The use of computer graphics for problem solving was investigated during use by individual learners and by distant pairs collaborating by interaction through the screen. In both investigations learners had to correct computer graphics representing population models, either by graphical manipulation or by written explanation. In both situations positive effects of manipulable graphics on problem solving performance were found. However, a detailed analysis of the interaction processes in the collaborative situation showed that, comparing both experimental conditions, static graphics led to a higher proportion of correct content‐related statements and explanations of system relations. No differences were observed between both conditions on the participation structure within the learning dyads. The results suggest that for the effective use of graphics in computer‐supported collaborative learning, structural support devices that encourage adequate processing and focus learners on the problem solving process seem to be necessary.