The Lisrel-model with structured means was used to study similarities and differences in the development of mathematical ability between two student groups as measured by two tests on three successive occasions. Procedures for testing models to assess the contribution on an individual parameter to the goodness of fit are described. (Author/BS)