Keywords

Due to technological advances, imaging methodology, such as functional magnetic resonance imaging (fMRI), positron emission tomography (PET), magnetoencephalography (MEG), Transcranial Magnetic Stimulation (TMS), and Diffusion Tensor Imaging (DTI), has been utilized recently in reading science. Neuroimaging research allows us to identify brain regions that become active upon reading. One important contribution of neuroimaging research to reading science is that it provides a useful tool for our understanding of neuronal processing and the brain mechanisms of human cognition and learning. It also offers insights into how efficient reading is accomplished. Evidence converges on the localization of neuronal networks in the left fusiform gyrus or the visual word form areas by going through a genetically constrained circuit (Perfetti, Liu, Fiez, Nelson, Bolger, & Tan, 2007). Perfetti et al. (2007) summarize the brain regions that are generally active for the three constituents of reading (i.e., orthography, phonology, and meaning) as follows:

The reading network includes posterior visual regions (occipital areas and the left mid-fusiform gyrus) for orthographic processes, temporal/parietal and anterior areas (superior temporal sulcus and inferior frontal sulcus/insula) for phonology, and both posterior (anterior fusiform) and anterior regions (inferior frontal gyrus) for meaning (p. 133).

Building upon research findings that both writing systems and readers’ proficiency are likely to yield different activations in brain areas, Perfetti and Liu (2005) propose the (writing) system accommodation hypothesis. The hypothesis posits that reading processes as well as the neural architecture and networks accommodate the specific visual and structural features of a given writing system for reading. As in Chapter 8, this theory is used as a theoretical framework for this chapter with a focus on neuronal networks and circuits, along with Dehaene’s (2009) neuronal recycling hypothesis. Since different brain activities can be explained within and between languages, this chapter reviews the extant literature within each language of Korean, Chinese, and Japanese as well as between languages as the first language (L1) and a second language (L2). This chapter begins with a brief discussion of theoretical considerations. The major brain circuits involved in reading among typical readers are reviewed. Next, empirical brain imaging studies in alphabetic Korean as well as in nonalphabetic Chinese and Japanese are reviewed. Finally, this chapter concludes with a discussion of how this line of empirical evidence extends to script relativity.

1 Theoretical Considerations

Dehaene (2009) notes that reading requires an activation of a universally involving area in the left hemisphere ventral occipito-temporal cortex known as the visual word form area, which processes orthographic stimuli independent of writing systems. He proposes the neuronal recycling hypothesis, positing that certain brain circuits evolve to adapt to variability within strong genetic constraints. Since writing was invented only about 5,000 years ago, there has not been sufficient time for the brain to evolve to form a brain circuitry reserved for visual word recognition. Therefore, the brain adapts to recycle the existing neuronal network in the brain. Accordingly, learning to read is constrained by the mechanism that is specified by the brain architecture (Dehaene, 2009). Dehaene (2009) also asserts that a different circuit of neurons is involved in our culture-specific activities, depending on the way in which our brain networks are connected to and support these activities. Reading behaviors are shaped by the fundamental workings of our nervous system. He notes that the brain not only governs a direct connection between our native neural structure and our acquired abilities, but also recruits different neuronal circuits according to different writing systems, although there is, in general, a biological unity that the same specialized cortical mechanism works for reading in the same brain regions. Reading is a useful vehicle for demonstrating how our brain organization and our learning are inextricably linked to each other and for testing the neural architecture and networks of the brain to understand the mechanism of cognitive functions.

Drawing upon the notions of universal principles and writing system variations that word identification occurs upon the activation of phonology at the moment of orthographic input, Perfetti, Liu, and Tan (2005) propose the lexical constituency model as a general theory of reading across writing systems (in relation to reading Chinese characters). Irrespective of writing systems, word identities are defined by the three interrelated consituents of orthography, phonology, and morphology. Based on the simulations of these three constituents’ priming effects, Perfetti et al. (2005) point out both universal reading processes and writing system constraints involved in reading. They indicate that lexical thresholds determine phonological and semantic effects, but not graphic (orthographic) effects. According to Perfetti et al. (2005), despite the universal phonological involvement in reading, the activation process of phonology is dependent upon the way in which the writing system structures its graphic units.

Perfetti et al. (2007) reviewed ERP and fMRI studies of Chinese-English bilinguals and learners of Chinese as a foreign language to elucidate the reading networks of the brain involved in reading in L1 and L2 in terms of the brain’s accommodation and assimilation. According to the system accommodation hypothesis (Perfetti & Liu, 2005), the neural circuits and networks of reading acquired in L1 become modified to accommodate and adapt to the linguistic demands of L2 script in reading. The brain’s assimilation occurs when using the existing L1 networks to process L2, while accommodation is involved when recruiting an additional network for L2 (Perfetti et al., 2007). fMRI studies show that learners of Chinese activate bilateral occipital-temporal and middle frontal areas when reading Chinese (e.g., Liu, Dunlap, Fiez, & Perfetti, 2007; Nelson, Liu, Fiez, & Perfetti, 2009). This is similar to the pattern of native speakers of Chinese but is different from the patterns of alphabetic reading. There seems to be an asymmetry between alphabetic readers and Chinese readers when they read in L2 Chinese and L2 English, respectively. Specifically, alphabetic readers tend to have neural circuits that accommodate the demands of L2 Chinese by engaging in neural networks that are not likely to be utilized for alphabetic reading. In contrast, Chinese natives tend to have neural circuits that assimilate L2 English into the Chinese writing system by recruiting neural networks that are used for reading Chinese characters (e.g., Cao, Tao, Liu, Perfetti, & Booth, 2013; Cao et al.).

2 Major Reading Circuits among Typical Readers

Studies using various imaging technologies, such as fMRI, PET, or MEG, show that word and pseudoword reading activates the left hemisphere posterior region that is associated with both the ventral circuit and dorsal circuit (Pugh et al., 2000). The left ventral occipito-temporal cortex (the junction of the occipital and temporal lobes) associates lateral extra-striate areas and a left inferior occipito-temporal area that are activated upon reading. The dorsal circuit (temporo-parietal region) includes the angular gyrus and supramarginal gyrus in the inferior parietal lobule as well as the posterior area of the superior temporal gyrus called Wernicke’s Area (Pugh et al., 2000). The region of the angular gyrus is associated with the mapping of orthography and phonology. Silent reading and naming are carried out in the anterior circuit that centers around Broca’s area in the inferior frontal gyrus (Pugh et al., 2000).

