Meta-analysis of neural correlates in Chinese reading and math across cognitive processing levels: A cross-domain...

Reading and math, despite their apparent differences, share overlapping neural substrates and involve multi-level organized cognitive processes that engage common brain networks. This study investigates the cross-domain neural substrates for different levels of cognitive processing in reading and math, focusing on the shared symbolic processing mechanisms underlying both domains. By analyzing Chinese reading and math, we examine how the cognitive processes from perceptual recognition to conceptual integration are organized in the brain. In the reading domain, the three levels include orthography, semantics, and sentence reading; in the math domain, the three levels include number sense, symbolic comparison, and calculation. Using 54 Chinese reading- and 65 math-related fMRI studies, we applied activation likelihood estimation (ALE) and meta-analytic connectivity modeling (MACM) to uncover brain activation patterns related to different levels of cognitive processing in reading and math and their dynamic interactions with functional networks (e.g., multiple-demand network and language network). The ALE results showed that, in reading, orthographic processing primarily engages the left middle frontal gyrus, the fusiform gyrus, and the parietal regions, whereas semantic and sentence reading increasingly recruit left-lateralized language networks (e.g., the left middle temporal gyrus and the inferior frontal gyrus). In math, number sense predominantly activates the right inferior parietal lobule and the frontal regions; symbolic comparison engages the bilateral parietal and the right frontal regions; and calculation mainly involves the bilateral parietal regions, the insula, and the frontal regions. Our MACM results indicated that the fronto-parietal multiple-demand system is more modulated by task demands, while the interaction with the language network progressively strengthens as the processing level increases in both domains. Based on our findings, we propose a Hierarchical Interactive Model to highlight the commonalities between reading and math from the perspective of the interaction patterns between regions and large-scale networks. This framework provides new insights for future investigations into the specific deficits in learning disabilities, such as dyslexia and dyscalculia.