This study uses network analysis to examine Twitter's level of autonomy from external influences, being the political and the media field. The conceptual framework builds upon Bourdieu's field theory, appropriated on social media as mediated social spaces. The study investigates conversation patterns on Twitter between political, media and citizen agents during election times in Belgium. Through the comparison of conversational practices with the positions users hold as political, media or citizen agents, we understand how the former is related to the latter. The analysis of conversation patterns (based on replies and mentions) shows a decentralized and loosely knit network, in which primarily citizen agents are present. Nonetheless, the prominence of citizens in the debate, mentions or replies to political and media agents are significantly higher, placing them more centrally in the network. In addition, politicians and media actors are closely connected within the network, and reciprocal communication of these established agents is significantly lower compared to citizen agents. We understand different aspects of autonomy related to the presence, positions and practices of the agents on Twitter and their relative positions as politicians, media or citizens. To conclude, we discuss the promises of Bourdieu's relational sociology and the limitations of our study. The approach proposed here is an attempt to integrate existing work and evolve towards a systematic understanding of the interrelations between political, media and citizen agents in a networked media environment.