing violent scenes. This finding is particularly significant, as facial expressions have been posited to reflect emotional responses to socioaffective signals (Ekman, 1993; Hess & Fischer, 2022; Seibt et al., 2015), with several authors establishing a connection between heightened facial mimicry and a greater degree of empathy (Holland et al., 2021; Rymarczyk et al., 2016). Based on this assumption, the control group, by exhibiting greater facial mimicry in response to others' suffering, appears to more extensively share the emotional states of the victims during the task, thus demonstrating an enhanced empathic response. Conversely, the diminished facial mimicry observed in IPVAW perpetrators would support a potential misaligned emotional response to others' suffering. Our results are consistent with prior studies on violent individuals that have shown reduced facial mimicry in response to others' negative emotions, linking it to emotional disengagement (Fanti et al., 2016, 2017; Frick et al., 2014). Thus, the pattern of emotional facial expression of IPVAW perpetrators might indicate a lack of emotional response when observing others' negative emotions, potentially hindering their socioaffective and behavioral regulation (Drimalla et al., 2019; Kyranides et al., 2022). Additionally, IPVAW perpetrators exhibited a tendency towards angry expressions specifically to IPVAW scenes, an expression that has been associated with poor emotional regulation (Fanti et al., 2017). This aligns with prior research indicating a greater tendency to express hostility in IPVAW perpetrators when confronted with IPVAW-related content (Birkley & Eckhardt, 2015; Gracia et al., 2015). Taken together, the current findings reinforce and extend the perspective of facial mimicry as a significant emotional response that influences both adaptive and maladaptive social interactions (Drimalla et al., 2019; Hess & Fischer, 2013).

Both groups also differed in their self-report emotional response. Compared with controls, IPVAW perpetrators were found to exhibit lower self-reported sadness, anger and tenderness during the task. The findings remained consistent when comparing IPVAW and general violence scenes. As stated, an adjusted emotional engagement requires the generation of an emotional response that aligns with what is observed in another person (Shamay-Tsoory, 2011). Kalawski (2010) specifically related emotional engagement with others' suffering to the experience of sadness and tenderness. According to this author, sadness would be congruent with the perceived negative state, while tenderness would provide an other-oriented emotional component. Similar to the results observed in our control group, prior research has reported an increase in self-reported sadness and compassion following negative-valence emotional videos among the general population, linking it with higher empathy (Barraza & Zak, 2009; Stellar et al., 2020). On the other hand, previous studies had documented a lack of self-reported emotional responsiveness to others' suffering in IPVAW perpetrators (Comes-Fayos et al., 2022; Romero-Martínez et al., 2021). Therefore, both the emotional facial response pattern and self-reported emotions provide further evidence of specific alterations in socioaffective functions relevant to empathy in IPVAW perpetrators. Notably, there was a divergence in the results associated with the anger response to the task. While a tendency toward angry facial expressions was registered in IPVAW perpetrators during the IPVAW scenes, they informed lower self-reported anger throughout the entire task. These results could be grounded in two emotional patterns previously found among IPVAW perpetrators: an increased tendency to express anger externally during distressing situations (Birkley & Eckhardt, 2015; Sesar et al., 2018) and a high prevalence of alexithymic traits (Comes-Fayos et al., 2022; Romero-Martínez et al., 2021; Strickland et al., 2017). Indeed, this alexithymic symptomatology would be also consistent with the overall impoverished self-reported emotional responsiveness to others ‘ suffering found in IPVAW perpetrators.

