The valid provision of subscores from an item response theory-based test implies a multidimensional test structure. Assuming, in the construction of a new test, that the test features required for a valid and reliable total test score have been specified already, this article describes the resulting subscore performance and the resulting degradation of the total score performance caused by multidimensionality. Subscore and total score error variances for both maximum likelihood and expected a posteriori estimators were determined for a typical test as a function of the test dimensionality (i.e., the number of subscores) and the level of correlation among the subscore abilities. The hit rates for detecting true differences among subscore abilities of practical importance are presented.