Researchers generally have acknowledged that sufficient orthographic knowledge benefits the processing of word spelling and reading comprehension (Perfetti and Hart, 2002; Ehri, 2005, 2014; Koda, 2007). However, as a morpheme-syllabic writing system, the orthography of Chinese is noticeably different from grapheme-phoneme writing systems such as English. This difference becomes one of the main challenges in learning Chinese (Shen, 2010; Hsiao et al., 2015; Zhang and Reilly, 2015; Li, 2020). There are two main difficulties that may be caused by this L1–L2 orthographic variance in CSL acquisition.

The first difficulty is due to complicated components of Chinese orthography, which requires lots of cognitive effort to decode, memorize, and distinguish (Qian et al., 2018). Chinese characters can be roughly classified into two categories: compound and integral (Wang et al., 2003; Shen and Ke, 2007; Du, 2015). In general, an integral character has only one radical on its own, and a compound character contains at least two or more radicals. The orthography of almost all Chinese characters consists of three elements: stroke, radical, and whole character (Shen and Ke, 2007). Usually, several strokes constitute a radical and several radicals form a character. Statistically, in modern Chinese, 214 radicals with 32 strokes that convey both sematic and pronounceable information appear in about 80% of 7,000 frequent Chinese characters (McNaughton and Ying, 1999; Chinese Language Committee, 2009).

The second difficulty is the weak correspondence between form and pronunciation of Chinese characters (Shen, 2005; Sung and Wu, 2011). Especially, majority of Chinese characters are homophones. For example, the phonogram “” (pronounced “zhu1,” spider), which consists of the semantic radical “” meaning “insect,” and the phonetic radical “” (pronounced “zhu1”) (the numbers are for indicating tones). Likewise, “,” “,” “,” and “” are all pronounced “zhu1” based on their phonetic radical “.” However, the connection among pronunciation, meaning, and character form is not always regular. In other words, it is diverse for the correspondence between the pronunciation of a phonogram and its combination of radicals. For instance, “” (pronounced “shu1,” different) is not dependent on its phonetic radical “” (zhu1) or its semantic radical “,” meaning “evil.” An indexical hypothesis (Glenberg and Robertson, 1999) claims that the most important language comprehension process is indexing of words or scripts to corresponding objects or mental representations. Together, these examples and studies indicate that understanding the underlying schema of Chinese characters could contribute to further Chinese reading comprehension of learners.

In summary, Chinese seems more linguistically complex than other languages might be on account of the unique combination rules of character orthography. However, the complicated features are almost certainly results of the historical evolution of the Chinese language for over thousands of years. As a basic unit of the Chinese language, character, which heavily carries culture, is a semiotic tool to communicate and a medium of cognition and thought.

Since the 1980s, the idea of embodiment has received increasing emphasis in cognitive sciences. According to the ECT, the medium through which one knows the world is the “body” (e.g., Clark, 1997; Lakoff and Johnson, 1999; Glenberg and Robertson, 2000; Streeck et al., 2011; Glenberg and Gallese, 2012; Wilson and Golonka, 2013). This “body” not only refers to one’s physical organisms but also bodily contact with the world (Inui, 2006). Namely, cognition is assumed to be grounded in embodied experience (embracing bodily activity and perceptual experience) and context. So far, scholars have a different understanding of embodiment. Based on previous literature, the implications of the embodiment of Chinese characters can be discussed in three aspects.

First, handwriting is an embodied activity. The ECT claims that cognitive processing is associated with the particularity of one’s physiological body (Shapiro, 2007; Streeck et al., 2011; Glenberg and Gallese, 2012). In other words, bodily activity might be a way of knowing. It has previously been observed that the perceptuo-motor process of bodily activity offers information and potentially influences the construction of knowledge (Streeck et al., 2011; Lawrence, 2012). For example, study of Strati (2007) described how a building worker skillfully walked on a roof using the tactile experience of his feet.

