Activation of L1 orthography in L2 word reading: Constraints from language and writing system

Abstract

When reading in a second language, a reader’s first language may be involved. For word reading, the question is how and at what level: lexical, pre-lexical, or both. In three experiments, we employed an implicit reading task (color judgment) and an explicit reading task (word naming) to test whether a Chinese meaning equivalent character and its sub-character orthography are activated when first language (L1) Chinese speakers read second language (L2) English words. Because Chinese and English have different spoken and written forms, any cross language effects cannot arise from shared written and spoken forms. Importantly, the experiments provide a comparison with single language experiments within Chinese, which show cross-writing system activation when words are presented in alphabetic Pinyin, leading to activation of the corresponding character and also its sub-character (radical) components. In the present experiments, Chinese–English bilinguals first silently read or made a meaning judgment on an English word. Immediately following, they judged the color of a character (Experiments 1A and 1B) or named it (Experiment 2). Four conditions varied the relation between the character that is the meaning equivalent of the English word and the following character presented for naming or color judgment. The experiments provide evidence that the Chinese meaning equivalent character is activated during the reading of the L2 English. In contrast to the within-Chinese results, the activation of Chinese characters did not extend to the sub-character level. This pattern held for both implicit reading (color judgment) and explicit reading (naming) tasks, indicating that for unrelated languages with writing systems, L1 activation during L2 reading occurs for the specific orthographic L1 form (a single character), mediated by meaning. We conclude that differences in writing systems do not block cross-language co-activation, but that differences in languages limit co-activation to the lexical level.

Keywords

bilingual lexical access orthography writing system

I Introduction

Writing systems reflect general mapping principles, varying in the level of language to which basic graphic elements are connected. Learning to read includes acquiring these connections and thus coming to implicitly know the principles of the writing system. Thus, in an alphabetic writing system, the learner connects written graphs (letters) to phonemes and, to a variable extent, to morphemes. In the morpho-syllabic system of Chinese, the learner connects the written units (characters and sub-character radicals) to syllabic morphemes. Not only are the mapping principles different, but the scripts – the formation of the graphs that give rise to different visual appearances – are highly contrastive. This high contrast between Chinese and alphabetic writing in both mapping principles and scripts leads to questions of comparison, not only for learning to read in a first language, but also for learning to read in a second language.

Here, we address the question of the effects of cross language and cross script relations for reading second language (L2) English by Chinese first language (L1) readers. Such a comparison involves unrelated languages, highly contrastive writing systems, and scripts of high contrast. The script factor in cross-language comparisons is important, although somewhat neglected in comparative reading research. Chinese writing has, by far, the largest number of orthographic units (characters) and thus the most complex graphs of any of the world’s orthographies (Chang et al., 2018). It is useful to keep in mind that, although scripts and writing systems are correlated for a good reason (Chang et al., 2018), they are separable. For example, Korean and English use different scripts, but they both belong to the alphabetic writing system. Arabic and Hebrew use the same abjad writing system, but they use different scripts. Before explaining how we address these cross-language issues in the case of L1-Chinese–L2-English, we provide a brief review of the key elements of Chinese compared with alphabetic writing that are important in comparisons of L1 reading.

1 Writing systems and L1 reading

In the alphabetic writing system of English, despite its spelling regularities, the graphic units (graphemes) correspond to phonemes. As a morpho-syllabic system, written Chinese maps its graphic units (characters) to a syllable that is also a morpheme. For example, ‘马’ corresponds to the syllable ‘mǎ’ and the morpheme meaning ‘horse’. Because of its morphemic mapping, Chinese characters can have orthographic overlap but with different pronunciations. For example, 马 (mǎ) is not only the character meaning ‘horse’, it also can stand as a left-position radical in the character 骄 (pronunciation: jiāo, meaning: ‘proud’). In Chinese, the sub-lexical orthographic constituent, a radical, is often assumed to be a basic sub-lexical component in models of Chinese reading (Ding et al., 2004) and has been found to be functional in Chinese character reading (Feldman and Siok, 1997, 1999). This structural feature allows ‘sub-lexical’ (radical-based) processes to co-occur with character processing. In the ‘proud’ example, neither the pronunciation (mǎ) nor the ‘horse’ meaning of 马 is a valid cue to the pronunciation or meaning of 骄. In other cases, a semantic radical (often the left radical in a horizontal character) can provide a cue to meaning, and a phonetic radical (usually the right radical in a horizontal character) can give a cue to pronunciation. For example: in character ‘晴’ (pronunciation: qíng; meaning: ‘sunny’), the semantic and phonetic radicals are 日 (meaning: ‘sun’; pronunciation: rì, which is related to the meaning of ‘sunny’) and 青 (pronunciation: qīng; meaning: ‘green’, which is related to the pronunciation of ‘晴’), respectively.

Empirical evidence from comparative L1 reading research points to differences as well as similarities in reading in Chinese compared with alphabetic reading. One similarity is that phonology is activated as part of word or character reading for native speakers of both English and Chinese. A difference is hypothesized to lie in the details of phonological activation. In English, phonology is activated immediately but incrementally with the decoding of orthography (Perfetti et al., 1988; Van Orden, 1991). In Chinese, the activation of lexical (or character) phonology is also rapid and automatic, but has been found to occur as processing reaches the threshold for orthographic recognition rather than ‘pre-lexically’ (Perfetti and Tan, 1998). The sub-lexical components, however, are activated as well and can play a role in the activation of character-level phonology and reading (Lee et al., 2006a). Models of Chinese reading capture the identification of characters along parallel links from graphs to meaning and to pronunciation (Perfetti et al., 2005) and establish a foundational role for radicals in character identification (Taft et al., 1999).

Another difference, more a matter of degree, is that, compared to English and other alphabetic writing, the large number of characters required by the morpho-syllabic structure of Chinese puts a premium on visual processing and visuo-spatial analysis (McBride-Chang and Ho, 2005; Perfetti et al., 2013; Tan et al., 2001). Chinese readers must learn to distinguish thousands of Chinese characters during L1 Chinese learning. With increased experience in reading Chinese, the involvement of visuo-orthographic analysis may increase (Cao et al., 2009). For skilled Chinese readers, those who attain high quality orthographic representations of characters, fine-grain representations of sub-lexical phonetic and semantic radicals are acquired and can function in character reading (Ding et al., 2004; Feldman and Siok, 1997, 1999; Lee et al., 2006a, 2006b; Su and Weekes, 2007; Tsang and Chen, 2009; Zhou and Marslen-Wilson, 1999; Zhou et al., 2013). Thus, the general picture is that with experience, native Chinese readers come to acquire detailed orthographic representations of thousands of specific characters and their sub-lexical constituents.

2 A within-language comparison: Alphabetic Pinyin reading in Chinese

Beyond its high contrast with alphabetic writing, Chinese itself is written in two different writing systems with different scripts. This allows an interesting comparison within the same language, as opposed to the typical cross language comparison in which the bi-orthographic reader reads different languages in the two writing systems.

Chinese also has an alphabetic system, Pinyin, an alphabetic transcription of spoken Chinese. Any Chinese word can be written in both morpho-syllabic Chinese characters and alphabetic Pinyin. To illustrate with our previous example, the English word ‘horse’ corresponds to the Chinese single-character word ‘马’.¹ Written alphabetically in Pinyin, the single syllable word for horse is written as ‘mǎ’. These different orthographic forms share the same pronunciation and at least one meaning, as shown in Table 1. In the initial phase of Chinese literacy education, the alphabetic writing system (Pinyin) is first introduced to facilitate later learning the morpho-syllabic system (character), which quickly becomes the primary written language unit. As characters become the dominant orthography, the use of Pinyin becomes limited, used as input to produce Chinese characters on computers and electronic devices and for introducing foreign names. The relative dominance of character reading to Pinyin reading for adults allows a perspective that treats characters as a functionally ‘first orthography’ and Pinyin as a functionally ‘second orthography’, although Pinyin was acquired first.

