همدلی و تشخیص حالت چهره از احساسات در متجاوزان جنسی، مجرمان غیر جنسی و افراد بهنجار

Abstract Research conducted on empathy and emotional recognition in sex offenders is contradictory. The present study was aimed to clarify this issue by controlling for some affective and social variables (depression, anxiety, and social desirability) that are presumed to influence emotional and empathic measures, using a staged multicomponent model of empathy. Incarcerated sex offenders (child molesters), incarcerated non-sex offenders, and non-offender controls (matched for age, gender, and education level) performed a recognition task of facial expressions of basic emotions that varied in intensity, and completed various self-rating scales designed to assess distinct components of empathy (perspective taking, affective empathy, empathy concern, and personal distress), as well as depression, anxiety, and social desirability. Sex offenders were less accurate than the other participants in recognizing facial expressions of anger, disgust, surprise and fear, with problems in confusing fear with surprise, and disgust with anger. Affective empathy was the only component that discriminated sex offenders from non-sex offenders and was correlated with accuracy recognition of emotional expressions. Although our findings must be replicated with a larger number of participants, they support the view that sex offenders might have impairments in the decoding of some emotional cues conveyed by the conspecifics' face, which could have an impact on affective empathy.

. Introduction The construct of empathy has received much attention in different domains of research (e.g., psychology, ethology, cognitive neuroscience, and psychiatry) (see Eisenberg and Strayer, 1987, Ickes, 1997, Preston and de Waal, 2002, Decety and Jackson, 2004 and Blair, 2005). Although there is little consensus among theorists regarding its definition and its constitutive components, most agree that empathy is a multidimensional phenomenon that involves an understanding of a person's subjective experience, a perspective taking, and a vicarious sharing of emotional states in response to another's affective cues, often resulting in feelings of concern or compassion for this person (i.e., sympathy), although it also can lead to self-oriented feelings (i.e., personal distress) (Davis, 1983, Hoffman, 1984 and Decety and Jackson, 2004). Although it is unclear whether all the components are required for the unfolding of the empathetic process, it is usually acknowledged that they play important functions in social bonds (Anderson and Keltner, 2002), prosocial behaviors (Eisenberg and Miller, 1987), altruism (Batson, 1991) and moral judgment (Hoffman, 1987). In contrast, empathy impairments have been related to aggressive, delinquent and antisocial behaviors (Miller and Eisenberg, 1988, Jolliffe and Farrington, 2004 and Lovett and Sheffield, 2007), and deficits in specific components of empathy have been proposed as a feature of autism or psychopathy (Blair, 2005). The inverse relationship typically found between empathy and aggression led a number of researchers and clinicians to suggest that sex offenders (e.g., child molesters, rapists) are likely deficient in their capacity to empathize with their conspecifics (e.g., Marshall and Barbaree, 1990). More specifically, it was assumed that because of a lack of compassion for the victim's distress, a child molester or a rapist does not inhibit an attack when he becomes sexually aroused toward a potential victim. However, empirical research on empathy in sexually aggressive persons conducted during the last three decades provided inconsistent findings (for reviews, see Marshall et al., 1995, Geer et al., 2000 and Covell and Scalora, 2002). Some studies reported lower levels of generalized empathy or for some of its components (e.g., empathic concern, perspective taking) in sexual offenders compared with non-sexual offenders (Rice et al., 1994, Lisak and Ivan, 1995, Lindsey et al., 2001 and Marshall and Moulden, 2001) or their non-offending counterparts (Hudson et al., 1993: study 2; Chaplin et al., 1995 and Burke, 2001), whereas other studies have found no differences between sexual offenders and other groups of participants (Hoppe and Singer, 1976, Langevin et al., 1988 and Hudson et al., 1993: study 1; Hanson and Scott, 1995), or showed empathy deficits in sex offenders only in specific situations or toward their own victims (Pithers, 1999 and Fernandez and Marshall, 2003). Marshall et