Lexical Representations in the Common and Specific Neural Networks for Visual, Phonological, and Semantic Processing in Chinese Reading

Abstract

ABSTRACT

Previous studies have investigated the common and specific neural correlates underlying visuo‐orthographic, phonological, and semantic processing in word reading. However, it remains unclear how those networks represent different types of lexical information and how such representations and the interactions between networks are modulated by task‐induced processing demands. To address this issue, 32 native Chinese participants were scanned with fMRI while performing a localizer task, and two reading tasks designed to elicit high demands on visuo‐orthographic processing (i.e., structural judgment task) and semantic processing (i.e., familiarity judgment task). Activation analyses identified both common and specific neural networks involved in visual, phonological, and semantic processing. Representational similarity analysis (RSA) further revealed that the common network represented multiple types of lexical information, whereas the specific networks selectively represented particular lexical information corresponding to their respective processing type. Moreover, processing demands modulated lexical representations of common and specific networks in distinct ways: the common network exhibited flexible representational patterns, representing task‐relevant lexical information under high processing demands, whereas the specific networks showed process‐dependent selectivity, representing corresponding lexical information only under high processing demands. Functional connectivity analyses further indicated that processing demands could modulate connectivity patterns among networks, particularly between the common and specific networks. These findings highlight the distinct functional roles of common and specific networks, providing a new perspective on the complementary contributions of functionally overlapping and specialized systems in word reading.

Introduction

As a fundamental aspect of language processing, word reading plays a crucial role in understanding written language (Feng et al. 2020; Zhou et al. 2021). Existing cognitive models of reading (e.g., dual-route model and triangle model) have proposed that reading is a complex process requiring the coordinated integration of multiple critical linguistic components (Coltheart et al. 2001; Seidenberg and McClelland 1989), including orthographic, phonological, and semantic processing. The efficient integration of these components serves as the foundation for successful word recognition. Using neuroimaging techniques, much research has revealed that reading relies on a widely distributed neural network, primarily involving prefrontal, temporoparietal, and occipitotemporal cortices (Dehaene et al. 2010, 2015; Guo et al. 2022; Lu et al. 2021; Price 2012; Price and Devlin 2011; Saur et al. 2008; Siok et al. 2020; Wu et al. 2012). With respect to the three key components, studies have revealed that visuo-orthographic processing primarily engages the occipitoparietal cortex and the left ventral occipitotemporal regions (Gu et al. 2024; Kuo et al. 2004; Mei et al. 2013; Paz-Alonso et al. 2018; Price and Devlin 2011; Purcell et al. 2017), phonological processing mainly involves the temporoparietal regions (Cao et al. 2017; Li, Yang, et al. 2022; Li et al. 2025; Luo et al. 2024; Paz-Alonso et al. 2018; Wagley and Booth 2022; Weiss et al. 2018; Yen et al. 2019), while semantic processing mainly relies on the frontotemporal regions (Jia et al. 2022; Li et al. 2023; Pattamadilok et al. 2017; Paz-Alonso et al. 2018; Weiss et al. 2018; Zhang et al. 2019).

Although previous studies have revealed the distributions of brain networks associated with visuo-orthographic, phonological, and semantic processing, few studies have focused on the common and specific neural mechanisms underlying the three linguistic components. Notably, some meta-analytic studies (Cattinelli et al. 2013; Vigneau et al. 2006, 2011; Wu et al. 2012), complemented by empirical evidence (Liu, Tao, et al. 2022; Paz-Alonso et al. 2018), have attempted to characterize the neural networks underlying visuo-orthographic, phonological, and semantic processing, providing converging evidence that these processes were supported by both common and specific networks. Specifically, several brain regions, such as the left middle frontal gyrus and fusiform gyrus, have been consistently activated during visuo-orthographic, phonological, and semantic tasks (Guo et al. 2022; Li, Wu, et al. 2022; Liu, Tao, et al. 2022; Wu et al. 2012). This convergence of activation implies that these regions may constitute a common sub-network that supports the integration of multiple components during reading. In contrast, other brain regions are selectively engaged in visuo-orthographic, phonological, or semantic processing. For instance, the superior parietal lobule has been found to play a critical role in supporting visuospatial processing (Lin et al. 2021; Lobier et al. 2012; Reilhac et al. 2013). The visual word form area (VWFA), located in the left ventral occipitotemporal cortex, has been found to be selectively engaged in orthographic processing (Cohen and Dehaene 2004; Dehaene and Cohen 2011; Glezer et al. 2009, 2015, 2016; Liu et al. 2008; Martin et al. 2019), and this function appears to be highly consistent across different language systems (Krafnick et al. 2016; Szwed et al. 2014). Moreover, the left dorsal inferior frontal gyrus, left precentral gyrus, and superior temporal gyrus are demonstrated to be critical for converting visual symbols into phonology and mapping phonology onto meaning (Cao et al. 2017; Deng et al. 2011; Li, Yang, et al. 2022; Liu, Tao, et al. 2022; Tan et al. 2005; Wu et al. 2012; Zhu et al. 2014). Similarly, the left ventral inferior frontal gyrus, angular gyrus, and lateral temporal cortex are associated with semantic retrieval and integration (Branzi and Lambon Ralph 2023; Jackson et al. 2015; Paz-Alonso et al. 2018; Price et al. 2015; Weiss et al. 2018; Yang et al. 2024; Zhang et al. 2019; Zhu et al. 2009). These findings suggest a functional specialization in the reading network, whereby specific brain regions are specialized for supporting particular linguistic processes.

Despite increasing evidence for the involvement of common and specific neural networks underlying visuo-orthographic, phonological, and semantic processing, the functional roles of those networks in word reading remain unclear. Two reasons may contribute to this lack of clarity. First, prior studies have predominantly focused on mapping the spatial distribution of common and specific networks, without examining how those networks are associated with different types of lexical information (Li, Wu, et al. 2022; Liu, Tao, et al. 2022; Paz-Alonso et al. 2018; Wu et al. 2012). As a widely used multivariate method, representational similarity analysis (RSA) allows for the comparison of neural activation patterns with models derived from linguistic feature spaces, providing a more fine-grained understanding of how lexical information is represented in the brain (Haxby et al. 2014; Kriegeskorte 2008; Kriegeskorte and Kievit 2013). This method provides an approach for elucidating the functional roles of common and specific networks in word reading. Building on this, an increasing number of studies have employed RSA to investigate how reading-related brain regions represent different types of lexical information. For example, the posterior VWFA represents orthographic information, whereas the anterior VWFA and the supramarginal gyrus are associated with phonological information (Fischer-Baum et al. 2017; Qu et al. 2022; Staples and Graves 2020; Zhao et al. 2017). In contrast, the angular gyrus and anterior temporal lobe have been implicated in representing semantic information (Chen et al. 2016; Graves et al. 2023; Staples and Graves 2020). However, most of these studies have primarily focused on isolated reading-related regions, and employed relatively small stimulus sets (typically ranging from 30 to 90 items) (Liu, Wisniewski, et al. 2022; Purcell et al. 2017; Qu et al. 2022; Wang, Wang, and Bi 2023; Zhang, Wang, et al. 2024; Zhao et al. 2017), leaving it unclear how the common and specific networks represent lexical information during word reading. Therefore, the present study combined RSA with a large stimulus set to investigate how common and specific networks represent lexical information during word reading.

Second, the modulation effect of task-induced processing demands on the common and specific networks remains unclear. Prior studies have shown that processing demands are a key factor in influencing the function of brain regions, affecting not only local activity (Mano et al. 2013), but also the pattern similarity and neural representations as measured by RSA (Bailey et al. 2025; Li et al. 2023; Nastase et al. 2017; Qu et al. 2022; Zhang, Liu, et al. 2020). For example, Bailey et al. (2025) adopted RSA to compare lexical representational patterns between aloud and silent reading, and they found that aloud reading enhanced visual, phonological, and articulatory representations, whereas silent reading elicited stronger orthographic representations, highlighting that processing demands can dynamically reweight representational content in the reading network. Importantly, such effects are not uniform across different brain networks, but rather show network-specific patterns. For example, task-relevant networks (e.g., frontoparietal control networks) typically exhibit increased neural activity under high processing demands, whereas the default mode network is generally deactivated as processing demands increase (Cocchi et al. 2013; Fox et al. 2005; Menon 2023; Raichle et al. 2001). Given the distinct involvement patterns of the common and specific networks in word reading (Liu, Tao, et al. 2022; Wu et al. 2012), it is reasonable to assume that processing demands would modulate their lexical representations in distinct ways. Compared to the specific networks, the common network may flexibly represent different types of lexical information depending on processing demands. However, empirical evidence for this hypothesis remains scarce. Therefore, it is necessary to examine how task-induced processing demands differentially modulate lexical representations in common and specific networks to deepen our understanding of their functional roles in word reading.

Third, the interactions between the common and specific networks and their modulation by processing demands remain poorly understood. Previous studies have demonstrated that key brain regions involved in visuo-orthographic, phonological, and semantic processing exhibit significant functional connectivity both during resting state (Chai et al. 2016; López-Barroso et al. 2020; Stevens et al. 2015), and reading-related tasks (López-Barroso et al. 2020; Wang, Joanisse, and Booth 2023; Wang et al. 2019). In addition, the strength of these connections has been found to be associated with individuals' reading performance (Benischek et al. 2020; Huang et al. 2024). Importantly, comparisons across different tasks have further shown that processing demands can modulate functional connectivity between distinct brain regions (Li et al. 2024; Planton et al. 2022). For example, Li et al. (2024) found that, compared to the non-verbal task, verbal tasks significantly enhanced the functional connectivity between the VWFA and other language-related brain regions (e.g., the left orbital frontal cortex). Despite these findings, it remains largely unexplored how the common and specific networks interact with each other, and how varying processing demands modulate their interactions. Considering the involvement of common networks in multiple processes, we speculate that processing demands may flexibly enhance the functional connectivity between common networks and specific networks. Exploring this issue is important for understanding the interactive mechanism between common and specific networks in word reading.

