MULTINOMIAL PROCESSING MODELS OF SOURCE MONITORING

This article presents a family of processing models for the source-monitoring paradigm in human memory. Source monitoring and the special case of reality monitoring have become very popular as paradigms to assess memory deficits in various subject populations. The paradigm provides categorical data that satisfy product-multinomial constraints, and this lends it nicely to multinomial modeling with processing-tree structures as described in Riefer and Batchelder (1988). The models developed herein are based on ideas from high-threshold signal-detection models, and they involve item-detection parameters, source-identification parameters, and various parameters reflecting guessing biases. The purpose of the models is to provide separate, theoretically based measures of old-item detection and source discrimination. The models may strengthen traditional analyses that are based on ad hoc statistics, as well as avoid flawed interpretations that the traditional analyses may produce. The usefulness of the models is revealed by analyzing published data sets from the areas of reality monitoring and bilingual memory.