al. (1995) proposed an appealing staged multicomponent model of empathy that was designed to examine potential deficiencies in sexual offenders at each stage of the process. According to this model, the ongoing process of empathy involves four discriminable steps: (1) emotional recognition, (2) perspective taking, (3) emotion replication, and (4) response decision. Stage 1, the emotion-recognition stage, is needed to decode or read emotional signals of others and is considered necessary for the unfolding of subsequent stages. Such signals, more specifically those conveyed by the face, constitute a primary way for communicating basic emotional states or current intentions and play a powerful role in the regulation of social interactions (e.g., Fridlund, 1994 and Ekman and Rosenberg, 2005). Stage 2, perspective taking, which is the ability to understand another person's viewpoint, is thought to be an essential component in the mechanisms that account for the theory of mind. Stage 3, emotion replication, involves some level of sharing or similarity in the feelings experienced by self and other. Finally, the last stage, response decision, refers to the observer's decision to exhibit or not, on the basis of his/her feelings, a socially oriented behavior (e.g., helping or comforting behaviors). The present study aims to focus on the first step of empathy process (recognition of emotional cues) in different groups of participants (sex offenders, non-sex offenders, and normal controls) and investigates the relationship between the ability to recognize emotional facial expressions and self-reported empathy. The few available findings suggest that sex offenders might be less accurate than controls or non-sex offenders in recognizing the emotional states of others (Hudson et al., 1993 and Lisak and Ivan, 1995), although some negative findings have also been reported in the form of dissertation abstracts (Franklin, 2000 and Cuevas, 2004). To date, the most detailed findings reported by Hudson et al. (1993) provided evidence that sexually aggressive men (child molesters and rapists) were less accurate than were violent non-sex offenders at recognizing emotional expressions, with confusions between fear and surprise on the one hand, and disgust and anger on the other hand. Such confusions between fear and surprise, in return, have been hypothesized to lead sex offenders to have difficulties in reading the distress in the “eyes” of another. However, because sex offenders often display affective disturbances (i.e., anxiety and depression) (Becker et al., 1991, Ahlmeyer et al., 2003 and Dunsieth et al., 2004), which were not controlled for in previous studies, it is unclear whether these negative effects have contributed to the variance of data and, thus, may explain inconsistency among findings. More specifically, because substantial evidence indicated that depression or anxiety in non-offenders may bias the recognition of emotional facial expressions (Gur et al., 1992, Surguladze et al., 2004 and Montagne et al., 2006), it is important for studies examining emotion recognition in sex offenders to partial out the effects of depression and anxiety. In the present study, such variables were entered as covariates in a design examining the ability of sex offenders (child molesters) to decode facial expressions of basic emotions compared with non-sex offenders and normal controls. Furthermore, because self-report measures may be affected by self-presentation, and because socially desirable responding has been reported in studies of sex offenders (Curwen, 2003), various scales presumed to assess distinct constitutive components of empathy (perspective taking, affective empathy, empathy concern, personal distress) entered self-reported social desirability as a covariate. By partialling out potential confounding effects such as depression, anxiety, and social desirability, we expected that our study should provide a more stringent assessment of the empathy process in both sex and non-sex offenders. The ability to decode distinct intensity levels of emotional expressions was also investigated in order to assess whether sex offenders were less accurate in recognizing emotions when affective cues expressed by others were reduced. Finally, by examining relationships between recognition accuracy of emotions and distinct components of empathy in offenders and non-offenders, our study was also designed to explore the relevance of the multidimensional staged process of empathy suggested by Marshall et al. (1995).