Converging evidence shows that the left ventral occipito-temporal cortex and nearby white matter tracts are the brain regions that are essential for reading. Yeatman, Rauschecker and Wandell (2013), using a combination of fMRI and white matter tractography, have found that three different pathways of white matter tracts are engaged in reading: (1) the inferior longitudinal fasciculus is connected to the occipital cortex in the anterior and medial temporal lobes; (2) the inferior fronto-occipital fasciculus is connected to the occipital cortex in the ventrolateral prefrontal cortex; and (3) the vertical occipital fasciculus is connected to the dorsal circuit in the lateral occipital parietal junction with the posterior angular gyrus and lateral superior occipital lobe.

Although there are major reading circuits commonly involved in reading, a mapping process from orthography to phonology can vary across writing systems as well as the characteristics of spoken languages. Bolger, Perfetti, and Schneider (2005) conducted a meta-analysis, including studies on English and European alphabetic languages, Chinese, and Japanese. Results showed that word recognition involved a common network of gross cortical regions in the brain regardless of script differences. However, some levels of localization or variation within those regions were observed depending on the script being read, suggesting that localizations might differ according to writing systems. The visual word form area showed consistent localization across tasks involved in reading and across writing systems. Bolger et al. (2005) concluded that the visual word form area in the left mid-fusiform gyrus was essential to word recognition across writing systems.

Of interest is to understand how the brain regions involved in reading vary within and across individuals as a result of reading experience, strategies used, and readers’ proficiency. Developmental variations in reading activation are observed in children and adults (Olulade, Flowers, Napoliello, & Eden, 2013). The effect of experience seems to be consistent with an inverted U-shaped function (Price & Devlin, 2011) given that an increased activation in the mid-regions of the left ventral occipito-temporal cortex is observed in emergent readers’ learning to read. A robust activation in the region is consistent with adults’ learning to read a new script regardless of the use of lexical or sublexical strategies (Mei et al., 2013). The low part of the inverted U shaped function is associated with reading familiar words, as decreased activation in the middle part of the left ventral occipito-temporal cortex is observed with familiar words (Twomey et al., 2013) among skilled adult readers (Olulade et al., 2013). Adults’ efficient reading appears to require less activation but increased anatomical connectivity within the reading network (Lebel et al., 2013). This seems to be related to the automaticity of reading, which is the foundation of script relativity.

Research shows that the brain’s blood oxygenation varies in regions according to the proficiency level of linguistic skills. In an fMRI study examining the neural correlates of learning to use pitch patterns of words by English-speaking adults with no previous exposure to the pitches, Wong, Perrachione, and Parrish (2007) measured blood oxygenation levels while participants discriminated pitch patterns of words before and after training. Participants who had mastered the learning program showed increased activation in the left posterior superior temporal region. Participants who had not mastered the pitch discrimination showed increased activation in the right superior temporal region and right inferior frontal gyrus (which were associated with nonlinguistic pitch processing) as well as prefrontal and medial frontal areas (which were associated with increased working memory and attentional efforts). These results indicated a relationship between the range of neural changes and language proficiency, suggesting the physiological contribution of the left dorsal auditory cortex to successful speech and word learning among adults.

3 Neuroimaging Studies of Reading Alphabetic Hangul in Relation to L2 English Reading

This section begins with a review of alphabetic Hangul reading because reading models and theories have originated from alphabetic orthographies. Since South Koreans unofficially use Chinese-derived characters (i.e., Hanja—traditional Chinese characters used in Korea), reading Hangul and Hanja is reviewed, and then the review moves on to L2 or L3 English reading in conjunction with L1 Korean.

3.1 Reading in Hangul and Hanja

Wolf (2018) makes a claim that “with no genetic bluerint for reading, there is no one ideal reading circuit” (p. 18, emphasis in original). Due to the lack of the optimal rading circuit, neural specialization and neural adaptation occur differently depending on the script being read, which is the tenet of the system accommodation hypothesis. The hypothesis is useful for Hangul reading because neuroimaging data allow us to better understand biological unity, script diversity, and brain network diversity, due to Hangul’s alphabetic nature and the use of non-Roman script in character-like syllabic blocks.

Based on the fact that the appearance of the Hangul script resembles Chinese characters, Yoon, Cho, and Park examined the brain activation of reading Korean words and recognizing pictures among native Korean speakers. They used 120 items consisting of 60 two- or three-syllable nouns in Hangul and 60 corresponding images that shared the same semantics (e.g., {고양이} <cat> vs. the image of a cat). Results showed that both reading Hangul and recognizing images activated the areas of the occipito-temporal region bilaterally. However, reading Hangul activated the frontal and temporal region as well, which was not activated with image stimuli. Reading Hangul also activated other areas, including the left middle frontal region (related to phonological and semantic processing), the right anterior cingulate (BA 32; related to language and sound organization), the superior temporal area (BA 29; related to phonological system), and the right medial frontal area (BA 8). Based on the activation of the right medial frontal region, the researchers indicated that reading Hangul involved nonverbal visual higher order control or the visuospatial analysis of the Hangul script due to the unique syllabic structure (Yoon, Cho, & Park). This finding suggests that Hangul’s square-block format might be the source of the departure point from European alphabets; that is, the unique visual configuration of Hangul seems to yield a unique processing pattern found in brain imaging studies.

Due to the use of both Hangul and Hanja in Korea, researchers have taken advantage of the biscript use in the reading science of Korean Hangul. Lee (2004) compared, using fMRI, brain activation in the reading of Hangul and Hanja among native Korean readers, along with a comparison group that read Hangul and English. An interaction analysis between the two groups showed that the right fusiform gyrus and the adjacent temporo-occipital region were more involved in reading Hanja than Hangul. In contrast, the regions in the bilateral inferior parietal lobules were more active in reading Hangul than Hanja. Lee (2004) indicated that both assembled phonological route and addressed lexical route seemed to be involved in reading Hangul, whereas reading Hanja might not require the assembled phonology route.