To address the above issues, the present study (1) precisely identified common and specific networks involved in visual, phonological, and semantic processing of Chinese reading; (2) conducted RSA with a relatively larger sample of stimuli (198 Chinese characters, larger than stimuli typically used in most prior RSA studies) to systematically investigate the lexical representations in those networks; and (3) examined how task-induced processing demands modulated both the lexical representations and functional connectivity among these networks. Specifically, 32 native Chinese speakers were recruited. To identify the common and specific networks, participants were instructed to perform a localizer task, which consisted of five subtasks, including radical, rhyme, semantic, perceptual, and consistency judgment tasks. Next, two reading tasks (i.e., structural and familiarity judgment tasks) with a relatively larger sample of Chinese characters were used to further examine the lexical representations of those networks and how the processing demand affects their representations. RSA was then performed on the two reading tasks by correlating neural representational dissimilarity matrices (RDMs) with model-based RDMs of the Chinese characters. In addition, functional connectivity analyses were conducted to characterize the interactions among the identified networks. Based on the different involvement patterns of common and specific networks (Liu, Tao, et al. 2022; Wu et al. 2012), as well as the modulatory effects of processing demands on the local activity and connectivity of reading-related brain regions (Li et al. 2024), we hypothesized that (1) common networks would represent multiple types of lexical information, whereas the specific networks would be specialized for representing particular lexical information; (2) processing demands would modulate the lexical representations of the common and specific networks in different ways. The common network would flexibly represent task-relevant lexical information depending on processing demands, while the specific networks would only represent their corresponding lexical information under high processing demands; and (3) processing demands would also modulate the functional connectivity between the common and specific networks. The common network would strengthen its connectivity with the corresponding network depending on processing demand.

Methods

Participants

Thirty-two native Chinese students (15 males; mean age = 22 ± 2.1 years, range 18–27) were recruited to participate in our study. All of them were right-handed as assessed by the Edinburgh Inventory (Snyder and Harris 1993), had normal or corrected-to-normal vision, and reported no history of neurological disorders. Written informed consent was obtained from each participant before the experiment, and they received compensation upon completion. This study was approved by the Institutional Review Board of the School of Psychology at South China Normal University.

Materials and fMRI Tasks

The Localizer Task

To localize the common and specific networks underlying visual, phonological, and semantic processing, participants were asked to perform five subtasks during fMRI scanning, including radical, rhyme, semantic, perceptual, and consistency judgment tasks. The radical, rhyme, and semantic judgment tasks were designed to probe brain activations associated with visual, phonological, and semantic processing of Chinese characters, respectively. The consistency judgment task served as a baseline for the radical judgment task to eliminate the activations arising from low-level visual information, whereas the perceptual judgment task served as a baseline for the rhyme and semantic judgment tasks to control activations arising from visuospatial information of Chinese characters.

Specifically, in the radical judgment task, participants were instructed to determine whether the two sequentially presented Chinese characters shared the same radical (e.g., “扮” and “份” = yes; “划” and “决” = no). In the rhyme judgment task, participants determined whether the two characters rhymed (e.g., “后/hou4” and “肉/rou4” = yes; “百/bai3” and “守/shou3” = no). In the semantic judgment task, they evaluated whether the two characters were semantically related (e.g., “米/rice” and “面/noodles” = yes; “时/time” and “弓/bow” = no). Additionally, in the perceptual judgment task, participants judged whether the two sequentially presented characters were the same size (e.g., “风” and “表” = yes; “句” and “亚” = no), while in the consistency judgment task, they were asked to decide whether two presented nonwords were identical (e.g., “” and “” = yes; “” and “” = no). All characters used in the localizer task had moderate to high frequency, and the frequency and stroke numbers of the Chinese characters and nonwords were matched across the five subtasks (Table 1).

TABLE 1 The mean number of strokes and frequency of the Chinese characters or nonwords used in the localizer task (Cai and Brysbaert 2010).

Task	Number of strokes	Frequency (per million)
Mean	SD	Mean	SD
Radical judgment	7.15	2.17	94.44	224.87
Rhyme judgment	6.74	1.80	106.49	141.14
Semantic judgment	7.03	2.59	106.01	162.20
Perceptual judgment	6.96	1.89	104.26	115.10
Consistency judgment	6.69	2.01	—	—

As shown in Figure 1A, the localizer task employed a block design comprising three runs. Each run included 10 blocks, with two blocks for each of five subtasks (radical, rhyme, semantic, perceptual, and consistency judgment). The subtask order was counterbalanced across runs using a Latin square design. Each block lasted 24 s, with an inter-block interval of 16 s. During the inter-block interval, a 12-s fixation was presented, followed by a 4-s task cue indicating the task type of the upcoming block. Each block consisted of six trials. Each trial began with a 500-ms fixation, followed by the first stimulus for 500 ms, a blank screen for 500 ms, and then the second stimulus for 2500 ms, during which participants were instructed to respond as quickly and accurately as possible. The entire experiment lasted approximately 21 min.

[IMAGE OMITTED. SEE PDF]

Word Reading Tasks

Based on the localization of common and specific networks, we employed word reading tasks (i.e., the structural judgment and familiarity judgment tasks) to further (1) examine the lexical representations of those networks, and (2) explore the effects of processing demands on their lexical representations. These tasks were selected because Chinese character reading relies heavily on visuo-orthographic and semantic processing, and previous studies have shown that structural and familiarity judgments effectively engage visuo-orthographic processing and semantic processing, respectively (Fernandino et al. 2022; Hyde and Jenkins 1973; Zhang, Yang, and Jiang 2020). Thus, adopting these tasks allows us to directly examine how processing demands modulate lexical representations in both common and specific networks. The two tasks were matched in terms of both materials and procedures to ensure that any observed differences could be attributed solely to processing demands.

Specifically, we selected a relatively large sample of 198 medium- to high-frequency Chinese characters as materials to construct a representative dataset, thereby enhancing the ecological validity of the analysis. The selected characters had an average stroke count of 10.09 (SD = 2.81, range: 4–19) and a medium-to-high frequency, with an average of 217.05 occurrences per million (SD = 907.72).

Both the structural judgment and familiarity judgment tasks used the event-related design (Figure 1B). Each task consisted of 6 runs, with 66 trials in each run. Each trial began with a 1-s fixation, followed by a Chinese character for 0.5 s, and then a blank screen for 2.5 to 11.5 s (M = 4.5 s). In each task, every Chinese character was presented twice, once in the first three runs and once again in the last three runs. During the structural judgment task, participants had to determine whether the presented character had a left–right structure (e.g., “酸”, where “酉” is on the left and “夋” is on the right) or a non-left–right structure (e.g., “悲”, where “心” is below “非”), with an equal number of characters in each structure type. During the familiarity judgment task, participants were instructed to evaluate how frequently the concept represented by the character appeared in their daily lives (i.e., “rarely”, “occasionally”, or “frequently”). The order of the tasks and the response key mappings were counterbalanced across participants to control for potential biases. The total scanning duration was approximately 80 min, with each task lasting about 40 min.

Image Acquisition

All MRI data were acquired using a 3.0 T Siemens scanner at the Brain Imaging Center of South China Normal University. Functional images were obtained with a T₂*-weighted single-shot echo-planar imaging (EPI) sequence using the following parameters: TR = 2000 ms, TE = 30 ms, θ = 90°, field of view (FOV) = 224 × 224 mm, matrix size = 112 × 112, voxel size = 2 × 2 × 3 mm³, slice thickness = 2 mm, number of axial slices = 58. Structural images were acquired with a T₁-weighted three-dimensional gradient echo pulse sequence with the following parameters: TR = 2530 ms, TE = 1.94 ms, θ = 7°, FOV = 256 × 256 mm, slice thickness = 1 mm, matrix size = 256 × 256, number of sagittal slices = 176.

Data Preprocessing

Data preprocessing was conducted using FSL's FEAT 6.00 (). The first six volumes of each run were automatically discarded to stabilize the MRI signal. Standard preprocessing steps were then applied, including motion correction, alignment, smoothing, and registration. Specifically, spatial smoothing was performed using a 5-mm Gaussian kernel, followed by temporal filtering with a 100-s non-linear high-pass filter. Finally, a two-step registration process was performed: functional images were first registered to the individual's structural images and then normalized to the Montreal Neurological Institute (MNI) template (Jenkinson and Smith 2001).

Activation Analysis

To identify common and specific networks underlying visual, phonological, and semantic processing during Chinese character reading, we conducted a whole-brain activation analysis for the localizer task. First, for each participant and each run, preprocessed data were entered into a general linear model (GLM), in which task onset times and durations were convolved with a standard double-gamma hemodynamic response function (HRF). Task cues were also convolved as events to ensure that the fixation served as the baseline. Six motion parameters and temporal derivatives were included as covariates to account for head motion and improve statistical sensitivity. Importantly, three contrasts were constructed to obtain activations involved in visual, phonological, and semantic processing of Chinese characters: (1) visual processing: radical judgment > consistency judgment, (2) phonological processing: rhyme judgment > perceptual judgment, and (3) semantic processing: semantic judgment > perceptual judgment. Then, first-level data were concentrated across runs using a fixed-effects model and subsequently entered into a group-level random-effects analysis. Statistical significance was determined using a threshold of Z > 2.6 and a cluster-corrected p < 0.05 based on Gaussian random field (GRF) theory.

