Purchase this article with an account.
John Spencer, Aaron Buss, Vince Magnotta; Testing Hemodynamic Predictions of a Dynamic Neural Field Model of Visual Working Memory. Journal of Vision 2013;13(9):9. doi: https://doi.org/10.1167/13.9.9.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Efficient visually-guided behavior depends on our ability to form, retain, and compare visual representations separated in space and time. This ability relies on visual working memory (VWM). Although research has begun to shed light on the neuro-cognitive systems subserving VWM, few theories have addressed these processes in a neurally-grounded framework. Here, we describe a layered dynamic neural field (DNF) architecture that captures the cortical population dynamics that underlie VWM, including the encoding, maintenance and comparisons operations involved in change detection. We then test this model using fMRI. Recent work has shown that the BOLD response is strongly correlated with local field potentials (LFPs). An analog of LFPs can be estimated from DNF models. This estimate can be convolved with an impulse response function to yield hemodynamic predictions. Using this approach, we show that the DFN model quantitatively captures fMRI data from recent studies probing changes in the BOLD response in the intraparietal sulcus (IPS) as set size increases in change detection. We also test a novel hemodynamic prediction regarding the source of errors in change detection. The BOLD responses from different neural fields in the model revealed a complex relationship between false alarms and misses. Some fields showed larger BOLD responses for false alarms versus misses due to the increased synaptic activity associated with the failure to consolidate items in working memory as well as the incorrect detection of new features. Other fields showed larger BOLD responses on misses, resulting from robust maintenance of information which suppresses the detection of changes. Using fMRI, we identified cortical regions that mirrored these patterns. These data run counter to classic explanations of errors in change detection which assume perfect maintenance at low set sizes with errors reflecting guessing. Results show that the DNF model can effectively bridge the gap between brain and behavior.
Meeting abstract presented at VSS 2013
This PDF is available to Subscribers Only