How do humans categorize items and events in their world, and what role does language play in this process? Such questions have been at the center of long-standing debate among philosophers and scientists from many eras, cultures, and disciplines. Fundamental to this debate are difficult questions of what is meant by the concept of a linguistic symbol. For instance, philosophers and scientists alike have articulated a conceptual distinction between the relations involved in rich linguistic symbols (true words) as compared to restricted paired associates (rote-learned associations). Yet it remains difficult to specify behaviorally what actually makes these two types of relations different. In this article, we describe methodology that may offer an operationalized model that allows empirical analysis of paired associate versus symbolic learning. Like any methodology, our approach alone does not pretend to fully resolve age-old questions. However, we intend to illustrate distinct advantages offered within this approach to scholars interested in symbolic functioning and its development.