The variations across language and writing systems have important implications for reading science. Perfetti and Verhoeven (2017, Table 19.1) present an extended summary of reading development across languages. Some conclusions are specific to writing systems and languages (e.g., phoneme awareness is more important for learners of alphabetic than nonalphabetic writing); some are applicable broadly within a writing system (e.g., phoneme awareness in alphabetic orthographies is not dependent on mapping consistency); some apply across all writing systems (e.g., children’s linguistic awareness emerges first at the syllable level).
One consequence of variation in mapping principles is variation in visual complexity. The number of graphs (the basic visual symbols of writing) depends on the number of linguistic units at the level where mapping occurs. In turn, the number of graphs determines their visual complexity: More graphs, more average complexity because the graphic features sufficient to distinguish among few graphs cannot distinguish among many graphs. The result is that abjads and alphabets, which typically have fewer than 40 graphs (letters), have less visual complexity than syllabaries and alpha‐syllabaries, which typically have more than 400 graphs. All systems are visually simpler than the Chinese basic morpho‐syllabary of more than 3 000 graphs (characters). In a study of graphs from different writing systems, Chang et al. (2017) reported that simple perceptual judgments of graphs vary with their complexity. Thus, visual complexity cannot be ignored in considering the challenges of learning to read. The long learning course required for Chinese and the many South Asian alpha‐syllabaries (Nag, 2017) is partly a reflection of the number of graphs and the resulting visual complexity of these orthographies.
Comparative research has also stimulated the extension of models of alphabetic reading to nonalphabetic reading. Li and Pollatsek (2020) presented an integrated model of word reading and eye movement control for Chinese, applying the Interactive Activation model (McClelland & Rumelhart, 1981) for word identification while also implementing word segmentation. Segmentation is needed because spaces separate characters but not words. PDP models have also been extended to reading Chinese (Yang et al., 2009; Zevin, 2019) and to morphological effects in Hebrew (Plaut & Gonnerman, 2000).
The brain’s reading network (revisited)
A universal brain reading network is strongly predicted by the fundamental principle of reading: that it converts systematic visual input (structured by writing systems) into language‐mediated meaning. Early comparisons confirmed this prediction across alphabetic languages, while also showing variation of the reading network in relation to consistency of letter‐phoneme mappings. English reading showed more use of a ventral pathway that includes the inferior temporal gyrus (IFG) compared with the more consistent Italian (Paulesu et al., 2000) and Spanish (Jamal et al., 2012). However, testing universality requires comparisons beyond alphabetic systems, and Chinese provides a high‐contrast comparison with alphabetic reading.
Early neuroimaging studies of reading Chinese produced evidence for both a universal network and writing‐system specific variations (Bolger et al., 2005; Tan et al., 2005), as does a more recent review (Xu et al., 2019). Universal areas include the left fusiform gyrus, highlighting its function in coding orthography regardless of the visual forms and mapping levels. However, Chinese shows more bilateral activation in posterior areas that support visual‐orthographic processing and a less prominent role in some (but not all) studies for the inferior frontal gyrus. Another difference is the more prominent role of the left middle frontal gyrus (LMFG) in Chinese. The LMFG’s location near a motor area involved in handwriting (Exner’s area) suggests that its prominence in Chinese reading reflects an effect of character writing on character reading, consistent with the importance of writing practice in Chinese literacy instruction. Evidence for this comes from greater overlap of passive recognition and imagined writing in the LMFG for Chinese than for English (Cao & Perfetti, 2017). The greater writing practice in learning to read Chinese may help secure long‐term orthographic memories for characters, consistent with conclusions from behavioral research (McBride‐Chang, Chung et al., 2011). Although writing seems especially important in Chinese reading, a study by Nakamura et al. (2012) comparing French and Chinese on recognizing handwriting suggested the writing‐reading role of the LMFG is shared across writing systems.
The significance and robustness of these Chinese‐alphabetic differences across different word reading paradigms remains an issue. In comparing meaning judgments made to speech and print, Rueckl et al. (2015) found the shared areas of print‐speech convergence across English, Spanish, Hebrew, and Chinese. These results affirm the universal connection of reading with spoken language. However, the brain networks for reading also reflect experience‐based accommodations to the orthography‐language connections required by the writing system (Cao et al., 2015).
Disruptions in the word‐identification system (revisited)
We should expect universal neural patterns associated with disruption in the word‐identification system for two related reasons: 1) the apparent existence of brain reading networks that include universal components; and 2) the language constraint that all writing systems map graphs to language. However, manifestations of word‐reading problems, including dyslexia, may vary with how the writing system makes demands on phonology. Such variation may depend on the level of phonological mapping – the grain size, phoneme or syllable (Wydell, 2019) – and the extent to which meaning encoded in morphology‐preserving orthography can compensate for a phonological deficit.
Chinese provides both of these. It maps syllables rather than phonemes and it has meaning cues in its morphological orthography that may further reduce the demands of phonology. The abjads of Hebrew and Arabic and the alpha‐syllabaries make additional demands on orthographic‐morphological processing, seemingly without substantially reducing the demands of phonology. Indeed, Chinese seems to require a multiple‐cause model that includes nonphonological sources. Phonological problems are found (Ho et al., 2000), but so too are associations of reading problems with rapid naming and orthographic knowledge (Ho et al., 2002) and morphology (Shu et al., 2006). Underactivation in the left middle frontal gyrus in Chinese readers with dyslexia appears more common than in alphabetic readers with dyslexia (Siok et al., 2004). If the LMFG supports neural‐motor preparation for character writing as part of character recognition (Cao & Perfetti, 2017), this may suggest an orthographic factor in Chinese dyslexia.
Visual‐orthographic processing challenges may be expected in Chinese, given the demands of learning around 3,000 characters in the first six years of school (see McBride‐Chang et al., this volume). Visual attention and copying skills have been found to predict reading ability of children in Hong Kong (Liu et al., 2015). In the multicause analysis, Chinese reading has phonological dyslexia, but fewer cases compared with alphabetic reading and even fewer cases with phonology as the only factor. Both visual‐orthographic processes and knowledge of Chinese compounding morphology may be important factors (McBride‐Chang, Lam et al., 2011). Interestingly, modeling of Chinese dyslexia (Yang et al., 2009; Zevin, 2019) suggested that either a morpho‐semantic or phonological disturbance produced wide‐ranging character reading problems in Chinese; in contrast, a semantic disturbance in English affected only identification of exception words.
The conclusion across writing systems might be that Chinese requires explanations of reading problems based on multiple factors, more than other systems. Phonological, morphological, and visual‐orthographic factors have been identified in behavioral research and inferred from brain imaging. However, these factors are likely to play a role in reading and