Thus, a handwritten script is “an imprint of action” (Longcamp et al., 2006). Writing by hand is a direct way by which bodily movements can interact with cognitive processing. In addition, handwriting mobilizes perceptuo-motor processing (Addy, 1996), and then the perception during writing contributes to cognitive processing (Longcamp et al., 2003; Tan et al., 2005b; James, 2017). Interestingly, brain neuroimaging research has found that some specific Brodmann’s areas were strongly activated for Chinese characters (Tan et al., 2001, 2005a; Hu et al., 2018). The findings from these studies suggest that the processing Chinese characters may be related to the experience of bodily activity. If so, handwriting, as an embodied action, might contribute to the cognitive processing of characters.

Second, learning a new language is learning a new conceptual system. Traditional dualism claimed that a physical substance (mind) had no bearing on a mental substance (body). Unlike the traditional view, the ECT supports the view that the experienced sensations and perceptions in the past are still work on one’s cognitive processing in the moment (Lakoff and Johnson, 1980). In short, an embodied experience might shape one’s way of thinking. Couched in the ECT framework, Lakoff and Johnson (1980) proposed that the way one conceptualizes the world is metaphorical. Likewise, there is a strong possibility that language is metaphorical. Therefore, to learn a new language² is likely to learn a new system of metaphors and conceptions apart from language features. Research has also shown that comprehension cannot be devoiced from the bodily sensory-motor system (Desai et al., 2010; Emmorey et al., 2014; Slepian and Ambady, 2014). For example, through tasks on building bodily metaphor, Slepian and Ambady (2014) found interactions among sensorimotor schema, comprehension and estimation, and abstract conception in reading.

Chinese, to some extent, is a representation of the ECT, with enormous metaphors involving embodied experience. Although simplification results in weaker link among radicals, the Chinese written language still represents the linguistic, psychological, and cultural features of the Chinese nation (Pae, 2020). According to Quinn (1987), a cultural model was critical for social members’ understanding of the world and their behavior. Cultural models underlying the conceptualization of Chinese characters reflect culture-specific concepts constructed in Chinese context. So far, researchers have proven the effect of past language usage of bilinguals on cognitive categorization in several conceptual domains, such as speech, gestures, memory for motion, and posture (Brown and Gullberg, 2008, 2011; Filipović, 2011; Smithson and Nicoladis, 2013), objects and substances (Cook et al., 2006; Pavlenko and Malt, 2011), emotions (Pavlenko and Driagina, 2007), motion events (Bylund et al., 2013; Park, 2020), and event construals (Wang and Wei, 2019). Moreover, behavioral and event-related potential (ERP) research has found that various language speakers have different ways of perceiving and conceptualizing the world (Liu et al., 2010; Jared et al., 2013; Wang et al., 2018). For example, Liu et al. (2010) investigated English and Mandarin speakers through category judgment tasks. English speakers showed larger N300 and N400 ERP component differences. In contrast, Chinese speakers showed no such differences in processing atypical and typical items (e.g., to judge the membership of “train” in English or huo3che1 “” in Chinese [atypical] and “car” in English or jiao4che1 “” in Chinese [typical] pictorial exemplars of a category “vehicle” in English or che1 “”in Chinese). The researchers further found that the Chinese speakers elicited moderate N300 and N400 effects on Chinese orthographically transparent items (e.g., the radical “bug” chong2 “” in character for the noun “butterfly” hu2die2 “”), while the English speakers showed such ERP effects on English morphologically transparent items (e.g., catfish). These findings indicate dramatic differences in the English and Chinese speakers’ processing of category judgments.

Therefore, one of the essential preconditions for understanding a script or a word is to underlie its metaphors and conceptions. If not, it may cause misunderstanding. Carrying over “” exemplifies the metaphor of Chinese characters. The metaphor “” originated from a Chinese schema that autumn (“”) is in one’s heart (“”). In detail, “” traditionally represents the physical heart and is the core of mind, thought, and mood [see Yu (2009) for more detailed explanations]. In Chinese culture, autumn is often a metaphor for mixed emotions of happiness of the harvest and sadness of the forthcoming desolate scene. Thus, “” generally means a complicated feeling of happiness, worry, solitude, and sadness. Compared to the integral characters stated before, “” is beyond simply describing the schema of concrete objects. All of these language idioms or conventions represent the perception and conceptualization of Chinese people in daily life.

