Collaborative problem-solving (CPS) has become an essential component of today's knowledge-based, innovation-centred economy and society. As such, communication and CPS are now considered critical 21st century skills and incorporated into educational practice, policy, and research. Despite general agreement that these are important skills, there is less agreement on how to capture sociocognitive processes automatically during team interactions to gain a better understanding of their relationship with CPS outcomes. The availability of naturally occurring educational discourse data within online CPS platforms presents a golden opportunity to advance understanding about online learner sociocognitive roles and ecologies. In this paper, we explore the relationship between emergent sociocognitive roles, collaborative problem-solving skills, and outcomes. Group Communication Analysis (GCA) -- a computational linguistic framework for analyzing the sequential interactions of online team communication -- was applied to a large CPS dataset in the domain of science (participant N = 967; team N = 480). The ETS Collaborative Science Assessment Prototype (ECSAP) was used to measure learners' CPS skills, and CPS outcomes. Cluster analyses and linear mixed-effects modelling were used to detect learner roles, and assess the relationship between those roles on CPS skills and outcomes. Implications for future research and practice are discussed regarding sociocognitive roles and collaborative problem-solving skills.