To identify the common network, a conjunction analysis was conducted at the group level to determine overlapping brain regions across the activation maps of visual, phonological, and semantic processing. In parallel, we identified the specific networks associated with each processing type. To do this, we first used activation maps obtained for each processing type as inclusive masks, thereby constraining subsequent analyses to regions specifically engaged in each processing type. Next, at the individual level, we constructed the following contrasts to identify the specific networks: (1) visual-specific: radical judgment > (rhyme judgment + semantic judgment), (2) phonological-specific: rhyme judgment > (radical judgment + semantic judgment), and (3) semantic-specific: semantic judgment > (radical judgment + rhyme judgment). Finally, at the group level, we excluded regions overlapping with the common network, as well as those shared by all three specific processes. The remaining regions were considered to be visual-specific, phonological-specific, and semantic-specific networks. Notably, a relatively liberal threshold of Z > 2.3 and a cluster-level correction of p < 0.05 were adopted to enhance sensitivity in identifying both common and specific networks.

Whole-brain activation analyses were also performed for the word reading tasks. In the first-level analysis, the onset and duration of each stimulus in each run were modeled as predictors for each task and each participant. The contrast of Chinese characters > fixation was constructed to obtain activations elicited by Chinese characters. Subsequently, first-level contrast images were concentrated across runs using a fixed-effects model to generate second-level estimates. To examine the activation differences between the two tasks, two contrasts were created: structural judgment > familiarity judgment to identify regions more strongly activated during the structural judgment task, and the reverse contrast familiarity judgment > structural judgment to identify regions more strongly activated during the familiarity judgment task. Finally, second-level contrast images were entered into group-level random-effects analyses. Statistical significance was determined using a height threshold of Z > 2.6 and a cluster-level threshold of p < 0.05, corrected for multiple comparisons using GRF theory.

Representational Similarity Analysis

To further investigate the lexical representations of common and specific networks and the effect of processing demands, we conducted RSA on the word reading tasks. The specific steps of the RSA were as follows:

First, neural RDMs were constructed for 198 Chinese characters to characterize the activation patterns of common and specific networks during the word reading tasks (i.e., structural and familiarity judgment tasks). Data preprocessing and first-level modeling followed the same procedures as those used in the activation analyses, except that no spatial smoothing was applied. For each participant, activation patterns for each character were estimated using least-squares estimation with ridge regression, and then the resulting beta maps were normalized to MNI space. Activation patterns for all characters were then extracted from the common and specific networks identified in the localizer task. Given the large stimulus set (198 Chinese characters) and only two repetitions per stimulus, which may reduce the stability of single-stimulus responses, we averaged activation patterns across participants to enhance representational robustness (Fernandino et al. 2022). This group-average approach has been shown to better replicate the representational geometry (Nili et al. 2014) and has been adopted in recent RSA studies (Fernandino et al. 2022). Subsequently, Pearson correlations were computed between the activation patterns of all character pairs, resulting in a 198 × (198–1)/2 similarity matrix. Dissimilarity values were then calculated as one minus the correlation coefficients, followed by Fisher-Z transformation to normalize the distribution.

Second, three model-based predicted RDMs were constructed to characterize the visual, phonological, and semantic features of 198 Chinese characters. For the visual RDM, each character was converted into a binary image in which background pixels were assigned a value of 0 and character pixels a value of 1. Pearson correlations were computed between all characters to generate a 198 × (198–1)/2 similarity matrix. Dissimilarity values were then derived by subtracting the correlation coefficients from 1, followed by applying Fisher-Z transformation. The phonological RDM was constructed based on the lexical constituency model (Perfetti et al. 2005), in which each character was decomposed into 23 consonants, 34 vowels, 5 tones, and an additional unit, forming 63 binary units. Pearson correlations were then performed between the units of all pairs of Chinese characters, with dissimilarity values subsequently derived by subtracting the correlation coefficients from 1 and applying Fisher-Z transformation. The semantic RDM was constructed using six major semantic dimensions (i.e., vision, motion, socialness, emotion, time, and space) from a large database called the Six Semantic Dimension Database (SSDD) (Wang, Zhang, et al. 2023), which provided estimated ratings for commonly used Chinese words. These ratings were generated by a computational model trained to map subjective ratings of the six dimensions onto word2vec embeddings, allowing it to predict semantic ratings for other words. The six dimensions were selected because they reflect embodied semantic representations essential for reading comprehension (Davis and Yee 2021; Ralph et al. 2017) and have been shown to capture semantic representations more accurately than taxonomic or distributional models (Chemero 2023; Fernandino et al. 2022; Xu et al. 2025; Zhang, Wu, et al. 2024). The dissimilarity values between all pairs of characters were calculated as the sum of squared differences across the six dimensions. To assess the potential interdependence among the dissimilarity matrices, we computed Spearman rank correlations between the three model-based predicted RDMs. The results revealed no significant correlation between any pair of RDMs: visual versus phonological, r = −0.0015, p = 0.834; visual versus semantic, r = −0.0043, p = 0.547; phonological versus semantic, r = 0.0001, p = 0.967. These findings indicate that the three model-based RDMs are largely independent, with each RDM capturing distinct lexical information. Additionally, a supplementary semantic RDM was constructed based on BERT-derived embeddings to further validate the results. Each Chinese character was represented by a 768-dimensional embedding vector, and Pearson correlations were computed between all pairs to generate a 198 × (198–1)/2 matrix. Dissimilarity values were then derived by subtracting the correlation coefficients from 1, followed by applying Fisher's Z transformation. We also constructed an orthographic RDM to characterize orthographic features among the 198 characters. Each character was represented as a 270-dimensional binary vector capturing structural features, radical-level orthographic features, and stroke-level orthographic properties (Yang et al. 2009). Pairwise dissimilarities were computed using the same procedure as in the visual and phonological models.

Finally, for each network, partial Spearman's rank correlations were computed between its neural RDM and each of the model-based predicted RDMs. It is worth noting that when examining a predicted RDM (e.g., visual), the other RDMs (e.g., phonological and semantic) were included as covariates. All correlation coefficients were transformed into Fisher's Z-scores. To assess the statistical significance of these correlations, non-parametric permutation tests were performed. In each permutation, the values within the model-based predicted RDMs were randomly shuffled, and the partial Spearman's rank correlation between the permuted predicted RDM and the neural RDM was recomputed, while the other two predicted RDMs were controlled as covariates. This procedure was repeated 5000 times to generate a null distribution. Nonparametric statistical tests were then performed to compare the observed correlation coefficients against null distributions. Moreover, to rule out potential effects of task-related reaction times, we constructed additional RDMs based on reaction times for each task. These RDMs were included as covariates in a supplementary analysis to further validate the robustness of the findings. Specifically, for each task, we first averaged the reaction times across all participants for each character, and then computed the Euclidean distances between all pairs of characters to construct the RDMs. In addition, to control for the potential confounding effect of word frequency on familiarity judgments, we also included it as a covariate in a supplementary analysis in the familiarity judgment task. Specifically, a frequency-based RDM was constructed by computing the Euclidean distance between the word frequencies of all possible pairs of the 198 Chinese characters.

Furthermore, given the well-established role of the VWFA in orthographic processing (Cohen et al. 2002; Dehaene and Cohen 2011; Glezer et al. 2009, 2015, 2016), we examined its orthographic representations using the same procedure as above. Specifically, the VWFA was defined as a spherical region with a 9-mm radius centered on the coordinates (−43, −61, −10) from White et al. (2023).

Functional Connectivity Analyses

To further examine how processing demands modulate the interactions among the common and specific networks, we conducted functional connectivity analyses between networks in both the structural judgment and familiarity judgment tasks. Specifically, for each participant and each task, we first extracted the mean time course from the common and specific networks and then conducted Pearson correlations between these networks. The correlation coefficients were transformed into Fisher's Z-scores and subsequently averaged across runs. Task-related differences in connectivity strength were assessed using paired-sample t-tests for six network pairs: common–visual (CV), common–phonological (CP), common–semantic (CS), visual–phonological (VP), visual–semantic (VS), and phonological–semantic (PS). p-values were corrected for multiple comparisons using the false discovery rate (FDR) method.

Results

Behavioral Performance

For the localizer task, as shown in Figure 1C, the accuracy for all tasks was above 0.97, indicating that participants performed the tasks diligently. One-way repeated-measures ANOVAs were conducted on both accuracy and reaction times. For accuracy, results revealed a significant main effect of task, F (4, 124) = 3.31, p = 0.013, η²_p = 0.097. Post hoc comparisons found that accuracy in the semantic judgment task was significantly lower than that in the consistency judgment task (p = 0.007). For reaction times, a significant main effect of task was also found, F (4, 124) = 75.39, p < 0.001, η²_p = 0.709. Post hoc comparisons revealed that reaction times in the radical, rhyme, and semantic judgment tasks were significantly longer than those in the baseline tasks (i.e., consistency and perceptual judgment; p < 0.001). Moreover, the rhyme judgment task elicited significantly longer reaction times than both radical and semantic judgment tasks (p < 0.001). Reaction times in the perceptual judgment task were also significantly longer than those in the consistency judgment task (p < 0.001). No significant difference was observed between the radical and semantic judgment tasks (p = 0.298).

For the structural judgment task, the mean reaction time and accuracy were 679.79 ms (SD = 138.33 ms) and 0.98 (SD = 0.02), respectively. For the familiarity judgment task, the mean reaction time was 1082.45 ms (SD = 228.50 ms). A paired sample t-test showed that reaction times in the familiarity judgment task were significantly longer than those in the structural judgment task, t (31) = 9.32, p < 0.001, Cohen's d = 1.65.

