Scholastic achievement tests and mental ability tests normally consist of a set of multiple choice items, all of which are assumed to measure school-relevant cognitive abilities. The presumption, in a given test situation, is that the answers/solutions to the given tasks represent cognitive capabilities on the part of the examinees. The purpose of this paper is to show that this assumption does not always hold. Analyzing simulated and empirical data it is proved that, based on the mixed Rasch model (Rost, 1990), it is possible to identify those examinees who have applied a guessing strategy to solve multiple choice items. As an empirical example the results in a biology test consisting of 23 items, each having 5 choices were analyzed. On the basis of the responses from 5,641 7th grade students a guessing class was identified. Further analyses provided information indicating that guessing behavior is shown by students with lower-level cognitive abilities, who might have used the "random strategy" to cope with the items that were too difficult. Five tables and four figures provide study data. (Contains 21 references.) (Author/AA)