The Effects of Age and Reading Experience on the Lifespan Neurodevelopment for Reading Comprehension

Reading comprehension is an essential cognitive skill, equipped for accessing information, acquiring knowledge, and interpreting authors' intentions. This skill is learned and honed during childhood, reaching proficiency in early adulthood, and continues to be utilized in middle and late adulthood, remaining relevant throughout an individual's lifespan (Locher & Pfost, 2020). Several studies have investigated the neural mechanisms of phonological and semantic processing in word-level reading across the lifespan (Jia, Liu, Tan, & Siok, 2023; Siok, Jia, Liu, Perfetti, & Tan, 2020). However, reading comprehension extends beyond word-level reading, involving various processes (Zhou, Cui, Shi, Su, & Cao, 2021; Martin, Schurz, Kronbichler, & Richlan, 2015), such as the recognition and comprehension of words, phrases, sentences, as well as texts. Nevertheless, the neurobiology of reading comprehension across the lifespan has not yet been well investigated.

According to the cognitive view of reading comprehension (Kendeou, Van Den Broek, Helder, & Karlsson, 2014), the process of reading comprehension can be roughly divided into lower-level (LL) and higher-level (HL) processes. The LL process involves the conversion of written codes into meaningful language units (e.g., decoding) and the HL process involves the integration of these language units into a meaningful and coherent mental representation (e.g., semantic integration; Kendeou et al., 2014). The LL process crucially relies on word-level morphological-semantic mapping, whereas semantic integration, syntactic analysis, and cognitive monitoring are crucial to the HL process. Neuroimaging studies have investigated the neural mechanism underlying the above two-level cognitive processes during reading comprehension. Specifically, the LL process involves the decoding of visual word form. Because written Chinese is a logographic script that contains a large amount of visuospatial information, the processing of visual word form requires the involvement of the right middle occipital gyrus (MOG), left superior parietal lobule (SPL), as well as left fusiform gyrus (FG; Xu, Wang, Chen, Fox, & Tan, 2015; Tan, Laird, Li, & Fox, 2005; Tan et al., 2001). In addition, the left middle temporal gyrus (MTG) is usually reported to be responsible for the semantic processing in word-level Chinese reading, which is particularly dependent on semantic processing because of the opacity of orthography (Taylor, Rastle, & Davis, 2013; Wu, Ho, & Chen, 2012; Booth et al., 2003). More than that, the left middle frontal gyrus (MFG) is considered to be a brain region specific to Chinese reading, serving a crucial role in the coordination of visuospatial and semantic analysis (Xu et al., 2015; Tan et al., 2001, 2005; Siok, Perfetti, Jin, & Tan, 2004). However, the HL process encompasses multiple processes at the sentence or text level, such as the processing of semantic control and semantic representation that recruits the frontotemporal neural network (Jedidi et al., 2021; Ralph, Jefferies, Patterson, & Rogers, 2017; Noppeney & Price, 2004), the retrieval and integration of semantic representations involving the left anterior temporal lobe and left inferior frontal gyrus (Ralph et al., 2017; Rice, Lambon Ralph, & Hoffman, 2015; Huang et al., 2012), the syntactic analysis involving the frontoparietal lobes (Ralph et al., 2017; Chou, Lee, Hung, & Chen, 2012; Nieuwland, Martin, & Carreiras, 2012; Ye & Zhou, 2009), and the cognitive control process involving frontal regions such as medial prefrontal lobes (Jobson, Hase, Clarkson, & Kalaria, 2021; AbdulSabur et al., 2014; Amodio & Frith, 2006). Therefore, the two processes of reading comprehension seem to be supported by different brain regions and networks. However, how these two processes develop and the underlying neural mechanisms remain unknown.

The highly influential Piagetian views that maturity and experience have a significant impact on cognitive development (Piaget, 1972). Maturation of the neural system that is accompanied by age is the fundamental guarantee of cognitive development, which is further shaped by experience. Although it has been well established that both maturation and experience contribute to the brain development of cognition, the role of maturation and experience in brain development has been debated. In the field of developmental neuroscience, the maturational perspective for functional brain development suggests that the functional specialization of brain regions arises through the maturation of particular brain regions (Johnson, 2001). However, the skill-learning hypothesis suggests that human functional brain development is shaped by postnatal experience, and the changes in the neural basis of a behavior result from the acquisition of specialized knowledge (Johnson, 2001).

In terms of reading development, there exists a similar debate as the above. For the word-level reading, a study on the phonological processing of visual Chinese words detected declining brain activation with age across the lifespan, but the differences in brain activation between children and adults disappeared when controlling for the accuracy of in-scanner phonological judgment in both groups (Siok et al., 2020). This finding emphasized the deterministic role of reading experience on the development of brain function. However, a separate study on semantic processing of visual Chinese word found age-related increases in functional connectivity (FC) across the lifespan, with little correlation between FC and the performance in semantic judgment, implying a stronger influence for maturation than experience on the reading brain (Jia et al., 2023). Moreover, the brain function of different processes may be influenced differently by maturation and experience. For example, age-related changes have always been found in word-level reading (Jia et al., 2023; Siok et al., 2020; Cao et al., 2010; Booth et al., 2003). However, for reading comprehension beyond word level, brain activation in the language comprehension network has been found to be stable in adulthood (Fitzhugh, Braden, Sabbagh, Rogalsky, & Baxter, 2019; Zhuang, Johnson, Madden, Burke, & Diaz, 2016). In addition, the brain function of reading comprehension is more likely modulated by reading experience than by age during childhood (Meyler et al., 2007). Therefore, for the debated findings and the possibility of different maturation and experience effects on the different reading processes, further studies specifying how maturation (here, age) and experience (here, reading performance) influence the brain functional network of the two-level processes of reading comprehension across the lifespan are highly needed.