An fMRI study showed that, while L1 Korean word reading activated a different part of the brain from the area that was activated when Chinese characters were read, the activation pattern in reading Korean words was similar to that in reading L2 English words on the global level (Yoon, Cho, Chung, & Park). This result suggests that Korean is closer to English than Chinese in terms of orthographic distance. However, the strong activations of the posterior part of the right dorsolateral prefrontal cortex and the right hemispheric dominance of the occipital lobe were particularly observed in reading Korean words, compared to reading English, suggesting that reading Korean might be slightly different from reading English, based on the visuospatial analysis involved in reading Korean. Yoon, Cho, Chung, and Park also examined the neural mechanism of brain activation patterns of reading Korean words in Hangul and Hanja using fMRI. The brain localization of native Korean speakers’ reading Chinese characters was similar to that of native Chinese speakers’ reading their own L1 characters such that the left-lateralized middle frontal cortex was strongly activated. Reading in Hangul showed activation in the areas of the bilateral fusiform gyrus, left middle frontal gyrus, left superior temporal gyrus, right mid-temporal gyrus, precentral gyrus, and insula. These results corroborated the collective findings that different activation patterns were observed in reading alphabetic scripts and the Chinese logographic script. Given the strong activation of the posterior part of the right dorsolateral prefrontal cortex in reading, which belonged to the visual higher order control area, the researchers argued that the area of the right dorsolateral prefrontal cortex was responsible for the processing of visuospatial information of Korean words, because the surface form of Hangul was associated with architectural balance in the syllabic unit.

Similarly, Cho et al. (2014) carried out two studies to examine native Korean speakers’ word reading of two different scripts of Hangul and Hanja using fMRI. Their first study compared the pattern of cortical activation in reading Hangul and Hanja to find that Hanja reading would cause more activation in the larger areas of the brain than Hangul reading. The second study used Koreans’ popular Hangul and Hanja namesFootnote 1 to assess recognition memory in light of morphemic clarity in each character of the two-syllable name (i.e., the degree to which each character of the name preserves clear morphemic information; e.g., in Hangul {현자} <wise + offspring> vs. {동은}---vague meaning) and semantic transparency of the two-syllable name as a whole (i.e., the degree to which the combination of two characters in the name delivers a clear meaning; e.g., in Hanja {美玉} <beautiful jade> vs. {貞玉}--vague meaning). The results showed that high morphemic clarity in each letter of the two-syllable name yielded larger effects than those of high semantic transparency as a whole in recognition memory. In terms of the particular areas activated (Lee, 2004; Yoon, Cho, Chung, & Park), the brain activation seems to be different in reading Hangul and Hanja.

Kim and colleagues (Kim, Kim, Kang, Park, Lim, Kim, & Bak) investigated brain mapping and the neurolinguistic circuitry of visual script familiarity for cortical representation in reading Hangul and Hanja among adult native Korean readers with two groups of Hanja proficiency levels. Based on previous findings of different neural pathways engaged in reading according to a given script’s orthographic regularity, these researchers also examined, using an implicit word reading task for fMRI, the effect of script familiarity according to different orthographic regularity in Hangul (more familiar script) and Hanja (less familiar script). The fMRI blood-oxygen level showed that reading Hanja involved the ventral pathway, whereas reading Hangul was associated with the dorsal pathway. Both the right superior parietal lobule area and the left supplementary motor area were more stimulated in reading Hanja and Hangul for the lower-proficiency group than the high proficiency counterpart.

3.2 Reading in L1 Hangul and L2 or L3 English

Kim et al. (2016) investigated how linguistic distance between L1 and L2 affected the pattern of brain activation among Korean–Chinese–English trilinguals. Linguistic distance is defined as the degree to which two languages are different according to the nature of writing systems of the given languages. Since English and Korean are alphabetic scripts, their linguistic distance is closer than that between English and Chinese or between Korean and Chinese. Using a visual rhyming judgment task, Kim et al. (2016) examined fMRI of Korean trilinguals’ reading in Korean (KK), Chinese (KC), and English (KE), along with two control groups of native Chinese (CC) and English (EE) speakers. The results of fMRI showed that the pattern of brain activation of KC was more akin to that of CC than KK. This suggests the brain’s neural accommodation. The KC group showed higher accuracy rates with decreased activation in the regions of the KK network. This suggests a reduced assimilation. On the contrary, the brain activation pattern of KE was more similar to that of KK than EE. This suggests neuronal assimilation. The KE group showed higher accuracy with decreased activation in the regions of the EE network. This suggests reduced neuronal accommodation. An analysis of brain regions of interest in the left middle frontal gyrus showed greater activation for the KC group than the KE group. This suggests selective involvement in L2 reading depending on the script being read. Kim et al. (2016) indicated that the brain network involved in L2 reading made use of brain networks established in L1 through an assimilation process when linguistic distance between L1 and L2 was narrow. When linguistic distance is huge between L1 and L2, a significant modification of the neural network seems to take place.

Another study by Kim, Liu, and Cao examined L1 influences on L2 reading among Korean-English and Chinese-English bilinguals, along with control groups of Korean monolinguals and Chinese monolinguals, using a visual word rhyming judgment task in L2 English. Results showed that brain activation upon L2 processing was similar to that of L1 processing for both Korean and Chinese bilinguals. Both Korean monolinguals and bilinguals showed more activation in the right inferior frontal gyrus and medial frontal gyrus regions than Chinese monolinguals and bilinguals, suggesting that the processing of Korean and Chinese might be different from each other. For bilinguals, Chinese bilinguals showed greater activation in the left middle frontal gyrus area than Korean counterparts. Overall, similar brain networks were recruited for L1 and L2 activation within each language group. However, the language difference between Chinese and Korean seemed to remain the same in L2 processing, indicating solid L1 influences on L2 processing.

4 Neuroimaging Studies of Reading Non-Alphabetic Chinese and Japanese Scripts

Chinese characters and Japanese Kanji as well as the co-use of Pinyin for Chinese and Kana for Japanese offer a useful means to investigate commonalities and differences in reading due to the use of different scripts than European alphabetic scripts. Research findings of these languages and scripts undoutedly facilitate our understanding of the reading mechanism as well as brain activation and circuitry involved in reading.

