DESCRIPTORS

The use of item response theory (IRT) in the Medical College Admission Test (MCAT) testing program has been limited. This study provides a basis for future IRT analyses of the MCAT by exploring the dimensionality of each of the MCAT's three multiple-choice test sections (Verbal Reasoning, Physical Sciences, and Biological Sciences) and the implications of that dimensionality for IRT calibration and scoring. Three sets of data were used: an MCAT form administered on 2 occasions (16,520 and 3,638 observations) and a second, content-equivalent form administered on the second occasion (12,625 observations). Dimensionality analyses suggested that while the items within each of the science test sections are not completely homogeneous, nor are they measuring distinct constructs corresponding to the disciplines. As is consistent with this finding, when confirmatory factor analyses were performed, the Biological Sciences disciplines were found to be correlated about 0.85, and the Physical Sciences about 0.95. Comparison of the results of the IRT calibration and scoring of the data by test and by disciplin