Third, cultural background influences cognition of Chinese characters. The ECT believes that interactive relationships exist among our mind, action, body, and context (Lakoff and Johnson, 1980, 1999; Hutchins, 1995; Barsalou, 2008; Spackman and Yanchar, 2014).

In the first place, the development of Chinese characters is embedded in culture and history, and, in turn, characters that carry cultural and linguistic information influence surrounding regions. Historically, Chinese characters were cornerstones of East Asian culture development. Characters documented Chinese culture and then spread throughout other regions for millennia. These regions are considered as a cultural sphere, Sinosphere (also known as the East Asian cultural sphere, Chinese cultural sphere, or Sinic/Sinitic world). On the one hand, there are notable linguistic similarities among Sinosphere languages (Matisoff, 1973, 2001; Enfield, 2011; Brunelle and Kirby, 2016). On the other hand, Sinosphere regions, such as Japan, Korea, and Vietnam, share similar conceptions, which can reflect in Chinese characters.

In the second place, Chinese characters acquisition is affected by bilingual learners’ L1 context. Contextual factors affect and constitute cognition. ERP research has found that bilinguals’ cultural context impacts their semantic (Jared et al., 2013) and conceptual representations (Berkes et al., 2018) in lexical access. To elaborate, we continue with the example “” (heart). Matisoff (1986) discussed the concept of heart and mind in Southeast Asian languages with a comparative perspective in English. The research found that many Sinosphere languages prefer to describe psychological phenomena using metaphors of a bodily organ. For Chinese characters, radical usage, which represents bodily organs, like “” (heart), gives morpheme a psychological allusion. As mentioned above, the Chinese “heart” covers the meaning of both heart and mind, regarded as the center of both emotion and mood. However, like other European languages, English concepts of the heart are separated from the mind. This difference reflects a culturally specific concept developed in the Anglo-Saxon context (Wierzbicka, 1989, 1992; quoted in Yu, 2009). Therefore, individuals, especially bilinguals (Jared et al., 2013; Berkes et al., 2018; Wang and Wei, 2019), with different L1 sociocultural backgrounds are likely to have different cognitions for characters.

The role of Chinese characters in reading has received increased attention across a number of disciplines in recent years. Unlike an alphabetic language, the orthographic form of Chinese characters is more effective than phonological representations in reading comprehension because of weak regular corresponding rules between the form and phonology of characters (Xu et al., 2013). Some research has investigated the role of characters by stroke (e.g., Zhang, 2014; Chang et al., 2015; Dall et al., 2021), pinyin (e.g., Zhang et al., 2019; Xiao et al., 2020; Ju et al., 2021), character copy skills (e.g., Ye et al., 2021), phonological awareness (e.g., Siok and Fletcher, 2001; McBride-Chang et al., 2004; Zhang and Roberts, 2019, 2021a), morphological awareness (Nagy et al., 2002; Liu and McBride-Chang, 2010), radical awareness (Shen and Ke, 2007; Wong, 2017; Zhang and Roberts, 2019), and orthographic awareness (Siok and Fletcher, 2001; Loh et al., 2018; Tong et al., 2019; Chen et al., 2021; Zhang and Roberts, 2021a). However, differences in testing methods may lead to different results. For instance, some research (e.g., Tan et al., 2005b; Wong, 2017) employed a copy task in which materials were self-designed to the character writing ability of test participants, and the results indicated close relationships between Chinese handwriting and reading. However, other research (e.g., Zhai and Fischer-Baum, 2019) used a visual same/different judgment task on pairs of characters, and the results failed to find this significant contribution of character writing to reading.

