In this article, the effect of functional roles on group performance and collaboration during computer-supported collaborative learning (CSCL) is investigated. Especially the need for triangulating multiple methods is emphasised: Likert-scale evaluation questions, quantitative content analysis of e-mail communication and qualitative analysis of open-ended questions were used. A comparison of 41 questionnaire observations, distributed over 13 groups in two research conditions - groups with prescribed functional roles (n = 7, N = 18) and nonrole groups (n = 6, N = 23) - revealed no main effect for performance (grade). Principal axis factoring of the Likert-scales revealed a latent variable that was interpreted as perceived group efficiency (PGE). Multilevel modelling (MLM) yielded a positive marginal effect of PGE. Most groups in the role condition report a higher degree of PGE than nonrole groups. Content analysis of e-mail communication of all groups in both conditions (role n = 7, N = 25; nonrole n = 6, N = 26) revealed that students in role groups contribute more 'coordination' focussed statements. Finally, results from cross case matrices of student responses to open-ended questions support the observed marginal effect that most role groups report a higher degree of PGE than nonrole groups.