4.1 Word Reading in Chinese

As the visual word form area within the left hemisphere ventral occipito-temporal cortex is known to be dominantly engaged in reading across scripts, script invariance has been assumed (Krafnick et al., 2016). Using fMRI, Krafnick et al. (2016) found that both English-speaking and Chinese-speaking monolingual first graders in the U.S. and China, respectively, showed activation in the left ventral occipito-temporal cortex when reading, with a significant overlap in the visual word form area, which suggests script invariance. Krafnick et al. (2016) further examined the left ventral occipito-temporal cortex region to find that Chinese children responded to object stimuli (line drawings) in the same way as that of reading Chinese characters. In contrast, English-reading children showed that the left ventral occipito-temporal cortex was more activated when objects were shown rather than English words. Collectively, these results endorsed script invariance in the visual word form area and indicated that the left hemisphere ventral occipito-temporal cortex was the area involved in character or word processing.

Based on the findings of recent fMRI studies showing that lexical processing in alphabetic languages took place in both ventral and dorsal neural pathways originating from the visual cortex, an fMRI study was conducted to identify the effective connectivity of brain regions in reading Chinese (Xu, Wang, Chen, Fox, & Tan, 2015). Xu et al. (2015) examined how the neural systems interacted with one another in reading Chinese by testing the multiple pathways model. Dynamic causal modeling showed that visual word recognition in Chinese involved the ventral pathway from the visual cortex to the left ventral occipito-temporal cortex. However, the activation of the dorsal pathway from the visual cortex to the left parietal region was not observed. The ventral pathway was linked to the superior parietal lobule and the left middle frontal gyrus. A dynamic neural network was formed with information flowing from the visual cortex to the left ventral occipito-temporal cortex to the parietal lobule and then to the left middle frontal gyrus. These findings suggest that the differences in the way in which orthography represents phonology across writing systems are likely to constrain the cortical dynamics connected to the brain regions.

An examination of the co-use of characters and Pinyin allows for a better explanation of brain functions in reading. In order to identify similarities and differences in reading the two scripts, Chen, Fu, Iversen, Smith, and Matthews (2002) tested the dual brain processing routes in reading Chinese characters and Pinyin using an fMRI. They compared the patterns of brain activity to see whether the same or different cognitive mechanisms were engaged in reading the two different scripts, using a block design of phonological and lexical tasks. Common brain areas, including the inferior frontal, middle, and inferior temporal gyri, the inferior and superior parietal lobules, and the extrastriate areas, were activated in reading Chinese characters and Pinyin with some variations across the regions depending on the script difference. Reading Pinyin yielded a greater activation in the inferior parietal cortex bilaterally, the precuneus, and the anterior middle temporal gyrus. Character reading was associated with the activation in the areas of the left fusiform gyrus, the bilateral cuneus, the posterior middle temporal, the right inferior frontal gyrus, and the bilateral superior frontal gyrus. Chen et al. (2002) concluded that both alphabetic and nonalphabetic scripts activated a common brain network for reading with no differences in terms of hemispheric specialization. However, some specialized areas were activated according to the script being read; that is, the inferior parietal cortex was involved for Pinyin via predominantly assembled processes, while the fusiform gyrus was engaged for Chinese characters via predominantly addressed processes.

Cao, Vu, Chan, Lawrence, Harris, Guan, Xu, and Perfetti examined the effect of instructional methods on brain activation. They have trained college students in a character-writing class (more focus on visual-spatial structure) and a Pinyin-writing class with a control group of English readers. fMRI showed that different networks were engaged in reading Chinese characters and English words, supporting the view that the brain’s accommodation occurred according to the script being read. The instructional effects were robust such that the character-writing condition yielded greater activation in the bilateral superior parietal lobules and bilateral lingual gyrus than the Pinyin-writing condition in both lexical decision and implicit writing tasks. A greater involvement of bilateral sensorimotor cortex was found for character-writing than Pinyin-writing in the lexical decision task. The recognition accuracy was related to the activation in right superior parietal lobule, right lingual gyrus, and left sensorimotor cortex. Consistent with previous behavioral studies, these researchers found that character-writing training facilitated connections with semantics through producing greater activation in bilateral middle temporal gyri, whereas Pinyin-writing training facilitated connections with phonology through producing greater activation in the right inferior frontal gyrus. The fact that the short-term training resulted in different connections in the brain has a significant implication for script relativity.

Similarly, research on the neural correlates of reading shows that the left middle frontal gyrus, which is typically involved in writing, is more active in reading Chinese than English. Cao and Perfetti (2016) assumed that the writing region would be more activated in Chinese reading due to the learning-to-read practices of copying characters. To test this hypothesis, they tested English speakers who had learned Chinese as a foreign language. Participants performed both reading and writing tasks in English and Chinese with one group learning Chinese characters by writing/copying characters and the other group by learning phonological properties to examine the effect of writing (copying characters) on reading. Results showed that the left middle frontal gyrus was more activated in writing than in reading regardless of English or Chinese, which confirmed that the left middle frontal gyrus was associated with writing. The left middle frontal gyrus was more activated in Chinese than English regardless of tasks performed. The group that had learned Chinese characters through character-writing showed more activation in the left middle frontal gyrus than the comparison group who had learned through phonological learning. The same results were found with native Chinese speakers, which ruled out the possibility that the above findings stemmed from language proficiency. These findings suggest that the reading-writing connection is modulated by learning experience.

Word recognition research in alphabetic scripts has revealed a possible facilitatory neighborhood size effect (i.e., facilitatory neighborhood size effects mean that the recognition of words with more orthographic neighbors is faster than that of words with fewer neighbors, whereas an inhibitory neighborhood size effects mean slower responses) in low frequency words. Li, Bi, and Zhang (2010) examined neural correlates of the orthographic neighborhood size effect in Chinese. Previous studies showed that reading Chinese characters invoked both facilitatory and inhibitory neighborhood size effects, depending on the frequency of neighbors. Li et al. (2010) also found, using fMRI, the facilitatory contributions of neighborhood size to orthographic activation in silent naming depending on the frequency of neighbors. Results identified greater activation in the left middle frontal gyrus for smaller neighborhood size than larger neighborhood size and activations in the bilateral inferior frontal region for high-frequency neighbors. Greater activation was found in the right middle occipital gyrus for larger neighborhood size than smaller neighborhood size when high frequency neighbors were absent; however, null neighborhood size effects were found in the presence of high frequency neighbors. These results suggest that different neural correlates are involved in reading according to neighborhood size.

