Empathy is defined as the ability to perceive and understand another person’s emotional state and the competence to appropriately respond to others’ emotions [2,3]. It is needed to induce prosocial behavior: free-willing behavior to benefit others [4]. Therefore, empathy is often referred to as ‘social glue’ in relationships [4–6].

From a developmental perspective, empathy has been divided into different layers: affective empathy, cognitive empathy and prosocial motivation. Affective empathy, also known as emotional contagion, is the process in which the emotional states of others cause a level of arousal in the observer. It consists of non-conscious behavioral mimicry of others’ facial, vocal, and bodily expressions [7]. This ‘mirroring of emotions’ is thought to be present at birth and originates from the Mirror Neuron System (MNS) in the brain. Through functional Magnetic Resonance Imaging (fMRI) studies, neuroscientific research has shown these neural networks. For instance, making a sad face or observing a sad face both activate the MNS via the amygdala and the anterior insula of the brain. This motor activation is then associated with an emotion representation; the person acknowledges a sad feeling [8–10]. These patterns strongly suggest that the formation of cortical representations about one’s own feelings is a necessary condition to engage in vicarious predictions about the emotions of others [5] (see Lamm & Majdandžić for an in-depth discussion of the plausibility of this assumption [11].

Whereas young children become upset and need comforting themselves through affective empathy, also referred to as ‘contagious crying’ or ‘emotional sharing’, around the age of two children change from a self-focused perspective towards another-focused perspective. Consequently, children gradually become to understand that their sad feelings are caused by another person in distress. This evokes an urge to support or comfort that person, as to relief their distress [3,5,12].

Cognitive empathy develops as children grow older and involves a more sophisticated comprehension of the other person’s emotional state [13]. The child starts to understand why the other is upset. Understanding emotions in others serves different goals. First, the observer is capable to distinguish between its own and the other’s emotions and thereby decreasing their own feelings of distress. Second, understanding the other leads to an increased tendency to support the other, and to care for the other [11]. This intrinsic prosocial motivation is essential because it signals to the person in distress that the observer pays attention to and highly values their emotions. He or she understands what is happening and wants to support. Moreover, a stronger level of cognitive empathy can also help to overcome in-group preferences [11]. The long-term purpose of cognitive empathy and prosocial motivation is to induce and maintain good social relationships [12]. As such, the development of cognitive empathy is largely dependent on social learning. fMRI studies have indirectly shown this as the relation between the MNS and self-reported cognitive empathy is less clear than for affective empathy [7,11]. For social learning to develop, this requires incidental learning skills; unplanned and unorganized learning abilities, with no educational intentions. Social learning takes place while interacting with others, and by trial-and-error.

A lack of empathy is associated with violence, aggression, criminality, and insensitive and unemotional behavior [14]. Empathic dysfunction has been associated with several psychiatric disorders such as psychopathy, autism spectrum disorders [15], conduct disorder, acquired sociopathy [16], and schizophrenia [17]. Children and adolescents who show little or no empathy are deemed to fail in our social world, and are put aside as having antisocial behavior. These behavioral problems may lead to the development of an antisocial personality disorder later in life [18]. Hence, it is of major importance for children to adequately develop empathic skills.

Little is known about the development of empathy in DHH children. However, certain prerequisites for successful empathic maturation, such as emotion recognition and regulation together with development of a Theory of Mind (ToM), have recently been addressed in this population. Studies show lower levels of emotion recognition and labelling of emotions in deaf preadolescents than in NH peers. In this population the onset of deafness was related to the ability to recognize emotions. Prelingually deaf preadolescents were more vulnerable than their postlingually deaf peers [19]. Regarding emotion awareness, DHH children were found to be less able to address multiple emotions in the negative domain simultaneously (e.g., anger and sadness) than NH peers during several emotion tasks. In the same study, children had to focus on approaching strategies towards an emotion-evoking situation. The results show less effective emotion regulation in DHH children than in NH peers [20]. ToM has been measured in toddlers with cochlear implants (CIs). Initially children with CIs were found to perform as well as NH children. However, at an older age they fell behind on more advanced ToM abilities such as false belief tasks [21]. Regarding their empathic behavior, no differences were found between young children with a CI and NH peers [5]. Yet, because of the young age of these children (1–5 years), only the affective domain of empathy could be taken into account in this study.

