The neural coding of tonal working memory load: An functional magnetic resonance imaging study

Abstract

Tonal working memory load refers to the number of pitches held in working memory. It has been found that different verbal working memory loads have different neural coding (local neural activity pattern). However, whether there exists a comparable phenomenon for tonal working memory load remains unclear. In this study, we used a delayed match-to-sample paradigm to evoke tonal working memory. Neural coding of different tonal working memory loads was studied with a surface space and convolution neural network (CNN)-based multivariate pattern analysis (SC-MVPA) method. We found that first, neural coding of tonal working memory was significantly different from that of the control condition in the bilateral superior temporal gyrus (STG), supplement motor area (SMA), and precentral gyrus (PCG). Second, neural coding of nonadjacent tonal working memory loads was distinguishable in the bilateral STG and PCG. Third, neural coding is gradually enhanced as the memory load increases. Finally, neural coding of tonal working memory was encoded in the bilateral STG in the encoding phase and shored in the bilateral PCG and SMA in the maintenance phase.

SF-MVPA: A from raw data to statistical results and surface space-based MVPA toolbox

“… Surface space-based multivariate pattern analysis (MVPA) stream of SF-MVPA, which is adopted from Li et al (2022b) . (A) Surface dividing method.…”

“…Surface space-based multivariate pattern analysis SF-MVPA adopts the surface space-based MVPA algorithm of Li et al (2022b). The analysis stream is shown in Figure 3.…”

“…BOLD signals within each cell were averaged and formed an n1 × n1 image. A circle with a radius of r (half the length of the grid side) was applied to Surface space-based multivariate pattern analysis (MVPA) stream of SF-MVPA, which is adopted from Li et al (2022b). (A) Surface dividing method.…”

“…To address these issues, researchers proposed surface space-based MVPA methods ( Oosterhof et al, 2010 ; Chen et al, 2011 ; Li et al, 2022a , b ). These MVPA methods obtain circular searchlights directly from the surface space (although there are differences in details of acquiring searchlights).…”

“…In the field of image recognition, convolutional neural network (CNN) is a popular classifier and has been proven to have good image recognition ability. SF-MVPA uses CNN as an optional classifier and adopts the surface space-based MVPA algorithm of Li et al (2022b) . The algorithm is different from the other surface space-based MVPA algorithms in that it uses grid-like searchlights.…”