Predicting Naming Latencies with an Analogical Model

Skousen's (1989, Analogical modeling of language, Kluwer Academic Publishers, Dordrecht) Analogical Model (AM) predicts behavior such as spelling pronunciation by comparing the characteristics of a test item (a given input word) to those of individual exemplars in a data set of previously encountered items. While AM and other exemplar-based models enjoy continuing success in their ability to predict what a participant's response to a given task will be, it does not yet include a widely tested mechanism for extending its predictions to other measures of interest in psycholinguistics such as response time (RT). This article reports the results of applying a formula derived in Estes (1959, in: Koch, Psychology: A study of a science, McGraw-Hill Book Co., Inc.) for approximating "mean predicted latency" in decision tasks to the alternative responses and their associated probabilities predicted by AM. The model is tested against six sets of data from previously published naming studies.