Résumé : In this dissertation, we aimed at better understanding the role of learning in shaping the contents of consciousness. To capture variations in consciousness, we took subjective measures as a starting point, and we performed several studies measuring their relation to human behavior, peripheral physiology and brain physiology in about 160 participants. We first focused on the learning mechanisms implicated in learning with feedback from the body, and then investigated how autonomic responses related to several aspects of awareness in associative learning tasks. Our results provide evidence that people can improve in their sensitivity to discriminate between mental states, while their confidence in doing so is unaffected. Our results also indicate no evidence for the malleability of phasic heart rate response by implicit knowledge. Taken together, these results suggest that consciousness is not easily influenced by learning with external feedback from the body. At a more abstract conceptual level, we explored several methodological considerations when interpreting changes in subjective reports and separated the potential contributions of knowledge and direct perception. Finally, we presented a novel taxonomy for categorizing neurofeedback paradigms, which may be of help to disentangle the learning process that is implicated in neurofeedback.