A case study in which dyads of subjects (both a pupil‐pupil and a teacher‐pupil dyad) use a microcomputer computer assisted learning (CAL) program is discussed. We argue that an approach involving the videotaping of pupils' interactions around the computer, microanalysing the resulting dialogue, and relating the results from the dialogue analysis to learning outcomes is a potentially very powerful method for evaluating the educational efficacy of CAL software. The present case study describes a technique for collecting appropriate data, indicates the richness of such dialogue data and illustrates some of the hidden problems in CAL program use which other evaluation methods are unlikely to expose.