Phonological and morphological literacy skills in English and Chinese: A cross‐linguistic neuroimaging comparison of Chinese–English bilingual and monolingual English children

Abstract

Over the course of literacy development, children learn to recognize word sounds and meanings in print. Yet, they do so differently across alphabetic and character‐based orthographies such as English and Chinese. To uncover cross‐linguistic influences on children's literacy, we asked young Chinese–English simultaneous bilinguals and English monolinguals (N = 119, ages 5–10) to complete phonological and morphological awareness (MA) literacy tasks. Children completed the tasks in the auditory modality in each of their languages during functional near‐infrared spectroscopy neuroimaging. Cross‐linguistically, comparisons between bilinguals' two languages revealed that the task that was more central to reading in a given orthography, such as phonological awareness (PA) in English and MA in Chinese, elicited less activation in the left inferior frontal and parietal regions. Group comparisons between bilinguals and monolinguals in English, their shared language of academic instruction, revealed that the left inferior frontal was less active during phonology but more active during morphology in bilinguals relative to monolinguals. MA skills are generally considered to have greater language specificity than PA skills. Bilingual literacy training in a skill that is maximally similar across languages, such as PA, may therefore yield greater automaticity for this skill, as reflected in the lower activation in bilinguals relative to monolinguals. This interpretation is supported by negative correlations between proficiency and brain activation. Together, these findings suggest that both the structural characteristics and literacy experiences with a given language can exert specific influences on bilingual and monolingual children's emerging brain networks for learning to read.

Cross‐linguistic comparison revealed that tasks that best characterize literacy success in each language, a phonological task in English and a morphological task in Chinese, yielded less activations and stronger brain‐behavior correlations. Bilinguals' English‐dominant literacy and Chinese heritage language experiences were associated with stronger left hemisphere engagement during literacy tasks in English and stronger right hemisphere engagement during literacy tasks in Chinese.

1. INTRODUCTION

Across languages, spoken words are made of units of sound and meaning. Yet, there are important cross‐linguistic differences in how children learn to map units of spoken words onto print. In English and other alphabetic languages, children learn to map units of sounds or phonemes onto letters. In Chinese, children learn to map lexical morphemes, or the smallest units of meaning, onto characters. In other words, whereas children's phonological and morphological abilities are foundational for literacy success across languages, there are important cross‐linguistic differences in how these abilities support the development of the “reading brain” or neural systems that make literacy possible (Cao et al., 2014, 2015). One way to uncover these literacy universals and their cross‐linguistic differences is through the study of bilingual learners with early and simultaneous exposure to their two languages. The present study examines young Chinese–English bilinguals and English monolinguals to shed new light on the influence of cross‐linguistic effects on learning to read.

2. CROSS‐LINGUISTIC DIFFERENCES IN THEORETICAL MODELS OF WORD PROCESSING

Theoretical perspectives on word reading, notably the Lexical Quality Hypothesis (LQH), posits that word reading proficiency builds upon sound‐to‐print and meaning‐to‐print associations (Perfetti & Hart, 2002; Verhoeven & Perfetti, 2021). Importantly, the nature of these associations is thought to vary across languages starting early in reading development (McBride et al., 2021). For instance, cross‐linguistic comparisons of second‐grade monolinguals suggest that young English speakers build stronger associations between word reading and phonological awareness (PA), whereas Chinese speakers show stronger associations between word reading and morphological awareness (MA; McBride‐Chang et al., 2005). Phonological and MA are the abilities to manipulate the smallest units of sound (phonemes) or smallest units of meaning (morphemes), respectively (Goodwin et al., 2020). Neurobiological perspectives on the reading brain are generally consistent with the LQH suggesting that neural systems for learning to read include the dorsal/phonological and ventral/semantic paths (Davis & Rastle, 2017; Dickens et al., 2019; Hickok & Poeppel, 2007; Liebenthal et al., 2013; Pugh et al., 2001). To understand cross‐linguistic literacy effects, we aim to complement the behavioral data with neuroimaging data by examining the brain basis that supports children's emerging phonological and morphological literacy skills.

2.1. Phonological awareness

Phonological awareness helps children discover the sounds of words in their printed forms. Sound‐to‐letter correspondences vary in consistency across alphabetic languages. In English, these correspondences are inconsistent relative to other alphabetic orthographies, at about 70% (Share, 2008). However, compared to Chinese, English sound‐to‐letter correspondences are considered predictable (Geva & Wang, 2001). In Chinese, most characters (80%–90%) include a semantic radical and a phonetic radical, cuing their meaning and pronunciation of the character. However, linguistic analyses of Chinese characters suggest that the predictive accuracy of the pronunciation from the phonetic radicals is only 23%–26% (Shu et al., 2003). In other words, Chinese orthography has poor phonological predictability (Geva & Wang, 2001). These orthographic differences in phonological predictability appear to play a role in learning to read. Previous research has shown that while PA is essential for learning to read in both English and Chinese, PA typically explains more variance in literacy gains in English than in Chinese learners (McBride et al., 2021). These differences in orthographies and learning to read have led to targeted hypotheses on cross‐linguistic effects on the development of neural literacy systems.

Neurobiological theories of literacy development have largely been built on evidence from alphabetic languages (Ip et al., 2019; Yan et al., 2021). Much of this evidence comes from studies using phonological tasks that have linked phonology to literacy success and dyslexia (Maisog et al., 2008; Martin et al., 2016; Saygin et al., 2013; Zhang & Peng, 2022). At the core of these theoretical perspectives of the reading brain and resulting evidence is the functionality of the left superior temporal gyrus (STG) for recognizing word sounds in speech and print (Binder, 2017; Chyl et al., 2018). The left STG region is an element of the dorsal neural network that also includes the dorsal left inferior frontal gyrus (IFG; Branum‐Martin et al., 2015; Hickok & Poeppel, 2007; Matchin & Hickok, 2020). As children become older and more proficient readers, they begin to exhibit more robust brain activations as well as stronger functional connectivity along the dorsal network (Cao et al., 2011; Chyl et al., 2018; Martin et al., 2016; Norton et al., 2014; Preston et al., 2016).

Research studies on PA and reading in Chinese highlight the role of left frontal regions, including left IFG and middle frontal gyrus (MFG) regions that support phonological analyses and verbal working memory (Brennan et al., 2013; Cao et al., 2011, 2020; Feng et al., 2017; T. Liu & Hsiao, 2012; Siok et al., 2004, 2008; Yang & Tan, 2020). In particular, Siok et al. (2008) have found stronger left frontal activation in typically developing readers relative to those with dyslexia, whereas no significant group differences were found in temporal regions. Moreover, a direct comparison between English‐ and Chinese‐speaking children and adults (Brennan et al., 2013; Cao et al., 2011) revealed that whereas both groups showed a developmental increase in left IFG activation, only English speakers demonstrated a developmental increase in left STG activation. In sum, there is a general agreement on the critical role of the dorsal/phonological neurocircuitry in learning to read. Yet, research with alphabetic languages has emphasized the role of temporal regions, while research with Chinese has emphasized the role of frontal regions (Yan et al., 2021), possibly due to the more analytically demanding nature of sound‐to‐print correspondences in Chinese.