In addition, sad facial expressions were positively correlated with self-reported sadness, fear, and tenderness, regardless of group. The importance of sadness and tenderness has been particularly emphasized for adjusted emotional engagement (Kalawski, 2010; Shamay-Tsoory, 2011). On the other hand, the facial expression of sadness has also been associated with both empathy and compassion (Baránková et al., 2019; Drimalla et al., 2019). Hence, it seems particularly relevant that, in the context of an empathic induction, the most notable association was observed between these variables. This might suggest that a consistent correspondence between sad facial expressions and self-reported sadness and tenderness could serve as an indicator of aligned emotional engagement to others' negative emotions. In contrast, no significant relationship was found between happy facial expressions and self-reported emotions, although a higher prevalence of this expression was observed in IPVAW perpetrators. One possible explanation for the absence of this relationship could be the susceptibility of self-reported questionnaires to bias (e.g., social desirability or manipulation), which could prevent the recognition of happiness in the presence of others' suffering (Romero-Martínez et al., 2016; Visschers et al., 2017). However, further investigation on the subject is required.

Finally, as expected, IPVAW perpetrators exhibited lower RMET scores compared to the control group. This result is in line with previous studies, indicating the presence of deficits in emotional decoding in IPVAW perpetrators, especially when recognizing others' facial expressions (Babcock et al., 2008; Nyline et al., 2018; Romero-Martínez et al., 2016). The social information processing theory has provided a framework for how social cues may be decoded in IPVAW contexts (Murphy, 2013). Adaptive social behavior relies on effective emotional decoding that enables interpretation of others' behavior, which helps formulate a coherent response in order to achieve healthy relational goals. Thus, dysfunctional processing of social information, including a reduced ability to recognize others' facial expressions, could partly explain the reduced ability to address conflict through conciliatory mechanisms, predisposing to IPVAW (Romero-Martínez et al., 2019, 2022).

Congruently, having worse RMET scores and belonging to the IPVAW group predicted lower sad facial expressions and higher happy expressions. Facial emotion recognition is thought to influence emotional empathy through its perceptual inputs and inferential processes (Holland et al., 2021). Consequently, deficiencies in emotional decoding are likely to imply poor emotional processing and responding (Feldman et al., 1991; Hess & Kafetsios, 2022). Thus, it seems logical that inadequate emotional interpretation of emotional cues (poor emotional decoding) potentially accounted for a mismatched emotional facial response during the empathic induction task, which may be indicative of maladaptive emotional responsiveness. In parallel, prior research has provided evidence that an inappropriate affective response to others' emotions is strongly linked to IPVAW (Godfrey et al., 2020). In fact, it has been underscored that a deficit in emotional decoding processes, together with other criminogenic characteristics of IPVAW perpetrators, may result in difficulties when responding to the feelings of others (Romero-Martínez et al., 2016). Therefore, our findings seem to be consistent with this literature and suggest that the distinct facial response pattern found in this study could be better understood due to the lower emotional decoding scores reported in IPVAW perpetrators (Murphy, 2013; Nyline et al., 2018).

This study provides novel information about the emotional response pattern of IPVAW perpetrators. However, it is not without limitations, which are outlined as follows. First, the methodology employed to induce empathy does not account for control videos or an empathy-related questionnaire, which could reduce the validity of our findings. Despite this, it is noteworthy to highlight that the clips employed in our task have proven effective in separately eliciting negative affect (Fernández-Megías et al., 2011). Moreover, a comparable design has previously been employed to assess the participants' response to negative emotional stimuli, including facial expressions registration (Stellar et al., 2020). Secondly, participants were instructed to complete the RMET task at the conclusion of the experimental protocol. While it is true that Block IV (i.e., the recovery period and sociodemographic data collection) falls between both tasks, the empathic induction may potentially influence the RMET performance. Therefore, the findings regarding emotional decoding should be approached with due caution. Another limitation might be the inclusion of IPVAW perpetrators as a unique group. IPVAW is a heterogeneous behavior with several potential mediators, including personality traits or the severity of violence. However, socioaffective functioning has been recognized as a core, cross-cutting factor for IPVAW (Birkley & Eckhardt, 2015; Expósito-Álvarez et al., 2023). Even so, further research is needed to replicate these results in diverse samples of IPVAW perpetrators. Finally, our experimental design lacks a multi-method approach that integrates facial EMG. EMG is a traditional measurement for studying emotional facial response in violent populations (Fanti et al., 2016; de Wied et al., 2012). Nevertheless, FaceReader is a reliable facial coding software that accurately records emotional facial expressions and strongly correlates with EMG outputs (Fanti et al., 2017).