Table 1.

Examples of Chinese word and Pinyin correspondence.

English translation	Chinese word character²	Pinyin	Pronunciation (IPA)
horse	马	mɑˇ	mᴀ
sky	天	tiān	tʻiɛn
knowledge	知识	zhī shi	tʂʅʂʅ

With this perspective, Chen et al. (2014b) hypothesized that character orthography would function in the reading of Pinyin because of its status as the dominant orthography. Chen et al. tested this hypothesis in task that required reading Pinyin immediately before making a color judgment on a Chinese character. If reading Pinyin activates the corresponding character, the time to judge the color of this character should be affected. Skilled Chinese readers read a word in Pinyin script (‘bēi zi’ that corresponded to the two character word 杯子) presented with a very brief exposure (300 ms). They then immediately judged the color of a target character (杯) that, on some trials, corresponded to the preceding Pinyin word; participants were to focus on the color of the character and ignore its form. Chen et al. (2014b) found that color judgment times for the target character ‘杯’, which corresponded to the presented Pinyin, were shorter than for a O-S- character ‘炒’, which is unrelated in orthography, phonology, and meaning (pronunciation: chǎo; meaning: ‘fry’) to the Pinyin word ‘bēi zi’. This result suggests that, for skilled native Chinese readers, reading a Pinyin word activates its corresponding character.

As we described above, Chinese radicals can play a ‘sub-lexical’ role in character reading. Accordingly, additional studies by Chen et al. (2019a) examined whether the activation of the sub-lexical level components, the phonetic and semantic radicals, occurs with the character equivalent during Pinyin reading. Native Chinese readers made a semantic judgment on a word in Pinyin and then a color judgment on (implicit reading), or named (explicit reading), a target character. In both cases, the presented character sometimes shared a sub-lexical component (semantic radical or phonetic radical) with the character that corresponds to the preceding Pinyin word. The results were that reaction times of characters that shared radicals with the corresponding characters of the Pinyin word were longer than unrelated characters in both the color judgment task and the naming task. This implies that the sub-lexical component (radical) was activated with the character equivalent when the participants read Pinyin for meaning. The more general conclusion is that the character has become the gateway to reading through experience. Further, skilled readers have developed well-specified representations of characters that include their radical components, creating a single representation of high lexical quality (Perfetti, 2007) and allowing interactions among phonological and meaning constituents to occur automatically and relatively synchronously.

If reading alphabetic Pinyin activates characters for Chinese readers, reading alphabetic English may also. However, the difference between within-language and cross-language activation, especially for unrelated languages, may limit this process. In particular, within Chinese the character orthographies and alphabetic Pinyin are associated with identical meanings and pronunciations, whereas across Chinese and English, the two orthographies are associated with similar (sometimes identical) meanings and different pronunciations. Before providing an overview of our experiments, we first review the main research findings on L1 activation in reading a second language with a different script or a different writing system, especially studies of L1 Chinese and L2 English.

3 L1 activation in reading a second language

Issues of L1–L2 co-activation are addressed in models of bilingual representations, including the BIA+ model (Dijkstra and Van Heuven, 2002), Multilink model (Dijkstra et al., 2019) and Revised Hierarchical Model (RHM; Kroll and Stewart, 1994; Kroll et al., 2010). For example, the BIA+ model and Multilink models propose an integrated L1–L2 lexicon that allows lexical candidates from both languages to be activated (Dijkstra and Van Heuven, 2002). These models represent both lexical and sub-lexical components, but have focused more on alphabetic languages (Dutch and English), with limited applicability for languages with different writing systems and scripts (Goral, 2019; Jiang, 2019; Mishra, 2019). More related to our issue is a computational model specifically for Chinese–English by Wen and Van Heuven (2018) that addresses across-writing system issues, but is restricted to the character level. More useful for the sub-lexical level is a conceptual model proposed by Degani et al. (2018) for different-script bilinguals. This model represents variations in orthographic overlap between L1 and L2 on both lexical and sub-lexical levels, although it does not make explicit predictions for how the writing system or script differences affect the orthographic activation at these two levels.

Most relevant for our study, which focuses on across language and across writing system co-activation, is the experimental evidence for L1–L2 interactions between unrelated languages and writing systems. Indeed, cross-language interactions have been widely reported even when the languages have either different writing systems and scripts (Chen et al., 2014a; Miwa et al., 2014; Nakayama et al., 2012; Nakayama et al., 2016; Smith et al., 2019) or different scripts only (Degani et al., 2018; Kim and Davis, 2003; Peleg et al., 2019). (For a meta-analysis, see Wen and Van Heuven, 2017.) For example, in a comparison that involves the same writing system but a different script, Arabic–Hebrew bilinguals performed a semantic relatedness task for Hebrew word pairs. When the Hebrew prime word had a ‘false cognate’ in Arabic, i.e. a word with similar phonology but different meaning, decision times were slowed, suggesting interference in L2 processing based on shared L1 phonology (Degani et al., 2018).

In the case of L1-Chinese–L2-English reading, evidence that the cross-language interactions involve automatic L1 activation comes from studies using semantic relatedness judgments (Thierry and Wu, 2007; Wu and Thierry, 2010, 2012) and lexical decisions (Zhang et al., 2011). For example, in an implicit priming paradigm, Chinese–English bilinguals judged whether two English words (e.g. train–ham), each requiring two characters for their Chinese equivalents, are related in meaning (Thierry and Wu, 2007). When the Chinese equivalents of the unrelated English words shared a character, as they do in train–ham (火车–火腿), the amplitude of the N400 (an event related potential, or ERP, component that responds to word repetition, among other things) of the semantically unrelated English words was reduced compared to the unrelated words whose Chinese equivalents that do not share a character as is the case for apple–table (苹果–桌子). Comparable results with behavioral measures were found in a masked priming paradigm (Zhang, Van Heuven, and Conklin, 2011) in which Chinese–English bilinguals made lexical decisions on an English word that was preceded by a meaning unrelated English word. Participants responded faster to an English word pair (e.g. east–thing) whose Chinese equivalents shared a character (东–东西) than to an English word pair whose Chinese equivalents do not share a character.

In a subsequent study, Wu and Thierry (2010) experimentally separated the contribution of orthography and phonology to the Chinese L1 effects of Thierry and Wu (2007). Participants judged whether a pair of English words (e.g. account–conference) were meaning related. The Chinese translation equivalents of the English word pair either shared orthography but with different pronunciations (会计–会议: kuǎi jì and huì yì) or shared pronunciation but with different characters (经验–惊讶: jīng yǎn and jīng yà). The N400 showed reduced amplitude in response to shared pronunciations but not in response to shared orthographies (Wu and Thierry, 2010). This suggests that, in this paradigm, it is the character-level phonology of the L1 equivalent rather than the character orthography that is activated in L2 English reading.

Ma and Ai (2018) reported an orthographic effect in a translation recognition task. Chinese–English bilinguals judged whether a Chinese character is the correct translation of a preceding English word. Participants were slower and less accurate in rejecting a character (坏, meaning: ‘bad’) that shared a component with the translation equivalent (杯) of the preceding English word (‘cup’). However, it is not clear that translation effects arise only from automatic activation of L1. The shared component instead could be noticed when the participant views the character and compares it to the character that would correctly translate the English word.