2.2. Morphological awareness

MA helps children recognize the smallest units of meaning (morphemes) in their printed form. In English and Chinese, lexical morphemes include free‐standing root morphemes (e.g., snow + man) and bound‐derived morphemes (un + like + ly). The importance of morphological awareness for learning to read has been well‐documented cross‐linguistically (Deacon et al., 2019; Levesque et al., 2021). In English and other alphabetic languages, MA helps children read phonologically opaque and polysyllabic words (e.g., care + fully; amuse + ment). Lexical root compounding, a process that produces polymorphic words composed of two or more free morphemes, as in star + fish or snow + man, is far more prevalent in Chinese than in English (Shu et al., 2006). In speech, nearly all Chinese words are polysyllabic, with each syllable typically representing a root morpheme. Of note is that phonologically, Chinese syllables that are expressed with the same tone can also have multiple meanings (Taft & Chen, 1992). Orthographically, individual morphemes map onto unique characters, which helps disambiguate homophony akin to I versus eye. Such morphological reciprocity among spoken and written words yields stronger associations between MA and character reading in emerging readers of Chinese compared to alphabetic languages like English. These stronger associations give rise to the crucial role of MA in Chinese language and literacy development (Ho et al., 2003; Tong & McBride‐Chang, 2010).

Relatively fewer studies have examined the neurobiology of MA, especially in reading‐age children (Marks et al., 2021). The multicomponent nature of morphemic analysis typically incurs the engagement of both the dorsal/phonological as well as the ventral/semantic neural pathways (Arredondo et al., 2015; Ip et al., 2019; Yablonski et al., 2019). The latter includes the ventral aspect of the left IFG (BA 45/47) and the middle temporal gyrus (MTG; Hagoort, 2019; Hickok & Poeppel, 2007). Most of the available developmental studies have highlighted the role of the left IFG region during morphological processes. This may be due to its critical role in linguistic synthesis, as it would be necessary for bringing together the phonological, lexico‐semantic, and rule‐governed aspects of polymorphic words (Arredondo et al., 2015; Bitan et al., 2017; L. Liu et al., 2013; Marks et al., 2021). For example, a study by Marks et al. (2021) asked English monolingual children to complete an auditory MA task during which children heard a target word (e.g., winner), followed by two other words (winning, window), and decided which of the two words was a better semantic match to the target word. The task included conditions with free root and bound derivational morphemes (derived example: dancer, waiter, corner). Whereas both conditions revealed robust engagement of both the dorsal and ventral networks during each experimental task relative to the control condition, the derivational condition incurred stronger left IFG activation, likely because the processing of bound morphemes posed stronger analytical demands on the language system.

Cross‐linguistically, some studies of MA suggest that the higher prevalence of morphological compounding in Chinese words may result in stronger finetuning of the left temporal regions for the automatic processing of these structures in Chinese relative to English (Ip et al., 2017, 2019). For instance, adult Chinese speakers showed stronger left temporal (STG/MTG) activation during morphological than phonological tasks but a reverse pattern for left IFG (stronger during phonology than morphology). In other words, it is possible that the more predictable aspect of a language in its orthography (phonology in English, lexical morphology in Chinese) becomes more finetuned within the literacy network.

In sum, successful word reading builds upon children's emerging phonological and MA literacy skills. As alphabetic letters map onto sounds and Chinese characters map onto morphemes, cross‐linguistic differences in the extent to which young readers rely on these skills emerge early on (McBride et al., 2021). Some have suggested that the phonological systems may undergo a greater neurodevelopmental change in English than in Chinese (Brennan et al., 2013; Su et al., 2018). Others have suggested that the two neural circuits might undergo a more balanced change in Chinese than in English (Su et al., 2018). Yet, current inferences about the cross‐linguistic differences are limited by the paucity of studies, especially those that involve direct cross‐linguistic comparisons within young bilingual developing readers.

2.3. Bilingualism

Early in life, the developing brain is sensitive to variations in language experiences which can influence perisylvian pathways' finetuning for language processing (Skeide & Friederici, 2016; Werker, 1984). Bilingualism is one of the most common variations in early‐life experiences and offers a unique window for uncovering neural plasticity in brain organization for language function, as children often vary in their dual‐language experiences and use. For instance, work with bilingual infants has shown that their neural specialization for phonological processing, as measured through neural response to native and non‐native phonological contrasts, may lag behind that of monolingual infants so as to allow dual‐language learners to develop complex dual‐language representations (Garcia‐Sierra et al., 2011). At the same time, the researchers observed that bilingual infants' neural response was stronger for the dominant language of experience relative to their secondary, less utilized language. The authors interpreted this finding as suggesting stronger “neural commitment” on the part of the brain tissue to children's dominant language of use. This interpretation is based on earlier monolingual findings of stronger neural phoneme discrimination response in infants who later show better vocabulary scores as toddlers. In other words, phonological and concomitant morphophonological processes that underlie literacy success may differ between bilingual and monolingual learners, as well as between the two languages of the bilingual—both as a function of cross‐linguistic differences as well as differences in learning experiences (Brice et al., 2021).

A recent meta‐analysis of neuroimaging research with bilinguals has made several general observations about the effects of bilingualism on brain function (Cargnelutti et al., 2019). First, the second language or L2 generally elicits a stronger neural response than the L1, with differences increasing with the age of acquisition, especially in the frontal lobe. The frontal lobe is considered key in controlling dual language switching and use, in addition to its support in analytically complex operations that underly the processing of languages of lower proficiency. Notice that this finding is different from the one observed in infants, possibly because infants are experiencing a passive listening oddball task in which a deviant stimulus elicits a response only if the infant is proficient enough to notice it; older children and adults are often made to do an active language task during neuroimaging, which is more taxing for those with lower proficiency. Second, in early exposed bilinguals, their L1 elicited a stronger response in left temporo‐parietal regions classically associated with lexical retrieval and processing, possibly explained by the role of this region in facilitating young children in gaining foundations of language function (Skeide & Friederici, 2016). Yet, this review was limited to the bilingual populations and thus addressed questions about L1 relative to L2 rather than bilinguals relative to monolinguals. Moreover, the review was predominantly based on adult late second language learners as literature on bilingual children and adult early L2 learners are generally more limited.

The few studies that focused on early exposed bilinguals also generally find stronger activation in classic language regions as well as their right hemisphere homologs in bilinguals relative to monolinguals, even when experimental groups are matched for their target language proficiency and task performance (Enge et al., 2020; Jasińska & Petitto, 2014; Jasínska & Petitto, 2013; Rodriguez‐Fornells et al., 2002; Román et al., 2015). In a study of L2 processing, Jasińska and Petitto (2014) asked child participants who were either monolinguals or bilinguals with varied ages of L2 (English) acquisition to complete a sentence judgment task in English. Bilinguals of all ages of acquisition exhibited stronger bilateral engagement of the perisylvian regions relative to monolinguals. In the early‐exposed group, this increase was mostly in the bilateral temporo‐parietal regions with a more moderate increase in the left frontal, whereas the later‐exposed group also showed a stronger and more bilateral increase in activation in frontal regions. In a study of bilingual L1 processing, Román et al. (2015) similarly found that early‐exposed adult bilinguals exhibited stronger bilateral engagement of the left temporo‐parietal regions relative to monolinguals during a sentence judgment task as compared to monolinguals, even in their L1. Similar L1 effects were found in a word reading study with early‐exposed bilingual adults who were compared to monolinguals in their shared L1 (Rodriguez‐Fornells et al., 2002). Researchers have generally attributed these findings of greater bilingual activation to the idea of added cognitive load in selecting between language systems. It is under the assumption that the two languages of the bilingual are often co‐active, such that individuals might simultaneously consider the words and rules of their languages even if only using one of their two languages (Kroll & Tokowicz, 2005). Moreover, the early‐exposed bilingual relative to monolingual group differences often appear to include temporo‐parietal regions critical to supporting language and brain development in early life (Enge et al., 2020) and thus likely to reflect early neurodevelopmental plasticity in dual language acquisition.

