To rule out the need for a two-parameter item response theory (IRT) model during item analysis by Rasch models, it is important to check the Rasch model's assumption that all items have the same item discrimination. Biserial and polyserial correlation coefficients measuring the association between items and restscores are often used in an informal way for this purpose, with the expectation that correlations will be stronger for items with a high degree of item discrimination and weaker for items where item discrimination is low. Assessment of these coefficients is subjective and often difficult because Rasch models are nonlinear models that do not require biserial and polyserial correlations to be constant across items. To make assessment easier and less uncertain, the authors propose an alternative way to measure the association between item scores and restscores under the Rasch model and show how to assess the significance of the difference between observed and expected associations. (Contains 3 tables.)