The impact of test item multidimensionality was examined as it affected fitting items on a teacher licensure test to an item response theory (IRT) model. Test item data from the 1990 study of G. W. Guiton and G. Delandshere are used. The Connecticut Elementary Certification Test (CONNECT) is a licensure examination designed to assess the subject matter knowledge of a beginning elementary teacher in the state. Nine content areas and pedagogy are covered. Items selected for analysis were taken from two test forms of the CONNECT administered prior to the date when it became a requirement for certification. In all, 826 responses were received from 681 candidates (some took both test forms). Classical item statistics indicated that test items were relatively heterogeneous, with the highest frequency of point biserials in the 0.10 to 0.30 range. These statistics confirm the previous (1990) finding of more than one underlying construct. However, successful attempts to fit the data to one-parameter and two-parameter models support the contention that a dominant underlying factor is sufficient to meet the assumption of unidimensionality required for item response analyses. An appendix summarizes the relationship between content knowledge and pedagogical knowledge for teachers. Three tables present study data. (SLD)