In the current study, the effects of age and reading experience were examined across the lifespan for both brain activation and FC of reading comprehension within the framework of the two-level processes. Three age groups covering the entire span of life (7–77 years old) were recruited, namely, children, young adults, and middle-aged and older adults. The functional neuroimage for the two-level processes of reading comprehension was established through the in-scanner word viewing task and sentence comprehension task. The reading experience was assessed based on behavioral performance in the Chinese character reading (ChR) and word reading efficiency (WRE) for LL process, and in silent sentence reading (SSR) and in-scanner sentence comprehension for HL process. We expected that the brain function of the LL process of reading comprehension across the lifespan is more likely influenced by age. Age effect is likely to be manifested by an increased dependency on the occipitotemporal lobe related to orthographic and semantic processing with age. Whereas we expected the HL process to be more likely shaped by reading experience, as indicated by an enhanced reliance on the frontotemporal semantic network with better behavioral performance.

We recruited the participants through advertising in the community of Haidian District in Beijing. One hundred fifty-eight participants from Beijing participated in the current study. We excluded six participants with Raven reasoning scores lower than the 25th percentile, two participants with reading scores lower than 2 SDs (one was in ChR test, the other was in SSR test), four participants with excessive head movement (>3 mm/deg), and six participants with accuracy of the in-scanner tasks below 60%. Ultimately, 140 participants remained in the final analysis, including 51 children (23 girls, 10.89 ± 1.96 years of age, 4.98 ± 1.87 years of education), 33 young adults (20 women, 21.24 ± 1.85 years of age, 15.24 ± 1.85 of education), and 56 middle-aged and older adults (36 women, 53.99 ± 14.02 years of age, 14.82 ± 3.19 years of education). For our adult participants, 78.65% had received college education. The final sample size of each group was in line with a sample size greater than 30 per group in previous fMRI studies of reading development (Zhou et al., 2021; Kersey, Wakim, Li, & Cantlon, 2019; Zhao, Bi, & Coltheart, 2017). All participants were right-handed native speakers of Mandarin Chinese with normal or corrected-to-normal vision and hearing and no history of neurological diseases according to their self-reports. Written informed consent was obtained from all participants and from the parents of child and adolescent participants. This study was approved by the research ethics committee at Beijing Normal University. See Table 1 for detailed demographic information.

Intelligence quotient (IQ) was measured by Raven's Standard Progressive Matrices (Raven, 1960). All participants included had normal intelligence (Raven reasoning score > 25th percentile of the peers). Reading performance was evaluated by three tests, namely, ChR (word level), WRE (word level), and SSR test (sentence level), assessing the reading comprehension ability at both the word and the sentence levels. First, the ChR test was used to measure the accuracy of the single Chinese character reading. Two different versions were employed: one for children and one for adults. All participants were instructed to read aloud each character from a list of 150 characters list arranged from easy to difficult, and the test was terminated if they failed for 15 consecutive characters. The only difference between the versions was that the child version selected 150 words from 3500 commonly used Chinese characters (Shu, Chen, Anderson, Wu, & Xuan, 2003), whereas the adult version selected 150 characters from 3500 less frequently used Chinese characters (Liu et al., 2013). The number of correctly read characters indicates the accuracy of the ChR. The second test was the WRE test with 180 high-frequency, two-character words (Xia, Hoeft, Zhang, & Shu, 2016; Zhang et al., 2012). Participants were instructed to read those words aloud as correctly and quickly as possible. The time spent completing the test was recorded. The score was calculated as the number of correctly read words per minute. The last test was the SSR test, which consisted of 100 single sentences or short paragraphs with the number of characters increasing from 7 to 159 (Lei et al., 2011). Participants were asked to read the sentences silently as quickly as they could within 3 min and judge whether the meaning of the sentences was correct by drawing “√” or “×” in the parentheses following each sentence. The score was calculated as the sum of characters in the correctly answered sentences.

During the MRI sessions, participants were asked to complete a sentence comprehension task and a passive word viewing task. For the sentence comprehension task, each sentence was divided into lexical chunks of one to three characters presented on five screens to avoid head movement. After the last lexical chunk, there was a full stop to indicate that the participants ought to respond. The participants needed to press a key with the index finger of their right hand if the sentence meaning was consistent with common sense (e.g., Dropping litter everywhere is uncivilized.), or the middle finger if the sentence was not common sense (e.g., Staying up late is good for our health.). The other task was passive word viewing, where the chunks of one to three characters were presented matching those used in the former task but arranged in a random order to prevent forming a coherent sentence. All chunks were presented in a pseudorandom way, with half being presented first in the sentence task and the other half being presented first in the passive reading task. The participants were instructed to read these lexical chunks passively and judge whether the characters in the last chunk were italic to stay focused. The index finger indicated “yes,” and the middle finger indicated “no.” The fixation was used as a baseline. Previous studies have shown that normal readers can automatically carry out the processing of word recognition when passively viewing words (Liuzzi, Ubaldi, & Fairhall, 2021; Lu et al., 2021; Dehaene, Le Clec'H, Poline, Le Bihan, & Cohen, 2002; Georgiewa et al., 1999). Therefore, by contrasting passive word viewing with fixation, we tapped into the LL process of reading comprehension, including visual word form processing and orthography-semantic mapping. In addition, by contrasting sentence comprehension with passive word viewing, we tapped into the HL process of reading comprehension, involving semantic integration, syntax, and cognitive control as well.