Debates on the role of character writing have gained fresh prominence, with many arguing whether handwriting contributes to Chinese reading. There is a vast amount of research showing that strong knowledge of and skills in character writing facilitate orthographic processing in Chinese reading for children (Tan et al., 2005b; Wang et al., 2014, 2015; Chung et al., 2018; Zhang J. et al., 2020) and adults (Kuo et al., 2015; Tong and Yip, 2015; Tong et al., 2016; Zhang J. et al., 2020). However, literature has emerged that offers contradictory findings on strong effects of character writing on reading (e.g., Bi et al., 2009; Zhai and Fischer-Baum, 2019). For instance, a case study on a brain-impaired adult patient with writing deficit found no such effect of writing through reading and character writing tasks (Bi et al., 2009). Although he lost the ability to write by hand, this patient could still recognize Chinese and read it aloud. This finding displayed a clear dissociation rather than a strong relationship between Chinese writing and reading. However, conflicting results from these studies suggest the need to further investigate if CSL learners with different Chinese character writing skills also have variances in reading acquisition or, more likely, if are there factors that moderate the reading differences and, if so, what are the factors?

The importance of handwriting is now well established. While some research has been carried out on character recognition, there is very little scientific understanding of character writing. For instance, studies have found that visual-orthographic knowledge can help learners more quickly and easily recognize characters (e.g., Tong and McBride-Chang, 2010; Huang et al., 2020). Nevertheless, neuroimaging studies have found that reading Chinese characters activates motor-related zones apart from visual zones in the brain (Longcamp et al., 2003, 2005; Tan et al., 2005b; Yin and Zhang, 2021). In addition, research has also found a moderate effect of writing motor execution on deep orthographic processing (Yin et al., 2020). However, the influence of character writing on reading comprehension in CSL reading acquisition has remained unclear.

Embodied cognition theory research opens an avenue for a better understanding of cognitive processing and the three elements (bodily activity, language learning experiences, and context). Reading comprehension is one of the advanced cognitive processes closely associated with linguistic knowledge and cognitive skills (Stanovich, 2000; Koda2005a,2007; Jeon and Yamashita, 2014). Recent L1 and L2 reading research has extensively applied the component-skills approach to reading. One much-debated question addressed by Alderson (1984) in reading research was whether L2 language knowledge (e.g., orthographic, vocabulary, and phonological knowledge) was related to reading proficiency or cognitive processing (e.g., attention, working memory and metacognitive awareness). The existing body of research on language experience suggests that prior literacy experience influences literacy development in SLA (Hamada and Koda, 2008; Koda, 2008; Tong et al., 2016), and that language learning experience benefits in the early formation of perceptual organization (Kimchi and Hadad, 2002; Yeh et al., 2003). For example, Yeh et al. (2003) investigated the effect of language learning experience on the perceived graphemic similarity of Chinese characters with two shape-sorting tasks. The results showed that Chinese and Japanese undergraduates categorized characters based on configurational structures. In contrast, Chinese illiterate adults and kindergarteners classified characters by strokes or components. However, what is not yet understood is the relationship between language experience and CSL reading comprehension.

Language distance refers to the extent of similarity or differences between two languages. In other words, language distance is a measurement of classifying languages by their linguistic features. The current understanding of language distance in SLA suggests a positive cross-linguistic influence when L1 and L2 are linguistically similar or overlapped (Sharwood Smith and Kellerman, 1986; Jarvis and Pavlenko, 2008), and negative influence more likely occurs in beginners (Odlin, 1989). This influence may happen at a conceptual level (Jarvis and Pavlenko, 2008). According to Koda’s 2005a,b Transfer Facilitation Model, language transferring ability can provide top-down help for reading development and other associated abilities in another language. Because of overlaps and differences among all languages, language distance might be a main factor of the transfer.

Evidence from several studies suggest that language distance plays a moderate role in L2 reading acquisition (Melby-Lervåg and Lervåg, 2011; Jeon and Yamashita, 2014). A meta-analysis research (Jeon and Yamashita, 2014) study showed that linguistically similar languages (between Indo-European L1s and L2s) were more closely related to reading comprehension than linguistically distant languages. Interestingly, in the study of Jeon and Yamashita, language distance significantly affected reading comprehension but had no moderating effects on vocabulary or grammar knowledge, thereby rejecting their hypothesis. The authors explained that observing the cross-linguistic influence might be easier at a complex variable level (e.g., reading comprehension) than at a single variable level (e.g., vocabulary and grammar). Meanwhile, Tong et al. (2016) proposed a non-native Chinese character processing (NCCP) model to explain how semantic and phonetic radical information was accessed when learners with different L1 orthographies were processing Chinese characters. In this model, L1–L2 (Chinese) orthographic distance and context modulated the activation of word identification in two language systems. However, there has been little agreement on how L1–L2 distances of learners moderate the relationship between Chinese character writing and reading comprehension.

