The item analysis model by Rasch is presented and examples are given from investigation of the validity of the Danish version of MCMI. It is shown, using new procedures for examining the adequacy of the model, that the syndrome P subscale measuring paranoia (after elimination of certain heavily biased and heterogeneous items) of MCMI possesses the property of local homogeneity: Items are homogeneous and responses are distributed according to the Rasch model for persons belonging to groups characterized by high raw scores. The cutting lines between domains of homogeneous and heterogeneous items, in principle corresponding to the diagnostic cutting lines associated with the original version of MCMI, are presented.