We are motivated in our work by the following question: what factors influence individual participation in social media conversations? Conversations around user posted content, is central to the user experience in social media sites, including Facebook, YouTube and Flickr. Therefore, understanding why people participate, can have significant bearing on fundamental research questions in social network and media analysis, such as, network evolution, and information diffusion. Our approach is as follows. We first identify several key aspects of social media conversations, distinct from both online forum discussions and other social networks. These aspects include intrinsic and extrinsic network factors. There are three factors intrinsic to the network: social awareness, community characteristics and creator reputation. The factors extrinsic to the network include: media context and conversational interestingness. Thereafter we test the effectiveness of each factor type in accounting for the observed participation of individuals using a Support Vector Regression based prediction framework. Our findings indicate that factors that influence participation depend on the media type: YouTube participation is different from a weblog such as Engadget. We further show that an optimal factor combination improves prediction accuracy of observed participation, by ∼9-13% and ∼8-11% over using just the best hypothesis and all hypotheses respectively. Implications of this work in understanding individual contributions in social media conversations, and the design of social sites in turn, are discussed.