Common and Specific Networks Underlying Visual, Phonological, and Semantic Processing

Whole-brain activation analyses revealed that, during the localizer task, visual, phonological, and semantic processing activated a broad reading-related brain network, including the bilateral prefrontal cortex, temporoparietal cortex, and temporo-occipital cortex (Figure S1A–C). More importantly, we identified common and specific networks underlying visual, phonological, and semantic processing during Chinese character reading (Figure 2). Specifically, the common network primarily involved the left prefrontal cortex, posterior inferior temporal gyrus, medial superior frontal gyrus, and bilateral insula. The visual-specific network was mainly localized in the bilateral superior parietal lobule and inferior temporal gyrus. The phonological-specific network mainly included the left precentral gyrus, supramarginal gyrus, and bilateral insula. The semantic-specific network primarily included the bilateral orbitofrontal cortex, left inferior frontal gyrus, medial superior frontal gyrus, and middle temporal gyrus.

[IMAGE OMITTED. SEE PDF]

For the structural judgment and familiarity judgment tasks, reading Chinese characters activated a wide range of reading-related brain regions, including the bilateral prefrontal, occipitoparietal, and temporo-occipital cortices (Figure S1D,E). Further comparisons showed that the structural judgment task elicited greater activation in the bilateral lateral occipital cortex, supramarginal gyrus, insula, and left middle frontal gyrus (Figure S1F). In contrast, the familiarity judgment task elicited greater activation in the bilateral orbitofrontal cortex, paracingulate gyrus, as well as the left prefrontal cortex and angular gyrus (Figure S1G).

The Lexical Representations in the Common and Specific Networks Were Modulated by Processing Demands

To further investigate how common and specific networks represented lexical information of Chinese characters, and whether processing demands modulated their lexical representations, we conducted RSA analyses for each network under both structural judgment and familiarity judgment tasks.

As shown in Figure 3, during the structural judgment task, significant visual representations were found in the common network (r = 0.020, p = 0.014) and visual-specific network (r = 0.027, p < 0.001). Neither the phonological nor the semantic representations were significant (the smallest p = 0.05).

[IMAGE OMITTED. SEE PDF]

As shown in Figure 4, during the familiarity judgment task, significant semantic representations were found in the common network (r = 0.054, p < 0.001) and semantic-specific network (r = 0.134, p < 0.001). Neither the visual nor the phonological representations were significant (the smallest p = 0.05). Detailed values can be found in Table S1. The results remained consistent with the above findings after controlling for both reaction times (in both tasks) and word frequency (in the familiarity judgment task) (Figures S2–S4).

[IMAGE OMITTED. SEE PDF]

However, as shown in Figure 5, no significant orthographic representations were observed in common and specific networks. In contrast, the VWFA exhibited significant orthographic representations during the structural judgment task (r = 0.168, p = 0.020), whereas no significant orthographic representations were found in the familiarity judgment task.

[IMAGE OMITTED. SEE PDF]

Taken together, the RSA results suggested distinct functional roles for the common and specific networks. First, the common network was capable of representing multiple types of lexical information (e.g., visual and semantic), whereas specific networks exhibited selective representations, representing only the type of information corresponding to their respective linguistic processing type. Second, the representational patterns of common and specific networks were modulated by processing demands. During the structural judgment task, which placed high demands on visuo-orthographic processing, both the common and visual-specific networks represented visual information. In contrast, during the familiarity judgment task, which imposed high demands on semantic processing, both the common and semantic-specific networks represented semantic information.

The Functional Connectivity Among the Common and Specific Networks Was Modulated by Processing Demands

As shown in Figure 6, paired-sample t-tests showed that the functional connectivity between the common network and visual-specific network in the structural judgment task was significantly stronger than in the familiarity judgment task (t (31) = 3.96, p < 0.001, Cohen's d = 0.70). In contrast, the functional connectivity between the common network and semantic-specific network (t (31) = 5.40, p < 0.001, Cohen's d = 0.95), as well as between the phonological-specific and semantic-specific networks (t(31) = 2.77, p = 0.018, Cohen's d = 0.49) in the familiarity judgment task was significantly stronger than in the structural judgment task. No other network pairs showed significant differences between tasks (ps > 0.05). These results indicated that processing demands selectively modulated the functional connectivity among the common and specific networks.

[IMAGE OMITTED. SEE PDF]

Discussion

The present study aimed to precisely identify common and specific neural networks underlying visual, phonological, and semantic processing, and to examine how those networks represented different types of lexical information, as well as how such representations were modulated by processing demands, and how processing demands affected functional connectivity among networks. Activation analyses revealed that visual, phonological, and semantic processing of Chinese characters recruited both common and specific networks. Critically, RSA revealed that the common network represented different types of lexical information (i.e., visual and semantic information), whereas the specific networks selectively represented the lexical information relevant to their corresponding processing type. Moreover, the lexical representations in both common and specific networks were modulated by task-induced processing demands, but in distinct ways. The common network exhibited representational flexibility, representing visual information during the structural judgment task and semantic information during the familiarity judgment task. By contrast, the specific networks exhibited representational selectivity, representing their corresponding lexical information only under high processing demands. Specifically, the visual-specific network represented visual information exclusively during the structural judgment task, whereas the semantic-specific network represented semantic information only during the familiarity judgment task. Similarly, the modulatory effect of processing demands was also observed in functional connectivity. Specifically, during the structural judgment task, connectivity between the common and visual-specific networks was increased, whereas during the familiarity judgment task, connectivity between the common and semantic-specific networks and between the phonological and semantic networks was enhanced.

The results of our study made several key contributions to the understanding of the common and specific neural mechanisms underlying visual, phonological, and semantic processing in Chinese reading. First, our study precisely distinguished the common and specific networks associated with visual, phonological, and semantic processing. In contrast to previous studies that typically focused on a particular processing type using only one reading task (Guo et al. 2022; Tan et al. 2005; Wu et al. 2012), our study comprehensively adopted a within-subject design to examine all three processes within the same group of participants. This design enabled a more accurate dissociation of the neural substrates underlying each processing type (Li, Wu, et al. 2022; Liu, Tao, et al. 2022; Paz-Alonso et al. 2018). Building on this approach, our activation results identified a common network that was activated across visual, phonological, and semantic processing during Chinese character reading. This network primarily included the left prefrontal cortex, posterior inferior temporal gyrus, medial superior frontal gyrus, and bilateral insula. These findings were consistent with prior findings (Liu, Tao, et al. 2022; Wu et al. 2012), indicating that these regions formed a common network jointly supporting multiple types of lexical processing. Notably, the middle frontal gyrus in the common network has been consistently recognized as a critical region involved in Chinese character reading, serving as an integration hub that coordinates inputs from visuospatial regions as well as phonological and semantic networks (Guo et al. 2022; Liu, Tao, et al. 2022; Siok et al. 2004; Tan et al. 2005). Consistent with this view, the present study found that the middle frontal gyrus was activated across visuospatial, phonological, and semantic processing of Chinese characters, further suggesting its general role that supports the integration and coordination of multiple types of lexical information. In addition to the common network, we also identified distinct brain regions that were selectively involved in the three processing types. Specifically, visual processing selectively engaged the bilateral superior parietal lobule and inferior temporal gyrus; phonological processing selectively engaged the left precentral gyrus, supramarginal gyrus, and bilateral insula; and semantic processing primarily recruited the bilateral orbitofrontal cortex, left inferior frontal gyrus, medial superior frontal gyrus, and middle temporal gyrus. Consistent with prior studies (Liu, Tao, et al. 2022; Reilhac et al. 2013; Weiss et al. 2018; Wu et al. 2012), these findings suggest word reading engages not only a common network but also distinct networks specialized for visual, phonological, and semantic processing.

Secondly, our study was the first to reveal how common and specific networks represented different types of lexical information, thereby providing new insights into the functional roles of those networks in word reading. RSA results revealed distinct representational patterns of the common and specific networks. Specifically, the common network was capable of representing different types of lexical information (i.e., visual and semantic information), indicating that it may play a central role in word reading by supporting the processing and integration of multiple lexical information (Liu, Tao, et al. 2022). Critically, these results provided empirical support for the view of distributed representations, wherein lexical information is not represented in isolated brain regions, but rather in a distributed and overlapping manner across the brain (Binder et al. 2009; Haxby et al. 2001; Huth et al. 2016; Price 2012). In contrast, the specific networks exhibited functionally specialized representational patterns. The visual-specific network selectively represented visual information, while the semantic-specific network selectively represented semantic information. These results suggested a relatively modular organization of word reading, where distinct networks are engaged in processing different types of lexical information (Binder et al. 2009; Dehaene and Cohen 2011; Price 2012). Taken together, the above findings, viewed from a multivariate analytical perspective, suggest that word reading relies on both distributed overlapping and functionally specific networks to support efficient word recognition. Moreover, we also examined orthographic representations using a computational orthographic model (Yang et al. 2009; Zhao et al. 2017) in both the common and specific networks, but no significant orthographic representations were observed. This may be largely attributed to the nature of the radical judgment task. Although this task has been widely used to investigate visuo-orthographic processing in Chinese reading (Cao et al. 2010; Li, Wu, et al. 2022; Liu, Tao, et al. 2022; Liu et al. 2025; Wei et al. 2024), it primarily involves fine-grained, sub-lexical visuospatial analysis of character components rather than abstract, lexical-level orthographic processing. In contrast, the orthographic model proposed by Yang et al. (2009) encodes character features that are important for orthography-to-phonology and other mappings (Yang et al. 2013), mainly capturing abstract orthographic information rather than sub-lexical visuospatial information (Zhao et al. 2017). In line with this distinction, significant orthographic representations were observed in the VWFA, which is consistent with previous studies and highlights the critical role of the VWFA in orthographic processing during visual word reading (Dehaene and Cohen 2011; Glezer et al. 2009, 2016; Guo et al. 2022). Together, these findings further suggest that the VWFA is sensitive to abstract orthographic representations, whereas the common and visual-specific networks may primarily reflect visuospatial information of Chinese characters. Additionally, it was noteworthy that when using a BERT-based semantic model, significant semantic representations were observed only in the semantic-specific network, while the common network showed no significant representations (Figure S5). This may be attributed to differences between semantic models. The six-dimensional model reflects experience-based semantic information of individual characters, whereas BERT embeddings primarily capture co-occurrence-based contextual regularities, encompassing semantic, syntactic, and other linguistic features (Rogers et al. 2020). Such characteristics may introduce additional variance, reducing sensitivity to semantic representations in distributed networks like the common network. Nevertheless, the significant semantic representations in the semantic-specific network support the robustness of our findings.

