Concern has been expressed over the item response theory (IRT) assumption that a person's ability can be estimated in a unidimensional latent space. To examine whether or not the response to an item requires only a single latent ability, unidimensional ability estimates were compared for data generated from the multidimensional item response theory (MIRT) compensatory and noncompensatory models. It was hypothesized that as the correlation between the two dimensional abilities increased, the response data would essentially become unidimensional and thus the confounding of difficulty and dimensionality would have little effect in either model. This study examined the unidimensional estimates of matched compensatory and noncompensatory data in which difficulty was confounded with dimensionality for different levels of correlation between two dimensional abilities. Eight data sets, four compensatory, and four noncompensatory, were generated. Each set was calibrated using the IRT calibration programs LOGIST and BILOG. BILOG calibration of response vectors generated to the matched MIRT item parameters appeared to be more affected than LOGIST by the confounding of difficulty and dimensionality. As the correlation between the generated two dimensional abilities increased, the response data appeared to become more unidimensional. Six figures are included. (Author/MDE)