Behavioral prioritization enhances working memory precision and neural population gain

Abstract

Humans allocate visual working memory (WM) resource according to behavioral relevance, resulting in more precise memories for more important items. Theoretically, items may be maintained by feature-tuned neural populations, where the relative gain of the populations encoding each item determines precision. To test this hypothesis, we compared the amplitudes of delay-period activity in the different parts of retinotopic maps representing each of several WM items, predicting amplitude would track with behavioral priority. Using fMRI, we scanned participants while they remembered the location of multiple items over a WM delay, then reported the location of one probed item using a memory-guided saccade. Importantly, items were not equally probable to be probed (0.6, 0.3, 0.1, 0.0), which was indicated with a pre-cue. We analyzed fMRI activity in ten visual field maps in occipital, parietal, and frontal cortex known to be important for visual WM. In early visual cortex, but not association cortex, the amplitude of BOLD activation within voxels corresponding to the retinotopic location of visual WM items increased with the priority of the item. Interestingly, these results were contrasted with a common finding that higher-level brain regions had greater delay-period activity, demonstrating a dissociation between the absolute amount of activity in a brain area, and the activity of different spatially-selective populations within it. These results suggest that the distribution of WM resources according to priority sculpts the relative gains of neural populations that encode items, offering a neural mechanism for how prioritization impacts memory precision.

Common neural mechanisms control attention and working memory

“…Eleven neurologically healthy participants (ages 23-53; 6 females) with normal or corrected-to-normal vision participated in this experiment. The sample size was determined using previous fMRI studies comparing selection on perceptual and WM representations (Nobre et al, 2004; Tamber-Rosenau et al, 2011), and is equal to or larger than previous studies that compared within- and across-condition decoding performance of classifiers trained on fMRI data (Jerde et al, 2012; Kwak & Curtis, 2022; Rademaker et al, 2019), as well as those that used population receptive field-weighted reconstruction analysis (Kwak & Curtis, 2022; Yoo et al, 2022). Participants provided written informed consent in accordance with procedures approved by the Institutional Review Board at New York University.…”

“…To estimate the response of the neural populations at the selected location, we reconstructed a pRF-weighted map using the selection-related activity. This procedure essentially projects voxel activity in each ROI to visual space in screen coordinates (Kok & de Lange, 2014; Kwak & Curtis, 2022; Yoo et al, 2022). In each ROI, for every selected location (left, right, bottom) in each task condition (pre-cue, retro-cue), we created a reconstructed map using where w i (( x, y ), I σ) is the weight associated with the i th voxel at location ( x, y ), and β i is the averaged GLM-acquired selection period β at voxel i in trials with the same selected location.…”

Common Neural Mechanisms Control Attention and Working Memory

“…Eleven neurologically healthy participants (ages 23-53; 6 females) with normal or corrected-tonormal vision participated in this experiment. The sample size was determined using previous fMRI studies comparing selection on perceptual and WM representations (Nobre et al, 2004;Tamber-Rosenau et al, 2011), and is equal to or larger than previous studies that compared within-and across-condition decoding performance of classifiers trained on fMRI data (Jerde et al, 2012;Kwak & Curtis, 2022;Rademaker et al, 2019), as well as those that used population receptive field-weighted reconstruction analysis (Kwak & Curtis, 2022;Yoo et al, 2022).…”

“…To estimate the response of the neural populations at the selected location, we reconstructed a pRFweighted map using the selection-related activity. This procedure essentially projects voxel activity in each ROI to visual space in screen coordinates (Kok & de Lange, 2014;Kwak & Curtis, 2022;Yoo et al, 2022). In each ROI, for every selected location (left, right, bottom) in each task condition (pre-cue, retro-cue), we created a reconstructed map using 𝑤 ((𝑥, 𝑦), 𝐈𝜎)𝛽…”

“…Behaviorally, the quality of memory is better for the retroactively cued item compared to non-cued items (Griffin & Nobre, 2003;Landman et al, 2003;Li et al, 2021;Souza et al, 2016). Neurally, WM representations are enhanced within the neural populations encoding the selected WM item (Ester et al, 2018;Lepsien et al, 2011;Sprague et al, 2016;Yoo et al, 2022).…”