In Virtual Reality (VR), users provided with a virtual body and congruent multi-sensory stimulation usually experience a feeling of embodiment of the virtual body, which can be interpreted as a transfer of bodily self-consciousness toward the avatar's body. This subjective experience depends on factors such as agency over the avatar's movements and visual similarity between virtual and physical bodies. As the user-avatar relationship has already been extensively investigated from the perspective of the self-avatar being essentially a virtual duplicate of the user, we propose to reverse this point of view and to investigate how the avatar's actions and appearance shape the user's experience. To explore the complex mutual relationship between VR users and their full-body avatars, we first investigate its motor aspect, focusing on VR users' tendency to follow their avatars' movements in case of visuo-motor incongruencies. We show that, for some repetitive motion and complex motor tasks, users tend to act with their avatar when it acts by itself, revealing a possible drive to maintain the alignment with the virtual body even when control is lost. Our results emphasize the implicit guidance of the self-avatar and suggest embodiment is an expected state that, when disrupted, calls for a restoration of the coherence between the physical and virtual selves. Second, motivated by the rise of social VR experiences, we also investigate self-perception and embodiment in shared virtual environments. To do so, we propose an embodied morphology assessment method exploring how VR users perceive their virtual and real body morphologies, how the avatars' morphology affects embodiment, and how exposure to virtual bodies with various body shapes influences the perception of the self. Our findings highlight the specifics of our method and emphasize the tight link between body perception, task performance, and motor control. Our research presents methodological contributions, including a standardized virtual environment with a virtual experimenter to ensure consistency and reproducibility. Overall, this work contributes to a better understanding of the mutual user-avatar influences, suggesting that the user-avatar relationship is a partnership in which the self-avatar is expected to be an ally. Leveraging this privileged relationship could lead to new lines of investigation on the cognitive mechanisms of embodiment and to the design of more efficient VR simulations, with applications for both professional and recreational purposes.