Bilingual literacy theories, such as the Interactive Transfer Framework (Chung et al., 2019), posit that bilinguals' two languages interact and that this cross‐linguistic interaction is influenced by factors such as language proficiency, similarities of linguistic features that interact and the complexity of those interacting features. It is now well established that PA is a literacy skill that appears to be most similar across languages and thus easily shared within the bilingual dual‐language system (Luo et al., 2014; Sun, Zhang, Marks, Karas, et al., 2022). In contrast, morphological skills appear to be more language‐specific than PA skills due to the lexical and grammatical characteristics of individual languages (Chung et al., 2019). Behavioral evidence that supports these theoretical assumptions shows stronger associations (or correlations) between bilingual children's phonological than MA skills, especially in children learning structurally distinct languages such as English and Chinese (Sun, Zhang, Marks, Karas, et al., 2022). From the neurological perspective, skills that are more practiced and automated, tend to incur less activation, especially in the frontal lobe, than the skills that are more effort demanding. We can therefore predict that PA skills jointly reinforced through dual‐language literacy practices may result in lesser engagement or a lower degree of activation in the left frontal lobe in bilinguals than in monolinguals. In contrast, this pattern may be reversed for the MA skills for which the bilinguals must adjudicate cross‐linguistic differences and build language‐specific representations, thereby resulting in stronger left frontal activation in bilinguals relative to monolinguals (Sun, Marks, et al., 2022).

2.4. The present study

Phonological and morphological abilities are foundational for learning to read across languages but have been revealed to have differential associations with literacy gains and dyslexia across languages. Although PA has been studied extensively, it has rarely been considered in relation to MA (Bitan et al., 2017; Ip et al., 2019; Su et al., 2018), and no study has yet considered those two abilities at the same time in a cross‐linguistic context during the key periods of brain development for learning to read. The present study thus aimed to uncover both the cross‐linguistic and bilingual effects on the mechanisms that support children's phonological and MA abilities for learning to read.

The study thus examined young bilingual heritage Chinese–English speakers and young monolingual English speakers. Heritage speakers are those who speak a home/community language that is different from the official language of the country (Benmamoun et al., 2013). As is typical of many heritage language speakers in the United States, bilinguals in the present study were exposed to both of their languages early in life, attended English‐only schools, and received varied amounts of Chinese literacy at home and community heritage language schools. Monolingual participants were recruited from the same schools, neighborhoods, and school districts as the bilinguals. The study used child‐friendly and silent functional near‐infrared spectroscopy (fNIRS) imaging to study phonological and MA skills in the auditory modality that precedes and predicts reading development. The analytical approach included whole‐brain analyses that considered all channels as well as the exploratory region of interest (ROI) analyses that examined the relationship between children's language proficiency and patterns of neural activity along the left IFG, STG, and MTG regions classically associated with phonological and lexico‐semantic processing.

The study aimed to advance LQH (Perfetti & Hart, 2002; Verhoeven & Perfetti, 2021) to better explain the bilingual literacy phenomena. The Lexical Quality Framework, built on monolingual and cross‐linguistic evidence, posits that cross‐linguistic experiences play a role in how we process phonological, lexical, and orthographic representations of words and their intersections at the morpho‐phonological level (Tan et al., 2005). To address our cross‐linguistic inquiry, we compared the bilinguals' two languages. We predicted that word processing should differ between English and Chinese, potentially yielding a stronger engagement of the phonological network in English, relative to Chinese, during the phonology task and a stronger engagement of the semantic network in Chinese, relative to English, during the morphology task. To address our bilingual aim, we compared bilinguals and monolinguals in their shared language of English. Prior bilingualism research, such as Jasińska and Petitto (2014), suggests that bilinguals show greater engagement and more variability of neural activation in bilateral frontal and temporal–parietal regions than monolinguals. Therefore, we predicted that there would be stronger activation in bilateral frontal and temporal–parietal regions as well as more channels reaching significance in bilinguals than in monolinguals. Furthermore, guided by the Interactive Transfer Framework, we predict the transfer is more likely to occur in the PA task. Furthermore, guided by the Interactive Transfer Framework, we predict that cross‐linguistic transfer is more likely to occur in the PA task, thereby enhancing bilingual children's processing efficiency during PA tasks. As greater processing efficiency is commonly associated with reduced levels of brain activation, especially in the frontal lobe (Nęcka et al., 2021), we predicted that during the PA task, the bilinguals may show lower activation in periyslvian language regions relative to the monolinguals.

3. METHOD

The sample was drawn from a larger project. The initial sample included 109 Chinese–English bilingual and 97 English monolingual children. The inclusion criteria included (1) right‐handedness, (2) normal hearing/vision, (3) no known history of head injury or developmental delays, (4) not being treated with psychotropic medications at the time of testing, (5) demonstrating a standard score of 85 or above during the English vocabulary test (Peabody Picture Vocabulary Test—Fifth Edition [PPVT‐5]; Dunn, 2019) and above 65% in accuracy for neuroimaging tasks.

The final sample included 119 children were included in this study. Among these, 69 children were Chinese–English bilingual (52% male M _age = 7.91, SD = 1.63), and 50 were English‐speaking monolingual children (53% male, M _age = 7.73, SD = 1.06; Table 1). Both bilingual and monolingual children attended English‐only speaking schools and were recruited from the same school districts.

TABLE 1.

Participants' language proficiency and experimental task performance.

				Age‐matched
	Bilingual (N = 69)			Bilingual (N = 50)	Monolingual (N = 50)
Task performance, M (SD)	English	Chinese	Comparison (t)	English		Comparison (t)
Standard scores
Vocabulary	108.04 (18.68)	92.51 (16.98)	−5.01***	109.07 (14.62)	113.78 (11.40)	−1.77
Word reading English standard score	116.54 (13.98)		—	118.49 (12.46)	110.20 (12.02)	3.32**
Character identification raw score		39.49 (21.46)	—	—	—
Bilingual assessments (% correct)
MA (ELMM)	53.89 (29.15)	61.57 (24.51)	1.67	49.11 (28.75)	57.65 (24.39)	−1.59
PA (CTOPP)	69.65 (22.33)	64.37 (27.15)	−1.24	67.78 (23.17)	69.43 (17.22)	−0.41
fNIRS task accuracy (% correct)
MA task	87.59 (11.20)	75.91 (13.81)	−5.32***	86.35 (14.47)	83.50 (16.11)	0.93
PA task	84.21 (18.04)	74.38 (16.58)	−3.20**	84.01 (18.70)	85.50 (13.15)	−0.46
MA control	97.06 (5.45)	96.37 (5.13)	−0.74	96.17 (6.36)	94.25 (5.67)	0.85
PA control	96.60 (5.76)	96.98 (5.46)	0.38	97.07 (5.86)	95.63 (5.54)	1.26
fNIRS task reaction time (ms)
MA task	4788 (273)	4697 (247)	−1.97	4771.52 (266)	4784.53 (301)	−0.23
PA task	4403 (344)	4618 (341)	3.59***	4375.71 (356)	4425.62 (346)	−0.69
MA control	4428 (258)	4143 (357)	−4.74***	4381.39 (251)	4439.39 (256)	−1.13
PA control	4155 (306)	4172 (378)	−0.20	4134.72 (319)	4269.46 (312)	−1.91

Abbreviations: CTOPP, Comprehensive Test of Phonological Processing; ELMM, Early Lexical Morphology Measure; fNIRS, functional near‐infrared spectroscopy; MA, morphological awareness; PA, phonological awareness.