The entire experiment contains two runs, and each run contains three blocks of sentence comprehension task and three blocks of passive word viewing task. Inspired by the experimental design of previous studies of sentence-level processing (Matchin, Hammerly, & Lau, 2017; Richardson, Seghier, Leff, Thomas, & Price, 2011; Kuperberg, Sitnikova, & Lakshmanan, 2008), a 20-sec fixation was set as a baseline, and the colors of the fixation were set as task cues. Two tasks alternated in turn with a 20-sec fixation as the interval, and its colors are indicated. If the fixation was white, the participants performed the sentence comprehension task. If the fixation was yellow, then the participants performed the passive word viewing task. Each block had five trials that began with an 800-msec fixation, followed by five 800-msec screens of word chunks, and ended with a 2500-msec reaction screen. The setting of the time per screen referred to the study of children's reading speed in Chinese (Pan, Liu, Li, & Yan, 2021), and a pilot experiment was carried out on primary school children to ensure that children could successfully complete the experiment. The experimental procedure is shown in Figure 1.

All neuroimaging data were acquired on a SIEMENS PRISMA 3-T scanner at the Brain Imaging Center of Beijing Normal University. The foam cushion provided by the manufacturer was used to reduce head movement. The functional image data were measured using a T2-weighted gradient EPI sequence with the following parameters: repetition time = 2000 msec, echo time = 30 msec, flip angle = 90°, slice number = 33, field of view = 200 × 200 mm², and voxel size = 3.1 × 3.1 × 3.5 mm³. High-resolution, T1-weighted images were acquired for each participant (repetition time = 2530 msec, echo time = 2.27 msec, flip angle = 7°, axial slices = 208, slice thickness = 1 mm, field of view = 256 × 256 mm², voxel size = 1 × 1 × 1 mm³).

The functional data were preprocessed using Statistical Parametric Mapping (12) with the following steps: head movement correction, realignment, coregistration, segmentation, normalization, and smoothing. All functional images were spatially realigned and coregistered to their corresponding anatomical images. Then, those functional images were normalized to the MNI (Montreal Neurological Institute) space by the segmentation of anatomical images (resampling to 3 × 3 × 3 mm³) and were further spatially smoothed with an 8-mm FWHM Gaussian kernel.

The generalized linear model (GLM) was constructed to obtain the BOLD signals of each subject under each condition (sentence comprehension, passive word viewing, and fixation). To minimize the influence of the subject's head movement, six head movement parameters were added to the model as covariates. Then, two contrasts of interest were set and calculated: The first was lower-level contrast (LL contrast), by contrasting the passive word viewing with baseline, indicating word-level processing; the second was higher-level contrast (HL contrast), by contrasting the sentence comprehension with passive word viewing, indicating sentence-level processing. We tested the whole-brain activation maps for both LL and HL contrasts for each age group (p < .05 FWE corrected at the voxel and cluster level) and for all participants (p < .001 FWE corrected at the voxel and cluster level) using a one-sample t test.

To determine the effects of age and reading experience on brain activation during reading comprehension across the lifespan, we employed a whole-brain regression analysis with age and reading performance as linear continuous predictors for both contrasts. Word-level reading tests (ChR and WRE) and sentence-level reading tests (SSR along with the ratio of accuracy and RT [acc/rt] on the in-scanner sentence comprehension task) were employed to assess reading performance for the LL contrast and HL contrast, respectively. To isolate the age effect, IQ and in-scanner performance were controlled. When examining the effect of reading performance, IQ and age were controlled. Of note, because of the use of different versions of the ChR test (child vs. adult version), we grouped participants into children and adults (merged young adults and middle-aged and older adults) when conducting all related regression analyses with ChR performance as a linear continuous predictor variable. In addition, considering educational level is also an important experiential factor and it may correlate with reading experience, we conducted an analysis to investigate the impact of education on whole-brain activation while controlling for age and IQ. The results of the age effect, reading experience effect, and educational-level effect were thresholded with an AlphaSim-corrected threshold of p < .001 (with a combined threshold of p < .001 at the voxel-level and a minimum cluster size of 40 voxels).