In future research, it would be valuable to replicate the current results by using EMG in conjunction with other physiological indicators of emotional reactivity, such as autonomic nervous system activity, and employing empathic induction stimuli that integrate positive and neutral affective stimuli. Due to the inherent limitations of each method, adopting a multimodal assessment approach to an empathic induction task will offer a more comprehensive understanding of variations in emotional response patterns among IPVAW perpetrators. Additionally, it is also necessary to demonstrate whether these results are cross-cutting for all IPVAW perpetrators and whether they could be extended to other offenders. It would be particularly valuable to expand the current design to include additional samples, such as IPVAW perpetrators with severe prison sentences, men convicted of other crimes and female offenders. Finally, exploring the impact of IPVAW intervention programs on their affective response to an empathic induction task, as well as its long-term stability, could provide a deeper understanding of the impact of addressing emotional processing in IPVAW risk and recidivism.

Finally, this study may have important applications for practice. It has been underscored that intervention programs that address well-established IPVAW risk factors report greater efficacy (Eckhardt et al., 2013). Our findings are consistent and broaden the understanding of maladaptive emotional processing as a potential criminogenic factor in IPVAW perpetrators (Brem et al., 2018; Spencer et al., 2022), with a particular emphasis on poor emotional engagement and decoding. By delving into the emotional processing patterns of IPVAW perpetrators, valuable insights can be obtained to develop strategies in IPVAW intervention programs. These strategies could include actions such as raising awareness of the impact of inappropriate emotional responses or enhancing emotional decoding skills. This approach would be in line with the Good Lives Model (Santirso et al., 2020), which stresses that intervention programs should focus not only on minimizing manifest risk factors, but also on supporting protective factors.

In conclusion, the results of the current study suggest that IPVAW perpetrators display a misaligned facial expression pattern, and stated less prevalence of self-reported emotionality, when viewing other people suffering. Hence, our results highlight an alternative affective response among IPVAW perpetrators that does not accurately reflect the observed affective state of others. Considering these findings, the variation in their facial expressions could be relevant when adapting interventions. For example, IPVAW perpetrators could benefit from interventions that emphasize emotional processing, focusing on adjusting their emotional response and enhancing their emotional decoding abilities. Ultimately, enhancing socioaffective functioning may serve as a mechanism to reduce the incidence of IPVAW.

CRediT – Contributor roles

Project administration

Participating authors: Javier Comes Fayos, Ángel Romero Martínez and Luis Moya Albiol

Conceptualization and Methodology

Participating authors: Javier Comes Fayos, Ángel Romero Martínez, Manuela Martínez Ortíz, Marisol Lila Murillo and Luis Moya Albiol

Resources

Participating authors: Javier Comes Fayos, Ángel Romero Martínez, Manuela Martínez Ortíz, Marisol Lila Murillo and Luis Moya Albiol

Search and sample selection

Participating authors: Javier Comes Fayos and Marisol Lila Murillo

Investigation and acquisition of data

Participating authors: Javier Comes Fayos and Ángel Romero Martínez

Data curation

Participating authors: Javier Comes Fayos, Ángel Romero Martínez and Luis Moya Albiol

Formal analysis

Participating authors: Javier Comes Fayos, Ángel Romero Martínez and Luis Moya Albiol

Writing—original draft

Participating authors: Javier Comes Fayos, Ángel Romero Martínez and Luis Moya Albiol

Writing—review & editing

Participating authors: Javier Comes Fayos, Ángel Romero Martínez, Manuela Martínez Ortíz, Marisol Lila Murillo and Luis Moya Albiol