In general, studies of L1 Chinese–English bilinguals have provided evidence for L1–L2 effects, but only a few of these effects can be interpreted clearly as due to automatic activation of characters during L1 reading (Thierry and Wu, 2007; Wu and Thierry, 2010; Zhang et al., 2011). Further, at least some studies designed to reveal such automatic effects report null results (Wen and Van Heuven, 2018). Finally, these studies have focused on character-level effects. Whether the sub-character components are part of the cross language interaction across writing systems, as has been found within Chinese (Chen et al., 2014b, 2019a), has not been addressed. With this background, the current study addresses both lexical (here the character level) and sub-lexical (radical) activation in L2 English reading in a paradigm shown to produce cross-writing system effects within Chinese.

4 The current study

There are two goals of the current study. First, we ask whether both the L1 Chinese meaning equivalent character and its sub-character orthography are involved in L2 English word reading in tasks that produce positive results within-language for the two writing systems of Chinese (Chen et al., 2014b, 2019a). Second, we address the role of language and writing system factors in bilingual activation by comparing the results of our cross-language studies with results from within-language experiments using alphabetic Pinyin reading.

The findings of Pinyin reading allow the suggestion that the character becomes the gateway (i.e. the functional L1 orthography) to access the Chinese lexicon, even for reading alphabetic input Pinyin. It is possible that reading alphabetic L2 English to meaning automatically activates both the lexical and sub-lexical level representations of the character equivalent as in Pinyin reading. Alternatively, within-language activation of the ‘first’ writing system should be stronger than cross-activation of L1 in a second writing system. Within Chinese, the Pinyin and the character represent the same language and have the same meaning and phonology, whereas across languages, the English word and the character represent different languages and, importantly, are not necessarily identical in meaning, but may only partially share meaning.

Two different tasks were used across the experiments, one that implicitly involves the L1 word – i.e. judging the color of the character – and one that explicitly involves it, i.e. naming the character. Experiments 1A and 1B used the color judgment task. Participants in Experiment 1A first silently read a briefly exposed (300 ms) English word, and immediately following, reported the color of a character that varied in its relation to the character equivalent of the English word. In Experiment 1B, instead of silent reading, participants first made a semantic judgment on the English word. Experiment 2 was similar to Experiment 1B, except that following the semantic judgment of the English word, participants were asked to name the character that followed. These varied tasks can test the activation of L1 Chinese translation equivalent across different tasks and provide direct comparisons with Pinyin reading (Chen et al., 2014b, 2019a), in which activation of the character equivalent was found on both the lexical level and sub-lexical level.

II Experiment 1A

1 Method

a Participants

Thirty-three Chinese–English bilinguals (24 females) with an age range 18–32 (M = 24, SD = 2.8) from the University of Pittsburgh were paid for their participation. Participants were all from mainland China and proficient in English as a second language. All participants had intensive in-class English instruction in China for at least 10 years prior to coming the USA and had lived in the USA for 16.3 months on average (SD = 10.8) at the time of the study. The participants in all three experiments had to reach a TOEFL (Test of English as a Foreign Language) score of at least 80. In Experiment 1A, the scores ranged from 90 to 114. They were right-handed with normal or corrected-to-normal vision. Each participant signed the consent form before the experiment. All procedures were approved by the University of Pittsburgh Institutional Review Board.

b Stimuli and design

Eighty sets of English word and character pairs were created for the experiment. Each set had four versions, one for each of the four conditions. To illustrate the four conditions, the ‘target’ character (嘴: mouth) that followed the English word was the same across the four conditions:

Equivalent condition: The ‘target’ character (嘴 ‘mouth’) presented for color judgment is the meaning equivalent character of English word ‘mouth’.

O+S+ condition: The target character (嘴 ‘mouth’) is meaning related to the English word ‘sing’ and shares a semantic radical (口) with the first character (唱) of the two-character meaning equivalent (唱歌) of the English word ‘sing’.³

O+S- condition: The target character (嘴) shares the semantic radical (口) with the first character (喷) of the two-character meaning equivalent (喷泉) of the English word ‘fountain’, but the character and the English word are unrelated in meaning.

O-S- condition: The target character (嘴) has a meaning that is unrelated to the English word ‘railway’ and does not share any orthographic information with the meaning equivalent of the English word (铁路).

In the O+S+, O+S-, and O-S- conditions, the target character does not have any phonological overlap with the two-character meaning equivalent of the English word. The design and sample stimuli are presented in Table 2. The full list of stimuli is presented in Appendix 1.

Table 2.

Example stimuli of each condition in Experiment 1.

Condition	English Word	Chinese equivalent of the English word with pronunciation (Pinyin and IPA)	Target character
Equivalent	Mouth	嘴 (zuǐ, ʦuei)	嘴 (zuǐ, ʦuei)
O+S+	Sing	唱歌 (chàng gē, tʂʻɑŋkɤ)	嘴
O+S-	Fountain	喷泉 (pēn quán, pʻənʨʻyn)	嘴
O-S-	Railway	铁路 (tiě lù, tʻiɛlu)	嘴

We will explain the rationale of these four conditions. The Equivalent condition tests whether the Chinese meaning equivalent is activated in L2 English word reading; if so, the color-judgment latency of the character in the Equivalent condition should be shorter than in the O-S- condition. The O+S- condition tests the activation of sub-lexical orthography in L2 English word reading, because the equivalent character of the English word shares a radical (sub-character) with the colored target character, but not a meaning. If the sub-character orthography is activated, this should result in longer color judgment latencies in the O+S- condition than the O-S- condition. This is because the orthographic activation of the target character would draw more attention from color judgment and increase the color judgment latency (Klein, 1964; MacLeod, 1991; Warren, 1974). The comparison between O+S+ and O+S- condition tests the meaning effect of the radical beyond its orthographic form. In the O+S+ condition, the equivalent character of the English word shares both meaning and orthography with the colored target character, whereas the O+S- condition shares only orthography. Thus, if there is a meaning effect, separable from the character form, the color judgment reaction times in the O+S+ condition should be longer than the O+S- condition. One thing to note is that we expect shorter, not longer, color judgment latency, for the Equivalent condition, compared with the O-S- condition based on the findings of the parallel Pinyin study (Chen et al., 2014b).

To counterbalance the relations between English words and target characters, there are 76 pairs of English words and characters as fillers, in which the Chinese translation equivalent of the English word does not share orthographic, semantic, or phonological information with the target character. The frequency of English words and their Chinese meaning equivalents was balanced across conditions, ps > 0.1. Word frequencies shown in Table 3 for English and Chinese were based on the frequency dictionary of Brysbaert and New (2009) and Cai and Brysbaert (2010), respectively.

Table 3.

Word frequency (per million) of stimuli (range in parentheses).

Condition	English word	Chinese equivalent
Equivalent	62 (2–289)	74 (0.2–417)
O+S+	90 (0.3–4525)	45 (0.5–1607)
O+S-	44 (0.1–454)	51 (1.5–662)
O-S-	58 (0.9–1830)	45 (0.2–402)

c Meaning relatedness assessment for word pairs

Another 32 participants from the same bilingual population did the meaning related assessment of the English word and the target character on a 7-point scale with 7 as maximum semantic relatedness. The average word meaning relatedness was 6.8, 4.2, 1.6, and 1.5 in the Equivalent condition, O+S+ condition, O+S- condition and O-S- condition, respectively. The word meaning relatedness in the Equivalent condition and O+S+ condition was significantly higher than the O-S- condition, ps < .001 and in the Equivalent condition it was significantly higher than the O+S+ condition, p < .001. The O+S- condition did not show differences from the O-S- condition, p = .27.