To further investigate the effects of age and performance on lifespan brain function development in the classic reading network, we conducted ROIs analysis for brain activation and FC. The ROIs in the LL contrast were defined based on previous meta-analysis literature of Chinese word reading (Nie, Wei, Zhu, Zhou, & Niu, 2015; Wu et al., 2012), including the opercular part of the left inferior frontal gyrus (IFGoper), left MFG, left MTG, left FG, left SPL, and right MOG. The ROIs in the HL contrast were defined based on meta-analysis and empirical studies of language comprehension involving semantic and syntactic processing (Rodd, Vitello, Woollams, & Adank, 2015; Ye & Zhou, 2009; Ferstl, Neumann, Bogler, & Von Cramon, 2008), including the orbital part of the left inferior frontal gyrus, the triangular part of the left inferior frontal gyrus (IFGtri), left medial superior frontal gyrus (mSFG), left anterior temporal lobe (aTL), left MTG near the superior temporal sulcus (STS/MTG), and left precuneus (preCUN). The coordinates of these ROIs were determined from the above literature. Each ROI was defined as an 8-mm sphere centered at the corresponding coordinates. See the locations and the detailed information of ROIs in the Figure 2 and Table 2.

Six ROIs for LL contrast and six ROIs for HL contrast were used as masks to extract the activation value (percentage of BOLD signal change) of each ROI in the corresponding condition for each participant through a Marsbar of spm12. We tested the age effect by calculating the partial correlation (two-tailed) between age and ROI activation, controlling for IQ and in-scanner performance. Likewise, the reading experience effect was tested by calculating the partial correlation (two-tailed) between the ROI activation and reading performance (LL contrast: ChR and WRE and HL contrast: SSR and acc/rt of the in-scanner sentence comprehension task), controlling for IQ and age in all participants. We also investigated the educational-level effect by calculating partial correlation (two-tailed) between the ROI activation and years of education, controlling for IQ and age in all participants. We employed false discovery rate (FDR) correction for multiple comparisons with an alpha level of .05. The p values reported in the results were corrected.

Furthermore, we tested whether the reading experience effect was age-dependent by conducting the partial correlation analysis (two-tailed) between activation of ROIs and reading performance (LL contrast: ChR and WRE and HL contrast: SSR and acc/rt of the in-scanner sentence comprehension task) within each age group. IQ and age were controlled for all age groups, and year of education was additionally controlled for the middle-aged and elderly group. FDR correction p < .05 for multiple comparisons was adopted. The p values reported in the results were corrected.

To investigate the effects of age and reading experience on FC in the classical reading network, we conducted an ROI-to-ROI psychophysiological interaction analysis. The ROIs (six ROIs in the LL contrast, six ROIs in the HL contrast; see locations of the ROIs in Figure 2) same as those in the ROI analysis of activation were defined as seeds, and their deconvolved time series were extracted for each participant in each group. Standard GLM analysis was conducted to model the contribution of experimental conditions to the time course of each seed. For the first-level GLM regression analysis, the regressors included a psychological variable that reflected the experimental design (passive word viewing contrast fixation, sentence comprehension contrast passive word viewing), a time series variable that represented the time course of each single seed, and the interaction between the experimental design and time course of the seed. White matter, cerebrospinal fluid, and six head movement parameters were denoised as confounds.

For the second-level analysis, we first conducted partial correlation analysis (two-tailed) between FC value and age, FC value and reading performance, as well as FC value and years of education in all participants. Finally, to examine whether the effect of reading experience depends on age, the partial correlation analysis between FC value and performance was conducted in each of the three groups separately. All control variables in the partial correlation analysis are identical to those in the ROI analysis of brain activation. FDR correction p < .05 for multiple comparisons was adopted. The p values reported in the results were corrected.

The behavioral performance of the three groups is shown in Table 1. All performance measures except the accuracy of the passive word viewing task showed significant differences among the three groups (ps < .05). The results of the intergroup comparison in Table 1 were corrected by Bonferroni when the variance was homogeneous or Games-Howell when the variance was uneven. Overall, the performance in reading tests and in-scanner tasks showed that young adults performed the best, followed by middle-aged and older adults, and finally children.

In addition, the correlations between education and reading performance were examined while controlling for IQ and age to investigate whether education and reading performance interrelated. A significant positive correlation was demonstrated between years of education and WRE (r = .68, p < .001) as well as between years of education and SSR (r = .70, p < .001) in all the participants. However, no significant correlation was demonstrated between years of education and ChR in either children (r = −.02, p = .89) or adults (r = .07, p = .55).

For LL contrast, participants in the three age groups primarily activated the bilateral inferior temporal-occipital cortex, left mid-inferior frontal regions, left inferior parietal lobe, left MTG, left precentral gyrus, SMA, and some subcortical nucleus (see Figure 3A and Table 3). Overall, the three groups showed very similar brain activation patterns, except that the children and middle-aged and older adults groups seemed to have more widespread activations than the young adults group. Particularly, the middle-aged and older group showed more extensive brain activation in the bilateral frontoparietal regions.

For HL contrast, all participants mainly activated the left inferior frontal areas, left mSFG, and left middle and anterior temporal lobe (see Figure 3B and Table 3). The three groups showed similar activation patterns with relatively limited clusters in the young adults group.

In terms of the age effect, extensive brain areas mainly including the bilateral occipital lobes, SPL, MFG, IFG, and pre- and postcentral gyrus were significantly positively correlated with age in the LL contrast (Figure 4A1 and Table 4). Negative correlations with age in the LL contrast were found in the left middle-posterior superior temporal gyrus (STG), angular gyrus (AG), the right MTG, AG, and preCUN. Although only a few small clusters showed significant positive correlations with age in the HL contrast (Figure 4A2 and Table 4), including the left MFG, right opercular part of IFG, and bilateral cerebellum, no negative correlation was found in the HL contrast.

