Four experiments tested whether self- and friend-biases in perceptual matching are modulated by whether stimuli are presented aligned with the participant’s body and seen from the same perspective (the embodied perspective). Participants associated three colours (blue, green, and red) with three people (self, friend, and stranger) and then judged if a pairing of a colour and a personal label matched. The colour was painted on the T-shirt of an avatar. We modulated the perspective of the avatar along with its alignment with the participant’s body. In Experiment 1 a single avatar appeared. In Experiments 2–4 there were two avatars, and we varied the social communicative environment between the two avatars (social vs. non-social in Experiments 2/4 vs. 3) and the distance between the two avatars and fixation (close, far, or equal in Experiment 2, 3 or 4). With a single avatar, performance on friend-match trials selectively improved when the avatar was aligned with patient’s body and viewed from the participant’s (first-person) perspective. The self-bias effect was unaffected by the perspective/embodiment manipulation and it was strong across all conditions. However with two avatars performance on both self- and friend-match trials improved when the target stimulus appeared on the avatar adopting a first person perspective and aligned with the participant’s body, when two avatars were shown in a social-communicative context. These selective improvements disappeared when two avatars turned their back on one another in a non-communicative setting. The data indicate that self- and friend-biases in perceptual matching are modulated by both how strongly stimuli align with the participant’s perspective and body, and the social communicative situation. We suggest that self-biases can reflect an embodied representation of the self coded from a first-person perspectives.