As a typical form of empathy, empathy for pain refers to the perception, judgment of others’ pain, as well as corresponding affective responses. Deficit in empathy for pain may lead to withdrawal, indifference to other’s pain or aggressive behavior.Some contextual factors, especially the observers’ traits, may have an influence on empathy for pain. Previous studies have focused on empathy for pain of special groups like physicians and offenders, but most of these studies are questionnaire surveys and there is little research on the cognitive neural mechanisms.Moreover, observing stimuli in different perspectives also affects empathy for pain, and whether this effect exists in special groups remains unknown. Investigating the capacity of empathy in violent offenders can deepen our understanding about the mechanisms of empathy and provide some supports for their training program. Using a picture-priming paradigm, the present studypresented painful and non-painful stimuli toviolent offenders and ordinary (control) participants and simultaneously recorded their ERP data. The experiment consisted of two blocks and the participants were asked to view stimuli in first-person and other-person perspectives, respectively. Painful or non-painful pictures were presented for 1000ms, which werefollowed by a fixationlasting randomly for 1500-1700ms. In order to make sure that the participants viewed the pictures carefully, only 10% pictures were randomly chosen and needed participants’ responses. These pictures were accompanied by a response screen showing that “Painful picture: F; Non-painful picture: J”. The screen disappeared after a button pressing.For the remained trials, the participants just needed to view pictures carefully. The results showedthat the latencies and amplitudes of N110, P2 and P300 were different between violent offenders and control group. When viewing painful pictures, violent offenders had longer latency of N110 and P2 than what, and the amplitude of P2 was larger than what. When they viewed pictures under other-person perspective they also had longer latency of N110 than what?. Furthermore, for the amplitude of P300, control grouphad smaller amplitude when viewing painful pictures in both perspectives, while for violent offenders this effect only existed under other-person perspective. Our results indicated that violent offenders might have lower capacity of empathy for pain than control group. The present study supports the Communications Model of empathy for pain and expands previous studies to some extent. This may also provide some empirical support for the training program of offenders. However, we also have some limitations that need further investigations in the future.