Because of the continuous development of cognitive empathy in childhood and preadolescence we are interested whether empathic abilities in DHH children and adolescents differ from those of their NH peers. Therefore, the aim of this study was to determine the differences in the levels of self-reported and observed empathy between DHH on the one hand, and NH children and adolescents on the other. To identify those factors that may be most influential for the levels of empathy in DHH children we also investigated the influence of several audiological factors on empathic abilities, such as language development, intelligence, degree of hearing loss, age at intervention of hearing loss, type of device, mode of communication, and educational setting.

On the basis of the research mentioned above we expected to find equal levels of affective empathy in DHH and NH children. However, regarding the development of cognitive empathy and prosocial motivation we expected DHH children to fall behind as a consequence of, among other things, their impaired ToM development. Concerning several audiological variables such as type of hearing amplification, it has been reported that DHH children wearing CIs experience lower levels of behavioral problems than children wearing Hearing Aids (HAs) [22]. Therefore, we expected to find differences in empathic ability between these two groups.

Gender differences have been described frequently in the literature. Girls consistently report higher levels of affective empathy and prosocial behavior than boys. Some researchers doubt these conclusions. They hypothesize that the reported differences are a result of differences in social desirability between boys and girls [8,23]. If true, we would find higher levels of self-reported affective empathy and prosocial motivation in girls, regardless of their hearing status but equal levels of empathy and supportive behavior during observations.

Due to the improved developmental outcomes after early intervention programs as reported by Yoshinaga-Itano et al. [24], we expected a relation between age at detection and intervention of hearing loss, and empathic abilities. Educational placement (mainstream or special schools) and mode of communication (spoken or sign language) have been reported to be related to levels of psychopathology in DHH children [25–28]. We therefore expected that children attending mainstream education and using spoken language as their preferred mode of communication show higher levels of empathy.

We recruited 122 DHH children and a control group consisting of 162 NH children from all over The Netherlands and the Dutch-speaking part of Belgium to participate in this study. All children were between 9 and 16 years of age at time of assessment. The age of 9 as a cut-off point was chosen because the children needed to be able to reflect on their own emotions and behavior [29]. All children had an IQ of 80 or higher and no other known disabilities besides their hearing loss. Of all DHH children, 52 were fitted with a CI and 70 children wore conventional HAs. Hearing impairment was defined as experiencing a loss of ≥40 dB in the best ear that was detected pre- or perilingually. Children with postlingual onset or detection of hearing loss were excluded. The NH group was matched with the DHH group on sex and mean age. As can be seen in Table 1, gender, intelligence, socio-economic status (SES), and age did not differ between the groups. No differences were found in type of school and mode of communication when comparing children wearing a CI with children using HAs. The onset of hearing impairment differed between the two groups; χ² (1, n = 115) = 3.92, p<.05. The HA-group presented more perilingual onset of hearing impairment than the CI group. As expected the degree of hearing loss differed between the two groups χ² (2, n = 114) = 73.62, p<.001. Children with a CI mainly experienced profound losses whereas children with HAs showed more moderate to severe hearing losses. Permission for this study was granted by the Medical Ethics Committee of the Leiden University Medical Center under number P10.137.

To increase the external validity of our findings, we tried to ensure diversity in our study population and recruited children all over The Netherlands and the Dutch-speaking part of Belgium via hospitals, speech- and hearing centers, primary and secondary schools, and special schools for the deaf. Written parental informed consent was obtained for all participating children. The assessment was carried out in a quiet room. Before starting the tests, children were assured that their answers would remain anonymous. Questions appeared one by one on a laptop. Depending on their preferred mode of communication, DHH children could choose between two versions of the questionnaires: a written text, or a version in which this text was simultaneously accompanied by sign language. The questionnaires were assessed as part of a larger study on the socio-emotional development of DHH and NH children. In between several tests, the experimenter acted live emotions to observe empathic reactions and supportive behavior during the test session. Parents completed questionnaires at home, they were also asked to complete a list of background variables such as net income and level of education. A socioeconomic status (SES) score was calculated using the net income of the family, job and level of education of both parents. Audiological variables were extracted from the child’s medical and/or audiological notes.