Sociocultural context is pivotal in SLA research (Vygotsky, 1978; Ellis, 2008; Green and Abutalebi, 2013; Tong et al., 2016). There are some interfaces between sociocultural theory (Vygotsky, 1978) in SLA and ECT in cognition: they both focus on the interaction between individuals and environments (Lan et al., 2015). According to the theory of Vygotsky, relationships between individuals and social context are inseparable; context and interaction with the context are two critical mediating factors in language learning (Ellis, 2008). Interacting in the social context provides essential SLA scaffolding (Vygotsky, 1978). This theory emphasizes the importance of external mediation, from which internal activity originates (Swain, 2000; Ellis, 2008). However, only few studies have investigated the role of character writing in Chinese reading acquisition based on these two crucial theoretical frameworks.

Theories on L2 reading have focused on predictors of social environment. For example, the Component Model of Reading (Joshi and Aaron, 2000, 2012) integrated individual and contextual factors. This model proposed that ecological and psychological surroundings had an impact on reading proficiency and cognitive skills. Moreover, recent bilingualism research has suggested ascribing language processing, cognitive processing, and brain organization to the experience of language learning and using and social factors surrounding this experience (Gullifer et al., 2018, 2021; Anderson et al., 2020; DeLuca et al., 2020; Tiv et al., 2020).

When it comes to Chinese characters, it is hard to overstate the importance of Sinosphere, because Sinosphere reflects the continuous improvement of East and Southeast Asian civilization around Chinese language and culture. Many previous studies have pioneered the effect of language background and compared CSL characters learners in various language contexts (Ke, 1998; Jiang, 2003; Xu, 2007; Zhang, 2008; Lin and Collins, 2012; Li et al., 2014; Ke and Chan, 2017; Tang and Chan, 2021; Zhang and Roberts, 2021b). Of these studies, Zhang (2016) compared the development of character orthographic awareness between ethnic and non-ethnic Chinese international students from Southeast Asian countries. The results failed to showed advantages of ethnic Chinese in the development of CSL character orthographic awareness but of character component and position awareness. Ke and Chan (2017) explored relationships among CSL reading strategies, L1 background, and L2 proficiency by examining participants from Chinese and non-Chinese cultural spheres. The results showed that intermediate learners with Chinese cultural background were more advantaged than the others, but this advantage disappeared when learners were more proficient in Chinese. These findings suggest a difference in L2 achievement in different L1 cultures.

Hanyu Shuiping Kaoshi (HSK, abbreviation of Chinese pinyin: Hànyǔ Shuǐpīng Kǎoshì; literally translated as Chinese Proficiency Test) is a standardized and the most widespread test for assessing the Chinese language proficiency of non-native speakers in the world. HSK development has gone through three stages (see review by Teng, 2017 and Meyer, 2014). The HSK Testing Center of Beijing Language Institute (BLCU) designed and developed this test in 1984. Then, in 2010, China’s Hanban/Confucius Institute Headquarters (i.e., Office of Chinese Language Council International) supported and reformatted the test as the New HSK. The designers kept all the three versions of HSK comparable (Hanban, 2011).

In this study, CSL reading and writing data came from the raw data of HSK (Elementary–Intermediate) (M = 50, SD = 15) in 2009.⁴ Why did we use the sample data of BLCU’s HSK rather than the New HSK? First, the BLCU’s HSK has a longer history. The BLCU’s HSK had advanced based on experience in CSL proficiency testing design and practice of over 26 years (1984–2010). During this time, researchers conducted many empirical investigations to improve the test (e.g., Chai, 2006; Cheng, 2006; Huang, 2006). Thus, these studies verified the BLCU’s HSK reliability and validity (Meyer, 2014). Second, the BLCU’s HSK mainly focuses on examinees’ integrative ability of four language skills (listening, speaking, reading, and writing) and communicative competence, which is more applicable for this study.