3. Results 3.1. Depression, anxiety and social desirability and their relations with other measures An analysis of variance (ANOVA) revealed that the sex offenders (SO) reported higher depression scores (M = 25.3, S.D. = 4.14) than the other participants (NSO: M = 13, S.D. = 2.21; controls: M = 0.9, S.D. = 1.28), F (2, 27) = 188.67, P < 0.001, with the non-sex offenders having more depressive mood than the controls (P < 0.001, Bonferroni test). The SO group also obtained higher S-anxiety scores (M = 67.9, S.D. = 3.93) than did the other groups (NSO: M = 61, S.D. = 2.79; controls: M = 28.4, S.D. = 3.50), and the NSO were more anxious than the controls [F(2, 27) = 376.38 and Bonferroni test, P < 0.0001]. When social desirability was entered as a covariate, the pattern of findings between the groups of participants did not change for depression and anxiety. BDI scores correlated significantly (P < 0.002) across groups with the following empathic measures: rs (28) = − 0.92 for EQ, − 0.92 for ES-IVE, − 0.93 for PT, − 0.55 for FS, − 0.93 for EC, and 0.83 for PD. Similarly, strong significant correlations were found between empathic measures and S-STAI scores (P < 0.0001): rs (28) = − 0.93 for EQ, − 0.82 for ES-IVE, − 0.91 for PT, − 0.70 for FS, − 0.91 for EC, and 0.71 for PD. Social desirability also correlated significantly (P < 0.01) with the following measures: rs (28) = 0.81 for EQ, 0.66 for ES-IVE, 0.79 for PT, 0.49 for FS, 0.78 for EC, and − 0.67 for PD. Furthermore, an ANOVA carried out on scores of social desirability revealed a main effect of Group, F(2, 27) =27.84, P < 0.0001. Controls had higher scores of social desirability (M = 23.3, S.D. = 3.62) than the other groups of subjects (P < 0.001, Bonferroni tests), while no significant difference was detected between sex offenders (M = 11.7, S.D. = 4) and non-sex offenders (M = 15.2, S.D. = 3.08), P > 0.05. 3.2. Accuracy rates of emotional recognition Table 1 shows mean rates of accuracy as a function of emotion category, group and intensity level of facial expressions. The MANCOVA indicated a significant main effect of BDI covariate on the accuracy rates of emotional recognition, F(1, 25) = 4.04, P = 0.05, ηp2 = 0.14. A main effect for Group was also found, F(2, 25) = 47.76, P < 0.0001, ηp2 = 0.79, indicating that the sex offenders (M = 0.51, S.D. = 0.096) were less accurate than the non-sex offenders (M = 0.93, S.D. = 0.03) and the controls (M = 0.96, S.D. = 0.03) for recognizing emotional expressions (Bonferroni test, P < 0.0001). Significant Emotion × Group, F(10, 44) = 4.16, P < 0.0001, ηp2 = 0.49, and Emotion × Intensity × Group interactions were detected, F(20, 34) = 2.56, P = 0.008, ηp2 = 0.60. One-way ANCOVAs conducted on each emotional expression showed that these effects mainly reflect the fact that the sex offenders recognized less accurately disgust, F(2, 25) = 65.83, P < 0.0001, ηp2 = 0.84, fear, F(2, 25) = 29.13, P < 0.0001, ηp2 = 0.70, and anger expressions, F(2, 25) =10.22, P = 0.001, ηp2 = 0.45, than did the other groups of participants. They also decoded less accurately surprise, F(2, 25) = 3.75, P = 0.04, ηp2 = 0.23, than the non-sex offenders (Bonferroni test, P < 0.05). Furthermore, the sex offenders were less accurate in decoding mild levels of disgust F(2, 25) = 5.75, P = 0.009, ηp2 = 0.315, and fear expressions, F(2, 25) = 4.12, P = 0.03, ηp2 = 0.25, than were the non-sex offenders (Bonferroni tests, all P < 0.05). It should be noted that the effect sizes are large, more particularly when the significant effects concerned the three groups of participants (i.e., see effect sizes for fear and disgust), suggesting that the effects can be accepted with confidence. Table 1. Accuracy rates of emotion recognition as a