The instruments used in this study are described here. Psychometric characteristics of all questionnaires are shown in Table 2.

Group demographics were compared using independent t-tests. To compare the levels of empathy (affective empathy, cognitive empathy and prosocial motivation) between the different subgroups repeated measures Multivariate Analysis of Variance (MANOVA) and Multivariate Analysis of Covariance (MANCOVA) were used. In case of differences between subgroups within the DHH children, when sample sizes were small (<40 children per group) the assumption of normality was violated. Therefore, to compare levels of empathy between these subgroups (e.g., uni- versus bilateral CI, pre- versus perilingual onset of hearing loss) a non-parametric test was chosen (i.e., Mann-Whitney U test). Correlations between the empathy subscales and audiological factors were calculated using Pearson’s correlations. These correlations were compared between the different groups using Fisher’s r-to-z transformations to be able to show significant differences between correlations. Statistical analyses were carried out using the program SPSS version 21.0 (IBM Corp., Armonk, NY).

To analyze the differences in self-reported empathy levels between children with a CI, those with HAs, and hearing children, we carried out a repeated measures MANOVA with Group (CI, HA, NH) as the between-subjects variable and self-reported empathy (affective empathy, cognitive empathy, prosocial motivation) as the within-subjects variable. The analysis showed a main effect for empathy (F_HF (1.97, 553.96) = 303.81, p<.001, η_p² = .52) and for group (F (2, 281) = 11.44, p<.001, η_p² = .08), which was qualified by an empathy x group interaction (F_HF (3.92, 553.96) = 2.46, p<.05, η_p² = .02). Post-hoc t-tests showed that on affective empathy children with CIs scored lower than the NH group. Scores on affective empathy by children with HAs did not differ from NH children. DHH children overall scored lower on cognitive empathy and prosocial motivation than NH peers, regardless of their type of hearing amplification.

Because of the known influence of language development and intelligence on the socio-emotional development of DHH children, these variables were added as covariates in the analyses. In a MANCOVA that corrected for language development and intelligence, the main effect for group remained (F (2, 236) = 6.30, p = .002, η_p² = .05), but the interaction effect was no longer significant (F_HF (3.95, 465.57) = 1.55, p = .19, η_p² = .01). Language development was significantly related to the levels of empathy (F(1, 236) = 5.25, p = .02) whereas intelligence was not (Fig 1).

A gender x self-reported empathy repeated measures MANOVA was conducted to define differences in self-reported empathic abilities between boys and girls, regardless of their hearing status. Results showed a main effect for empathy (F (1.97, 555.75) = 393.96, p<.001, η_p² = .58) and for gender (F (1, 281) = 11.10, p = .001, η_p² = .04), which was qualified by an empathy x gender interaction (F_HF (1.97, 555.75) = 4.33, p<.05, η_p² = .02). Post-hoc analysis revealed that girls scored higher on affective empathy and prosocial motivation than boys. Equal levels of cognitive empathy were reported. The aforementioned results were combined in a 3 (hearing group) x 2 (gender) x 3 (self-reported empathy) repeated measured MANCOVA with language development and intelligence as covariates. The main effect for group remained (F (2, 233) = 5.75, p = .004, η_p² = .05) whereas the results no longer showed a main effect for gender (F (1, 233) = 3.54, p = .06, η_p² = .02).