Third, our study provided empirical evidence for the differential modulation of lexical representations by processing demands in common and specific networks. In the present study, we observed that the modulation of lexical representations by processing demands differed between the common and specific networks. Specifically, the common network exhibited flexible representations of lexical information (i.e., visual or semantic) under high processing demands (i.e., visuo-orthographic or semantic). By contrast, for specific networks, lexical representations emerged only under high processing demands and were absent under low processing demands. This dissociation may be attributed to the distinct functional roles of common and specific networks in supporting lexical processing. The common network appears to act as a flexible integrative hub, capable of coordinating and integrating multiple types of lexical information according to the task-induced processing demands. In contrast, the specific networks exhibit stable selectivity for particular types of lexical processing and become engaged under high processing demands to support the processing of corresponding lexical information. This highlights the functionally specific nature of these networks in lexical processing. These findings were also consistent with prior studies, indicating the critical role of processing demands in shaping the function of brain regions (Chen et al. 2013; Li et al. 2024; Qu et al. 2022; Yang and Zevin 2014). In particular, task-induced processing demands dynamically modulate neural representations, allowing brain networks to flexibly configure information according to cognitive requirements (Bailey et al. 2025; Branzi et al. 2022; Nastase et al. 2017; Wang et al. 2018; Zhang, Liu, et al. 2020).

Finally, our study revealed that processing demands differentially modulated the functional connectivity between the common and specific networks. During the structural judgment task, which placed high demands on visuo-orthographic processing, connectivity between the common and visual-specific networks was enhanced. In contrast, during the familiarity judgment task, which placed high demands on semantic processing, connectivity between the common and semantic-specific networks, as well as that between the phonological and semantic networks, was enhanced. These findings were consistent with previous studies, suggesting that processing demands not only influence local activity and neural representational patterns but also modulate the functional connectivity among networks (Bailey et al. 2025; Li et al. 2023, 2024; Mano et al. 2013; Planton et al. 2022), thereby supporting visuospatial analysis or semantic retrieval. More importantly, these findings were also in line with our RSA results, indicating that the common network not only exhibited shared and flexible lexical representations, but also could adjust its connectivity patterns with specific networks according to task requirements. Taken together, the RSA and functional connectivity results highlight the adaptive role of the common network in response to processing demands.

Two limitations should be discussed. First, we did not observe phonological representations in either the common or specific networks. This is likely due to the lack of a task with high phonological demands. Considering that Chinese reading relies more heavily on visuo-orthographic and semantic processing (Perfetti et al. 2005; Tan et al. 2005), as well as the constraints on task duration, we adopted the structural judgment and familiarity judgment tasks to emphasize visuo-orthographic and semantic processing, respectively. However, both tasks placed relatively low demands on phonological processing. Future research could employ tasks with higher phonological processing demands to further examine how phonological information is represented in both the common and specific networks. Second, the inter-trial intervals (ITIs) in the word reading tasks were relatively short (mean = 4.5 s). Previous research has suggested that longer ITIs could further improve the accuracy of single-trial estimates, which is particularly important for analyses such as MVPA and RSA (Zeithamova et al. 2017). Future studies could consider using longer ITIs to further enhance the reliability of single-trial estimates and then increase the sensitivity of RSA analyses.

Conclusions

To summarize, our study precisely distinguished the common and specific networks for visual, phonological, and semantic processing in reading Chinese characters, and revealed how those networks represented lexical information, and how processing demands modulated both their lexical representations and functional connectivity. We found that the common network represented multiple types of lexical information, while the specific networks selectively represented information corresponding to their respective processing types. Moreover, processing demands modulated the lexical representations in both common and specific networks, but in different ways. Specifically, the common network exhibited representational flexibility, whereas the specific networks showed process-dependent selectivity. In addition, processing demands also modulated connectivity patterns among networks, particularly between the common and specific networks. These findings provide significant insights into the common and specific neural mechanisms underlying word reading.

Author Contributions

Yuan Feng: data curation, formal analysis, investigation, visualization, writing – original draft, and writing – review and editing. Aqian Li: data curation, writing – review and editing. Xinqi Su: writing – review and editing. Huihui Zhu: writing – review and editing. Yujie Cao: writing – review and editing. Leilei Mei: conceptualization, funding acquisition, methodology, supervision, and writing – review and editing.

Acknowledgments

We sincerely thank Professor Jianfeng Yang for assistance regarding the orthographic model described in Yang et al. (2009).

Funding

This study was supported by grants from the National Natural Science Foundation of China (32571226, 32271098), Research Center for Brain Cognition and Human Development, Guangdong, China (2024B0303390003), and the Guangdong Basic and Applied Basic Research Foundation (2024A1515011023), and Striving for the First-Class, Improving Weak Links and Highlighting Features (SIH) Key Discipline for Psychology in South China Normal University.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The data and codes are available upon reasonable request to the corresponding author.

References

Bailey, L. M., H. E. Matheson, J. M. Fawcett, G. E. Bodner, and A. J. Newman. 2025. “Differential Weighting of Information During Aloud and Silent Reading: Evidence From Representational Similarity Analysis of fMRI Data.” Imaging Neuroscience 3: imag_a_00428. https://doi.org/10.1162/imag_a_00428.

Benischek, A., X. Long, C. S. Rohr, S. Bray, D. Dewey, and C. Lebel. 2020. “Pre‐Reading Language Abilities and the Brain's Functional Reading Network in Young Children.” NeuroImage 217: 116903. https://doi.org/10.1016/j.neuroimage.2020.116903.

Binder, J. R., R. H. Desai, W. W. Graves, and L. L. Conant. 2009. “Where Is the Semantic System? A Critical Review and Meta‐Analysis of 120 Functional Neuroimaging Studies.” Cerebral Cortex 19, no. 12: 2767–2796. https://doi.org/10.1093/cercor/bhp055.

Branzi, F. M., and M. A. Lambon Ralph. 2023. “Semantic‐Specific and Domain‐General Mechanisms for Integration and Update of Contextual Information.” Human Brain Mapping 44, no. 17: 5547–5566. https://doi.org/10.1002/hbm.26454.

Branzi, F. M., C. D. Martin, and P. M. Paz‐Alonso. 2022. “Task‐Relevant Representations and Cognitive Control Demands Modulate Functional Connectivity From Ventral Occipito‐Temporal Cortex During Object Recognition Tasks.” Cerebral Cortex 32, no. 14: 3068–3080. https://doi.org/10.1093/cercor/bhab401.

Cai, Q., and M. Brysbaert. 2010. “SUBTLEX‐CH: Chinese Word and Character Frequencies Based on Film Subtitles.” PLoS One 5, no. 6: e10729. https://doi.org/10.1371/journal.pone.0010729.

Cao, F., R. Lee, H. Shu, et al. 2010. “Cultural Constraints on Brain Development: Evidence From a Developmental Study of Visual Word Processing in Mandarin Chinese.” Cerebral Cortex 20, no. 5: 1223–1233. https://doi.org/10.1093/cercor/bhp186.

Cao, F., X. Yan, Z. Wang, et al. 2017. “Neural Signatures of Phonological Deficits in Chinese Developmental Dyslexia.” NeuroImage 146: 301–311. https://doi.org/10.1016/j.neuroimage.2016.11.051.

Cattinelli, I., N. A. Borghese, M. Gallucci, and E. Paulesu. 2013. “Reading the Reading Brain: A New Meta‐Analysis of Functional Imaging Data on Reading.” Journal of Neurolinguistics 26, no. 1: 214–238. https://doi.org/10.1016/j.jneuroling.2012.08.001.

Chai, X. J., J. A. Berken, E. B. Barbeau, et al. 2016. “Intrinsic Functional Connectivity in the Adult Brain and Success in Second‐Language Learning.” Journal of Neuroscience 36, no. 3: 755–761. https://doi.org/10.1523/JNEUROSCI.2234‐15.2016.

Chemero, A. 2023. “LLMs Differ From Human Cognition Because They Are Not Embodied.” Nature Human Behaviour 7, no. 11: 1828–1829. https://doi.org/10.1038/s41562‐023‐01723‐5.

Chen, Y., M. H. Davis, F. Pulvermüller, and O. Hauk. 2013. “Task Modulation of Brain Responses in Visual Word Recognition as Studied Using EEG/MEG and fMRI.” Frontiers in Human Neuroscience 7: 376. https://doi.org/10.3389/fnhum.2013.00376.

Chen, Y., A. Shimotake, R. Matsumoto, et al. 2016. “The ‘When’ and ‘Where’ of Semantic Coding in the Anterior Temporal Lobe: Temporal Representational Similarity Analysis of Electrocorticogram Data.” Cortex 79: 1–13. https://doi.org/10.1016/j.cortex.2016.02.015.

Cocchi, L., A. Zalesky, A. Fornito, and J. B. Mattingley. 2013. “Dynamic Cooperation and Competition Between Brain Systems During Cognitive Control.” Trends in Cognitive Sciences 17, no. 10: 493–501. https://doi.org/10.1016/j.tics.2013.08.006.

Cohen, L., and S. Dehaene. 2004. “Specialization Within the Ventral Stream: The Case for the Visual Word Form Area.” NeuroImage 22, no. 1: 466–476. https://doi.org/10.1016/j.neuroimage.2003.12.049.

