Diagnosing learning behavior is one of teachers' most central tasks. So far, accuracy in teachers' judgments on students' achievement has been investigated. In this study, a new perspective is taken by developing and testing a three-dimensional model that describes the process of diagnosing learning behavior within a sample of N = 293 participants. The model is validated by participants' actual diagnoses in a case scenario. The postulated model provides a good fit and following its steps leads to more accurate diagnoses. Future research should aim to foster diagnostic competence based on the model. (Contains 5 tables and 2 figures.)