According to the Matthew effect model, interindividual differences in reading competence between poor and normal readers should become wider as students grow older. The second part of the model assumes that these differential pathways are mainly attributable to differential reading activities. The purpose of this study is to examine whether both assumptions can be verified in a sample of German elementary school students. Data from 1,124 students, participating in the BiKS longitudinal study with assessment starting in Grade 3 and two further points of measurement in Grade 4, were available for our analyses. Latent growth curve models showed a significant lower linear and quadratic trend on reading competence development for students with severe reading deficits in comparison with their better classmates. Further analysis indicates that differential reading behaviour seems to be a prominent factor in explaining these differential pathways whereas the students' general cognitive abilities seem to be of less importance.