Cohen, L., S. Lehéricy, F. Chochon, C. Lemer, S. Rivaud, and S. Dehaene. 2002. “Language‐Specific Tuning of Visual Cortex? Functional Properties of the Visual Word Form Area.” Brain 125, no. 5: 1054–1069. https://doi.org/10.1093/brain/awf094.

Coltheart, M., K. Rastle, C. Perry, R. Langdon, and J. Ziegler. 2001. “DRC: A Dual Route Cascaded Model of Visual Word Recognition and Reading Aloud.” Psychological Review 108, no. 1: 204–256. https://doi.org/10.1037/0033‐295X.108.1.204.

Davis, C. P., and E. Yee. 2021. “Building Semantic Memory From Embodied and Distributional Language Experience.” Wiley Interdisciplinary Reviews: Cognitive Science 12, no. 5: e1555. https://doi.org/10.1002/wcs.1555.

Dehaene, S., and L. Cohen. 2011. “The Unique Role of the Visual Word Form Area in Reading.” Trends in Cognitive Sciences 15, no. 6: 254–262. https://doi.org/10.1016/j.tics.2011.04.003.

Dehaene, S., L. Cohen, J. Morais, and R. Kolinsky. 2015. “Illiterate to Literate: Behavioural and Cerebral Changes Induced by Reading Acquisition.” Nature Reviews Neuroscience 16, no. 4: 234–244. https://doi.org/10.1038/nrn3924.

Dehaene, S., F. Pegado, L. W. Braga, et al. 2010. “How Learning to Read Changes the Cortical Networks for Vision and Language.” Science 330, no. 6009: 1359–1364. https://doi.org/10.1126/science.1194140.

Deng, Y., T. Chou, G. Ding, D. Peng, and J. R. Booth. 2011. “The Involvement of Occipital and Inferior Frontal Cortex in the Phonological Learning of Chinese Characters.” Journal of Cognitive Neuroscience 23, no. 8: 1998–2012. https://doi.org/10.1162/jocn.2010.21571.

Feng, X., I. Altarelli, K. Monzalvo, et al. 2020. “A Universal Reading Network and Its Modulation by Writing System and Reading Ability in French and Chinese Children.” eLife 9: e54591. https://doi.org/10.7554/eLife.54591.

Fernandino, L., J.‐Q. Tong, L. L. Conant, C. J. Humphries, and J. R. Binder. 2022. “Decoding the Information Structure Underlying the Neural Representation of Concepts.” Proceedings of the National Academy of Sciences of the United States of America 119, no. 6: e2108091119. https://doi.org/10.1073/pnas.2108091119.

Fischer‐Baum, S., D. Bruggemann, I. F. Gallego, D. S. P. Li, and E. R. Tamez. 2017. “Decoding Levels of Representation in Reading: A Representational Similarity Approach.” Cortex 90: 88–102. https://doi.org/10.1016/j.cortex.2017.02.017.

Fox, M. D., A. Z. Snyder, J. L. Vincent, M. Corbetta, D. C. Van Essen, and M. E. Raichle. 2005. “The Human Brain Is Intrinsically Organized Into Dynamic, Anticorrelated Functional Networks.” Proceedings of the National Academy of Sciences 102, no. 27: 9673–9678. https://doi.org/10.1073/pnas.0504136102.

Glezer, L. S., G. Eden, X. Jiang, et al. 2016. “Uncovering Phonological and Orthographic Selectivity Across the Reading Network Using fMRI‐RA.” NeuroImage 138: 248–256. https://doi.org/10.1016/j.neuroimage.2016.05.072.

Glezer, L. S., X. Jiang, and M. Riesenhuber. 2009. “Evidence for Highly Selective Neuronal Tuning to Whole Words in the “Visual Word Form Area”.” Neuron 62, no. 2: 199–204. https://doi.org/10.1016/j.neuron.2009.03.017.

Glezer, L. S., J. Kim, J. Rule, X. Jiang, and M. Riesenhuber. 2015. “Adding Words to the Brain's Visual Dictionary: Novel Word Learning Selectively Sharpens Orthographic Representations in the VWFA.” Journal of Neuroscience 35, no. 12: 4965–4972. https://doi.org/10.1523/JNEUROSCI.4031‐14.2015.

Graves, W. W., J. Purcell, D. Rothlein, D. J. Bolger, M. Rosenberg‐Lee, and R. Staples. 2023. “Correspondence Between Cognitive and Neural Representations for Phonology, Orthography, and Semantics in Supramarginal Compared to Angular Gyrus.” Brain Structure and Function 228, no. 1: 255–271. https://doi.org/10.1007/s00429‐022‐02590‐y.

Gu, L., Y. Pang, J. Yang, J. Qu, N. Gu, and L. Mei. 2024. “Orthographic and Phonological Processing in the Left Ventral Occipitotemporal Cortex During Chinese Word Reading.” Psychophysiology 61, no. 12: e14703. https://doi.org/10.1111/psyp.14703.

Guo, W., S. Geng, M. Cao, and J. Feng. 2022. “The Brain Connectome for Chinese Reading.” Neuroscience Bulletin 38, no. 9: 1097–1113. https://doi.org/10.1007/s12264‐022‐00864‐3.

Haxby, J. V., A. C. Connolly, and J. S. Guntupalli. 2014. “Decoding Neural Representational Spaces Using Multivariate Pattern Analysis.” Annual Review of Neuroscience 37, no. 1: 435–456. https://doi.org/10.1146/annurev‐neuro‐062012‐170325.

Haxby, J. V., M. I. Gobbini, M. L. Furey, A. Ishai, J. L. Schouten, and P. Pietrini. 2001. “Distributed and Overlapping Representations of Faces and Objects in Ventral Temporal Cortex.” Science 293, no. 5539: 2425–2430. https://doi.org/10.1126/science.1063736.

Huang, C., A. Li, Y. Pang, et al. 2024. “How the Intrinsic Functional Connectivity Patterns of the Semantic Network Support Semantic Processing.” Brain Imaging and Behavior 18, no. 3: 539–554. https://doi.org/10.1007/s11682‐024‐00849‐y.

Huth, A. G., W. A. De Heer, T. L. Griffiths, F. E. Theunissen, and J. L. Gallant. 2016. “Natural Speech Reveals the Semantic Maps That Tile Human Cerebral Cortex.” Nature 532, no. 7600: 453–458. https://doi.org/10.1038/nature17637.

Hyde, T. S., and J. J. Jenkins. 1973. “Recall for Words as a Function of Semantic, Graphic, and Syntactic Orienting Tasks.” Journal of Verbal Learning and Verbal Behavior 12, no. 5: 471–480. https://doi.org/10.1016/S0022‐5371(73)80027‐1.

Jackson, R. L., M. A. Lambon Ralph, and G. Pobric. 2015. “The Timing of Anterior Temporal Lobe Involvement in Semantic Processing.” Journal of Cognitive Neuroscience 27, no. 7: 1388–1396. https://doi.org/10.1162/jocn_a_00788.

Jenkinson, M., and S. Smith. 2001. “A Global Optimisation Method for Robust Affine Registration of Brain Images.” Medical Image Analysis 5, no. 2: 143–156. https://doi.org/10.1016/S1361‐8415(01)00036‐6.

Jia, F., C. Y. Liu, L. H. Tan, and W. T. Siok. 2022. “Lifespan Developmental Changes in Neural Substrates and Functional Connectivity for Visual Semantic Processing.” Cerebral Cortex 33, no. 8: 4714–4728. https://doi.org/10.1093/cercor/bhac374.

Krafnick, A. J., L.‐H. Tan, D. L. Flowers, et al. 2016. “Chinese Character and English Word Processing in Children's Ventral Occipitotemporal Cortex: fMRI Evidence for Script Invariance.” NeuroImage 133: 302–312. https://doi.org/10.1016/j.neuroimage.2016.03.021.

Kriegeskorte, N. 2008. “Representational Similarity Analysis – Connecting the Branches of Systems Neuroscience.” Frontiers in Systems Neuroscience 2: 4. https://doi.org/10.3389/neuro.06.004.2008.

Kriegeskorte, N., and R. A. Kievit. 2013. “Representational Geometry: Integrating Cognition, Computation, and the Brain.” Trends in Cognitive Sciences 17, no. 8: 401–412. https://doi.org/10.1016/j.tics.2013.06.007.

Kuo, W. J., T. C. Yeh, J. R. Lee, et al. 2004. “Orthographic and Phonological Processing of Chinese Characters: An fMRI Study.” NeuroImage 21, no. 4: 1721–1731. https://doi.org/10.1016/j.neuroimage.2003.12.007.

Li, A., C. Chen, Y. Feng, et al. 2025. “Functional Divisions of the Left Anterior and Posterior Temporoparietal Junction for Phonological and Semantic Processing in Chinese Character Reading.” NeuroImage 311: 121201. https://doi.org/10.1016/j.neuroimage.2025.121201.

Li, A., C. Chen, X. Wu, et al. 2024. “Processing Demands Modulate the Activities and Functional Connectivity Patterns of the Posterior (VWFA‐1) and Anterior (VWFA‐2) VWFA.” NeuroImage 303: 120923. https://doi.org/10.1016/j.neuroimage.2024.120923.

Li, A., R. Yang, J. Qu, J. Dong, L. Gu, and L. Mei. 2022. “Neural Representation of Phonological Information During Chinese Character Reading.” Human Brain Mapping 43, no. 13: 4013–4029. https://doi.org/10.1002/hbm.25900.

Li, H., Y. Cao, C. Chen, X. Liu, S. Zhang, and L. Mei. 2023. “The Depth of Semantic Processing Modulates Cross‐Language Pattern Similarity in Chinese–English Bilinguals.” Human Brain Mapping 44, no. 5: 2085–2098. https://doi.org/10.1002/hbm.26195.