2 Procedure

Participants were tested individually in a quiet experimental lab, seated approximately 50 cm from a computer screen. On each trial, a fixation signal (a black ‘+’) was first presented at the center of the computer screen for 500 ms, followed by the appearance of an English word in 24 Times New Roman. Each word was presented for 300 ms. After the English word disappeared, a colored target character was presented, and participants had 2,000 ms to respond by making a keyboard response to indicate whether the color was blue or red (‘f, j’). Participants were instructed that they would see an English word briefly followed by a character. They were asked to respond to the color of the character as quickly and accurately as possible. The keys indicating ‘red’ were counterbalanced among participants and in each condition the response numbers of red and blue were the same. A 1,000 ms blank followed each trial. The trials were presented randomly across conditions. The experiment lasted about 12 minutes. There were 20 practice trials before the experiment.

3 Results

The main results of interest were color judgment times, which are shown in Table 4. The key results were as follows:

Color judgment times for the Equivalent condition were shorter than for the O-S- condition.

Times in the O+S- condition did not differ from the O-S- condition.

Times in the O+S+ condition were longer than the O+S- condition.

Table 4.

Means and standard deviations (in parentheses) of target character reaction time (RT in ms) in Experiment 1A, 1B, and 2.

Conditions	Experiment 1A	Experiment 1B	Experiment 2
Equivalent (mouth–嘴)	552 (194)	577 (194)	532(158)
O+S+ (sing–嘴)	590 (260)	598 (210)	577(167)
O+S- (fountain–嘴)	570 (206)	574 (169)	577(154)
O-S- (railway–嘴)	577 (230)	573 (177)	572(148)

The following paragraphs report the statistical tests that support these conclusions. Color judgment accuracy is listed in Appendix 2.

We analysed the data using linear mixed-effects modeling (Baayen et al., 2008) with character color judgment times as the dependent variable. Trials with incorrect color judgment responses (3.1%) were excluded in the reaction time (RT) analysis. The base model included condition as the fixed effect and intercepts for participants and items as random effects. Model comparisons determined whether the final model included the random participant slope or random item slope for the condition. The final model had condition as a fixed factor, and random intercepts for participant and item and a by-participant slope for condition as random effects. To determine condition effects, we compared a model with condition as a fixed effect against a reduced model without condition. The two models differed, χ² = 42.72, p < .001, supporting the conclusion of a condition effect. Planned pair-wise comparisons showed that color judgment times in the Equivalent condition were significantly shorter than the O-S- condition, Estimate = −22.75 ms, SE = 11.26, t = −2.02, p < .05. The O+S- condition and O-S- condition did not show differences, Estimate = −8.50, SE = 9.97, t = −0.86, p = .40. The O+S+ condition was significantly longer than the O+S- condition on color judgment reaction times, Estimate = 21.90, SE = 10.48, t = 2.09, p < .05.

We used generalized logistic mixed models (GLMM) to analyse the color judgment accuracy. The final model for color judgment accuracy had condition as a fixed factor, and participant and item intercepts as random effects. No main effect of condition was found on color judgment accuracy, χ² = 3.70, p > .1.

4 Discussion

Experiment 1A, based on differences in color judgment times for characters that varied in their relation to the meaning equivalent characters of the presented English word, demonstrates activation of the Chinese meaning equivalent character during L2 English word reading. Color judgment times in the Equivalent condition were shorter than in the O-S- condition, suggesting that activation of the meaning equivalent character was facilitative in this case. However, there was no evidence for activation of the sub-character orthography (radical), because color judgment times in the O+S- condition did not differ from the O-S- condition.

Color judgment reaction times in the O+S+ condition were longer than the O+S- condition, suggesting interference when the English word and the target character are meaning related. The interference in the O+S+ condition is consistent with the assumption that increased meaning activation of the target character attracts attention from color judgment, thus increasing color judgment latency compared with O-S- condition (Klein, 1964; MacLeod, 1991; Warren, 1974). This contrasts with the Equivalent condition, where the lexical constituents are in full alignment. The facilitative effect of the meaning equivalent and interference from related meaning are consistent with the results in the Pinyin studies, which employed the same paradigms (Chen et al., 2014b, 2019a). In the Pinyin studies, the character equivalent showed facilitation and related meaning showed interference for color judgment as well. The combination of facilitation in the Equivalent condition and interference in the O+S+ condition indicates that the equivalent effect is not simple accumulation of the effects of meaning, orthography, and phonology. Otherwise, the equivalent character would show greater interference than the meaning related characters. Findings from an ERP study of Guo et al. (2012) suggest that the semantic effect and equivalent effect in L2 word reading are independent and occurred in different time windows. The facilitative equivalent effect reflects a unified effect of the three lexical constituents of L1. For skilled readers, the specific and redundant representations on the lexical level would encourage synchronous access and activation of the three lexical constituents of L1 words in L2 word reading, even though meaning overlap is the only association for the non-cognate words across two languages.

In summary, the results of Experiment 1A indicate that activation of the L1 Chinese equivalent occurs at the character level in L2 English word reading, mediated by meaning. There was no evidence that the activation of the equivalent extended to the sub-character level (radical). In Experiment 1B, we explored the activation of the L1 meaning equivalent character when readers have a more specific reading task for the L2 word. In Experiment 1B, the silent reading task for the English word in Experiment 1A was replaced by a semantic judgment task, which requires explicit attention to word meaning.

III Experiment 1B

1 Method

a Participants

Thirty-three Chinese–English bilinguals (21 females) with an age range from 18–30 (M = 22.7, SD = 3.1) from the same population of Experiment 1A participated in Experiment 1B. All participants had intensive in-class English instruction in China for at least 10 years prior to coming to the USA and had lived in the USA for 19.0 months on average (SD = 12.1) at the time of study. The TOEFL scores ranged from 80 to 111.

b Stimuli, design and procedure

The stimuli and design were the same as Experiment 1A. The procedure was adapted from Experiment 1A, with a major change in the English word task, for which participants were asked to make a semantic judgment of the English word, deciding whether the word referred to a person. They responded to the English word when it referred to a person (e.g. nurse) and made no response to all other English words (e.g. flower). The 80 sets of stimuli for four experimental conditions did not require a response. There were 36 word pairs as fillers, in which the English word referred to a person and thus required a response (SPACE BAR) within 1,000 ms. There were 40 other fillers to counterbalance the relations between English words and target characters, in which the Chinese translation for the English word did not share either a semantic radical or meaning with the target character. After the semantic judgment response to the English word, a colored target character was presented and participants had 2,000 ms to make a keyboard response that indicated whether its color was blue or red.

2 Results

The main result of interest was color judgment times. The key results were as follows:

Color judgment times for the Equivalent condition were not different from the O-S- condition.

The color judgment latency of the O+S- condition was not different from the O-S- condition.

The color judgment latency of the O+S+ condition was significantly longer than for the O+S- condition, replicating the meaning effect in Experiment 1A.

The following paragraphs report the statistical tests that support these conclusions. The decision times for the color judgments of target characters are shown in Table 4. Accuracies of semantic judgment and color judgment are listed in Appendix 2.

