The development of the between-class, within-class analysis has facilitated a more refined analysis of classroom processes and influences: those which affect a class as a whole, a student within a class, or an individual student independent of context. However, a problem arises in using a between-class, within-class analysis to infer a particular model of classroom processes. There are several competing classroom process models which can cause any particular pattern of between- and within-class findings. Several examples are given of the correspondence between classroom models and patterns of statistical findings. In one case it is assumed there are no competing effects (zero-order effects). In the other case, competing effects are dealt with. Most of the examples are hypothetical, but real data are given from the Teaching-Learning Interaction Study, where coping style is used to predict achievement gain. In this case, there is a positive within-class, but not between-class effect. The possible causes for this situation are discussed in terms of the models previously given. It is concluded that educational investigators need to use several sources of evidence to infer a particular model, and that any such model needs to be refined and tested. (Author/BW)