The HSK consists of three levels of CSL proficiency: beginning level (HSK Basic), intermediate level (HSK Elementary–Intermediate), and advanced level (HSK Advanced). Learners who have mastery of 2,000–5,000 Chinese words and certain grammar rules are at the HSK Elementary–Intermediate level. This level covers learners with a wider range of proficiency than other HSK levels. HSK Elementary–Intermediate consists of four subtests (i.e., listening [50 items], grammar structure [30 items], reading [50 items], and cloze [40 items]), with 170 items in total. Of these, 1–154 items are multiple choice and 155–170 items are fill-in-the-blanks. This study focused on the results of two sections: the reading section and character writing part of the cloze section.

The data on CSL reading proficiency came from the results of the HSK reading test. The whole HSK reading section lasted 60 min and consisted of two parts: word substitution and paragraph reading. The first part examined the ability of students to understand, recognize, and use words in specific contexts. Within the first part, the test gave 20 sentences with words underlined. Then, examinees chose the best words with the most similar meaning and usage from four options to replace the underlined word in each sentence. The second part examined the competence of students in paragraph reading comprehension and reading speed within a given time. For example, competence consisted of summarizing and interpreting texts, extracting keywords and sentences, deducing implicit information, and identifying attitudes, moods, opinions, or intentions of authors. The test contained six to eight short articles of varying lengths, difficulties, forms, and topics. In addition, examinees answered 30 multiple-choice questions (one point per item), Figures 1, 2, respectively, show sample tests in the first and second parts of the HSK reading test.

The study used the phrase “CCWP” to refer to Chinese character writing test scores that students got in the HSK test. The second part of the HSK cloze section was the basis of the CCWP data. This part mainly examined students’ CSL orthographic competence of mastery and usage of lexical words and writing. Test-takers filled in 16 blanks (one point per blank) within the orthography part with appropriate and correct handwritten Chinese characters based on three to four passages that were give. The whole cloze section lasted for a total of 30 min. Figure 3 shows a sample test in the HSK character writing test. The CCWP was divided into two groups with low and high writing proficiency (baseline = low level, score lower than an HSK character handwriting test score of 7/16).⁵

This study computed the language distance between participants’ L1s and Chinese through the Automated Similarity Judgement Program (ASJP) database. ASJP is an extensive cross-linguistic linked database belonging to the Max Planck Institute. The latest version (19th edition) of ASJP (Wichmann et al., 2020) covers 40-item word lists of 5,499 languages. These lists cover nearly 70% of the world’s extant languages.

With consideration for language diversity, this study adopted the L1s of participants rather than the official languages of their nationalities. By applying the ASJP program (Bakker et al., 2009), we calculated language distance values using the Levenshtein Distance method comparing phonetic forms of 40 core words (Holman et al., 2008; see Table 1). The less linguistically similar an L1 is to Chinese, the higher the ASJP value. The language distance ranged from 0 to 102.88 (M = 97.92, SD = 6.7).

Based on previous studies (Pae, 2020; Tang and Chan, 2021), Sinosphere contains five entities: China, Japan, North Korea, South Korea, and Vietnam. Some researchers prefer to divide Sinosphere into two categories: the “Chinese cultural sphere” and the “non-cultural sphere.” Others include South Asian countries (such as Laos and Singapore), with the limited L1 of learners influenced by the Chinese language and culture. For instance, they do not use Chinese characters in their language systems (B. Zhang, 2020). However, the dichotomy overlooks the individual differences caused by the influence of the L1 backgrounds of learners, CSL environments, and cultural factors (Li and Zhang, 2021). Thus, this study divided the sample by participants’ language context of current residence into three groups: the narrow-Sinosphere (NSG, i.e., learners from Japan, North Korea, South Korea, and Vietnam), broad-Sinosphere (BSG, i.e., learners from 10 Southeast Asian countries: Laos, Cambodia, Thailand, Burma, Malaysia, Singapore, Indonesia, Brunei, the Philippines, and East Timor) and non-Sinosphere (non-SG, i.e., learners from all other countries) groups (baseline = NSG).

