The theoretical contribution of this paper is to regard teachers' evaluations with a prognostic claim about students' future academic ability as a result of a special social situation in the classroom. We assume that after teachers have framed the social situation, particular scripts of action will determine the criteria on which teachers ground their evaluations. In concrete terms, we propose a theoretical approach that integrates existing meritocratic and 'habitus' explanations in the comprehensive framework of frame selection theory with its important distinction between a more automatic and a more rational type of information processing. Our empirical contribution is to test the hypotheses that we deduced from our theoretical assumptions in a set of structural equation models. Using data from the Cologne High School Panel (CHiSP), we find that even when controlling for the path structure of the model, indicators for both kinds of concepts are statistically significant. However, regardless of the underlying type of information processing, the predictive power of indicators operationalizing the meritocratic explanation is comparatively higher.