Mixed-effects modeling was used for data analysis. Trials with incorrect responses of semantic judgment or color judgment were excluded in RT analysis of color judgment (7.7%). The base model included condition as the fixed effect and intercepts for participants and items as random effects. Model comparisons determined whether the final model included the random participant slope or random item slope for the condition. The final model was the base model, which included the condition as a fixed factor and intercepts for participants and items as random effects. A model that included the condition factor has a better fit than a model without the condition factor, χ² = 8.92, p < .05. Planned pair-wise comparisons showed no differences between the Equivalent condition and the O-S- condition, Estimate = 5.16, SE = 9.80, t = 0.53, p = .60. The O+S- condition and the O-S- condition did not show differences in color judgment, Estimate = 0.88, SE = 9.72, t = 0.09, p = .93. No evidence was found that L1 sub-character radicals were activated by L2 English word reading. Color judgment reaction times of the O+S+ condition were longer than the O+S- condition, Estimate = 24.48, SE = 9.73, t = 2.52, p < .05, suggesting a meaning effect.

Generalized logistic mixed models were used to analyse semantic judgment accuracy and color judgment accuracy. The final models for both semantic judgment accuracy and color judgment accuracy included condition as a fixed factor and intercepts for participants and items as random effects. There was no main effect of condition on either English word semantic judgment accuracy (χ² = 2.6, p = .45) or color judgment accuracy (χ² = 3.6, p = .30).

3 Discussion

In contrast to Experiment 1A, there was no meaning equivalent effect in Experiment 1B, which required meaning judgment of the English word. This difference may reflect the attention demand differences of the two experiments. The greater demands of meaning processing in Experiment 1B may attract attention to meaning and away from the color judgment process, cancelling the facilitative effect of form priming provided by the activated character.

Consistent with Experiment 1A, Experiment 1B found no evidence for the activation of the sub-character radical during English word reading. Thus, whether the L2 task was silent word reading (1A) or specifically directed at word meaning (1B), activation of the sub-character radical level was not found to be an automatic part of reading L2 English. The results of these two experiments contrast with the result of Chen et al. (2019a), where activation of the character equivalent occurred on both the character level and sub-character (radical) level during Pinyin reading. This contrast suggests that equivalent activations that occur across different scripts involve more sub-lexical components when they encode the same language (Pinyin and character) than across different languages and writing systems (English and character).

In Experiment 2, we address a methodological factor that might be relevant for both the lack of a sub-character radical effect and for the failure to replicate the character level meaning equivalent effect of Experiment 1B. It is possible that both meaning equivalent and sub-character radical effects occur in L2 English reading, but not robustly enough to be detected consistently in a task that only implicitly involves lexical access, even though the color judgment task is able to detect both character level and sub-character level activation in Pinyin reading. To more explicitly require lexical processing, participants in Experiment 2 were required to name the target character instead of making a color judgment. Naming requires attention to the character form, at minimum the whole character, and can also involve the radical components and their functional representations (Zhou and Marslen-Wilson, 1999; Zhou et al., 2013). If the sub-character level radical of the L1 Chinese equivalent is activated in L2 English word reading, we expect the naming task would be sensitive to detect the activation.

IV Experiment 2

1 Method

a Participants

Thirty-four Chinese–English bilinguals (18 females), age 18–22 years (M = 19.41, SD = 1.64), from the University of Pittsburgh participated for course credit. Participants were all from mainland China and proficient in English as a second language. All participants had intensive in-class English instruction in China for at least 10 years prior to coming the USA and had lived in the USA for 20.6 months on average (SD = 12.7) at the time of the study. The TOEFL scores ranged from 80 to 109.

b Stimuli, design and procedure

The stimuli and design were the same as Experiment 1B. The only difference was that participants in Experiment 2 were requested to name the target character instead of making a color judgment after the semantic judgment of the English word. Participants had 2,000 ms to name the target character. A microphone recorded participants’ oral response. If the Chinese character equivalent is activated in the naming task, we expect the naming latency of the target character in the Equivalent condition to be shorter than the O-S- condition. If there is an independent meaning effect, the activated meaning in the O+S+ condition will facilitate naming of the target character, causing the naming latency in the O+S+ condition to be shorter than the O+S- condition. In contrast, if the sub-lexical level orthography of the L1 Chinese character equivalent is activated in L2 English word processing, we should see a longer character naming latency in the O+S- condition than the O-S- condition because the competition from similar orthography produces inhibition during character naming (Perfetti et al., 2005).

2 Results

The main result of interest is naming times, which are shown in Table 4. The key results were as follows:

The naming times of target characters in the Equivalent condition were significantly shorter than in the O-S- condition.

Naming times of the O+S- condition were not different from the O-S- condition.

No significant naming differences were found between the O+S+ condition and the O+S- condition.

The following paragraphs report the statistical tests that support these conclusions. (The naming accuracy of target characters and semantic judgment accuracy of English words are shown in Appendix 2.)

Only trials that produced correct responses for both semantic judgment and character naming were included in naming times analysis, thus excluding 6.8% of trials. In the final model, condition was a fixed factor, and participant and item intercepts were random effects. To determine whether there was an effect of condition, we compared a model with condition as a fixed effect against a reduced model without condition. The two models showed a significant difference, χ² = 52.91, p < .001, indicating an effect of condition. Planned pair-wise comparisons showed that the naming times of the Equivalent condition were significantly shorter than the O-S- condition, Estimate = −38.49 ms, SE = 7.09, t = −5.43, p < .001. O+S- was not different from the O-S- condition, Estimate = 5.70, SE = 7.08, t = 0.81, p = .42. O+S+ was also not different from O+S-, Estimate = −0.65, SE = 7.09, t = 0.09, p = .93.

Generalized logistic mixed models were used to analyse semantic judgment accuracy and naming accuracy. There was no main effect of condition on either semantic judgment accuracy (χ² = 1.69, p = .64) or character naming accuracy (χ² = 3.32, p = .34). Final models for both semantic judgment accuracy and character naming accuracy had condition as a fixed factor, and participant and item intercepts as random effects.

3 Discussion

Experiment 2 replicated the results of Experiment 1A, showing character level activation of the meaning equivalent with no evidence for the sub-lexical radical level activation during L2 English word reading. Because the character naming task of Experiment 2 requires explicit attention to character orthography, it should be optimal for exposing both character level and sub-character level effects; it produced evidence for the first, but not the second. Thus over the three experiments, we find evidence for character level effects of meaning equivalence in two experiments, but no evidence for sub-lexical level radical effects in any. Unlike Experiments 1A and 1B, which used color naming, the O+S+ condition did not differ from the O+S- condition in the naming times of Experiment 2. This difference may reflect the naming task’s greater involvement of phonological processing relative to semantic processing. The failure to find a meaning effect in a naming task was also reported by Kim and Davis (2003) in their study of Korean–English bilinguals, although the effect was found in lexical decision.

V General discussion

Three experiments employed an implicit reading task (color judgment) and an explicit reading task (naming) to explore the activation of L1 Chinese meaning equivalent character and their sub-character orthography in L2 English word reading. Across the three experiments, there were two main findings. The equivalence effect – i.e. the effect on color judgment or naming when the target character was the meaning equivalent character of the English word – indicated that the L1 Chinese meaning equivalent character was activated when reading the corresponding L2 English word. This effect appears to reflect automatic L1 activation during L2 word reading across two unrelated languages and writing systems. The experiments further suggest that this effect was restricted to the character level and no evidence was found to support the activation at the sub-lexical orthographic level.

1 Activation of L1 Chinese meaning equivalent in L2 English reading

Consistent with previous studies on Chinese–English bilinguals (Thierry and Wu, 2007; Zhang et al., 2011), we found cross-language and cross-writing system activation (here, at the character level) during L2 reading. The co-activation of Chinese and English indicates that shared language forms and writing systems are not necessary for L1 activation in L2 reading, suggesting a universal tendency for the stronger initial language to function implicitly in L2 reading.