Previous research has suggested that differences between males and females and children and adults may influence L2 reading and writing development (Phakiti, 2003; Jeon and Yamashita, 2014; Van der Slik et al., 2015; Vilas et al., 2019). However, reasons for the variation are still under investigation. This study controlled for the gender and age demographics of the examinees. Following previous research, two controlled variables were encoded as dichotomous, specifically, males as 1, females as 0, children (under 18 years old) as 0, and adults as 1.

The main purpose of the final model analysis was to explore the main and interaction effects of CCWP, language distance, and cultural background on the reading proficiency of the learners after controlling for gender and age. In Table 3, the L1 mean base value (γ₀₀ = 24.68, SE = 0.62, p < 0.001) indicates the average reading score for the reference student. This reference student was an adult female with a narrow Sinosphere cultural background and an average language distance between L1 and Chinese and mastered high Chinese character writing skills. The mean reading score of this student was 24.68.

H₁ addressed the relationship between CCWP and reading proficiency. The regression coefficient for CCWP was 12.48 (γ₃₀ = 12.48, SE = 0.34, p < 0.001). Since CCWP was coded as 0 = low and 1 = high, this meant that, on average, learners with high CCWP scored 12.48 points higher on the reading test. Accordingly, this finding supported H₁.

The language distance was negatively correlated to the average reading score (γ₀₁ = −0.06, SE = 0.01, p < 0.001). This result meant that the average reading scores should decrease by 0.06 scale points, each standard deviation higher on the language distance. This prediction does not seem very much, but the language distance in this study ranges from 0 to 102.88, so the predicted difference between the closest and most distant languages is (102.88–0) × 0.06 = 6.17 points on the reading test. This study also observed that, on average, the NSG learners outperformed their BSG and non-SG counterparts by 6.62 and 10.71 points, respectively. For the controlled variables, the results indicated a negative relationship between gender and reading score γ₁₀ (=−0.48, SE = 0.18, p < 0.01), but there was no significant difference between children and adult learners (γ₂₀ = 0.47, SE = 0.45, p > 0.05).

In this model, we also added interaction effects to test if the influence of CCWP still held. Table 3 also displays the unique effects associated with the interactions of CCWP on reading proficiency.

H₂ addressed the moderating effects of cultural background. Unexpectedly, this study found no interactions for cultural background. This result indicated that considering the main effects together, the beneficial effects of cultural background factors and CCWP were neither in conflict nor synergistic to learners’ improvement in reading proficiency (the Discussion section addresses this further). Therefore, the findings did not support H₂.

H₃ addressed the moderating effects of language distance. The results revealed that there was a negative interaction between language distance and CCWP (γ₃₁ = −0.06, p < 0.001). Figure 4 provides a visual representation of this interaction, signifying that the effect of CCWP on reading proficiency can range from curbing to favorable, depending on the language distance between L1s and Chinese. Specifically, each standard deviation increase in language distance predicted a slight score decrease by 0.06 points, which was small but significant, in reading text for learners with high handwriting skills. These results suggested that learners proficient in handwriting still held their advantage in reading when their L1s were more distant from Chinese. Thus, this finding supported H₃.

Moreover, outcomes of random effects showed the most variance between learners (σ²). This result indicated that individual differences of learners determined reading difficulty more than features of their L1s. This finding was not surprising, since learners succeed in reading development more through embodied activities, especially handwriting, than linguistic features. Interestingly, there was also variance between children and adults, although it was not significant. This finding may be due to the fairness of HSK design. Future research could explore more age-related factors to parse out this variance.

Finally, the model reported an σ² of 88.10 in the full model and 58.65 in the baseline model. According to these results, the effect size of explained variance (Cohen’s d = 0.33) for the full model was moderate (cutoff range: 0.15 < d ≤ 0.35) (Cohen, 1988). This finding meant that the slope variance reduced sharply after introducing the predictors of two levels and interactions in the null model.