Since multi-character words or compound words are prevalent in the Chinese lexicon, studies of multi-characers add new insights into reading mechanisms. Lin, Yu, Zhao, and Zhang (2016) examined the functional anatomy of the recognition of Chinese multi-character words in light of the effects of nonwords, lexicality, and word frequency. In order to rule out possible confounding effects (e.g., effects that are modulated by an interaction effect between different tasks) of reaction time, Lin et al. (2016) used transposable nonwords, regular nontransposable nonwords, and real words. They performed a conjuction analysis on contrasts beween transposable nonwords and regular nonwords and between words and regular nonwords in order to determine whether these different tasks activated the same areas of the brain. Both significant conjunctional effects and positive word-frequency effects were observed in the bilateral inferior parietal lobules and posterior cingulate cortex regions. Conjunctional effects were found in the anterior cingulate cortex area only.

Another study was conducted using multiple scripts. Xue, Jiang, Chen, and Dong (2008) examined how the writing system, stimulus length, and presentation duration affected visual word recognition using event-related potentials (ERPs). They compared early electrophysiological responses (i.e., the first negative peak; N1) to familiar and unfamiliar writings under different conditions in terms of lexicality (words vs. nonwords for familiar writings only), length (characters/letters vs. words), and presentation duration (100 ms vs. 750 ms). Native Chinese speakers with English as L2 participated in reading four types of scripts, including Chinese, English, Korean Hangul, and Tibetan. Results showed no significant differences found between words and nonwords. The language experience (familiar vs. unfamiliar) was significantly affected by stimulus length and writing systems and was affected by presentation duration to a smaller degree. Specifically, the language experience effect (i.e., a stronger N1 response to familiar writings than to unfamiliar writings) was significant for alphabetic letters only, but not for alphabetic words. The difference between Chinese characters and Hangul was significant in the condition of short presentation duration only, but not in the long presentation condition. Long stimuli elicited a stronger N1 response than did short stimuli in the familiar writings, suggesting that N1 response might not reliably differentiate the familiar script from the unfamiliar script being read. Overall, Xue et al. (2008) indicated that N1 was modulated by visual, linguistic, and task factors.

In summary, when reading Chinese characters, Pinyin, and English, common brain networks were involved in the left hemisphere ventral occipito-temporal cortex. This indicated a script invariance. However, specialized regions varied to the extent that scripts being read yielded different brain circuits that were recruited according to instructional methods (character-copying or phonological-learning), readers’ language and reading proficiency, neighborhood size, word frequency, and tasks to perform. The dorsal pathway from the visual cortex to the left parietal region was less likely to be active in Chinese reading than in English reading.

4.2 Word Reading in Japanese Kanji and Kana

Given that the Japanese use both Kanji (morphograms) and Kana (syllabograms) within one sentence, its complexities have drawn scientific interests in reading science. Research has demonstrated that Kanji and Kana are processed differently due to the difference in the nature of these two scripts. Kana are processed in a way similar to other phonetic languages such as English, while Kanji are processed in a similar way to Chinese characters (Nakamura, Dehaene, Jobert, Bihan, & Kouider, 2005; Sakurai et al., 2000; Thuy et al., 2004). Sakurai et al. (2000) found different cortical activities evoked upon reading of Kanji and Kana in a positron emission tomography study. Reading Kanji activated the lateral fusiform gyrus (BA 37), while reading Kana activated the middle and inferior occipital gyri (BAs 18 and 19) and the deep perisylvian temporo-parietal area (BAs 40/22 and 22/21). Nakamura, Dehaene, Jobert, Bihan, and Kouider (2005) also examined the functional architecture of visual word recognition in Kana and Kanji, using a subliminal priming method with fMRI. Participants were asked to perform semantic judgment of words, in which a subliminal presentation of either the same or a different word and in the same or a different script was followed by each target word. Word repetition (i.e., same word subliminal presentation) yielded a significant priming effect regardless of script presentation (i.e., Kanji or Kana). Results showed a shared visual occipito-temporal activation for words in Kanji and Kana. However, Kanji reading recruited slightly more mesial and right-predominant activation, while Kana reading was associated with greater occipital activation. These findings indicated that script-dependent and script-independent regions were engaged subliminally in the posterior temporal lobe for the different scripts of Kanji and Kana.

Coderre, Filippi, Newhouse, and Dumas (2008) examined the Stroop effects of color naming using fMRI to identify similarities and differences in brain processing in Kana and Kanji. Significant Stroop effects in reaction time were found within the Kana script and within the Kanji script, but there was no significant difference in reaction time between Kana and Kanji. The brain imaging data showed that the anterior cingulate gyrus, which was the area involved in inhibiting automatic processing, was activated for both Kana and Kanji. Although the Stroop effect was not significant in reaction time between Kana and Kanji, the two scripts showed the different areas of activation in fMRI. Specifically, the left inferior parietal lobule area was activated for the Stroop task in Kana, while the left inferior frontal gyrus region was activated for the Stroop task in Kanji. These results indicated that conflict detection and resolution occurred in the different brain regions according to script input, as evidenced by the activation of different brain areas depending on whether phonographic Kana or morphosyllabic Kanji were used.

