Search

Teachers' accuracy in judging students' achievement is often assumed to be a general ability of teachers. Based on this assumption, teachers should be at least consistent in their accuracy across different content domains within a school subject. Yet, this assumption has rarely been investigated empirically so far. Data from 54 mathematics teachers (N = 1170 students) and 55 language teachers (N = 1255 students) were analysed using a Bayesian multivariate multilevel modelling approach. Results indicate that latent accuracy measures across content domains indeed are substantially correlated within both investigated subjects, but may still be considered to represent different dimensions. • Explored content-specificity of teachers' judgment accuracy in mathematics and German language class. • Modelled accuracy and the relationships between content domains simultaneously. • Used Bayesian multivariate multilevel models with latent predictor variables. • Found low to medium average judgment accuracy in all content domains. • Content domains were substantially correlated, yet empirically distinguishable. [ABSTRACT FROM AUTHOR]