Social robotics continues to expand as a prominent area of research due to the increasing use of robots in settings involving social interactions. In this paper we study the social dimension of human robot interaction during a posture imitation game. We discuss how human personality traits influence a learning robot. A neural architecture is used to enable autonomous and interactive learning. We address this issue by linking the performance of the robot algorithms (learning trajectory, convergence, recognition score...) to the personality traits of its partners (extraversion and anxiety). As results of this work we noticed the emergence of pattern related to the extraversion trait of individuals interacting with the robot but no evidence for anxiety. By analyzing the robot’s learning with each partner, we observe that the acquisition time of the partner’s visual representation was correlated with the extraversion trait. The results show that even in a simple posture imitation game, we observe a fluctuation in a robot's learning based on the personality traits exhibited by its human partner.