A neurocognitive model of Chinese character processing : an fMRI study on its constituent processes

Abstract

Functional magnetic resonance imaging (fMRI) studies have revealed that processing alphabetic and logographic reading involves both overlapping and distinct neural mechanisms. In particular, the left fusiform gyrus (FFG) and the inferior frontal gyrus (IFG) are commonly recruited during reading of any language, but the left middle frontal gyrus (LMFG, BA 9/46) and the right FFG are more strongly activated during logographic reading. Studies reporting a common brain network usually tap into semantic processing, while those highlighting script-specific regions typically measure form-level processing. Hence, the discrepancy may lie in the constituent processes that are differentially examined by those studies. To have a comprehensive understanding of the neural mechanism of reading, we need to characterize it from its components separately. The aim of this study is to identify the neural networks for these constituents, namely orthography, phonology and semantics, and build a neurocognitive model that explains the functions served by the script-specific regions. This research is divided into three studies. In Study 1, the neurocognitive underpinnings of orthographic, phonological and semantic processing were examined. Native Mandarin speakers performed orthographic, phonological and semantic judgement tasks during fMRI scans. Whole-brain analysis revealed that the three reading constituents recruited a common neural network covering the left MFG, IFG, superior parietal lobule and bilateral FFG, while functional connectivity analysis revealed a posterior visuo-orthographic network and a frontal phono-semantic network, interfaced by the LMFG. This finding suggests that the MFG plays a coordinating role in Chinese reading. Study 2 was designed to delineate how various semantic relations are represented. The brain networks of two additional semantic tasks, namely semantic categorization and semantic association judgement, were compared with the synonym judgment task in Study 1. Whole-brain analysis found a common semantic network including the left MFG, middle temporal gyrus and bilateral IFG, but the activation levels of specific regions and the effective connectivity patterns differed among the three tasks. This study shows that subtle differences in semantic relation processing can be demonstrated through the cortical dynamics that are not apparent in the whole-brain activation patterns. The LMFG, which has been reported to play a key role in working memory functions, was consistently identified in Study 1 and 2. To examine whether the neural correlates of the domain-general working memory can predict Chinese literacy performance, participants performed visual and verbal working memory tasks in the final study. Their out-of-scanner behavioral performance in reading, writing and orthographic knowledge was measured. The activation level of the LMFG during a verbal memory task was found to contribute significantly to orthographic knowledge. The results provide an interpretation to the role of the LMFG as an interface between the orthographic and phono-semantic networks. A neurocognitive model of Chinese reading is proposed from the results of the three studies. Before orthographic information is passed to the phono-semantic network for pronunciation and meaning retrieval, it is temporarily stored and processed in the LMFG. This study implies that the neural mechanism of Chinese reading may differ from that of alphabetic languages due to differences in form-level processing.