Li, H., J. Wu, R. A. Marks, et al. 2022. “Functional Mapping and Cooperation Between the Cerebellum and Cerebrum During Word Reading.” Cerebral Cortex 32, no. 22: 5175–5190. https://doi.org/10.1093/cercor/bhac006.

Lin, Y. H., N. B. Dadario, J. Hormovas, et al. 2021. “Anatomy and White Matter Connections of the Superior Parietal Lobule.” Operative Neurosurgery 21, no. 3: E199–E214. https://doi.org/10.1093/ons/opab174.

Liu, C., W. T. Zhang, Y. Y. Tang, et al. 2008. “The Visual Word Form Area: Evidence From an fMRI Study of Implicit Processing of Chinese Characters.” NeuroImage 40, no. 3: 1350–1361. https://doi.org/10.1016/j.neuroimage.2007.10.014.

Liu, C. Y., R. Tao, L. Qin, S. Matthews, and W. T. Siok. 2022. “Functional Connectivity During Orthographic, Phonological, and Semantic Processing of Chinese Characters Identifies Distinct Visuospatial and Phonosemantic Networks.” Human Brain Mapping 43, no. 16: 5066–5080. https://doi.org/10.1002/hbm.26075.

Liu, X., D. Wisniewski, L. Vermeylen, A. F. Palenciano, W. Liu, and M. Brysbaert. 2022. “The Representations of Chinese Characters: Evidence From Sublexical Components.” Journal of Neuroscience 42, no. 1: 135–144. https://doi.org/10.1523/JNEUROSCI.1057‐21.2021.

Liu, X., Y. Yang, Y. Liang, et al. 2025. “Learning Chinese Induces Altered Resting‐State Functional Connectivity Relating to Orthographic and Phonological Processing in Adult Alphabetic Language Speakers.” Brain Structure and Function 230, no. 7: 118. https://doi.org/10.1007/s00429‐025‐02977‐7.

Lobier, M., C. Peyrin, J.‐F. Le Bas, and S. Valdois. 2012. “Pre‐Orthographic Character String Processing and Parietal Cortex: A Role for Visual Attention in Reading?” Neuropsychologia 50, no. 9: 2195–2204. https://doi.org/10.1016/j.neuropsychologia.2012.05.023.

López‐Barroso, D., M. Thiebaut De Schotten, J. Morais, et al. 2020. “Impact of Literacy on the Functional Connectivity of Vision and Language Related Networks.” NeuroImage 213: 116722. https://doi.org/10.1016/j.neuroimage.2020.116722.

Lu, C., H. Li, R. Fu, J. Qu, Q. Yue, and L. Mei. 2021. “Neural Representation in Visual Word Form Area During Word Reading.” Neuroscience 452: 49–62. https://doi.org/10.1016/j.neuroscience.2020.10.040.

Luo, Y., K. Wang, S. Jiao, J. Zeng, and Z. Han. 2024. “Distinct Parallel Activation and Interaction Between Dorsal and Ventral Pathways During Phonological and Semantic Processing: A cTBS‐fMRI Study.” Human Brain Mapping 45, no. 1: e26569. https://doi.org/10.1002/hbm.26569.

Mano, Q. R., C. Humphries, R. H. Desai, et al. 2013. “The Role of Left Occipitotemporal Cortex in Reading: Reconciling Stimulus, Task, and Lexicality Effects.” Cerebral Cortex 23, no. 4: 988–1001. https://doi.org/10.1093/cercor/bhs093.

Martin, L., C. Durisko, M. W. Moore, M. N. Coutanche, D. Chen, and J. A. Fiez. 2019. “The VWFA Is the Home of Orthographic Learning When Houses Are Used as Letters.” eNeuro 6, no. 1: ENEURO.0425‐17.2019. https://doi.org/10.1523/ENEURO.0425‐17.2019.

Mei, L., G. Xue, Z. L. Lu, et al. 2013. “Orthographic Transparency Modulates the Functional Asymmetry in the Fusiform Cortex: An Artificial Language Training Study.” Brain and Language 125, no. 2: 165–172. https://doi.org/10.1016/j.bandl.2012.01.006.

Menon, V. 2023. “20 Years of the Default Mode Network: A Review and Synthesis.” Neuron 111, no. 16: 2469–2487. https://doi.org/10.1016/j.neuron.2023.04.023.

Nastase, S. A., A. C. Connolly, N. N. Oosterhof, et al. 2017. “Attention Selectively Reshapes the Geometry of Distributed Semantic Representation.” Cerebral Cortex 27, no. 8: 4277–4291. https://doi.org/10.1093/cercor/bhx138.

Nili, H., C. Wingfield, A. Walther, et al. 2014. “A Toolbox for Representational Similarity Analysis.” PLoS Computational Biology 10, no. 4: e1003553. https://doi.org/10.1371/journal.pcbi.1003553.

Pattamadilok, C., V. Chanoine, C. Pallier, et al. 2017. “Automaticity of Phonological and Semantic Processing During Visual Word Recognition.” NeuroImage 149: 244–255. https://doi.org/10.1016/j.neuroimage.2017.02.003.

Paz‐Alonso, P. M., M. Oliver, G. Lerma‐Usabiaga, et al. 2018. “Neural Correlates of Phonological, Orthographic and Semantic Reading Processing in Dyslexia.” NeuroImage: Clinical 20: 433–447. https://doi.org/10.1016/j.nicl.2018.08.018.

Perfetti, C. A., E. W. Wlotko, and L. A. Hart. 2005. “Word Learning and Individual Differences in Word Learning Reflected in Event‐Related Potentials.” Journal of Experimental Psychology: Learning, Memory, and Cognition 31, no. 6: 1281–1292. https://doi.org/10.1037/0278‐7393.31.6.1281.

Planton, S., S. Wang, D. Bolger, M. Bonnard, and C. Pattamadilok. 2022. “Effective Connectivity of the Left‐Ventral Occipito‐Temporal Cortex During Visual Word Processing: Direct Causal Evidence From TMS‐EEG Co‐Registration.” Cortex 154: 167–183. https://doi.org/10.1016/j.cortex.2022.06.004.

Price, A. R., M. F. Bonner, J. E. Peelle, and M. Grossman. 2015. “Converging Evidence for the Neuroanatomic Basis of Combinatorial Semantics in the Angular Gyrus.” Journal of Neuroscience 35, no. 7: 3276–3284. https://doi.org/10.1523/JNEUROSCI.3446‐14.2015.

Price, C. J., and J. T. Devlin. 2011. “The Interactive Account of Ventral Occipitotemporal Contributions to Reading.” Trends in Cognitive Sciences 15, no. 6: 246–253. https://doi.org/10.1016/j.tics.2011.04.001.

Price, C. J. 2012. “A Review and Synthesis of the First 20years of PET and fMRI Studies of Heard Speech, Spoken Language and Reading.” NeuroImage 62, no. 2: 816–847. https://doi.org/10.1016/j.neuroimage.2012.04.062.

Purcell, J. J., X. Jiang, and G. F. Eden. 2017. “Shared Orthographic Neuronal Representations for Spelling and Reading.” NeuroImage 147: 554–567. https://doi.org/10.1016/j.neuroimage.2016.12.054.

Qu, J., Y. Pang, X. Liu, Y. Cao, C. Huang, and L. Mei. 2022. “Task Modulates the Orthographic and Phonological Representations in the Bilateral Ventral Occipitotemporal Cortex.” Brain Imaging and Behavior 16, no. 4: 1695–1707. https://doi.org/10.1007/s11682‐022‐00641‐w.

Raichle, M. E., A. M. MacLeod, A. Z. Snyder, W. J. Powers, D. A. Gusnard, and G. L. Shulman. 2001. “A Default Mode of Brain Function.” Proceedings of the National Academy of Sciences 98, no. 2: 676–682. https://doi.org/10.1073/pnas.98.2.676.

Ralph, M. A. L., E. Jefferies, K. Patterson, and T. T. Rogers. 2017. “The Neural and Computational Bases of Semantic Cognition.” Nature Reviews Neuroscience 18, no. 1: 42–55. https://doi.org/10.1038/nrn.2016.150.

Reilhac, C., C. Peyrin, J. F. Démonet, and S. Valdois. 2013. “Role of the Superior Parietal Lobules in Letter‐Identity Processing Within Strings: FMRI Evidence From Skilled and Dyslexicreaders.” Neuropsychologia 51, no. 4: 601–612. https://doi.org/10.1016/j.neuropsychologia.2012.12.010.

Rogers, A., O. Kovaleva, and A. Rumshisky. 2020. “A Primer in BERTology: What We Know About How BERT Works.” Transactions of the Association for Computational Linguistics 8: 842–866. https://doi.org/10.1162/tacl_a_00349.

Saur, D., B. W. Kreher, S. Schnell, et al. 2008. “Ventral and Dorsal Pathways for Language.” Proceedings of the National Academy of Sciences 105, no. 46: 18035–18040. https://doi.org/10.1073/pnas.0805234105.

Seidenberg, M. S., and J. L. McClelland. 1989. “A Distributed, Developmental Model of Word Recognition and Naming.” Psychological Review 96, no. 4: 523–568. https://doi.org/10.1037/0033‐295X.96.4.523.

Siok, W. T., F. Jia, C. Y. Liu, C. A. Perfetti, and L. H. Tan. 2020. “A Lifespan fMRI Study of Neurodevelopment Associated With Reading Chinese.” Cerebral Cortex 30, no. 7: 4140–4157. https://doi.org/10.1093/cercor/bhaa038.

Siok, W. T., C. A. Perfetti, Z. Jin, and L. H. Tan. 2004. “Biological Abnormality of Impaired Reading Is Constrained by Culture.” Nature 431, no. 7004: 71–76. https://doi.org/10.1038/nature02865.