function of group (SO = sex offenders, SNO = non-sex offenders, controls = non-offenders controls) and intensity levels (40% = mild, 70% = moderate, 100% = strong) of facial expressions Emotion intensity level (%) Group SO NSO Controls Total Mean S.D. Mean S.D. Mean S.D. Mean S.D. Neutral 0.82 0.13 0.86 0.12 0.94 0.08 0.87 0.12 Anger (40%) 0.70 0.26 1 0 1 0 0.90 0.20 Anger (70%) 0.75 0.35 1 0 1 0 0.92 0.23 Anger (100%) 0.90 0.21 1 0 1 0 0.96 0.13 Disgust (40%) 0.05 0.16 0.55 0.16 0.65 0.24 0.43 0.32 Disgust (70%) 0 0 1 0 1 0 0.67 0.48 Disgust (100%) 0.05 0.16 1 0 1 0 0.68 0.46 Fear (40%) 0.10 0.21 0.75 0.26 0.95 0.16 0.60 0.42 Fear (70%) 0 0 1 0 1 0 0.67 0.48 Fear (100%) 0.10 0.32 1 0 0.95 0.16 0.68 0.46 Happiness (40%) 0.90 0.21 1 0 1 0 0.97 0.13 Happiness (70%) 1 0 1 0 1 0 1 0 Happiness (100%) 1 0 1 0 1 0 1 0 Sadness (40%) 0.65 0.42 0.80 0.26 0.90 0.32 0.78 0.34 Sadness (70%) 0.60 0.46 1 0 1 0 0.87 0.32 Sadness (100%) 0.95 0.16 1 0 1 0 0.98 0.09 Surprise (40%) 0.40 0.32 0.80 0.26 1 0 0.73 0.34 Surprise (70%) 0.55 0.44 0.95 0.16 0.95 0.16 0.82 0.33 Surprise (100%) 0.50 0.41 0.95 0.16 0.95 0.16 0.80 0.34 Table options Significant effects for Group, F(2, 25) = 27.49, P < 0.001, ηp2 = 0.69, and Emotion × Group interaction were found for the emotional sensitivity index (d′), F(10, 44) = 3.01, P = 0.006, ηp2 = 0.41, with sex offenders being less sensitive (M = 0.041, S.D. = 0.78) than the other groups (controls: M = 4.18, S.D. = 0.49; NSO: M = 3.59, S.D. =0.47). One-way ANCOVAs, using Group as the between-subjects factor, were also conducted on d′ to specific target faces (e.g., surprise responses to fear faces). The analyses revealed that sex offenders falsely identified disgust expressions as anger expressions, F(2, 25) =22.633, P < 0.001, ηp2 = 0.64 (controls: M = 3.55, S.D. =0.87; NSO: = 3.00, S.D. = 0.87; SO: M = − 2.98, S.D. =1.32). Furthermore, sex offenders provided more false surprise responses when viewing fear faces, F(2, 25) =15.39, P < 0.001, ηp2 = 0.55 (controls: M = 4.33, S.D. =1.03; NSO: M = 3.42, S.D. = 1.36; SO: M = − 3.87, S.D. =1.74). No differences were detected between non-sex offenders and normal controls (P > 0.05, Bonferroni test). The bias index (C) submitted to the MANCOVA indicated an effect for depression, F(1, 25) = 6.22, P = 0.02, ηp2 = 0.20), and for Emotion × Group interaction, F(10,44) = 2.06, P = 0.049, ηp2 = 0.32). The ANCOVAs conducted on C yielded a significant main effect of Group only for the sadness expression, F(2,25) = 4.08, P = 0.03, ηp2 = 0.25). Post hoc comparisons revealed only that sex offenders tended to select less frequently the sad label than the non-sex offenders, P = 0.09 (controls: M = − 0.14, S.D. = 0.29; NSO: M = − 0.21, S.D. = 0.60; SO: M = 0.10, S.D. = 0.67). 3.3. Intensity ratings of facial expressions The MANCOVA failed to show significant effects for the distinct variables. Only the Emotion × Intensity × BDI interaction reached significance, F(10, 16) = 2.55, P < 0.05, ηp2 = 0.61, suggesting the importance of depression for explaining differences in intensity ratings of emotional expressions as a function of their intensity level. Significant effects were found for the distinct variables (Intensity, Group, Emotion, Emotion × Group, and Intensity × Group) only when the covariates were excluded from the analysis. 