CCN meeting | Brad Postle (Department of Psychology, University of Wisconsin–Madison, USA)

25-06-2019 from 15:00 to 16:00
Henri Dunantlaan 2, room 1.2
Add to my calendar

Neural codes underlying priority in visual working memory

An important intersection between theory and data in working memory research is priority. Neural data can help to address such questions as whether multiple items can be held in the focus of attention, or whether the features that make up an object remain bound when that object is outside the focus of attention (FoA). Here, I'll address the mechanisms that may underlie different states of priority. Whereas there's been considerable interest in recent  years about putative "activity-silent" mechanisms for holding information in working memory but outside the FoA, studies are now beginning to find evidence for active traces of unprioritized memory items (UMIs). Although some have proposed a "cortical specialization" scheme, with UMIs shuttled off to a "special place," this talk will consider evidence for priority-based recoding. Multivariate inverted encoding modeling of data from fMRI and EEG experiments suggest that the transition from FoA to UMI is accompanied by a recoding that is accomplished by a rotational transformation of the representation, such that the mappings between stimulus values (e.g., orientation of a grating patch) and neural states shift in a way that preserves the metric distances between neural states. Thus, priority-based recoding is accomplished by a remapping within the neural code, rather than by a transition to a new code. Furthermore, this is observed, in parallel, for different stimulus dimensions in their different specialized circuits (i.e., for the UMI's orientation in occipital cortex, and for it's location in parietal cortex). Neural network simulations suggest that such rotational recoding may be a general principle for shifts of priority in distributed representations.