^** p < .01;

^*** p < .001.

Bilingual participants' inclusion criteria included exposure to Chinese from birth and to English before age three. Ninety percent of the bilingual participants had both parents as native Chinese speakers, and the rest had one native Chinese‐speaking parent. Seventy‐four percent of the participants (N = 51) attended Chinese heritage schools, where they received Chinese literacy instruction once a week. In addition, the bilinguals were also required to have a standard score of 80 or above during the Chinese vocabulary test (PPVT—Revised, PPVT‐R, Lu & Liu, 1998). The study was approved by the institutional review board at the University of Michigan, and participants were paid for their participation.

All bilingual children were included in the within‐group cross‐linguistic comparison of the Chinese and English languages. A subset of 50 bilinguals was directly compared with the 50 monolinguals; the 2 groups were matched on age and English vocabulary. Parental reports revealed that most of the bilinguals received early exposure to both of their languages and started producing words in English (M = 1.67; SD = 0.68) and Chinese (M = 1.31; SD = 0.52) at the same time. At the time of testing, the bilinguals were reported to be spending more time reading in English (M = 64.15 mins/day; SD = 25.34) than in Chinese (M = 48.40 mins/day; SD = 29.80), as would be consistent with their English‐dominant schooling. Finally, the resulting subsample of bilinguals who were compared to the monolinguals showed better word reading but similar phonological and MA scores than the monolinguals. See Table 1 for detail.

Procedure. Following an initial screening protocol, eligible participants were invited to complete the behavioral and neuroimaging measures during one laboratory visit. Consent and assent from parents and participants were obtained at every testing session. All participants completed testing in one lab visit. Each testing session was about 2.5 h for monolingual participants and 4 h long for bilingual participants, including breaks and snacks to support the children's cooperation and alertness during the visit. Parents completed a bilingual language use background survey, including family demographic backgrounds and children's language history.

3.1. Behavioral assessments

During the testing sessions, participants completed a battery of behavioral language and literacy assessments in English for monolinguals and in English and Chinese for bilinguals (see below). The full description of each task and a complete list of stimuli can be assessed from previous publications (Sun, Marks, et al., 2022; Sun, Zhang, Marks, Karas, et al., 2022). In Chinese, we used mostly experimental language assessments that were designed to match the English tasks, except for the Chinese receptive vocabulary. Participants' task performances are summarized in Table 2.

TABLE 2.

Neuroimaging task accuracy and reaction time ANOVA analysis results.

	English versus Chinese in bilinguals
Accuracy	F	df	p	η 2	Reaction time	F	df	p	η 2
Task	1.95	252	.17	.03	Task	28.62	252	<.001	.33
Condition	248.24	252	<.001	.81	Condition	185.18	252	<.001	.76
Language	22.84	252	<.001	.29	Language	4.80	252	.033	.08
Task * Language	3.32	252	.22	.03	Task * Language	29.70	252	<.001	.34
Task * Condition	2.34	252	.13	.04	Task * Condition	10.48	252	.002	.16
Condition * Language	27.28	252	<.001	.32	Condition * Language	23.17	252	<.001	.29

	English in bilinguals versus monolinguals
Accuracy	F	df	p	η 2	Reaction time	F	df	p	η 2
Task	1.89	194	.14	.02	Task	1.77	194	.24	.03
Condition	198.32	194	<.001	.77	Condition	187.34	194	<.001	.76
Group	2.01	194	.23	.03	Group	1.98	194	.21	.02
Task * Group	2.22	194	.18	.04	Task * Group	2.18	194	.11	.03
Task * Condition	3.24	194	.20	.04	Task * Condition	2.96	194	.16	.03
Condition * Group	3.17	194	.14	.03	Condition * Group	3.02	194	.17	.03

Note: Neuroimaging task accuracy and reaction time were analyzed with two separate 2 (Languages: English, Chinese) * 2 (Tasks: PA Task, MA Task) * 2 (Conditions: Experimental, Control) ANOVAs for bilinguals' two languages, and two separate 2 (Group: bilingual, monolingual) * 2 (Tasks: PA Task, MA Task) * 2 (Conditions: experimental, control) ANOVAs for monolinguals and age‐and proficiency matched bilinguals.

Abbreviations: ANOVA, analysis of variance; MA, morphological awareness; PA, phonological awareness.

Background Questionnaire. Parents of bilingual participants completed a background questionnaire that included children's demographic information, language development, physical health, family, and educational history. Parents of bilingual participants also completed a, as well as a modified version of the Bilingual Input–Output Survey (BIOS; Peña et al., 2021). BIOS asked parents to report language exposure on a typical weekday and weekend day, which included the history of what language the child heard and used. Parents also reported the average reading time in each language on a typical day.

3.1.1. English measures

English receptive vocabulary. Participants completed the PPVT‐5 to assess their receptive vocabulary level (Dunn, 2019). In this task, participants were asked to pick the word spoken by the experimenter that best describes the choice in four pictures shown to them.

English word reading. Participants completed the Letter–Word Identification Woodcock–Johnson—III test to assess their ability to identify letters and words (Schrank et al., 2014). Participants read a list of words aloud.

English MA. Participants completed the Early Lexical Morphology Measure (ELMM; Marks et al., 2022) to assess their ability to manipulate morphemes. In this task, participants were asked to complete a sentence using a word derived from a cue word given by the experimenter. For example, the experimenter would cue the child with the word “Friendly” and say: “She is my best ___.” [friend].

English PA. Participants completed the Comprehensive Test of Phonological Processing 2 (CTOPP‐2) Elision tasks to assess their reading‐related phonological processing skills (Wagner et al., 1999). In this task, participants were asked to take away part of the word (either phoneme or syllable) and then say out loud the remaining parts (e.g., “Say baseball without saying base”—ball).

3.1.2. Chinese measures

Chinese receptive vocabulary. Chinese receptive vocabulary was measured by the Chinese version of the PPVT‐R (Lu & Liu, 1998).

Chinese character recognition and reading. The Chinese word recognition and the reading task is an experimental measure that was developed based on the Chinese curriculum that is commonly used in local heritage Chinese schools. The character recognition component consisted of 30 items. Participants were asked to pick the correct character that was read to them. The character reading component consisted of 60 items.

Chinese MA. The Chinese MA task was modified based on Song et al. (2015). Participants were given the meaning of a compound word and were asked to create a new word with the morpheme(s) in the given word. (e.g., Apple‐trees grow apples. What trees might grow bread?—breadtree, 一颗长着苹果的树, 叫苹果树, 一棵长着面包的树叫什么呢?—面包树).