A different activation in the brain regions according to varied scripts in size and scrambled-characters has also been investigated using fMRI. For example, Thuy et al. (2004) investigated the implicit and explicit processing of two-syllable Kanji and two-syllable Kana words (i.e., Kanji words were transcribed in Hiragana) and nonwords. In a task, rest (0), task 1 (Kanji), and task 2 (Kana) were alternately repeated as in 0102010201020102 in a time course block design. Each item was shown for 500 ms with a 500 ms inter-stimuli interval. The subject was asked to respond by judging which character was bigger between the first and the second. One of the two characters presented was 25% bigger than the other. The researchers operationally defined size judgment for character stimuli as an implicit language task for linguistic stimuli; operationally defined size judgment for scrambled-character stimuli as an implicit language task for non-linguistic stimuli; and operationally defined lexical decision as an explicit language task. The results of the fMRI study showed that the size judgment for scrambled-Kanji stimuli and scrambled-Kana stimuli led to activation in the bilateral lingual gyrus (BA 18), the bilateral occipito-temporal regions (BA 19/37), and the bilateral superior and inferior parietal cortices (BA 7/40). In addition to these areas, the left inferior frontal region (Broca’s area, BA 44/45) and the left posterior inferior temporal cortex (BA 37), which have been considered language areas, were also activated during size judgment for Kanji character stimuli (i.e., implicit language task). Size judgment for Kana character stimuli (i.e., implicit language task) also activated Broca’s area, the left posterior inferior temporal cortex, and the left supramarginal gyrus (BA 40). The lexical decisions for both Kanji and Kana (i.e., explicit language task) yielded activation in these language areas as well. These findings indicated that both implicit and explicit language processing was obligatory for both Kanji and Kana scripts. However, a comparison between the scrambled Kanji condition and the scrambled Kana condition showed no activation in the language areas. A comparison between Kanji and Kana scripts during size judgment and lexical decision showed a greater activation in the bilateral fusiform gyrus (dominant in the left). Thuy et al. (2004) used the common subtraction analysis, which is a method where the summed activation maps of all stimulation conditions are first rescaled based on average images and then the baseline activation maps are subtracted from the stimulation activation maps in order to specify the activation locations in the brain. A Kana-minus-Kanji analysis showed activation in the left supramarginal gyrus during size judgment. The Broca area and left middle/superior temporal junction were active during lexical decision. These results indicated that, despite largely overlapping cortical regions, implicit and explicit reading was different across Kanji and Kana. Thuy et al. (2004) concluded that Kanji reading seemed to be involved in more visual orthographic retrieval and the lexical–semantic system through the ventral route, whereas Kana reading required phonological recoding to access semantic information through the dorsal route. These findings are consistent with the concept of script relativity.

Differet tasks were also used in the comparison of the two scripts in the brain functions. Ino, Nakai, Azuma, Kimura, and Fukuyama (2009) conducted an fMRI study of Kanji and Kana to examine brain activation for processing words written in the two scriptal types using word recognition (pressing a button for the real word) and reading aloud. Brain activation was similar to each other for Kanji and Kana words in reading aloud tasks. However, the regions of bilateral frontal, parietal and occipito-temporal cortices (all of which are related primarily to visual word-form analysis and visuospatial attention) were activated in the word recognition task. Concerning the difference of brain activity between the two tasks, a differential activation was found only in the regions associated with task-specific sensorimotor processing for Kana. Greater activation was found for Kanji in the visuospatial attention network in the word recognition task than the reading aloud task. Ino et al. (2009) concluded that the differences in brain activation between Kanji and Kana were dependent upon the interaction between the script characteristics and the task demands.

In a similar vein, research showed that different neural circuits were activated in reading Kana and Kanji. Higuchi et al. (2015) examined the neural basis for the processing of the hierarchical visual form of Kanji characters using gradient stimuli from fragments of the character to real Kanji characters. Their stimuli included (1) real Kanji characters, (2) pseudo Kanji characters (subcomponents with partial characters), (3) artificial characters (character fragments), and (4) checkerboard (simple photic stimuli). Robust activation according to different stimulus types was found in the left occipito-temporal visual region along the posterior–anterior axis in the order of the structural complexity of the stimuli (i.e., from fragments to complete characters). Only Kanji characters produced functional connectivity between the left inferior temporal area and the language area (left inferior frontal triangulary). Pseudo Kanji characters produced connectivity between the left inferior temporal area and the bilateral cerebellum and left putamen. Higuchi et al. (2015) concluded that the visual processing of Japanese Kanji involved the left occipito-temporal cortex in a hierarchical structure within the region to the extent that the neural activation was sensitive to the hierarchical coding of the character from the fragment of the character to the full set of the character.

The different role of spatial frequency involved in reading Kanji and Kana has also been investigated. Using high-density ERP, Horie et al. (2012a) reported that the source of different reading in Kanji and Kana could be attributable to spatial frequency information, indicating that low spatial frequency information was associated with Kana, while high spatial frequency information was related to Kanji. In order to identify which brain areas were related to the difference between Kanji and Kana, Horie et al. (2012b) performed fMRI among native Japanese readers, presenting unfiltered and spatially filtered Kanji and Kana words. When either Kanji or Kana unfiltered stimuli were presented, the bilateral inferior temporal (BA 37) regions were activated, compared to the resting condition. Kana reading activated the bilateral inferior parietal lobules (BA 40), but Kanji reading did not activate this area. When Kanji and Kana reading was directly compared, Kanji reading yielded activation in the left inferior temporal region, while Kana reading activated the left inferior parietal lobules. For filtered high spatial frequency stimuli, the Kanji reading minus Kana reading comparison showed a significant activation of the left inferior temporal region only without activation in the inferior parietal lobules area. In contrast, for filtered low spatial frequency stimuli, the Kana reading minus Kanji reading comparison showed a significant activation of the left inferior parietal lobules only without activation in the left inferior temporal region. These results indicated that Kanji and Kana engaged in the relatively overlapping network, with more involvement of the left inferior temporal region for Kanji processing and with more involvement of the left inferior parietal lobules for Kana processing. Consistent with the results of Horie et al.’ (2012a) study with ERP, Horie et al.’s (2012b) fMRI results demonstrated that spatial frequency was the source of the dissociation found in reading Kanji and Kana (i.e., high spatial frequency in Kanji and low spatial frequency in Kana).

Koyama, Kakigi, Hoshiyama, and Kitamura (1998) conducted a magnetoencephalographic study to examine areas recruited for reading Kanji and Kana. Participant were asked to read 44 Kanji, 44 Kana, and 20 alphabet letters and to count the number of letters. The magnetic responses were recorded with dual 37-channel Superconducting Quantum Interference Device gradiometers from the temporal, parietal, and occipital areas of the brain. Similar magnetic responses were found for Kanji and Kana in the locations of equivalent current dipoles. The equivalent current dipoles in the posterior-inferior temporal areas were found, approximately corresponding to Brodmann area 37 in the latency range of 150-300 milliseconds. These activities were shown in both hemispheres without consistent laterality. As response time increases, the location of the equivalent current dipoles moved forward from posterior to anterior in the posterior-inferior temporal area. Given that activities in posterior-inferior temporal areas were also found in alphabet letters, the bilateral posterior-inferior temporal areas were considered to play an essential role in reading.