The possibly universal nature of L1 involvement in L2 reading reflects the important role of L1 in L2 learning. L2 learning involves a prolonged period of language interaction, as learners use L1 knowledge to acquire L2 language forms and usages. The legacy of these prolonged language interactions may be their relative permanence, leading to the kind of L1 activation during L2 language processing that we see in the present experiments and in other research (Degani et al., 2018; Thierry and Wu, 2007; Wu and Thierry, 2010, 2012; Zhang et al., 2011). Thus, even for highly proficient L2 learners, L1 may continue to be involved in many circumstances. Studies of other languages also have found L1 activation in L2 English word reading for highly proficient L2 learners. L1 Hindi was found to be activated in L2 English reading for highly proficient Hindi–English bilinguals (Mishra and Singh, 2014; Sunderman and Priya, 2012). In a semantic categorization task, translation priming effects were found for Japanese–English bilinguals (Finkbeiner et al., 2004). In a phoneme monitoring task, when participants were required to decide whether the name of a picture in English contains a target phoneme, Korean–English learners accessed L1 Korean (Moon and Jiang, 2012).

While L1 activation is widely found in L2 word reading, the Chinese–English comparison here allows some additional perspectives, because Chinese and English are unrelated linguistically and have different writing systems. Thus, the general result that L1 Chinese character equivalents were activated during English word reading is especially interesting in suggesting a fundamental experience-based source of the activation. It can occur across contrasting orthographies, because it does not depend on shared forms, but rather on the dominance of L1 experience and, perhaps, age of acquisition. However, this activation may not extend to the components of orthographic structure, i.e. the radicals that make up a character. When L1 and L2 share spoken and written forms, which is the case in related languages written alphabetically, sub-lexical L1 orthography is obligatorily accessed in L2 word reading. When an L2 word is encountered, orthographic candidates in both languages are active, and the neighborhood density of both languages affect L2 word recognition (Dijkstra and Van Heuven, 2002; Lemhöfer and Dijkstra, 2004; Van Heuven et al., 1998). For instance, when Dutch speakers read the English word ‘work’, orthographically or phonologically similar neighbors in both English (e.g. ‘cork’, ‘word’) and Dutch (e.g. ‘vork’, ‘werk’) are both active and the number of neighbors affects reading times (Van Heuven et al., 1998). This suggests that shared orthographic units trigger the cross-language activation that occurs when the two languages are alphabetic and can share spelling patterns that correspond to morphemes. In contrast, Chinese and English share no morphemic spelling patterns. There is nothing in the spelling of the English word ‘sunny’ to trigger the ‘spelling’ of the radical ‘日’, which is a component of the character equivalent of ‘sunny’.

To illustrate the cross-language interactions that we think characterize L1 Chinese reading of L2 English, Figure 1 shows an across-languages lexical network with connections between and across the two languages. In reading the English word ‘sunny’, activation spreads from the English spelling to its meaning. (Phonology is represented but was not examined in our experiments.) Because the English meaning of ‘sunny’ overlaps with the meaning of the Chinese morpheme/syllable (<sunny>, /qing/), the activation spreads from meaning to spoken and written Chinese forms, specifically the meaning equivalent at the character level. For skilled Chinese readers who have developed high quality representations of L1 words, interaction among three lexical constituents occurs automatically and relatively synchronously, creating a single word entity of three constituents (Perfetti, 2007). The activation of the three lexical constituents of the L1 translation equivalent may be routine and perhaps automatic for the Chinese bilingual reader of English. In Experiment 1A, the combination of a facilitative effect of the meaning equivalent character and the inhibitory effect of semantic overlap in a related character suggests the importance of the specific character – its unique combination of orthographic structure and meaning – in the cross-language effects. This interpretation is in line with the results of Guo et al. (2012), who found that semantic effects and character-equivalent effects in L2 word reading were independent and occurred in different time windows.

Figure 1.

Illustration of English and Chinese interaction when highly proficient Chinese–English bilinguals read English.

At the sub-lexical form level, there is no overlap between English and Chinese. Thus, whereas the Chinese character is activated through its meaning relation with the English word, its radicals are activated only in interaction with the character. This interaction can strengthen the activation of a phonological unit (the syllable) when the phonetic radical and the character have the same pronunciation and create competition when their pronunciations differ. This interaction is strong within-language (Chinese), because the Pinyin and the equivalent character activate exactly the same pronunciation and the same meaning, allowing the spread of activation to other characters sharing the radical forms and associated meaning and pronunciation. Across unrelated languages and written forms, our results suggest that this spread of activation is limited, allowing meaning-based co-activation that is restricted to a specific character.

Because cross-writing system activation occurs both within a language (Chinese) and across languages (Chinese-to-English), such activation, at least at the character level, is not fundamentally restricted by writing system differences. However, the nature of the co-activation may depend on the kind of convergence between forms and meaning that is possible within a language but not across languages. Thus, alphabetic writing can activate character orthography within Chinese more fully then reading an alphabetic L2 because the alphabetic orthography and the character orthography map onto exactly the same pronunciation and meaning. This allows a fuller activation of the character form that includes its orthographic components to spread activation to other characters. This degree of convergence is not possible across languages, although it can be approximated in the case of cognates within the same writing system.

These observations may be relevant for models of bilingual representations that distinguish between L1–L2 connections in terms of overlap at the form (phonology and orthography) level. Such connections are possible for related languages with related writing systems but not for unrelated languages with unrelated writing systems. In particular, the model of Degani et al. (2018) explicitly recognizes a two-layer (lexical and sub-lexical) representation of connections between L1–L2 representations. Our results, however, reflected in Figure 1, suggest that the lexical level is functional for unrelated languages with unrelated writing systems, but that the sub-lexical orthographic level may not be, at least in the case of Chinese–English. Because of the contrasting results in Pinyin, the factor controlling sub-lexical functionality is not due to differences in the writing system or the script, but differences in the language.

2 Implications for L2 word reading

English and Pinyin reading show differences in the activation of their character equivalents, especially at the sub-orthographic radical level. Pinyin produces wide spread activation of its corresponding character and other characters that share its sub-character components. English produces activation restricted to the character equivalent. These differences indicate that co-activation of L1 and L2 during word reading is constrained by language relatedness. Cross-writing system co-activation can be spread widely on the basis of sub-lexical units within a language, but not between languages that are unrelated.

To consider the implications of this account for L2 reading, it is useful to consider the case in L1 Chinese. Alphabetic Pinyin plays a prominent role in the initial phase of Chinese literacy and builds a foundation of orthography to phonology mapping, thus facilitating the learning of Chinese characters. With increased vocabulary, Pinyin becomes less reliable in accessing the word meaning, because of the large number of homophones in Chinese. The morpho-syllabic character system evades the homophony problem and appears to be, if not an optimal system (Frost, 2012), at least a highly adaptive system for Chinese (Perfetti and Verhoeven, 2017). With experience, the character becomes the gateway to the Chinese lexicon. As characters become the main orthographic form experienced by readers, they become activated even when Pinyin is read. Consistent with this suggestion, character activation during Pinyin reading is correlated with Chinese reading experience for Chinese learners (Chen et al., 2019b).

More generally, language and writing system differences not only constrain L1 orthography activation in L2, but also affect learning a second language. When L1 and L2 are from the same writing system, learners can easily transfer the mapping principle from L1 to L2. For example, Korean and English have different scripts but share the basics of an alphabetic writing system. Studies of Korean learning find that phoneme awareness is a strong predictor in Hangul word acquisition as it is in English (Cho and McBride-Chang, 2005). Comparative studies find that Korean–English bilinguals, whose native language of Korean is written in an alphabet, employ sub-lexical phonological coding procedures in English reading more than Chinese–English bilinguals did (Ben-Yehudah et al., 2019). Chinese–English bilinguals, whose native language is a morpho-syllabic writing system, appear to rely more on orthographic information and less on phonological information in English word processing than do Korean–English bilinguals (Wang et al., 2003). Some neuroimaging evidence suggests that Chinese–English bilinguals, when they read English, recruit the bilateral fusiform brain regions that are involved in Chinese reading, rather than mainly the left hemisphere fusiform that is typical in alphabetic reading. This suggests that Chinese–English bilinguals may assimilate English writing into their existing Chinese reading procedures (Nelson et al., 2009; Tan et al., 2003).