Chinese PA: The Chinese PA task was based on E. H. Newman et al. (2011) measure with the same paradigm as in English. The task has 36 items with 6 syllable‐level elisions and 30 phoneme‐level elisions (e.g., Can you say “/hé/”? Can you say “/hé/” without “/h/” sound? 请跟我念 “和/hé/” “和/hé/” 去掉 “/h/” 是什么呢?—鹅/é/).

3.2. Neuroimaging tasks and stimuli

The imaging task design and stimuli selection have been listed and validated in (Sun, Marks, et al., 2022; Sun, Zhang, Marks, Nickerson, et al., 2022), which also include extensive detail on stimuli selection. The data and experimental measures used in this study are freely available in the DeepBlue repository under the name “Morphological and phonological processing in English monolingual, Chinese–English bilingual and Spanish–English bilingual children: An fNIRS neuroimaging dataset” (Sun, Zhang, Marks, Nickerson, et al., 2022). The neuroimaging tasks were conducted auditorily. There were two tasks (MA and PA) per language. Within each task, there was an experimental condition of interest (either MA or PA) and a word recognition control condition. Across all tasks, children heard three consecutive words and were asked to select the matching word via button press (Figure 1). In addition, the PA and MA tasks employed a block design, given its advantage in its power to estimate the response that is averaged across the entire block. There were four blocks per condition. Each block contains four trials lasting 30 s, and there was a 6‐s rest between each block.

FIGURE 1.

This is an example trial of the fNIRS phonological processing task. The tasks were presented auditorily. Each trial began with a target word (e.g., “running”) and a white box on the screen. Next, participants heard two words of choice (e.g., “walking” and “raining”) and saw blue and yellow boxes, respectively. Responses were recorded using a button box. The morphological processing task followed the same experimental paradigm. fNIRS, functional near‐infrared spectroscopy.

In the MA task, the three consecutive words were comprised of a target word, followed by a word with a shared morpheme and a phonological distractor in random order (e.g., English task: classroom‐bedroom‐mushroom; Chinese task: 火车/huǒ chē/(train)—火箭/huǒ jiàn/(rocket)—伙伴/huǒ bàn/(pal)). Children had to attend to the words' meanings and choose the morphological match via a button press. This task also included a whole‐word recognition control condition, in which children heard two matchings (identical) words and one non‐matching (phonologically and semantically dissimilar word (e.g., English task: winter‐winter‐maybe; Chinese task: 冬天/dōng tiān/(winter)—猴子/hóu zi/(monkey)—冬天/dōng tiān/(winter)). The morphology task also included a derivation morphology condition (in English and Chinese), which is not included in the present study as it is a low‐frequency feature in Chinese and young heritage Chinese speakers found it to be disproportionately challenging.

In the PA task, the three words were a target word, followed by a phonologically matched word and a semantic distractor (e.g., running‐walking‐raining; 勺子/shaó zi/(spoon)—沙子/shā zi/(sand)—叉子/chā zi/(fork)). Children had to segment words by sounds and choose the phonological match. Similarly, children indicated which words had the same starting sound via a button press. This task also included the same whole‐word recognition control condition as in the morphology task.

Participants were trained on each task in each language (English and Chinese) before neuroimaging. First, participants received training on the task paradigm. Following the grasp of the task paradigm, participants performed additional computer‐based training items. The MA training task consisted of eight items in each language. Among these, the first three training items were with picture cues, while the remaining five were without picture cues, the same as in the experimental tasks. Similarly, in the phonological tasks, participants completed training with an experimenter and additional six training items on the computer in each language. The neuroimaging sessions started after the participants reached 100% accuracy in the training items to ensure a complete understanding of the task paradigm. The tasks were delivered via E‐Prime 2.0 software (Psychology Software Tools, Pittsburg, PA, http://support.pstnet.com/).

3.3. fNIRS data acquisition and processing

Data acquisition. The fNIRS probe set covers major language and literacy brain networks established in published literature, including the inferior frontal, superior temporal, and middle temporal regions (Hu et al., 2020; Marks et al., 2021; Sun, Marks, et al. 2022; Sun, Zhang, Marks, Karas, et al., 2022). The fNIRS probe set was embedded in caps made of silicone bands and included 12 light emitters and 18 detectors spaced about 2.7 cm apart in a grid‐like shape, yielding 36 data channels (18 per hemisphere, Figure 2c). The geometric structure of the probe set was transformed into Montreal Neurological Institute (MNI) stereotactic space. The MNI coordinates and regions covered for the fNIRS cap are listed in Table S1. More detailed spatial registration and MNI localization information are available from Hu et al. (2020).

FIGURE 2.

Children's brain activation during the phonological and morphological tasks (task > control, FDR corrected, q < .05, plotted using MNI 152 brain template). FDR, false discovery rate.

The fNIRS data were collected using the TechEN‐CW6 system with 690 and 830 nm wavelengths with a sampling frequency of 50 Hz. The signal‐to‐noise ratio thresholds were set to be in the standard 80 and 120 dB range. For each subject, the experimenters identified the nasion, ion, Fpz, and left and right pre‐auricular points and measured head circumferences before applying the appropriately sized fNIRS cap. The F7, F8, T3, and T4 were anchored to a specific source or detector. The optode placement quality was checked for each subject for the presence of a cardiac signal at each channel. The final dataset only included subjects with an in‐scanner accuracy greater than 65% in each task and with neuroimaging data that passed quality control.

Data processing. The fNIRS data analyses were done using the NIRS Brain AnalyzIR, a MATLAB‐based analysis toolbox (Santosa et al., 2018). The analysis used only oxygenated hemoglobin (HbO) data as it accounts for a greater portion of the fNIRS signal (HbO: 76% vs. HbR: 19%; Gagnon et al., 2012). The findings presented in this paper only showed effects that survived false discovery rate correction for multiple comparisons.

At the individual level, raw data were first trimmed to keep only 5 s before and after the experimental task baseline data. Second, data were down‐sampled from 50 Hz to 2 Hz, knowing that the fNIRS signal of interest is within the 0–1 Hz range. The optical density data were then converted to hemoglobin concentration data using the modified Beer–Lambert Law. Noise at the individual level was accounted for using the autoregressive‐iteratively reweighted least squares method (Barker et al., 2013; Caballero‐Gaudes & Reynolds, 2017). Temporal and dispersion derivatives, as well as the discrete cosine transform matrix, were included in the model to account for the signal drifts over time. The individual‐level data were analyzed using the general linear model (GLM; Friston, 2009). The GLM model yielded individual‐level regression coefficients for HbO signals for each channel.

At the group level, we used a linear mixed‐effects model for each data channel. In the linear mixed‐effect models, the task conditions (MA task, MA control, PA task, and PA control) were modeled as a fixed effect, with the participant as a random effect and the individual‐level beta values for HbO and HbR as predicting dependent variables.

Analytical formula. To examine the morphological and PA in English monolingual and Chinese–English bilinguals, we fitted the first linear mixed‐effect model and modeled the interaction between Tasks (PA, MA), Conditions (Task, Control), and Languages (English, Chinese) to predict the beta values for HbO (formula: HbO ~ Language * task * condition + age + maternal education + PPVT + (1 | subject)). We controlled for age, mother's education level, and language proficiency (PPVT). As is standard for block design analyses, brain activation was averaged across the entire block, across all trials within the block. Exploratory analyses that included task accuracy covariates yielded identical outcomes as the model without them. Therefore, we have presented a more parsimonious model.

