A model of the attribution process based upon an observer's probability estimates is presented. Four hypotheses were tested using hypothetical achievement situations. Subjects were presented with a student's present quiz grade, the class average, and the student's average
on past quizzes. Five levels for each of these three independent variables created a 5 × 5 × 5 factorial design. The model assumed that the results of achievement behavior can be attributed either internally or externally and either stably or unstably. The dependent variables were
subjects' ratings of the level and importance of ability, effort, task difficulty, and luck as possible causes for the student's present quiz grade. Three of the four hypotheses were strongly supported. The advantages and implications of this approach are discussed.