Since the Japanese multi-scripts allow for writing the same word in either Kanji and Kana, it is possible to differentiate a word’s lexical frequency from the visual form frequency. Making use of the Japanese scripts’ uniqueness, Twomey et al. (2013) examined the dissociation of visual forms of Kanji and Hiragana. Results showed faster responses to high frequency words than low frequency words and faster responses to visually familiar words than less familiar words in both Kanji and Hiragana. The brain imaging results showed that visual familiarity had a stronger effect on activation in the ventral occipito-temporal cortex than lexical frequency. Activation in the ventral occipito-temporal cortex was also greater for Kanji than Hiragana words, which was not due to their inherent differences in visual complexity. Twomey et al. (2013) explained these findings within a predictive coding framework in which the left ventral occipito-temporal cortex received the bottom-up encoding of complex visual forms and top-down predictions from regions encoding the non-visual attributes of the stimulus.

Ischebeck et al. (2004) examined the role of visual familiarity in brain function within the regular orthography of Japanese Kana using fMRI. They used two phonologically equivalent but visually dissimilar syllabaries, which allowed the writing of foreign loanwords to be written in two ways but only one of which was visually familiar. Three forms of Kana syllabaries included familiarly written words, unfamiliarly written words, and pseudowords (6 conditions in total: 2 types of foreign words x 3 forms of stimuli written in Kana) in a silent articulation task (Experiment 1) and a phonological lexical decision task (Experiment 2). In both experimental tasks, different brain regions were activated according to the three different stimulus types (familiar, unfamiliar, and pseudoword) especially surrounding the areas previously known for phonological encoding and word retrieval for meaning. Pseudowords and visually unfamiliar words, which were known to use phonological assembly, led to an increased brain activity in the left inferior frontal regions (BA 44/47), compared to visually familiar words. The two types of words (i.e., visually familiar words and unfamiliar words) activated the areas associated with lexico-semantic processing more than pseudowords, including the left and right temporo-parietal regions (BA 39/40), the left middle/inferior temporal gyri (BA 20/21), and the posterior cingulate (BA 31).

4.3 Word Reading in Chinese or Japanese in Relation to L2 English

Liu and Perfetti (2003) examined the time course of brain activity in reading both English and Chinese among Chinese-English bilinguals performing a delayed naming task. A principal component analysis of ERP from the onset of the stimulus indicated a different pattern of a temporal unfolding of graphic, phonological, and semantic processing depending on the script being read. Specifically, reading Chinese produced an earlier and higher amplitude shift (negativity 150 milliseconds; N150) than English at 150 milliseconds. Frequency effects were robust at 250 milliseconds for both Chinese and English. However, only the English frequency effect was reliable at 450 milliseconds. A source localization analysis by the Low Resolution Electromagnetic Topography showed that the visual recognition of Chinese characters involved the bilateral occipital regions (left BA 17, right BA 18). High-frequency English word recognition was involved in the left occipital region only (left BA 17), while low-frequency English words activated bilaterally in more diffused and extended temporal patterns. The right prefrontal area (BA 10) was strongly activated in the mid latency between 300 and 400 milliseconds periods of Chinese character naming, whereas English word naming showed more medial frontal (BA 8, and 10) activation. A post 450-milliseconds visual verification was general for both Chinese and English.

The neural strategies employed for the character-decoding and morphosyntax of Japanese and Chinese were also investigated using fMRI. Huang, Itoh, Kwee, and Nakada (2012) examined brain strategies for sentence reading among Japanese speakers who were literate in Mandarin and Mandarin speakers who were literate in Japanese. The activation pattern in the brain was distinctly different across the two groups. Irrespective of the participants’ native languages, Chinese reading activated more areas than Japanese reading, suggesting that Chinese reading was much more complex than Japanese reading. Chinese reading additionally activated the cortical areas in the right hemisphere. The activation pattern shown in Japanese reading by native Japanese speakers was highly consistent with previous reports, including the left inferior frontal gyrus, left posterior temporal lobe, and left ventral premotor cortex. The activation pattern associated with Chinese reading by native Chinese speakers was also highly consistent with previous reports, including the left inferior frontal gyrus, left posterior temporal lobe, left ventral premotor cortex, left anterior temporal lobe, and bilateral parieto-occipital lobes. The activation pattern shown in native Japanese speakers’ Chinese reading was identical to that shown by native Chinese speakers. However, native Chinese speakers’ reading Japanese yielded additional activation in the bilateral parieto-occipital lobes compared to native Japanese speakers. Huang et al. (2012) called the bilateral parieto-occipital lobes the “Chinese language area,” while the ventral premotor cortex was the “Japanese Kanji area.” This study suggests that the inferior frontal gyrus and posterior temporal lobe are universally involved as the language areas. The anterior temporal lobe seems to be essential for processing analytic morphosyntax in Japanese and Chinese.

The age of acquisition seems to be a contributing factor to brain networks as well. Kim, Relkin, Lee, and Hirsch (1997) examined how L1 and L2 were represented in the cortical areas by identifying the spatial relationship between L1 and L2 in the brain cortex. Neuroimaging results showed that L2 was processed spatially in separated regions from L1 within the frontal-lobe language regions (i.e., Broca’s area) for late bilinguals who had learned an L2 in adulthood. In contrast, L1 and L2 were processed in the common frontal cortical areas for early bilinguals who had learned an L2 at an early age. However, little or no difference was found in activation in the temporal-lobe language regions (i.e., Wernicke’s area) for both late and early bilinguals. These results suggest that L1 and L2 are processed in different cortical areas depending on the age of acquisition of an additional language.

