In recent years the estimation of unidimensional abilities for instruments with subtests has been a focus of item response theory. Different hierarchical models, which assume a common unidimen-sional latent trait and several subtest specific latent traits, have been proposed in order to cope with local item dependencies due to subtests. In contrast to these models, the generalized subdimension model (GSM) allows for the estimation of a latent mean ability based on multidimensional latent traits. Examining a small data set (n=72) this article examines the implicit weighting of the unidi-mensional model in contrast to the explicit weighting of the GSM to improve measurement preci-sion.