When using IRT models in Educational Achievement Testing, the model is as a rule too simple to catch all the relevant dimensions in the test. It is argued that a simple model may nevertheless be useful but that it can be complemented with additional analyses. Such an analysis, called profile analysis, is proposed and applied to the reading data of Pisa 2000. Participating countries are compared on three different categorizations of the item material. (Contains 5 figures and 8 tables.)