In an earlier study, we had tested if observing a collaboration model, or alternatively, following a collaboration script could improve students' subsequent collaboration in a computer-mediated setting and promote their knowledge of good collaboration. Both model and script showed positive effects. The current study was designed to further probe the effects of model and script by comparing them to conditions in which the learning was supported by providing elaboration support (instructional prompts and a reflective self-explanation phase). In addition, we applied a newly developed, innovative rating scheme to analyze the collaborative process: The rating scheme combines qualitative evaluation with quantitative assessment. Forty dyads were tested, eight in each of the following conditions: model plus elaboration, model, script plus elaboration, script, and control. Observing a collaboration model with elaboration support yielded the best results over all other conditions on measures of the quality of collaborative process and on outcome variables. Model without elaboration was second best. The results for the script conditions were mixed; on some variables, even below those of the control condition. The results of the current study lead us to challenge the positive view on collaboration scripts prevalent in CSCL research. We propose adaptive scripting as a possible solution.