This study explores the potential contribution of Eggins and Slade's (2004) Speech Functions as tools for describing learners' participation patterns in Synchronous Computer-Mediated Communication (SCMC). Our analysis focuses on the relationship between learners' self-efficacy (i.e. personal judgments of second language performance capabilities) and discourse roles displayed in the online medium. A small corpus of data was selected as a sample from a larger study, comprising one face-to-face interaction (FtF2) and one synchronous text-based chat interaction (SCMC2) between two participants: Celine, a high-self-efficacy (HSE) learner, and Concetta, a low-self-efficacy (LSE) learner. The chat-log and conversation transcript were analyzed by employing: (a) quantitative measures of participation; namely words and turns produced by the participants, (b) Dörnyei and Kormos's (1998) taxonomy of Communication Strategies, and (c) Eggins and Slade's (2004) classification of speech functions. Our results suggest that speech functions are indeed effective at describing the social roles enacted by learners during interaction across the two media, in terms of discourse dependence or independence, as well as dominance. Therefore, by complementing other methods, such as quantitative measures of participation and qualitative analyses of communication strategies, speech functions can contribute to providing a comprehensive picture of the relationship between SCMC, learners' self-efficacy, and participation patterns.