Georg Rasch has developed a new one-parameter latent trait model to explain the performance of examinees on achievement tests. The model can be viewed as a special case of Birnbaum's two-parameter logistic model where all items are assumed to have equal discriminating power. Birnbaum's model permits items to vary in discriminating power. Both models assume that guessing does not occur. This study compares how well the one- and two-parameter models fitted different sets of empirical data. For each model, a measurement of agreement was made between the expected and obtained distributions of ability estimates using three sets of data. The results revealed that the more general the model, the better the fit with real data. (Author/DG)