Qualitative analysis of neuropsychological instruments has been a long tradition in neuropsychological assessment. This study extended this type of analysis to the Matrix Reasoning (MR) subtest of the WAIS-III. The study compared the performance of TBI participants on the item types identified within the MR subtest (i.e., pattern completion, classification, analogy and serial reasoning) with a group of normal controls. MR items were classified into categories (as defined respectively by the Psychological Corporation and by our own research definition). Ninety-three non brain-injured control and 72 brain injured control participants were included in the study. One way analysis of variance indicated that the TBI group performed significantly worse than the non brain-injured group the MR performance overall as well as for both the Psychological Corporation classification and on the research defined categories. Within group analysis revealed that both groups performed significantly differently across the item categories with the most difficult categories being analogy and serial reasoning for the research defined categories and the classification and serial reasoning categories for the Psychological Corporation-defined groups. The results of the study indicate that an item type analysis of the MR performance may further contribute to the qualitative aspects of diagnostic formulation.