Snyder, P. J., and L. J. Harris. 1993. “Handedness, Sex, Familial Sinistrality Effects on Spatial Tasks.” Cortex 29, no. 1: 115–134. https://doi.org/10.1016/S0010‐9452(13)80216‐X.

Staples, R., and W. W. Graves. 2020. “Neural Components of Reading Revealed by Distributed and Symbolic Computational Models.” Neurobiology of Language 1, no. 4: 381–401. https://doi.org/10.1162/nol_a_00018.

Stevens, W. D., M. H. Tessler, C. S. Peng, and A. Martin. 2015. “Functional Connectivity Constrains the Category‐Related Organization of Human Ventral Occipitotemporal Cortex.” Human Brain Mapping 36, no. 6: 2187–2206. https://doi.org/10.1002/hbm.22764.

Szwed, M., E. Qiao, A. Jobert, S. Dehaene, and L. Cohen. 2014. “Effects of Literacy in Early Visual and Occipitotemporal Areas of Chinese and French Readers.” Journal of Cognitive Neuroscience 26, no. 3: 459–475. https://doi.org/10.1162/jocn_a_00499.

Tan, L. H., A. R. Laird, K. Li, and P. T. Fox. 2005. “Neuroanatomical Correlates of Phonological Processing of Chinese Characters and Alphabetic Words: A Meta‐Analysis.” Human Brain Mapping 25, no. 1: 83–91. https://doi.org/10.1002/hbm.20134.

Vigneau, M., V. Beaucousin, P. Y. Hervé, et al. 2006. “Meta‐Analyzing Left Hemisphere Language Areas: Phonology, Semantics, and Sentence Processing.” NeuroImage 30, no. 4: 1414–1432. https://doi.org/10.1016/j.neuroimage.2005.11.002.

Vigneau, M., V. Beaucousin, P.‐Y. Hervé, et al. 2011. “What Is Right‐Hemisphere Contribution to Phonological, Lexico‐Semantic, and Sentence Processing?” NeuroImage 54, no. 1: 577–593. https://doi.org/10.1016/j.neuroimage.2010.07.036.

Wagley, N., and J. R. Booth. 2022. “Neural Pathways of Phonological and Semantic Processing and Its Relations to Children's Reading Skills.” Frontiers in Neuroscience 16: 984328. https://doi.org/10.3389/fnins.2022.984328.

Wang, J., M. F. Joanisse, and J. R. Booth. 2023. “Learning to Read Strengthens Functional Connectivity Between the Ventral Occipitotemporal Cortex and the Superior Temporal Gyrus During an Auditory Phonological Awareness Task.” Journal of Speech, Language, and Hearing Research 66, no. 11: 4532–4546. https://doi.org/10.1044/2023_JSLHR‐23‐00337.

Wang, S., Y. Zhang, W. Shi, et al. 2023. “A Large Dataset of Semantic Ratings and Its Computational Extension.” Scientific Data 10, no. 1: 106. https://doi.org/10.1038/s41597‐023‐01995‐6.

Wang, X., B. Wang, and Y. Bi. 2023. “Early Language Exposure Affects Neural Mechanisms of Semantic Representations.” eLife 12: e81681. https://doi.org/10.7554/eLife.81681.

Wang, X., Y. Xu, Y. Wang, et al. 2018. “Representational Similarity Analysis Reveals Task‐Dependent Semantic Influence of the Visual Word Form Area.” Scientific Reports 8, no. 1: 3047. https://doi.org/10.1038/s41598‐018‐21062‐0.

Wang, Z., X. Yan, Y. Liu, G. J. Spray, Y. Deng, and F. Cao. 2019. “Structural and Functional Abnormality of the Putamen in Children With Developmental Dyslexia.” Neuropsychologia 130: 26–37. https://doi.org/10.1016/j.neuropsychologia.2018.07.014.

Wei, Y., J. Wang, H. Wang, and P. M. Paz‐Alonso. 2024. “Functional Interactions Underlying Visuospatial Orthographic Processes in Chinese Reading.” Cerebral Cortex 34, no. 9: bhae359. https://doi.org/10.1093/cercor/bhae359.

Weiss, Y., H. G. Cweigenberg, and J. R. Booth. 2018. “Neural Specialization of Phonological and Semantic Processing in Young Children.” Human Brain Mapping 39, no. 11: 4334–4348. https://doi.org/10.1002/hbm.24274.

White, A. L., K. N. Kay, K. A. Tang, and J. D. Yeatman. 2023. “Engaging in Word Recognition Elicits Highly Specific Modulations in Visual Cortex.” Current Biology 33, no. 7: 1308–1320. https://doi.org/10.1016/j.cub.2023.02.042.

Wu, C.‐Y., M.‐H. R. Ho, and S.‐H. A. Chen. 2012. “A Meta‐Analysis of fMRI Studies on Chinese Orthographic, Phonological, and Semantic Processing.” NeuroImage 63, no. 1: 381–391. https://doi.org/10.1016/j.neuroimage.2012.06.047.

Xu, Q., Y. Peng, S. A. Nastase, M. Chodorow, M. Wu, and P. Li. 2025. “Large Language Models Without Grounding Recover Non‐Sensorimotor but Not Sensorimotor Features of Human Concepts.” Nature Human Behaviour 9, no. 9: 1871–1886. https://doi.org/10.1038/s41562‐025‐02203‐8.

Yang, J., B. D. McCandliss, H. Shu, and J. D. Zevin. 2009. “Simulating Language‐Specific and Language‐General Effects in a Statistical Learning Model of Chinese Reading.” Journal of Memory and Language 61, no. 2: 238–257. https://doi.org/10.1016/j.jml.2009.05.001.

Yang, J., H. Shu, B. D. McCandliss, and J. D. Zevin. 2013. “Orthographic Influences on Division of Labor in Learning to Read Chinese and English: Insights From Computational Modeling.” Bilingualism: Language and Cognition 16: 354–366. https://doi.org/10.1017/S1366728912000296.

Yang, J., and J. Zevin. 2014. “The Impact of Task Demand on Visual Word Recognition.” Neuroscience 272: 102–115. https://doi.org/10.1016/j.neuroscience.2014.04.044.

Yang, Y., L. Li, S. De Deyne, B. Li, J. Wang, and Q. Cai. 2024. “Unraveling Lexical Semantics in the Brain: Comparing Internal, External, and Hybrid Language Models.” Human Brain Mapping 45, no. 1: e26546. https://doi.org/10.1002/hbm.26546.

Yen, M., A. T. DeMarco, and S. M. Wilson. 2019. “Adaptive Paradigms for Mapping Phonological Regions in Individual Participants.” NeuroImage 189: 368–379. https://doi.org/10.1016/j.neuroimage.2019.01.040.

Zeithamova, D., M. A. de Araujo Sanchez, and A. Adke. 2017. “Trial Timing and Pattern‐Information Analyses of fMRI Data.” NeuroImage 153: 221–231. https://doi.org/10.1016/j.neuroimage.2017.04.025.

Zhang, Q., H. Wang, C. Luo, J. Zhang, Z. Jin, and L. Li. 2019. “The Neural Basis of Semantic Cognition in Mandarin Chinese: A Combined fMRI and TMS Study.” Human Brain Mapping 40, no. 18: 5412–5423. https://doi.org/10.1002/hbm.24781.

Zhang, W., Y. Liu, X. Wang, and X. Tian. 2020. “The Dynamic and Task‐Dependent Representational Transformation Between the Motor and Sensory Systems During Speech Production.” Cognitive Neuroscience 11, no. 4: 194–204. https://doi.org/10.1080/17588928.2020.1792868.

Zhang, X., S. Yang, and M. Jiang. 2020. “Rapid Implicit Extraction of Abstract Orthographic Patterns of Chinese Characters During Reading.” PLoS One 15, no. 2: e0229590. https://doi.org/10.1371/journal.pone.0229590.

Zhang, Y., S. Wang, X. Dong, J. Yu, and C. Zong. 2024. “Navigating Brain Language Representations: A Comparative Analysis of Neural Language Models and Psychologically Plausible Models.” (arXiv:2404.19364). arXiv. https://doi.org/10.48550/arXiv.2404.19364.

Zhang, Y., W. Wu, D. Mirman, and P. Hoffman. 2024. “Representation of Event and Object Concepts in Ventral Anterior Temporal Lobe and Angular Gyrus.” Cerebral Cortex 34, no. 2: 1–14. https://doi.org/10.1093/cercor/bhad519.

Zhao, L., C. Chen, L. Shao, et al. 2017. “Orthographic and Phonological Representations in the Fusiform Cortex.” Cerebral Cortex 27, no. 11: 5197–5210. https://doi.org/10.1093/cercor/bhw300.

Zhou, W., X. Cui, B. Shi, M. Su, and M. Cao. 2021. “The Development of Brain Functional Connectome During Text Reading.” Developmental Cognitive Neuroscience 48: 100927. https://doi.org/10.1016/j.dcn.2021.100927.

Zhu, L., Y. Nie, C. Chang, J. H. Gao, and Z. Niu. 2014. “Different Patterns and Development Characteristics of Processing Written Logographic Characters and Alphabetic Words: An ALE Meta‐Analysis: Different Phonological Processing Patterns.” Human Brain Mapping 35, no. 6: 2607–2618. https://doi.org/10.1002/hbm.22354.

Zhu, Z., J. X. Zhang, S. Wang, Z. Xiao, J. Huang, and H. C. Chen. 2009. “Involvement of Left Inferior Frontal Gyrus in Sentence‐Level Semantic Integration.” NeuroImage 47, no. 2: 756–763. https://doi.org/10.1016/j.neuroimage.2009.04.086.