In the present paper, we develop and apply a technique, based on the logic of process dissociation, for obtaining numerical estimates of item and order information. Certain variables, such as phonological similarity, are widely believed to produce dissociative effects on item and order retention. However, such beliefs rest on the questionable assumption that item and order memory can be measured through performance on a particular kind of retention test (e.g., order can be measured through a reconstruction test). Retention measures are probably not process pure, but instead recruit multiple kinds of information and memory processing. Across four experiments, we show that our derived estimates conform generally to expected trends, although surprising results emerged in some instances. We discuss the implications of our analysis for traditional beliefs about the item-order distinction, and we show that a popular immediate memory model--the perturbation model--is capable of handling most of the empirical trends.