Extracted contrasts. The estimated group‐level channel‐based effects were extracted for the following contrasts within each language: experimental task condition > control. In addition, the estimated group‐level effects were also extracted between tasks: morphology task − control > phonology task − control within each language as well as between languages: English task − control > Chinese task − control.

Exploratory brain‐behavior correlations within ROI. To better understand sources of heterogeneity in participants' brain activity, we conducted exploratory brain‐behavior correlations between children's language proficiency and their brain activation in left IFG, STG, and MTG regions, for which we found consistent brain activations across groups and languages.

4. RESULTS

4.1. Neuroimaging task performance

As can be seen in Table 2, in English and Chinese, bilinguals showed comparable performance across MA and PA tasks. Similarly, the monolinguals also showed comparable performance across MA and PA, and these observations were confirmed by the analysis of variance (ANOVA) analyses for each group that did not find a main effect of task (Table 2). There were, however, language differences in the bilinguals who were significantly faster and more accurate in English than in Chinese, as confirmed by a significant main effect of language. Finally, there was a main effect of condition across languages and groups, as the children were faster and more accurate during the control condition relative to the MA and PA conditions.

4.3. Cross‐linguistic differences in bilinguals

4.3.1. Phonological awareness

As can be seen in Figure 2, panel a (see also Table S2 for more statistical details), within‐language analyses suggested that more left frontal and parietal as well as bilateral temporal channels reached significance in Chinese than in English. Between‐language comparisons showed there was stronger activation in the left inferior parietal lobule, supramarginal gyrus, and angular gyrus (IPL/SMG/AG) as well as the right sensory‐motor cortex in Chinese relative to English. In addition, the right parietal region revealed stronger activation in English relative to Chinese. Note that the left temporal occipital differences stem from differences in deactivations. In sum, bilingual English and Chinese PA tasks elicited comparable activations in the frontal and temporal regions. Nevertheless, relative to the English, the Chinese PA tasks elicited stronger activation in left parietal activations, while the English PA task elicited stronger right parietal activation than the Chinese.

4.3.2. Morphological awareness

As seen in Figure 2a (see also Table S3 for more statistical detail), in the bilingual group, more regions reached significance in English than in Chinese, including the left frontal and bilateral temporal regions. Direct comparisons revealed that the left IFG and bilateral IPL/SMG regions were significantly more active in English relative to Chinese. In contrast, right STG and post‐central regions were more active in Chinese relative to English. Examination of the beta values revealed that some of the regions seen as more active in English relative to Chinese were differences in deactivation, including one left MTG channel and the left post‐central regions (see Figure S2 for more details). In sum, during the MA task, the bilinguals showed stronger left IFG, bilateral IPL/SMG activation in English and stronger left MTG and right STG/post‐central activation in Chinese.

4.4. Bilinguals versus monolinguals

4.4.1. Phonological awareness

As can be seen in Figure 2b, English monolinguals showed activation in left IFG and temporal regions. Between‐group comparisons revealed stronger left IFG activation in monolinguals relative to the bilinguals and stronger right IPL/SMG activation in bilinguals relative to monolinguals. The comparison also revealed stronger inferior temporal gyrus (ITG)/occipital and sensorimotor activation in monolinguals, but this effect stemmed from differences in greater deactivation during the control condition in monolinguals (see Figure S1 and Table S4 for more details). In sum, the bilingual and monolingual children's activations during PA tasks in English suggest overall similarity in brain activations, with stronger left IFG in monolinguals and stronger right IPL/SMG activation in bilinguals.

4.4.2. Morphological awareness

As can be seen in Figure 2b, bilinguals showed stronger and more widespread activations relative to the monolinguals in the left frontal, STG, ITG/occipital, and right IPL/SMG regions. Direct comparisons between the two groups confirmed this observation. The direct comparison contrast also suggested several regions were more active in monolinguals relative to bilinguals. Yet, examination of the beta values for these regions revealed differences in deactivation in the control condition to be the source of those group differences (see Figure S2 and Table S4 for more details). In sum, during the MA task, the bilinguals showed stronger activation in the left IFG, STG, ITG/occipital, and right parietal regions relative to the monolinguals.

4.5. Phonological versus morphological awareness

Chinese in bilinguals. As can be seen in Figure 3, in Chinese, the bilinguals showed stronger and more widespread activation for the PA relative to the MA task. In looking at within‐task activations (Figure 2a), we can see that task difference in Chinese is confirmed by those observations.

FIGURE 3.

Comparison between phonology (red) and morphology (blue) conditions (task > control contrasts; FDR corrected, q < 0.05, plotted using MNI 152 brain template). FDR, false discovery rate.

English in bilinguals and monolinguals. Task‐related differences appeared to be less substantial in English across the two groups, especially in the left frontal lobe, in which no channel reached significance. Both groups showed stronger activation for morphology in the left MTG and IPL/SMG/AG regions. The bilinguals also showed stronger activation for morphology in left ITG/occipital and right IPL/SMG regions, whereas monolinguals showed additional activations for right STG and IFG. For the phonology condition, activations were stronger in left ITG and AG and right IFG in the bilinguals and in left ITG/occipital and right sensory motor regions in the monolinguals.

In sum, task comparison revealed stronger differences in Chinese of the bilinguals than in the English of either bilinguals or monolinguals. In Chinese, there was stronger and more widespread activation during phonology, whereas in English, both groups converged in showing stronger left temporo‐parietal activation for morphology with some group‐specific differences in phonology.

4.6. ROI analysis and brain‐behavioral correlations

PA analyses focused on the left IFG and STG regions. In the left IFG, the activations were positive in English across groups and languages (Figure 4). Direct t‐test comparison revealed stronger activation in Chinese than in English within bilinguals and stronger in monolinguals relative to bilinguals. In the left STG, there were no language or group differences. Brain‐behavior correlations further revealed negative associations between neural activity and both age and proficiency in English, across regions and across groups. In contrast, no meaningful associations were found in Chinese for either of the regions.

FIGURE 4.

Patterns of brain activity and brain‐behavior correlations for regions of interest (β values for phonology > rest in IFG and STG, and for morphology > rest in IFG and MTG regions). IFG, inferior frontal gyrus; MTG, middle temporal gyrus; STG, superior temporal gyrus.

MA analyses focused on the left IFG and MTG regions. In the left IFG, all the activations were positive and different between languages/groups. English in the bilinguals eliciting the strongest activation compared to the other language/group (Figure 4). In the left MTG, the activation was positive only in Chinese, and it was also stronger in Chinese than in English of the bilinguals. Moreover, deactivation in English of the bilinguals was stronger than in monolinguals. Similar to the left IFG in phonology, brain‐behavior associations for IFG in morphology also showed negative associations with age and proficiency in English in both groups. In the left MTG, there was a positive association with age and language proficiency in English across groups. There were no meaningful associations in Chinese for either of the regions.

In sum, brain‐behavior correlations revealed that in English, the children's primary language of reading instruction, there was a similar pattern of significant associations between brain activity, age, and language proficiency. Notably, in the frontal regions, there was a negative association and in the temporal regions, there was a positive association. In contrast, no meaningful brain‐behavior associations were seen in bilinguals' Chinese.

