Working memory – the ability to hold information “in mind” in the face of environmental interference – has traditionally been associated with the prefrontal cortices (PFC), based primarily on data from monkeys. High resolution functional imaging (such as fMRI) have revealed that PFC is just one part of a larger working memory network, notably including the parietal cortex, which has long been the focus of research in the visual domain, and is primarily thought to carry out spatial computations.
What role might such spatial computations have in working memory? Wendelken, Bunge & Carter describe (in their new Neuropsychologia article) how a region of the posterior parietal cortex may be involved in “working memory for organized content – by virtue of its rich representation of space and spatial relationships.” Specifically Wendelken et al argue that information is organized in a short-term state according to a spatial code in the superior parietal lobe (SPL). The authors note similarities between their theory and that of Marshuetz and colleagues, who proposed something similar based on the SPL’s apparent involvement in number and magnitude representations (with yet others arguing that space, number and magnitude are all enabled by the same underlying representation).
In a 2002 Neurocomputing article, Wendelken proposed a connectionist model in which ventrolateral regions of the prefrontal cortex (vlPFC) contain “pointers” to item-specific representations in the inferotemporal cortices, and that dorsolateral prefrontal areas (dlPFC) are able to dynamically “bind” or associate these pointers to spatial locations, as represented in the superior parietal lobe, using a temporal synchrony mechanism. The essential predictions of this model were twofold:
1) SPL and dlPFC areas should be most sensitive to demands for organizing information in working memory, and maintaining that organization
2) vlPFC should be most sensitive merely to the number of items which need to be maintained in working memory
The new article reports Wendelken et al.’s verifications of these predictions with neuroimaging. The authors conducted two experiments which directly manipulated working memory load in terms of the number of items to be remembered (3 vs. 6 items in experiment 1, and 4 vs 7 items in experiment 2) and organizational demand in terms of the grouping of items (deciding whether 2 probe letters occurred in one of 3 previously-presented sets of 2 letters each, in experiment 1; deciding whether a particular pair of letters could be generated from a previously-presented directed graph of 4 letters and 3 transitions between them). This design is illustrated below:
The results: as you would expect, subjects found the low working memory demand condition easier than the high demand or the organizational demand conditions (as indicated in accuracy and reaction times). But the neuroimaging showed something far more interesting: differences in activation between the high and low working-memory demand conditions activated left posterior VLPFC, as well as bilateral dlPFC, inferior frontal gyrus, and superior parietal lobe. However, only the superior parietal lobe was more active for the organizational demand condition than the high working memory load condition across both experiments. Finally, in the second experiment, activity in ventral dlPFC increasingly predicted that in the SPL with increasing memory load, and that in dorsal dlPFC increasingly predicted that in the SPL with increasing organizational demand. In the first experiment, such increased functional connectivity was only visible at a lower statistical threshold and even then only for organizational demand.
The take home: Wendelken et al. were able to verify the crucial prediction of their model, which is that superior parietal regions are particularly important when working memory has an organizational demand placed on it, and that this relies on interactions with dlPFC.
As the authors note, however, there are numerous alternative explanations:
A) It’s possible that by manipulating “organizational demand” the authors actually just manipulated the amount of working memory load. According to this hypothesis, there is no qualitative difference between the dlPFC/SPL circuit and the vlPFC/IT circuit, but rather one of generalized “task demands” or effort.
B) the inclusion of arrows in the second experiment may have directly triggered superior parietal activity, due to its obvious visuospatial component; similarly, the use of alphanumeric characters in the experiment might have led to phonological processing in the SPL or the IPL.
C) Subjects may have attempted to visually encode and maintain the items, leading to the appearance of parietal cortex involvement in active maintenance that was in fact specific to a particular strategy. (The authors argue this explanation is unlikely given the long cue period and perceptual mismatch between the cue and test arrays – I tend to agree).
D) There was no significant delay-period activity in dlPFC in the first experiment (although there was strong cue-period activity) and no difference in functional connectivity as a result of organizational demands. Wendelken et al. argue that this may indicate that dlPFC is specifically activated by relational organization (as in experiment 2) but not mere grouping (as in experiment 1).
E) It’s not clear to me how a temporal synchrony mechanism for binding would predict increased activity in any region as a function of task demands. As far as I know, coherence of neural firing has not been directly related to the BOLD response, whereas firing rate has. In addition, specifying the mechanism as one of temporal synchrony seems strange when predictions based on synchrony-based models have not been empirically verified.
Nonetheless, the general idea – that parietal cortex may be specialized for holding information temporarily online in some spatial form – is alluring, if only because it would seem to parallel other evidence that the parietal cortex has a much wider role than traditionally acknowledged.
Binding Solutions Converge on Parietal Cortex
Parietal Cortex in Filtering
When TMS Helps: Does Parietal Cortex Cause Binding Errors?
The Argument for Multiplexed Synchrony
Dopamine to the Rescue: The Binding Problem
Task-Switching: A Role for Inferior Parietal Cortex