Finally, we want to emphasize that cross-language activation may not be the whole story in L2 word reading. Costa et al. (2017, 2019) argued that language activation is restricted to one language once L2 learners achieve a sufficiently proficient level. On this view, the L1 translation equivalent effect does not come from co-activation of L1 translation equivalents, but from the relationships of words in L2 lexical organization. The development of L2 word organization ‘copies’ the structure of words in the L1 lexicon. L2 translation equivalents are more likely to be associated when words are related (meaning or form) in the L1 lexicon. In the introduction we described the ERP study of Thierry and Wu (2007), which found that when Chinese–English bilinguals made meaning judgments on a pair of English words that were different in meaning, the N400 was reduced when the Chinese equivalents shared a character without meaning overlap, e.g. train–ham (火车–火腿). In the framework of Costa et al. (2017, 2019), this effect is not due to co-activation of Chinese during English word reading, but due instead to a reduction in the semantic distance between train and ham that results in learning the translation equivalent. i.e. train is ‘火车’ and ham is ‘火腿’. Similar evidence that L1 shapes L2 word meaning representations comes from a study of Hebrew–English bilinguals, who rated English word pairs for their meaning similarity. When the word pairs (e.g. tool and dish) can be translated by the same Hebrew word (‘kli’ for tool and dish), they were rated as more similar in meaning than word pairs that do not share a translation in Hebrew (Degani et al., 2011). Such examples show the value of conceptualizing L2 learning, as well as bilingual processing, as involving dynamic cross-language interactions.

VI Conclusions

The present Chinese–English experiments show evidence for co-activation of Chinese L1 orthography during English L2 reading. Because this co-activation was limited to the character (lexical) level rather than extending to the sub-lexical level, in contrast to results across writing systems within Chinese, we conclude that lexical co-activation is not restricted by different writing systems and scripts, but that sub-lexical activation may be restricted by differences between languages.

Footnotes

Appendix

Appendix 2.

Accuracies of semantic judgment and color judgment.

Conditions	Experiment 1A	Experiment 1B		Experiment 2
Conditions	Color judgment accuracy (SD)	Semantic judgment accuracy (SD)	Color judgment accuracy (SD)	Semantic judgment accuracy (SD)	Naming accuracy (SD)
Equivalent	0.96 (0.19)	0.95 (0.22)	0.95 (0.22)	0.96 (0.19)	0.99 (0.09)
O+S+	0.97 (0.18)	0.96 (0.19)	0.96 (0.19)	0.96 (0.20)	0.99 (0.08)
O+S-	0.97 (0.17)	0.95 (0.22)	0.97 (0.17)	0.97 (0.18)	0.99 (0.12)
O-S-	0.98 (0.14)	0.96 (0.19)	0.96 (0.19)	0.95 (0.21)	0.99 (0.08)

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the Learning Research & Development Center and by the National Science Foundation (Grant #SBE-0836012).

ORCID iDs

Lin Chen

Li-Yun Chang

Notes

References

Baayen

Davidson

Bates

(2008) Mixed-effects modeling with crossed random effects for subjects and items. Journal of Memory and Language 59: 390–412.

Ben-Yehudah

Hirshorn

Simcox

, et al. (2019) Chinese–English bilinguals transfer L1 lexical reading procedures and holistic orthographic coding to L2 English. Journal of Neurolinguistics 50: 136–48.

Brysbaert

New

(2009) Moving beyond Kučera and Francis: A critical evaluation of current word frequency norms and the introduction of a new and improved word frequency measure for American English. Behavior Research Methods 41: 977–90.

Cai

Brysbaert

(2010) SUBTLEX-CH: Chinese word and character frequencies based on film subtitles. PloS one 5: e10729.

Cao

Lee

Shu

, et al. (2009) Cultural constraints on brain development: Evidence from a developmental study of visual word processing in Mandarin Chinese. Cerebral Cortex 20: 1223–33.

Chang

Chen

Perfetti

(2018) GraphCom: A multidimensional measure of graphic complexity applied to 131 written languages. Behavior Research Methods 50: 427–49.

Chen

Zhou

Gao

, et al. (2014a) Cross-language translation priming asymmetry with Chinese–English bilinguals: A test of the sense model. Journal of Psycholinguistic Research 43: 225–40.

Chen

Zhong

Leng

, et al. (2014b) The role of the character graphic information in different Pinyin processing tasks. Acta Psychologica Sinica 46: 1661–70.

Chen

Perfetti

Fang

, et al. (2019a) Reading Pinyin activates sublexcial character orthography for skilled Chinese readers. Language, Cognition and Neuroscience 34: 736–46.

10.

Chen

Perfetti

Leng

(2019b) Reading Pinyin activates character orthography for highly experienced learners of Chinese. Bilingualism: Language and Cognition 22: 103–11.

11.

Cho

McBride-Chang

(2005) Correlates of Korean Hangul acquisition among kindergartners and second graders. Scientific Studies of Reading 9: 3–16.

12.

Costa

Pannunzi

Deco

, et al. (2017) Do bilinguals automatically activate their native language when they are not using it? Cognitive Science 41: 1629–44.

13.

Costa

Pannunzi

Deco

, et al. (2019) Does bilingualism alter lexical structure?: Response to Oppenheim, Wu, and Thierry (2018) Cognitive Science 43: e12707.

14.

Degani

Prior

Hajajra

(2018) Cross-language semantic influences in different script bilinguals. Bilingualism: Language and Cognition 21: 782–804.

15.

Degani

Prior

Tokowicz

(2011) Bidirectional transfer: The effect of sharing a translation. Journal of Cognitive Psychology 23: 18–28.

16.

Dijkstra

Van Heuven

(2002) The architecture of the bilingual word recognition system: From identification to decision. Bilingualism: Language and Cognition 5: 175–97.

17.

Dijkstra

Wahl

Buytenhuijs

, et al. (2019) Multilink: A computational model for bilingual word recognition and word translation. Bilingualism: Language and Cognition 22: 657–79.

18.

Ding

Peng

Taft

(2004) The nature of the mental representation of radicals in Chinese: A priming study. Journal of Experimental Psychology: Learning, Memory, and Cognition 30: 530–39.

19.

Feldman

Siok

(1997) The role of component function in visual recognition of Chinese characters. Journal of Experimental Psychology: Learning, Memory, and Cognition 23: 776–81.

20.

Feldman

Siok

(1999) Semantic radicals contribute to the visual identification of Chinese characters. Journal of Memory and Language 40: 559–76.

21.

Finkbeiner

Forster

Nicol

, et al. (2004) The role of polysemy in masked semantic and translation priming. Journal of Memory and Language 51: 1–22.

22.

Frost

(2012) Towards a universal model of reading. Behavioral and Brain Sciences 35: 263–79.

23.

Goral

(2019) The bilingual mental lexicon beyond Dutch–English written words. Bilingualism: Language and Cognition 22: 680–81.

24.

Guo

Misra

Tam

, et al. (2012) On the time course of accessing meaning in a second language: An electrophysiological and behavioral investigation of translation recognition. Journal of Experimental Psychology: Learning, Memory, and Cognition 38: 1165–86.

