Education and cognition research today generally recognize the tri-dimensional nature of reasoning processes as involving cognitive, social and emotional phenomena. However, there is so far no theoretical framework articulating these three dimensions from a descriptive perspective. This paper aims at presenting a first model of how group emotions work in collective reasoning, and specifies their social and cognitive functions. This model is inspired both from a multidisciplinary literature review and our extensive previous empirical work on an international corpus of videotaped student debates. The cognitive function of emotions is defined in reference to the process of schematization (Grize 1996, 1997) and associated emotional framing (Polo et al. 2013). On the other hand, the social function of emotions refers to recognition-oriented behaviors that correspond to engagement into specific types of group talk (e.g. Mercer 1996), implying specific politeness rules or face-preservation systems (Brown and Levinson 1988). We believe that our multi-dimensional and multi-level approach to group reasoning, which mostly employs a linguistic perspective, can be applied to a diversity of contexts. We hope it will serve as a basis for further discussion on the role of emotions in reasoning among the interdisciplinary community of argumentation studies.