Delayed match-to-sample in working memory: A BrainMap meta-analysis

Abstract

Working memory (WM), or the ability to temporarily store and manipulate information, is one of the most widely studied constructs in cognitive psychology. Since its inception, it has become one of the leading explanations for how humans are able to operate on a cognitive level. The current study probed the neural networks underlying one of the most commonly used tasks, delayed match-to-sample (DMTS), to study WM. An activation likelihood estimation (ALE) analysis of 42 functional neuroimaging studies (626 participants) was conducted to demonstrate neural network engagement during DMTS. Results demonstrated strong convergence in brain regions commonly associated with the working memory construct (i.e., dorsolateral prefrontal cortex, fusiform gyrus, and posterior parietal cortex). However, neural activation in two regions frequently attributed to WM were absent from this meta-analysis: the anterior cingulate and the rostral prefrontal cortex, suggesting that these regions may be more sensitive to task or stimuli characteristics. In a post-hoc analysis, we deconstructed the DMTS meta-analysis to examine nonverbal versus verbal stimuli, and found notable neurofunctional differences such that DMTS using nonverbal stimuli consistently engaged the right middle frontal gyrus (BA 6/46) and precuneus (BA 7) more so than verbal stimuli based DMTS. These results provide a foundation for future models of functional connectivity that may elucidate subtle differences in working memory attributable to pathological processes.

Keywords: ALE; Activation likelihood estimation; Functional neuroimaging; Neural networks; fMRI.