Concerning attendance to others’ emotions, a 2 (DHH, NH) x 2 (boys, girls) one-way ANCOVA that corrected for language skills and intelligence revealed an effect for hearing group (F (1, 235) = 8.52, p<.01) and gender (F (2, 235) = 18.04, p<.001). NH children reported higher scores than DHH children and girls scored higher than boys. Language development was significantly related to the attendance towards others’ emotions (F(1, 240) = 4.80, p<.05). A one-way ANCOVA to compare the effect of hearing group and gender on emotion recognition as scored by parents corrected for language development and intelligence showed no differences between the hearing groups or gender (F (1, 182) = 0.03, p = .87 and (F (1, 182) = 0.065, p = .80, respectively).

Differences between gender and hearing status in observed empathic behavior during the live emotions tasks were assessed with language and intelligence as covariates. A 2 (DHH, NH) x 2 (boys, girls) mixed ANCOVA revealed an effect for hearing status and for gender; DHH children scored higher than their NH peers on emotion attention(F (1, 220) = 28.80, p<.001); regardless of their type of hearing amplification. Girls scored higher than boys (F (1, 220) = 10.94, p = .001). To compare DHH and NH boys and girls on their observed supportive behavior, a 2 (DHH, NH) x 2 (boys, girls) mixed ANCOVA was performed showing an effect for hearing status but not for gender (F (1, 220) = 16.03, p<.001 and F (1, 220) = .66, p = .42, respectively). Conversely to their ‘emotion attention’, NH children more often showed supportive behavior than DHH children.

In order to properly examine levels of empathy between DHH children at special education (for the deaf and hard of hearing child) and at mainstream education, a MANCOVA was performed with school-type (special or mainstream) as the between-subjects variable, the self-reported levels of empathy as the within-subjects variables and language development as a covariate. The two groups did not differ in background and audiological characteristics (e.g., age at detection of hearing loss, age at intervention, intelligence, SES). The analysis showed a main effect for empathy (F_HF (2, 204) = 9.16, p<.001, η_p² = .08) and for school-type (F (1, 102) = 4.38, p<.05, η_p² = .04). Post-hoc ANCOVA’s revealed higher levels of cognitive empathy in DHH children attending mainstream schools than in DHH children attending special schools, (F_HF (1, 102) = 7.89, p<.01), whereas for affective empathy and prosocial motivation no significant differences were found (F_HF (1, 102) = 1.61, p = .21. and F_HF (1, 102) = .91, p = .34, respectively). No significant differences were found in observed empathic reactions nor in parent reported emotion recognition or attendance to others’ emotions comparing DHH children in mainstream and special education when corrected for their language skills.

When comparing the child’s preferred mode of communication DHH children using sign (supported) language scored lower on self-reported prosocial motivation and on observed attention to emotions than DHH children who preferred to use spoken language (U = 986.5, z = -2.95, p = .003 and U = 802.5, z = -2.32, p = .021, respectively). Two participants solely communicated by sign-language. All analyses were rerun without these two participants. The results did not differ.

No significant differences were found between the levels of empathy in children regarding the moment of detection of their hearing loss (i.e., pre- or perilingual). When comparing within the CI group, parents reported higher levels of emotion recognition in unilaterally implanted children compared to bilaterally implanted children (U = 82, z = -2.54 p = .01).

In the DHH group, the relation between several continuous audiological variables (degree of hearing loss, age at detection of hearing loss, age at intervention of hearing loss, age at implantation) and the levels of empathy (self-report, parent-report and observed) were analyzed by means of Pearson’s correlations. No relations were found between these variables and the different levels of empathy.

The psychometric properties of the empathy questionnaire were satisfying with good reliability in both DHH children and their NH peers. However, to assure that DHH children are as capable as hearing children in understanding the items well, further psychometric properties will be useful to examine. Item response theory models can shed further light on issues such as measurement invariance, which includes differential item functioning. Because of power issues we were not able to perform this type of analyses. Future studies with a larger cohort of DHH children are needed to address these issues. Regarding the design of this study, we have to point out that cross-sectional data were used, which prevents us from drawing conclusions about causality. Therefore, we started longitudinal data collection to confirm the assumptions made here.