5. DISCUSSION

The present study examined the neurocognitive bases of morphological and phonological literacy skills in Chinese–English bilinguals and English monolinguals. Cross‐linguistic comparisons revealed that bilingual children showed less activation in left IFG and parietal regions during the tasks that were more central to reading in a given orthography, such as morphology in Chinese and phonology in English. Group comparisons between bilinguals and monolinguals in English—children's shared language of academic instruction—revealed that left IFG was less active during phonology but more active during morphology in bilinguals relative to monolinguals. Nevertheless, the two groups showed similar negative brain‐behavior associations in the left IFG, STG, and MTG regions in English, across tasks. In contrast, in Chinese, no meaningful brain‐behavior associations were found in these regions across tasks. This observation might reflect differences in dual‐language experiences and proficiency in bilinguals. These findings inform theories of literacy development by revealing the influences of language, bilingualism, and proficiency on children's emerging neural organization for learning to read in different languages.

5.1. Phonological and morphological awareness in bilinguals

Bilingual children who participated in our study showed overall better performance in English relative to Chinese across behavioral measures of language and literacy in and out of the scanner. This observation is consistent with parental reports suggesting that these children were receiving English‐dominant schooling and were generally spending more time speaking and reading in English than in their heritage language. Nevertheless, our bilingual participants had age‐appropriate Chinese vocabulary and the majority (74%) reported receiving some form of heritage language literacy instruction at home or/and at local heritage language schools.

During the PA task, the bilinguals showed greater engagement of the left IFG and parietal regions in Chinese than in English, as can be seen in Figures 2 and 3. Notably, the left parietal differences also reached significance during the direct cross‐linguistic comparison. Prior research suggested that in English and other alphabetic languages, older and more proficient readers often exhibit more robust brain activations as well as stronger functional connectivity along the dorsal phonological network (Cao et al., 2011; Chyl et al., 2018; Martin et al., 2016; Norton et al., 2014; Preston et al., 2016). The fact that our observations diverge from the prior findings (stronger activation for Chinese relative to English) could potentially stem from differences in our populations' dual language proficiency. Research typically finds that bilingual children and adults typically exhibit stronger activation along the perisylvian language networks, as well as in other brain regions, in their language of lower proficiency relative to the dominant language, to support additional language and cognitive demands (Cargnelutti et al., 2019). One possible explanation for our findings of stronger activation in left IFG and parietal cortex in Chinese relative to English may thus lie in children's lower Chinese proficiency. At the same time, note that the pattern is reversed for the morphology task, which yielded stronger left IFG and parietal activations for English relative to Chinese, as can be seen in Figures 2 and 4. Therefore, we suggest that cross‐linguistic differences between English and Chinese, rather than proficiency differences, are likely to explain the observed pattern of results. The left IFG region is generally thought to support analytically complex and attention‐demanding language processes (Friederici & Gierhan, 2013). Left parietal regions, including the SMG and AG regions, have been previously shown to support various elements of PA, including verbal working memory, phonological analyses, and sound‐to‐print mapping demands (Dehaene‐Lambertz et al., 2005; R. L. Newman & Joanisse, 2011). Differences in the strength of activation in these regions between English and Chinese may thus reflect differences in how phonological and morphological cues predict word recognition across the two languages. In Chinese, morphology is more predictable of the language's word structure and its orthographic characteristics relative to phonology, whereas the opposite is typically found in English (Chen et al., 2009; McBride‐Chang et al., 2005). It is, therefore, possible that the stronger engagement of the left IFG and parietal regions for phonology in Chinese and morphology in English reflect cross‐linguistic differences in the increased analytical effort that children are allocating to the lower frequency features in each of their languages.

Another notable difference between bilinguals' two languages is that of stronger left MTG activation during the morphology condition in Chinese relative to English, as seen in Figures 2 and 4. The left MTG region is generally associated with automated lexical retrieval and processing and has previously been shown to be active during morphology tasks in English, Chinese, and other languages (Acheson et al., 2011; L. Liu et al., 2009; Yang & Li, 2019). Compound morphology is a high‐frequency feature of Chinese as characters map onto morphemes, and almost all words in Chinese are morphological compounds. Learning to read in Chinese in the early years typically involves manipulating two or more morphemes and appears to develop naturally (Li et al., 2012). It is, therefore, possible that the high frequency of compounding in Chinese yields stronger engagement of automated lexical morphology processes linked to the MTG's functionality.

Direct comparisons between bilinguals' PA and MA tasks revealed more pronounced task differences in Chinese than in English (see Figure 3). In Chinese, the differences were widespread, incurring stronger bilateral activation during the phonology relative to the morphology condition. In contrast, in English, morphology elicited stronger activation in two left temporo‐parietal channels, whereas phonology elicited stronger activation in the left parietal, inferior temporal, and right frontal regions. We suggest that both cross‐linguistic characteristics and reading experience differences contribute to the observed differences. One possibility for the substantially stronger activation for phonology relative to morphology in Chinese is the cross‐linguistic difference in the predictability of the two features. Phonology, a lower predictability feature, may incur stronger activation in brain regions relative to morphology. Moreover, differences in activation between phonology and morphology were more widespread in Chinese relative to English. This second observation may stem from differences in reading experiences. Both phonological and MA skills are strengthened through reading experiences (Goswami & Bryant, 2016). Bilinguals' literacy experiences were generally more limited in Chinese relative to English. The strongest levels of activation for phonology in Chinese, relative to morphology in Chinese and English in general, may thus stem from a combined effect of low predictability of phonology in Chinese orthography and low levels of Chinese PA skill.

5.2. Bilinguals versus monolinguals

In their primary language of reading instruction in English, bilinguals and monolinguals showed overall similar brain activations across tasks (see Figures 2 and 3). This pattern is consistent with the overall similarity in groups' performance in and out of the scanner and their bilingual experience of early exposure to English being primarily educated in English. At the same time, some group differences emerged. During the phonology task, monolinguals showed stronger left IFG activation relative to bilinguals. During the morphology task, bilinguals showed stronger left IFG and STG/MTG activation relative to monolinguals (see Figures 2 and 4). Across PA and MA tasks, bilinguals generally showed stronger right IPL/SMG activation, especially in the parietal regions as can be seen in Figure 2. The right IPL has been previously shown to support attention and selective inhibition demands during task‐switching protocols (Singh‐Curry & Husain, 2009). Prior research finding stronger right IPL activation in bilinguals has attributed this effect to the attentional aspects of dual language use (Calabria et al., 2018). Therefore, one possible explanation for our findings is that the right parietal regions help support various attention demands for dual language use because bilinguals are often found to co‐activate both of their languages even when using only one of their languages (Arredondo et al., 2019). Variation in the left IFG as well as STG/MTG activation, across the phonology and morphology tasks may thus relate to bilinguals' cross‐linguistic transfer influences, whereas the right hemisphere differences might be more generalized across bilingual experiences.

Bilingualism frameworks generally posit that young bilinguals develop a singular PA capacity that subserves both of their languages (Bolger et al., 2005; Perfetti et al., 2007). In contrast, the MA capacity is typically considered to be more language specific to be able to account for cross‐linguistic variation in how morphemes combine into words across languages (Chung et al., 2019). The lower engagement of left IFG during phonology in bilinguals relative to monolinguals (see Figure 4) may therefore reflect the possibility that phonological processes are being jointly and similarly entrained across bilinguals' two languages, resulting in a more automated and less effortful process in the bilinguals. In contrast, stronger left IFG and STG/MTG activation during morphology in bilinguals may reflect that morphology computations are more language‐specific and may thus require a greater extent of neural resources for developing language‐specific knowledge. In turn, this development can aid in responding to the demands of the given language as well as adjudicating discrepancies between the two different morphological systems.