3.4. Empathy ratings Self-rated empathy scores from the three questionnaires are reported in Table 2. The ANCOVA and the Bonferroni test indicated that the controls had significantly higher EQ scores than the offenders, F(2,24) =7.18, P < 0.003, ηp2 = 0.37. Differences between SO and NSO approached significance, P = 0.053 (Bonferroni test). For the ES-IVE score, a group effect was also found, F(2,24) = 8.41, P = 0.002, ηp2 = 0.41; the SO group reported lower affective empathy scores than the other groups, P < 0.05, whereas no difference was found between the NSO and the controls. An ANCOVA was also applied on each IRI subscale. For PT, a significant difference was found among the three groups, F(2, 24) = 4.04, P = 0.03, ηp2 = 0.25. The controls had a higher score on the PT subscale than the other groups (P < 0.05, Bonferroni test), but no significant difference was found between SO and NSO. For the FS subscale, the Group factor did not reach significance, F(2, 24) = 1.82, P = 0.13. Differences between groups were also found for EC, F(2, 24) = 3.73, P = 0.04, ηp2 = 0.24. Post hoc analysis revealed that the controls had higher ratings of EC than the SO only, P < 0.05. Finally, a significant difference between the groups was observed for PD, F(2, 24) = 3.55, P = 0.04, ηp2 = 0.23. Bonferroni tests revealed that the SO had higher PD scores than the NSO, P = 0.05. The inspection of size effects for the IRI subscales indicates that they are rather small (< 0.3), suggesting that the relationships between empathy components and delinquency are not consistent. In contrast, only the ES-IVE scale revealed a moderate size effect (> 0.4). Table 2. Mean scores on Empathy Quotient (EQ) scale, and on subscales of Empathy–IVE and Interpersonal Reactivity Index (IRI) in sex offenders (SO), non-sex offenders (NSO) and controls SO NSO Controls Mean S.D. Mean S.D. Mean S.D. Empathy Quotient (EQ) 5.60 1.07 19.8 5.39 50.5 6.36 Empathy subscale-IVE 4.60 1.57 12.3 2.06 17.0 1.70 Perspective Taking (PT-IRI) 6.60 1.26 12.2 2.39 20.9 1.66 Fantasy (FS-IRI) 10.50 1.58 8.50 3.44 16.7 3.77 Empathic Concern (EC-IRI) 4.60 1.90 12.0 2.79 23.9 2.60 Personal Distress (PD-IRI) 18.10 1.20 12.7 1.49 10.3 2.58 Table options 3.5. Relationships between accuracy rates of emotion recognition and empathy scores Partial correlations (pr), using BDI and S-STAI scores as covariates, were performed between emotional recognition accuracy and empathy scores across participants. Positive correlations were found between ES-IVE scores and accuracy rates for the following emotions: disgust (pr = 0.56, P = 0.002), fear (pr = 0.45, P = 0.02), surprise (pr = 0.33, P = 0.08), and sadness (pr = 0.33, P = 0.08). In other terms, the ability to recognize some emotions was greater in participants who obtained high scores of affective empathy. No significant correlations, however, were detected between EQ, PT, FS or EC scores and accuracy rates of emotion recognition. Nevertheless, negative correlations were found between PD and recognition accuracy rates for the following emotions: disgust (pr = − 0.34, P = 0.08), fear (pr = − 0.364, P = 0.057), surprise (pr = − 0.43, P = 0.02), anger (pr = − 0.48, P = 0.01), and sadness (pr = − 0.36, P = 0.06). Partial correlations between empathic and recognition accuracy rates were also examined for the distinct levels of facial expressions of emotions. Recognition accuracy rates were positively correlated with ES-IVE scores for moderate (pr = 0.57, P = 0.001) and strong (pr = 0.50, P = 0.006) levels of fear, and for mild (pr = 0.36, P = 0.058) and strong (pr = 0.36, P = 0.058) levels of surprise. Finally, a positive correlation was found between the ability to decode moderate expressions of sadness and ES-IVE scores (pr = 0.42, P = 0.02). In contrast, PD scores were negatively correlated with recognition accuracy rates for moderate (pr = − 0.46, P = 0.01) and strong pr = − 0.39, P = 0.03) levels of disgust. PD scores were also negatively correlated with recognition accuracy rates for moderate expressions of anger (P = − 0.52, P = 0.006), fear (pr = − 0.46, P = 0.01), surprise (pr = − 0.46, P = 0.01) and sadness (pr = − 0.45, P = 0.02).