In terms of the reading experience effect, positive correlations between brain function and reading performance were only found in the HL contrast. Specifically, the activation of the right MTG, cingulate gyrus, and SMA was positively correlated with the scores of SSR (Figure 4B1 and Table 4). In addition, the activation of the right MFG, postcentral gyrus, left STG, and mSFG was positively correlated with the ratio of acc/rt of the in-scanner sentence comprehension task (Figure 4B2 and Table 4). No significant correlation was found in the LL contrast.

In terms of educational-level effect, negative correlations between years of education and the whole-brain activation were found in the LL contrast (Figure 4C1 and Table 4), specifically in the left middle cingulate gyrus and medial part of the SFG, whereas positive correlations between years of education and the whole-brain activation were found in the HL contrast (Figure 4C2 and Table 4), specifically in the bilateral insula, the right supramarginal gyrus (SMG) and paracentral lobule.

Mean value of BOLD activation for each ROI in LL and HL contrasts was shown in Table 2. The age effects were only found in the LL contrast (Figure 5A), where age significantly positively correlated with the brain activation of four ROIs after controlling IQ and the performance in the in-scanner passive word viewing task, namely, the left MFG (r = .24, p = .006), left IFGoper (r = .27, p = .003), left SPL (r = .53, p < .001), and left FG (r = .26, p = .004). The p values reported here and below were all corrected by FDR multiple comparisons. No significantly negative correlation was found.

Similar to the whole-brain analysis, the reading experience effects were only found in the HL contrast (Figure 5B). Specifically, the acc/rt of the in-scanner sentence comprehension task was significantly positively correlated with the left STS/MTG (r = .28, p = .006) after controlling for IQ and age. Of note, no significant correlations between years of education and activation of the ROIs were found either in the LL contrast (−0.20 < rs < .01, ps > .274) or in the HL contrast (−0.10 < rs < .03, ps > .924).

Because the experience effects may be dependent on age, we then analyzed the correlation between the activation of ROIs and behavioral performance in each of the three groups. Interestingly, we only found a positive correlation in the HL contrast in middle-aged and older adults (Figure 6B), that is, the acc/rt of the in-scanner sentence comprehension positively correlated with the activation in the left aTL (r = .32, p = .048) and the left STS/MTG (r = .44, p = .006) after controlling for age, IQ, and education.

The age effect was only found in the LL contrast (Figure 5A), with positive correlations between age and the FC between the left SPL and the left FG (r = .27, p = .015) after controlling for IQ and acc/rt of the in-scanner passive word viewing task. The p values reported here and below were all corrected by FDR multiple comparisons.

The reading experience effect was only found in the HL contrast (Figure 5B), with positive correlations between acc/rt of the in-scanner sentence comprehension task and the FC between the left IFGtri and the left STS/MTG (r = .25, p = .045) after controlling for IQ and age. Of note, no significant correlations between years of education and FC of the ROIs were found either in the LL contrast (−0.18 < rs < .06, ps > .473), or in the HL contrast (−0.183 < rs < .094, ps > .480).

To examine whether the reading experience effect differs among the three age groups, we further analyzed the correlations between the FC of ROI-to-ROI and reading performance in each of the three groups. In the LL contrast, correlations were only found in young adults after controlling for age and IQ (Figure 6A). Specifically, the scores of WRE were positively correlated with the FC between the left IFGoper and the left SPL (r = .41, p = .026), between the left MTG and the left FG (r = .41, p = .032), between the left MFG and the left FG (r = .41, p = .042), and between the left IFGoper and the right MOG (r = .43, p = .045). In the HL contrast, correlations were only found in middle-aged and older adults after controlling for age, IQ, and education (Figure 6B). Specifically, the scores of SSR were positively correlated with the FC between the left mSFG and the left STS/MTG (r = .27, p = .050), between the left IFGtri and the left preCUN (r = .34, p = .042), and between the left IFGtri and the left aTL (r = .31, p = .033).

The present study investigated the brain activation and FC within the context of two-level processes of reading comprehension across the lifespan. We found that: (1) in the LL word recognition process of reading comprehension, functional brain activation and connectivity across the lifespan were mainly influenced by age rather than reading experience. (2) In the HL sentence comprehension process, functional brain activation and connectivity across the lifespan were mainly influenced by reading experience rather than age. (3) The reading experience effect varied among different age groups.

During the word-level process (LL process) of reading comprehension, both positive and negative age effects on brain function were found. Positive age effects were observed in extensive brain areas in the whole-brain analysis, including the bilateral parieto-occipital lobe and frontal lobe (e.g., MOG, SPL, IFG, and MFG). Similarly, the activation of the preselected ROIs, such as the left SPL, FG, IFG, and MFG, also increased with age. Specifically, bilateral MOG and left SPL have been considered responsible for visual–spatial processing, especially in Chinese ChR (Xu et al., 2015; Tan et al., 2001, 2005). In addition, the left FG, often referred to as the visual word form area, is responsible for orthographic processing across languages (Feng et al., 2020; Cohen et al., 2002). Moreover, the left IFG and MFG may collaborate in the process of controlled selection and retrieval of lexical representations (Cao et al., 2010). The observed age-related increases in brain activation in the above reading-related areas have also been reported in previous studies on reading development (Zhu, Nie, Chang, Gao, & Niu, 2014; Cao et al., 2010; Booth et al., 2003). We extended this pattern to encompass lifelong development, suggesting greater involvement of orthographic-semantic mapping with age over the lifespan. Beyond the findings of regional brain activation, we also found that the FC between the left SPL and left FG increased with age, which is consistent with the findings of increased FC in the left FG throughout the lifespan (Jia et al., 2023). The FC results again underscore a greater reliance on interactions between regions related to orthographic processing in word reading as age increases.

Worth mentioning is that beyond the classical word reading network, there are additional regions that showed increased activation with age at the whole-brain level, such as the bilateral pre- and postcentral gyrus. These regions are supposed to be responsible for sensorimotor processing to assist the automated reading process (Ang et al., 2020; Koyama et al., 2010; Carreiras, Mechelli, Estévez, & Price, 2007). The increased activation of the bilateral pre- and postcentral gyri with age may primarily be driven by middle-aged and old participants in this study, as the activation map in Figure 3 and the results in Table 3 seemed to show a relatively wider range of activation areas in middle-aged and older adults in LL contrast. Previous studies have shown that additional activation in the frontoparietal lobe (like pre- and postcentral gyrus) may serve as a marker of healthy aging (Drag et al., 2016; Li et al., 2015). Thus, the increased activation in such areas for reading processing may indicate an increased demand for assistance when completing word-level reading tasks in middle-aged and older adults.

We also noticed negative age effects in the bilateral temporoparietal regions including the left posterior STG, the right preCUN, and the bilateral AG in the entire cohort. These regions have been thought to play critical roles in phonological processing (Wang, Joanisse, & Booth, 2020; Yi, Leonard, & Chang, 2019; Callan, Callan, & Masaki, 2005; Fiebach, Friederici, Müller, Von Cramon, & Hernandez, 2003; Booth et al., 2002). Specifically, the STG is supposed to represent the acoustic and perceptual features of the phonology (Wang et al., 2020; Yi et al., 2019; Booth et al., 2002). The right preCUN plays a special role for acoustic-phonological codes in the visual processing of early word learning (Fiebach et al., 2003). The AG is an area involved in visual-to-phonological conversion for written characters or words (Callan et al., 2005; Booth et al., 2002). The decrease of the activation in these regions with age indicates a reduced dependence on phonological processing during word-level reading comprehension development. This is consistent with previous studies on Chinese reading development, which showed developmental decreases in activation and FC across tasks in temporal phonological regions (Liu et al., 2018; Cao et al., 2010).

In contrast to the extensive age effects on parieto-occipital and frontal lobes, the reading experience did not show a significant impact on brain function during the word-level processing of reading comprehension. This is consistent with a previous lifespan neurodevelopmental study of visual-word phonological processing, which found that brain connectivity is more likely to be influenced by age than by experience (Jia et al., 2023). This might be because word recognition processing heavily depends on fluid intelligence (e.g., visual perception, speed of processing, and working memory; Christopher et al., 2012), which is greatly affected by the maturation of the nervous system, but less affected by experience.

However, our findings indicate a negative correlation between years of education and brain activation in middle cingulate gyrus and mSFG. These regions have been implicated in conflict monitoring or cognitive control, as they are more active when two words are inconsistent in terms of their phonology or orthography (Bolger, Hornickel, Cone, Burman, & Booth, 2008). Hence, the negative correlations may suggest that individuals with higher levels of education exhibit less cognitive control during word reading, possibly because of increased automaticity of reading processes with longer educational duration.

Collectively, during the LL process of reading comprehension, age significantly impacts brain activation and connectivity, as evidenced primarily by increased engagement of orthographic-semantic processing and decreased engagement of phonological processing across the lifespan. In addition, years of education also impacts brain activation in this process, as evidenced by decreased activation in cognitive control regions across lifespan.

Conversely, a reversed pattern was observed in the HL process of reading comprehension. We found that brain function in the lifespan development of the comprehension process was influenced mainly by reading experience. At the whole-brain level, better performance in the in-scanner sentence comprehension task was associated with greater activation of bilateral frontotemporal regions (e.g., left STG, SFG, and right MFG). Similarly, better performance in SSR was associated with greater activation of the right frontotemporal cortex (e.g., right MTG, cingulate gyrus, and SMA) when controlling for age. Our findings suggest that the bilateral frontotemporal cortex is shaped by the experience during the HL process of reading comprehension. The frontotemporal lobes have been considered as a control network and are critically involved in semantic and syntactic processing during language comprehension (Jedidi et al., 2021; Ding et al., 2020; Fitzhugh et al., 2019; Emmorey, McCullough, & Weisberg, 2015; Turken & Dronkers, 2011). The ROI-level findings of the positive correlation between the activation of the left STS/MTG and the FC of this region with the left IFGtri and the performance in the in-scanner sentence comprehension task further emphasize that left frontotemporal lobes and the interactions between regions within them are shaped by the sentence-level reading comprehension experience. This aligns with previous meta-analysis studies showing consistent activation of the left STS/MTG and IFGtri in syntactic and semantic processing during sentence comprehension (Rodd et al., 2015; Ferstl et al., 2008). In addition, our results of the activation in the right hemisphere are also consistent with a growing number of studies (Horowitz-Kraus, 2015; Diaz & Hogstrom, 2011; Turken & Dronkers, 2011; Robertson et al., 2000), suggesting the important role of the right hemisphere in reading comprehension. Taken together, our findings of increased activation and strengthened connectivity suggest that the lifespan development of brain function for comprehension is modulated by sentence comprehension proficiency.

Years of education also showed a positive impact on brain activation in several brain regions in this contrast, including the bilateral insula, the right SMG, and the paracentral lobule. Insula and SMG are known to be involved in verbal working memory (Llorens et al., 2023; Deschamps, Baum, & Gracco, 2014). Because the sentences in our in-scanner task were presented one word chunk at a time, participants had to continuously preserve and manipulate the stimuli. As a motor-related area (Tomasino, Bernardis, Maieron, D'Agostini, & Skrap, 2023; Levänen, Uutela, Salenius, & Hari, 2001), the paracentral lobule may aid in the verbal working memory of the stimuli through articulation and rehearsal. Interestingly, a recent study showed that long-term musical training can reshape the insula, SMG, and the paracentral lobule network, indicating that this network may be experience-dependent (Cheng et al., 2023). Overall, our study may suggest that education may reshape brain regions related to verbal working memory required by high-level language comprehension.

Of note, although we found significant correlations between years of education and brain activation in the whole-brain analysis, we did not find significant correlations between years of education and the activation of ROIs either for LL or HL contrast. This may indicate that education may primarily have an impact on brain regions related to general cognitive ability such as cognitive control or verbal working memory, rather than on brain regions specific to reading.

In contrast, we only noticed a few small clusters showing a positive age effect in whole-brain level activation, including bilateral cerebellum and middle-inferior frontal lobes. No activation and FC values were found to be significantly correlated with age at the ROI level. Studies dealing with representations of either auditory or written language comprehension have consistently found a shared language comprehension network: the lateral temporal lobe and the inferior frontal gyrus (see the introduction of Lindenberg & Scheef, 2007), which is similar to the network recruited in our HL process of reading comprehension. Previous studies have shown that language comprehension develops early in human life, even during infancy, manifesting a similar language network from infancy to adulthood (Monzalvo & Dehaene-Lambertz, 2013; Dehaene-Lambertz, Dehaene, & Hertz-Pannier, 2002), although it takes a long time before children master the complex grammar (Skeide, Brauer, & Friederici, 2016). In addition, previous studies have found that the brain network associated with language comprehension does not change with age during adulthood (Fitzhugh et al., 2019). Given the relatively basic sentences used in our in-scanner sentence comprehension task, a similar network was found for participants regardless of age.

Altogether, we found that the HL process of reading comprehension is mainly affected by reading experience and years of education. This may be because the HL cognitive process of reading comprehension heavily depends on crystal intelligence like world knowledge (Scarborough, 2001), which is highly experience-dependent.

Despite that we did not find a significant effect of reading experience in the LL process across all participants, we did find significant positive correlations between WRE performance and the FC of four connections in young adults. These connections included the right MOG – the left IFGoper, the left SPL – IFGoper, the left FG – MFG, and the left FG – MTG. Most of these connections connect the occipital and parietal lobes to the frontal lobe. These parieto-occipital regions serve a specific role in visual–spatial and orthographic processing during Chinese reading (Xu et al., 2015; Tan et al., 2001, 2005). The connections between frontal regions and these parieto-occipital regions may suggest cognition control regulating the orthographic processing (Deng, Guo, Ding, & Peng, 2012) or the cooperation of the posterior visuospatial network and frontal phonosemantic network (Liu, Tao, Qin, Matthews, & Siok, 2022). In addition, the connection between the left FG and MTG is supposed to be a classical ventral pathway that is always involved in orthographic-semantic mapping in the word reading (Molinaro, Paz-Alonso, Duñabeitia, & Carreiras, 2015; Taylor et al., 2013). Our findings indicate that the more word reading experience, the stronger the connectivity of regions within the word reading network in young adults.

Moreover, among three age groups, we found brain–behavior correlations in the HL process only in the middle-aged and older adults. The correlations mainly appeared in the frontal lobe and the anterior and posterior temporal lobe. Specifically, the activation of aTL and STS/MTG was positively correlated with acc/rt of the in-scanner sentence comprehension task. These two regions have been suggested to be involved in language comprehension as well as semantic integration in previous literature (Chang, Dehaene, Wu, Kuo, & Pallier, 2020; Turken & Dronkers, 2011). In addition, the connectivity of mSFG – STS/MTG, IFGtri – preCUN, and IFGtri – aTL were also found to be significantly positively correlated with the performance in SSR. Particularly, the connection between IFGtri and aTL has been suggested to be involved in semantic processing during language comprehension (Friederici, 2011). Furthermore, mSFG and preCUN have been considered critical regions of the theory of mind network (AbdulSabur et al., 2014; Amodio & Frith, 2006), and the IFGtri has been found to be more activated in sentence comprehension than in lexical level comprehension, and it may be associated with semantic integration (Chang et al., 2020). The connections between mSFG and STS/MTG, and between IFGtri and preCUN might indicate synergies between important processes in the HL process of reading comprehension, such as semantic integration and mentalizing. These findings are indicative of the more experienced in sentence-level reading, the stronger the dependency on regions related to HL comprehension in middle-aged and older adults.

Taken together, some reading experience effects were contingent on age in the two processes of reading comprehension. When examining the experience effect in separate age groups, word-level reading experience affected the brain function for LL word reading only in young adults, whereas sentence-level reading experience affected the brain function for HL reading comprehension only in middle-aged and older adults. As mentioned earlier, the LL process probably heavily depends on fluid intelligence components such as visual perception, processing speed, and working memory (Christopher et al., 2012), whereas the HL of reading comprehension heavily depends on crystalized intelligence components such as world knowledge (Scarborough, 2001). Fluid intelligence reaches its peak in youth, which may lead young people to rely more on fluid intelligence (Chen, Hertzog, & Park, 2017). Therefore, word-level reading experience may significantly affect the brain function of young adults. However, crystalized intelligence develops continuously throughout adulthood (Spreng & Turner, 2019; Lindenberger, 2014), which may lead middle-aged and older adults to rely more on crystalized intelligence and be mainly affected by experience during the high-level process of reading comprehension.

Some limitations of the current study should be noted. First, this is a cross-sectional study; thus, we cannot eliminate all individual differences between participants. Longitudinal research is needed to better characterize brain functional changes across the lifespan in future studies. Second, the LL process of reading comprehension may be mixed with other processing because participants were asked to judge whether the last word in a list of word chunks was in italics. For instance, by simply contrasting this task with fixation, more general visuospatial processing and motor processing may be involved in addition to visual word processing. However, selecting classical word reading regions as ROIs based on previous studies compensates for this limitation to some extent. Nevertheless, more rigorous experimental designs are needed for follow-up studies. Third, whether our findings can be reproduced in alphabetic languages remains to be investigated in the future.

The present study is the first to investigate the neural changes in reading comprehension throughout the lifespan from a perspective of maturation and experience influencing cognitive development. It has yielded two key conclusions. First, the brain activation and connectivity of the LL process in reading comprehension are more likely shaped by age, demonstrating an increased reliance on parieto-occipital visuospatial regions and inferior-middle semantic frontal regions, alongside a decreased dependence of temporoparietal phonological regions with age. By contrast, the HL process appears to be more likely shaped by reading experience, demonstrating greater dependency on frontotemporal lobe with better sentence reading performance. Second, the reading experience effect is contingent on age in different processes of reading comprehension. In the LL process, more reliance on occipital-parietal and inferior-middle frontal regions with better word reading was only evident in young adults. In the HL process, more reliance on frontotemporal semantic language comprehension network with better sentence reading was exclusive to middle-aged and older adults among the three groups. Finally, years of education seems to impact both LL and HL processes of reading comprehension; however, the effect appears to be primarily on brain regions associated with general cognitive ability rather than those specific to reading. Altogether, this study demonstrates that functional brain development in distinct reading processes is differently modulated by maturation and experience across the lifespan, with experience effects varying among different age groups. In conclusion, our findings augment our understanding of functional brain development in reading comprehension across the lifespan.

The authors thank Xitong Liang and Chaoying Xu for experimental data collection; Danqi Gao, Yin He, and Mingnan Cai for providing insightful comments; and Honghao Zheng for code support.

Corresponding author: Li Liu, C309, 8 Xinjiekou Outer St, Xicheng District, Beijing, China, 100088, or via e-mail: [email protected].

The data that support the findings of this study are available on request from the corresponding author.

Xinyang Liu: Conceptualization; Data curation; Formal analysis; Writing—Original draft; Writing—Review & editing. Lihuan Zhang: Methodology. Saiwen Yu: Methodology; Writing—Review & editing. Zilin Bai: Methodology. Ting Qi: Writing—Review & editing. Hengyu Mao: Writing—Review & editing. Zonglei Zhen: Writing—Review & editing. Qi Dong: Conceptualization; Writing—Review & editing. Li Liu: Conceptualization; Writing—Review & editing.

This research was funded by the Child Brain-Mind Development Cohort Study in China Brain Initiative, grant number: 2021ZD0200534; the National Natural Science Foundation of China (https://dx.doi.org/10.13039/501100001809), grant numbers: 31970977, 31571155; the Interdisciplinary Research Funds of Beijing Normal University; and the Fundamental Research Funds for the Central Universities, grant number: 2015KJJCB28.

A retrospective analysis of the citations in every article published in this journal from 2010 to 2020 has revealed a persistent pattern of gender imbalance: Although the proportions of authorship teams (categorized by estimated gender identification of first author/last author) publishing in the Journal of Cognitive Neuroscience (JoCN) during this period were M(an)/M = .408, W(oman)/M = .335, M/W = .108, and W/W = .149, the comparable proportions for the articles that these authorship teams cited were M/M = .579, W/M = .243, M/W = .102, and W/W = .076 (Fulvio et al., JoCN, 33:1, pp. 3–7). Consequently, JoCN encourages all authors to consider gender balance explicitly when selecting which articles to cite and gives them the opportunity to report their article's gender citation balance.