In summary, there are common language areas involved in the brain for reading Japanese, such as the left inferior frontal gyrus, left posterior temporal lobe, and left ventral premotor cortex. Research shows that different scripts recruit slightly different brain regions. Kanji are processed through the ventral route which is more related to lexico-semantic processing as well as the areas of bilateral frontal, parietal and occipito-temporal cortices which are related to visuospatial attention. In contrast, Kana reading requires the dorsal route which is more related to phonological recoding as well as the left inferior parietal lobule area. In general, reading Japanese seems to be less complex than reading Chinese characters, as reading Japanese Kanji does not activate the cortical areas in the right hemisphere, which activates for reading Chinese characters.

Table 9.1 compares the brain regions activated when reading across the three East-Asian scripts, along with the areas that are generally activated in reading. Note that the table is not based on an exhaustive review of brain-imaging studies, however, due to space constraints.

Table 9.1. Representative Brain Areas Associated with Each Script
Full size table

5 Toward the Script Relativity Hypothesis: Biological Unity, Scriptal Diversity, and Cognitive Diversity

Neuroimaging research has identified the visual word form area that is specialized for reading in the neuronal networks on the left fusiform gyrus in the left hemisphere. This common area engaged in reading across different scripts indicates a biological unity. However, differences in the localization of brain regions have also been found according to the script being read.

In short, reading is affected by spoken language, the writing system, and its orthographic characteristics. Neuroimaging studies reviewed so far point toward not only different circuits and networks specialized depending on the nature of the script being processed (i.e., alphabetic or nonalphabetic and logographic or phonographic), but also robust L1 effects on L2 processing. Depending on tasks, languages, and reader or learner characteristics, such as proficiency and the age of acquisition, neuroimaging results vary. Although the identification of precisely specialized regions activated in the brain is still an open question, converging evidence provides a largely consistent picture. Research collectively shows that brain specialization and networks are associated differently for alphabetic scripts and non-alphabetic scripts (Cao, 2018; Kim, Liu, & Cao; Perfetti et al., 2007). As Cao (2018) asserts, the brain networks and circuits involved in word reading are significantly different across languages. Cao (2018) further notes that the cross-language differences in brain activation get larger as language and reading skills improve, given that reduced neural specialization has been found in children with low reading proficiency. Although the brain’s assimilation and accommodation are engaged in L2 processing, the degree of each mode involved in reading L2 seems to be variable according to the linguistic characteristics of L2. Evidence shows that late unbalanced bilinguals tend to use specialized brain networks established for L1 when processing L2, indicating L1 effects on L2 processing (Cao, 2018; Kim, Liu, & Cao). In contrast, early balanced bilinguals tend to show an overlap in brain activation in both L1 and L2. These findings suggest that the brain can accommodate the linguistic properties of an additional language by recruiting required brain regions as necessary. This is consistent with the neuronal recycling hypothesis posed by Dehaene (2009).

A series of studies show that native English speakers tend to accommodate the demand of L2 characteristics, while native Chinese speakers tend to assimilate L2 into L1 (Cao, 2018; Perfetti et al., 2007). Kim and Wang (2018) note that the difference between accommodation and assimilation in the brain network can be attributable to the linguistic distance between L1 and L2. Specifically, when the two orthographies of L1 and L2 are similar to each other, the brain is likely to assimilate L2 processing to the brain’s acquired L1 mechanism. However, when the two orthographies are different from each other, the brain is likely to accommodate the linguistic demands of L2. Regardless of the mode of involvement (i.e., assimilation or modification), L2 lexical processing requires more cognitive and brain resources than L1 lexical processing (Ma, Ai, & Guo, 2018).

The different neural specialization and adaptation according to language input can be explained through the theory of the universal language constraints which implicates the universal reliance of reading upon spoken language to the extent that writing systems encode spoken language (Perfetti, 2003). This also extends to the system accommodation hypothesis which links the accommodation of the universal dependency to the intricacies of the writing system. Neuroimaging evidence indicates that neural specialization occurs according to the linguistic features of a given language, as the brain adapts to the linguistic demands of an additional script to be read. Why would this different specialization occur according to the script being read? Since reading is a cognitive process, reading activities are orchestrated by our cognition, attention, and brain activity, and, in return, habitual reading influences what we pay attention to, how to process incoming information, and how to form new knowledge. At the center of this process is the script. This is a manifestation that the script in which we read can change our reading and further our thinking and cognition. Since different cultures have different scripts and different thought patterns, the fact that the script in which we read can change the way we think is closely related to script relativity. In fact, it would be difficult to explain the unique specializations in the brain regions found in Chinese, Japanese, and Korean without taking the lens of script relativity. Likewise, it would also be difficult to explain L1 scriptal influences on L2 script processing without looking through the prism of script relativity, given that L1 scriptal properties and skills constrain neuronal circuits and networks for processing L2.

It should be noted that script relativity is a new hypothesis. For this reason, we cannot readily find research in the extant literature that has directly tackled the matter of script relativity. All I can do at this point is to connect dots among the findings of previous studies from the angle of script influences on our thinking. However, an interpretation from the angle of script relativity has been missing in the existing literature that has specifically addressed L1 scriptal influences on L2 script processing. Script relativity is one explanation that can collectively construe the scriptal influences found in a multitude of studies of cross-scriptal transfer.

It has taken more than a half a century going through the ebb and the flow for linguistic relativity to gain a solid ground in recent decades with supporting evidence. There are still some scholars who are not in favor of linguistic relativity in the face of the vast amount of research that endorses the hypothesis. Considering the history of linguistic relativity, it may take a long time, if not as long as linguistic relativity’s trajectory, to accept or dismiss script relativity. Despite the possible controversy, what is meaningful is that I am opening the door to a new debate of script relativity. With improved research methodology and statistical advances, it is possible to efficiently covary potential intervening or spurious variables to single out script influences on our cognition. As I indicated in Chapter 1, I intentionally leave out South-Asian alphasyllabaries and abjads of Arabic and Hebrew from this book because the contrast between English and the three East-Asian scripts serves its purpose well and because I personally do not know those scripts. Well-designed research by scientists who are well versed in those alphasyllabaries of South Asia and abjads of Arabic and Hebrew scripts to test script relativity is expected to be carried out in reading science in the very near future.

With the unprecedented use of digital text in the current digital era, the influence of scripts can take a different mode than traditional reading. In consideration of the recent vicissitudes, Part III covers the impact of online reading and the future directions of script relativity.