Similarity and Deviation in Event Segmentation and Memory Integration: Commentary on Richmond, Gold, & Zacks

In this commentary on Richmond, Gold, and Zacks (2017), we focus on two complementary processes that play critical roles in event segmentation theory, and thus figure prominently in the arguments of the authors. We argue that the processes of similarity and deviation are important not only to event segmentation, but to a process that may seem its polar opposite, namely, integration of separate episodes of experience. In event segmentation theory, the perception of similarity in the ongoing flow of activity is fundamental to behavioral control such that as long as elements of an activity are similar, the event is the “same” and the controlling schema continues to be valid. Equally important in the model is the complement of similarity-deviation is critical in that (a) within an episode, as long as there is no deviation, the current schema can control behavior; and (b) when predictions are violated (a deviation), event boundaries are imposed and the event model is updated. We illustrate these processes in the context of the ERISS model (Encoding, Reactivation, Integration, Selection, and Self-derivation), which was developed to understand derivation of new knowledge through integration of separate episodes, and is here extended to event segmentation.