Self Comes to Mind

unique-personal, say. That distinction is roughly comparable to the semantic/episodic distinction, or the generic/contextual distinction.
    It is also useful to preserve the distinction between factual memory and procedural memory because it does capture a fundamental divide between “things”—entities that have a certain structure, in repose—and the “movement” of things in space and in time. Even here, however, the distinction can get dicey.
    In the end, the validity of these categories of memory resides with whether the brain honors the distinction. By and large, the brain honors distinctions between unique and nonunique levels of processing at the level of recall, and between factual and procedural kinds of memory, both in the making of a memory and in the recall.
A Possible Solution to the Problem
     
    Reflection on these observations led me to propose a model of neural architecture aimed at accounting for recall and recognition. 4 What the model accomplished is as follows.
    Images can be experienced during perception and during recall. It would be impossible to store the maps that underlie all images one has experienced, in their original format. For example, the early sensory cortices are continuously constructing maps about the current environment and have no resources to store discarded maps. But in brains such as ours, thanks to the reciprocal connections between the map-making brain space and the dispositional space, maps can be recorded in dispositional form. In such brains, dispositions are also a space-saving mechanism for information storage. Finally, dispositions can be used to reconstruct the maps in early sensory cortices, in the format in which they were first experienced.
    The model took into account the neuropsychological findings described earlier and posited that the cell ensembles at the top levels of the processing hierarchies would not hold explicit representations of the maps for objects and events. Rather, the ensembles would hold know-how, that is, dispositions , for the eventual reconstruction of explicit representations when they become needed. In other words, I was using the simple disposition device that I introduced earlier, but this time, rather than commanding a trivial movement, the disposition was commanding the process of reactivating and putting together aspects of past perception , wherever they had been processed and then locally recorded. Specifically, the dispositions would act on a host of early sensory cortices originally engaged by perception. The dispositions would do so by dint of connections diverging from the disposition site back to early sensory cortices. In the end, the locus where memory records would actually be played back would not be that different from the locus of original perception.
Convergence-Divergence Zones
     
    The main piece of the proposed framework was a neural architecture of cortical connections that had convergent and divergent signaling properties relative to certain nodes. I called the nodes convergence-divergence zones (CDZs). CDZs recorded the coincidence of activity in neurons hailing from different brain sites, neurons that had been made active by, for example, the mapping of a certain object. No part of the overall map of the object had to be permanently re-represented in the CDZs, to be placed in memory. Only the coincidence of signals from neurons linked to the map needed to be recorded. To reconstitute the original map and thus produce recall, I proposed the mechanism of time-locked retroactivation . The term retroactivation pointed to the fact that the mechanism required a process of “going back” in order to induce activity; time-locked called attention to another requirement: it was necessary to retroactivate the components of a map roughly within the same time interval, so that what occurred simultaneously (or nearly so) in perception could be reinstated simultaneously (or nearly so) in recall.
    The other critical element in the framework consisted of positing a division of labor between two kinds of brain systems, one that managed maps/images and another that managed dispositions. As far as the cerebral cortices were concerned, I proposed that the image space consisted of several islands or early sensory cortices—for example, the ensemble of visual cortices that encircle the primary visual cortex (area 17 or V 1 ), the ensemble of auditory cortices, that of somatosensory cortices, and so forth.

     
Figure 6.1: