We have seen that language comprehension involves a complex set of processes in which listeners actively construct meanings by integrating linguistic input with background knowledge and local contextual information. Pursuing a constructivist approach, I have suggested that language cues the retrieval of frames from long-term memory for the construction of cognitive models in working memory. Language input, perceptual input, and current conceptual content all influence the recruitment of information as well as the construction of models. Comprehension proceeds by assembling a series of simple cognitive models and establishing the mappings, or correspondences, that exist between the elements and the relations in different spaces, or partitions of working memory.

When a word cannot be integrated into the model currently being built in the focus space, it may trigger frame-shifting, or reanalysis of information that has come before. Frame-shifting occurs when the currently activated frame does not adequately represent the relationship among two or more objects, actions, or events. While jokes provide a particularly salient example of the frame-shifting phenomenon, the need to construct novel frames to accommodate new information is quite common. Because sentential integration involves the coordination and accommodation of different frames, people presumably appeal to mechanisms of frame-shifting in the course of everyday language understanding.

However, as Churchland (1989) observes, frame-shifting can be accompanied by an experience of sudden insight. Can this remarkable event really be related to the more common processes of on-line meaning construction? Traditional models suggest the answer is no.