25.

Jiang

(2019) Phonology-based bilingual activation among different-script bilinguals? Bilingualism: Language and Cognition 22: 693–94.

26.

Kim

Davis

(2003) Task effects in masked cross-script translation and phonological priming. Journal of Memory and Language 49: 484–99.

27.

Klein

(1964) Semantic power measured through the interference of words with color-naming. The American Journal of Psychology 77: 576–88.

28.

Kroll

Stewart

(1994) Category interference in translation and picture naming: Evidence for asymmetric connections between bilingual memory representations. Journal of Memory and Language 33: 149–74.

29.

Kroll

Van Hell

Tokowicz

(2010) The Revised Hierarchical Model: A critical review and assessment. Bilingualism: Language and Cognition 13: 373–81.

30.

Lee

Tsai

Chiu

, et al. (2006a) The early extraction of sublexical phonology in reading Chinese pseudocharacters: An event-related potential study. Language and Linguistics 7: 619–36.

31.

Lee

Tsai

Huang

, et al. (2006b) The temporal signatures of semantic and phonological activations for Chinese sublexical processing: An event-related potential study. Brain Research 1121: 150–59.

32.

Lemhöfer

Dijkstra

(2004) Recognizing cognates and interlingual homographs: Effects of code similarity in language-specific and generalized lexical decision. Memory and Cognition 32: 533–50.

33.

(2018) Chinese learners of English see Chinese words when reading English words. Journal of Psycholinguistic Research 47: 505–21.

34.

MacLeod

(1991) Half a century of research on the Stroop effect: An integrative review. Psychological Bulletin 109: 163–203.

35.

McBride-Chang

(2005) Predictors of beginning reading in Chinese and English: A 2-year longitudinal study of Chinese kindergartners. Scientific Studies of Reading 9: 117–44.

36.

Mishra

(2019) A few suggestions on broadening the cross-linguistic relevance of the Multilink model. Bilingualism: Language and Cognition 22: 691–92.

37.

Mishra

Singh

(2014) Language non-selective activation of orthography during spoken word processing in Hindi–English sequential bilinguals: An eye tracking visual world study. Reading and Writing 27: 129–51.

38.

Miwa

Dijkstra

Bolger

, et al. (2014) Reading English with Japanese in mind: Effects of frequency, phonology, and meaning in different-script bilinguals. Bilingualism: Language and Cognition 17: 445–63.

39.

Moon

Jiang

(2012) Non-selective lexical access in different-script bilinguals. Bilingualism: Language and Cognition 15: 173–80.

40.

Nakayama

Ida

Lupker

(2016) Cross-script L2–L1 noncognate translation priming in lexical decision depends on L2 proficiency: Evidence from Japanese–English bilinguals. Bilingualism: Language and Cognition 19: 1001–22.

41.

Nakayama

Sears

Hino

, et al. (2012) Cross-script phonological priming for Japanese–English bilinguals: Evidence for integrated phonological representations. Language and Cognitive Processes 27: 1563–83.

42.

Nelson

Liu

Fiez

, et al. (2009) Assimilation and accommodation patters in ventral occipitotemporal cortex in learning a second writing system. Human Brain Mapping 30: 810–20.

43.

Peleg

Degani

Raziq

, et al. (2019) Cross-lingual phonological effects in different-script bilingual visual-word recognition. Second Language Research. Epub ahead of print 19 February 2019. DOI: 10.1177/0267658319827052.

44.

Perfetti

(2007) Reading ability: Lexical quality to comprehension. Scientific Studies of Reading 11: 357–83.

45.

Perfetti

Tan

(1998) The time course of graphic, phonological, and semantic activation in Chinese character identification. Journal of Experimental Psychology: Learning, Memory, and Cognition 24: 101–18.

46.

Perfetti

Verhoeven

(2017) Epilogue: Universals and particulars in learning to read across seventeen orthographies. In: Verhoeven

Perfetti

(eds) Learning to read across languages and writing systems. Cambridge: Cambridge University Press, pp. 455–80.

47.

Perfetti

Bell

Delaney

(1988) Automatic (prelexical) phonetic activation in silent word reading: Evidence from backward masking. Journal of Memory and Language 27: 59–70.

48.

Perfetti

Cao

Booth

(2013) Specialization and universals in the development of reading skill: How Chinese research informs a universal science of reading. Scientific Studies of Reading 17: 5–21.

49.

Perfetti

Liu

Tan

(2005) The lexical constituency model: Some implications of research on Chinese for general theories of reading. Psychological Review 112: 43–59.

50.

Smith

Walters

Prior

(2019) Target accessibility contributes to asymmetric priming in translation and cross-language semantic priming in unbalanced bilinguals. Bilingualism: Language and Cognition 22: 157–76.

51.

Weekes

(2007) Effects of frequency and semantic radical combinability on reading in Chinese: An ERP study. Brain and Language 103: 111–12.

52.

Sunderman

Priya

(2012) Translation recognition in highly proficient Hindi–English bilinguals: The influence of different scripts but connectable phonologies. Language and Cognitive Processes 27: 1265–85.

53.

Taft

Zhu

Peng

(1999) Positional specificity of radicals in Chinese character recognition. Journal of memory and Language 40: 498–519.

54.

Tan

Liu

Perfetti

, et al. (2001) The neural system underlying Chinese logograph reading. Neuroimage 13: 836–46.

55.

Tan

Spinks

Feng

, et al. (2003) Neural systems of second language reading are shaped by native language. Human Brain Mapping 18: 158–66.

56.

Thierry

(2007) Brain potentials reveal unconscious translation during foreign-language comprehension. Proceedings of the National Academy of Sciences 104: 12530–35.

57.

Tsang

Chen

(2009) Do position-general radicals have a role to play in processing Chinese characters? Language and Cognitive Processes 24: 947–66.

58.

Van Heuven

Dijkstra

Grainger

(1998) Orthographic neighborhood effects in bilingual word recognition. Journal of Memory and Language 39: 458–83.

59.

Van Orden

(1987) A ROWS is a ROSE: Spelling, sound, and reading. Memory and Cognition 15: 181–98.

60.

Wang

Koda

Perfetti

(2003) Alphabetic and nonalphabetic L1 effects in English word identification: A comparison of Korean and Chinese English L2 learners. Cognition 87: 129–49.

61.

Warren

(1974) Association, directionality, and stimulus encoding. Journal of Experimental Psychology 102: 151–58.

62.

Wen

Van Heuven

(2017) Non-cognate translation priming in masked priming lexical decision experiments: A meta-analysis. Psychonomic Bulletin and Review 24: 879–86.

63.

Wen

Van Heuven

(2018) Limitations of translation activation in masked priming: Behavioural evidence from Chinese–English bilinguals and computational modeling. Journal of Memory and Language 101: 84–96.

64.

Thierry

(2010) Chinese–English bilinguals reading English hear Chinese. Journal of Neuroscience 30: 7646–51.

65.

Thierry

(2012) Unconscious translation during incidental foreign language processing. NeuroImage 59: 3468–73.

66.

Zhang

Van Heuven

Conklin

(2011) Fast automatic translation and morphological decomposition in Chinese–English bilinguals. Psychological Science 22: 1237–42.

67.

Zhou

Peng

Zheng

, et al. (2013) Sub-lexical phonological and semantic processing of semantic radicals: a primed naming study. Reading and Writing 26: 967–89.

68.

Zhou

Marslen-Wilson

(1999) The nature of sublexical processing in reading Chinese characters. Journal of Experimental Psychology: Learning, Memory, and Cognition 25: 819–37.