5.3. Brain‐behavior correlations

In English, both monolingual and bilingual groups showed negative associations of age and language proficiency (as measured by vocabulary and word reading) with brain activation in the left IFG, left STG, and a positive association in the left MTG region. In contrast, no meaningful brain‐behavior associations were found in bilinguals' Chinese phonological and morphological processing. The brain‐behavior correlations observed in the left STG and left MTG regions for English are generally consistent with previously reported studies. Prior studies have shown that as children become older and more proficient in their language and literacy skills, they exhibit less neural recruitment in the left STG region (Bitan et al., 2007; Noble et al., 2006) and more activation in the MTG region (Chou et al., 2006; Cone et al., 2008; Turkeltaub et al., 2003). Turkeltaub et al. (2003) have interpreted this effect as relating to a reduced need for laborious phonological analyses and more automated access to meaning during word recognition tasks in older and more proficient learners. Neural activation in the left IFG region is often found to be positively associated with age and language proficiency (Bitan et al., 2007; Cone et al., 2008; Turkeltaub et al., 2003) in alphabetic languages. Nevertheless, we observed that older and/or more proficient children showed less left IFG activation across tasks. Our tasks involved both elements of phonology and morphology—during phonology trials participants had to pay attention to sound and ignore the semantic distractor, and vice versa for morphology trials. In sum, it is possible that the negative patterns of brain‐behavior associations across the left IFG and STG regions reflect that children with stronger language and reading proficiency exhibited greater automaticity during our experimental tasks, reflected in weaker activation.

In Chinese, studies have suggested that age and language proficiency are positively correlated with activations in the left middle occipital gyrus, superior parietal lobule, right middle occipital gyrus, and negatively correlated with activation in the right STG (Cao et al., 2009, 2014). Yet, the previously reported patterns in Chinese were not observed in the present study. Bilingual children's age did not significantly correlate with brain activity during the two tasks. Children in the present study were receiving systematic literacy instruction and had varied and sporadic Chinese literacy instruction. The lack of brain‐behavior association in bilinguals' Chinese is thus likely attributed to the inconsistency in bilinguals' heritage language learning experiences.

5.4. Theoretical contributions

Our neurocognitive findings support and extend the LQH, which conceptualizes word reading development as the product of emerging sound‐to‐print and meaning‐to‐print associations (Perfetti & Hart, 2002). Cross‐linguistically, orthographic characteristics of a given language are thought to play a role, resulting in stronger sound‐to‐print associations in English and stronger meaning‐to‐print associations in Chinese (McBride et al., 2021). Developmentally, these differences may be apparent in children's emerging phonological and lexical morphology abilities, as in alphabetic languages, sounds map onto letters and in character‐based languages, morphemes map onto characters. Our neurocognitive findings align with these behavioral observations by revealing stronger patterns of brain‐behavior associations between phonology than morphology in English and vice‐versa in Chinese, as well as weaker neural activity for phonology in English and morphology in Chinese. Our bilingual findings suggest that even within the same young learner, there can be a differentiation in the development of literacy systems. This differentiation may occur first, in response to cross‐linguistic differences, and second, in response to quantity and quality of literacy instruction as our participants were dominant readers of English.

Prior neuroimaging work have generally suggested that bilinguals should exhibit stronger brain activations than monolinguals, and that bilinguals' lower proficiency language should further incur stronger activation than the dominant language (Sulpizio et al., 2020). These differences are usually attributed to the greater cognitive effort required for the acquisition and processing of two different languages as well as the lower proficiency languages. Our work with early‐exposed heritage language speakers reveals a more nuanced pattern. In particular, the between‐group comparison reveals stronger left hemisphere activation in bilinguals than monolinguals only during the morphology condition that requires stronger differentiation between bilinguals' two languages, but not the phonology condition that is considered more language‐common. The bilinguals did exhibit stronger right IPL/SMG activation than monolinguals across tasks, a finding that is largely consistent with prior bilingual work (Jasińska & Petitto, 2014), but the nature of which requires further inquiry. Taken together, our findings advance theoretical perspectives that consider a bilingual language system as consisting of one common set of systems with subsystem differentiation (Marks et al., 2022; Satterfield, 1999) by revealing evidence of neurodevelopmental adaptation to shared and distinct characteristics of bilinguals' two languages.

6. LIMITATIONS

The present study has several limitations. First, bilingual children had stronger English than Chinese proficiency. Although our analyses controlled for vocabulary and our observations are generally consistent with those predicted by monolingual cross‐linguistic research, it remains possible that the observed cross‐linguistic differences are predominantly driven by differences in proficiency. Second, the experimental sample included early exposed heritage language learners and thus, the findings may not generalize to other types of bilingual learners.

7. CONCLUSION

This study explored the neurocognitive bases of learning to read in two typologically contrasting languages, English and Chinese, in young heritage language bilinguals and young English monolinguals. The findings revealed cross‐linguistic and bilingual effects on children's emerging neural architecture for learning to read. Within each language, the task that best characterizes literacy success of the given orthography (phonology in English and morphology in Chinese) elicited weaker brain activity in the left IFG and parietal regions, which may reflect greater automaticity for the given language. Compared with monolinguals, the bilinguals showed stronger left hemisphere activation during the condition that differentiated the most between the bilinguals' two languages (morphology), and lower activation/greater automaticity during the condition that is considered more language‐common (phonology). Finally, bilinguals and monolinguals showed similar brain‐behavior associations with the left IFG, STG, and MTG regions in their language of primary academic instruction. Taken together, the findings suggest that experiences with two typologically distinct languages can exert cross‐linguistic, proficiency, and dual‐language use influences on children's emerging architecture for learning to read.

FUNDING INFORMATION

This study is supported by the National Institute of Health (NIH), grant/award number: R01 HD092498.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

ETHICS STATEMENT

This study was approved by the Institutional Review Board (IRB) of the University of Michigan. All participants have assented, and parents consented to participate in this study.

Supporting information

Figure S1. Appendix figure.

Figure S2. Appendix figure.

Table S1. Estimated left hemisphere brain regions covered by the fNIRS probe set.

Table S2. Brain activations during phonological task for bilinguals' two languages.

Table S3. Brain activations during morphological task for bilinguals' two languages.

Table S4. Brain activations during phonological and morphological tasks for monolinguals.

Zhang, K. , Sun, X. , Yu, C.‐L. , Eggleston, R. L. , Marks, R. A. , Nickerson, N. , Caruso, V. C. , Hu, X.‐S. , Tardif, T. , Chou, T.‐L. , Booth, J. R. , & Kovelman, I. (2023). Phonological and morphological literacy skills in English and Chinese: A cross‐linguistic neuroimaging comparison of Chinese–English bilingual and monolingual English children. Human Brain Mapping, 44(13), 4812–4829. 10.1002/hbm.26419

DATA AVAILABILITY STATEMENT

This study's data and experimental measures are open to free access and stored in the DeepBlue repository under the name “Morphological and phonological processing in English monolingual, Chinese–English bilingual and Spanish–English bilingual children: an fNIRS neuroimaging dataset.” The description of the dataset is published in Data in Brief (Sun, Zhang, Marks, Karas, et al., 2022).