Emotions in online collaboration have been largely neglected or considered as a static phenomenon. Therefore, the present study aims to explore dynamics of emotional reactions in online collaboration of Communities of Learners (CoL) based on the Emotions as Social Information (EASI) Theory. In addition, relationships between emotional reactions and the learning outcomes of CoL are investigated. Data of 11 CoL with a total of 87 learners are collected in an online training programme. Qualitative content analysis was used to explore dynamics of emotional reactions and quantitative analysis was used to investigate relationships. Results showed that various positive and negative emotional reactions could be identified in the data. Further, by using a newly developed method displaying Emotional Reaction Threads (ERT) it was shown that emotional reactions unfold dynamically during online collaboration. Beyond that, results revealed relationships between the learning outcomes [(i) participation, (ii) exam grade] of CoL and positive [(i) p < 0.01] and negative emotional reactions [(i) p < 0.01, (ii) p < 0.05]. Based on the results of the qualitative content analysis explanations for the identified relationships are provided. Instructors of the online training programme can positively influence the learning outcomes of the CoL through expressing positive emotional reactions during online collaboration.