The Extended CLS Model: Mind-Wandering as Deep Learning
Our goal in this section is to use the McClelland et al. (1995) CLS architecture as a starting place to explain mind-wandering. To do this, we will need to propose several refinements and extensions to the CLS model, resulting in a new model that we call the extended CLS model. A number of these extensions concern psychological functions (e.g., episodic memory, conceptual learning systems, consciousness) that are likely to be uniquely human, or at least elaborated to a unique extent in humans.
Episodic Memory in the Surface System
In formulating the original CLS model, McClelland, McNaughton, and O'Reilly assumed the hippocampal surface learner records discrete semantic items (e.g., robins are birds, penguins cannot fly) and repeatedly replays these items to the deep neocortical system. In the extended CLS model, our focus is on a more complex type of memory—episodic memories (Allen & Fortin, 2013; Tulving, 2002). These are memories of personal experiences that have a rich quasiperceptual phenomenology. They contain conjoined information about the "W questions": What happened and who was involved? Where did it occur and what was the context? When did it occur and in what order did it unfold? There is extensive evidence that the hippocampal system does indeed play a critical role in encoding and recall of episodic memories (Burgess, Maguire, & O'Keefe, 2002; Squire, 1992), so our extension of the CLS model to emphasize this type of memory is plausible.