Shifting dominance: bilingual language skills in Spanish–English bilinguals’ English reading comprehension over time

Abstract

This longitudinal study investigated the relative importance of various bilingual language skills predictors for English reading comprehension over time in Spanish–English bilingual children in elementary grades. We examined the contributions of phonological awareness (PA), morphological awareness (MA), syntactic awareness (SA), and vocabulary in both English and Spanish at two time points. Dominance analysis was used to examine the shifting pattern of relative contributions of bilingual language skills to children’s English reading comprehension. Results revealed a shifting dominance pattern of predictor importance over time. At Time 1, English PA was the strongest predictor, followed by English MA. However, by Time 2, English MA emerged as the most important predictor, surpassing English PA. Spanish language skills, particularly MA and PA, showed notable contributions at Time 1 but decreased in importance by Time 2. English SA and vocabulary demonstrated increasing contributions over time, while their Spanish counterparts declined. These findings highlight the evolving nature of reading comprehension predictors in bilinguals’ reading development, suggesting a transition from reliance on both first and second language skills to predominantly second language-specific skills.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s11145-026-10762-z.

Introduction

Reading is considered fundamental for academic success, career attainment (World Literacy Foundation, 2015), and health-related outcomes (Batterham et al., 2016). As such, approaches to best support children’s reading development have long been discussed among researchers, practitioners, and stakeholders (Castles et al., 2018; Kim, 2008). Research generally concurs that reading comprehension is an intricate process involving both decoding and language comprehension, which are considered distinct yet interrelated components (e.g., the Simple View of Reading [SVR]; Gough & Tunmer, 1986; Hoover & Gough, 1990). Decoding refers to the ability to map letters to their spoken language environment, while language comprehension1 involves the ability to derive meaning from text (Gough & Tunmer, 1986). The SVR posits a multiplicative relation between decoding and language comprehension, such that mastery in one aspect but difficulty in another can hinder reading comprehension as a result. In the early stages of reading, decoding is often considered critical as children learn to sound out words, but as they progress, language comprehension becomes increasingly important as they become more efficient decoders and transition to “reading to learn” through increasingly difficult texts (Chall, 1983; Mancilla-Martinez, 2020).

However, extensive research attention has been focused on English monolinguals; less is known about this highly complex reading process of emergent bilinguals (EBs),2 who are simultaneously developing home language skills and English as additional language and learning to read in their additional language. In the U.S., Spanish–English EBs, the largest bilingual group, often face unique challenges as they acquire literacy skills in both languages. Their reading development may differ significantly from that of their English monolingual peers, as they are not only acquiring content knowledge but also developing English proficiency in tandem. Research to date has demonstrated that, in contrast to their English monolingual peers, EBs tend to develop adequate decoding skills but may struggle with language comprehension, and consequently reading comprehension (Mancilla-Martinez & Lesaux, 2017; Nakamoto et al., 2008). Further, the development of reading comprehension becomes more complex because EBs must navigate two linguistic systems, which can complicate the process of reading comprehension in a second language (L2; Koda, 2007), and this complexity is further influenced by contextual factors such as socioeconomic status, motivation, and the home literacy environment that shape opportunities for language exposure and literacy engagement (Camping et al., 2020; Goodrich et al., 2025; Shen, 2025; Shen & Wang, 2024). Thus, examining the reading development of EBs requires a nuanced understanding of unique language skills that contribute to reading success in both languages. Such understanding is critical for developing tailored instructional approaches and interventions to support the unique needs of EBs, particularly in the U.S. context where bilingual education policies and school environments vary significantly.

Bilingualism and English reading development

For bilingual children, the process of comprehending texts is shaped not only by the cognitive demands of reading itself but also by the need to simultaneously develop oral language skills in both their first language (L1) and L2. EBs are also faced with the additional challenge of learning to read in a language they may not speak at home. Research has shown that bilingualism affects various aspects of reading, due largely to the common underlying proficiency between the two languages (Cummins, 1979, 1991). However, this shared proficiency interacts with the language-specific skills that support comprehension, leading to developmental shifts in the relative importance of different predictors. According to the SVR (Gough & Tunmer, 1986), reading comprehension depends on decoding and linguistic comprehension; for bilingual children, both components are shaped by the dynamics of cross-language transfer and growth of proficiency in each language (Cummins, 1979; Durgunoğlu et al., 1993; Shen & Goodrich, 2024; Wagley et al., 2022).

Cross-language transfer is a key concept in understanding bilingual reading development, rooted in the framework of linguistic interdependence hypothesis (Cummins, 1979, 1991). The hypothesis suggests that bilingual children’s L1 competence can substantially influence their L2 skill development. To comprehend L2 texts, readers tend to leverage a wide range of linguistic knowledge and cognitive abilities in their L1 (Genesee et al., 2006). The Unified Model of Language Acquisition (MacWhinney, 2005) also supports this idea by emphasizing that bilinguals rely on interconnected linguistic systems, where the knowledge from both languages interacts and influences language processing. Extant research has demonstrated the significant contributions of L1 vocabulary (Proctor et al., 2006), and phonological (Durgunoğlu et al., 1993), morphological (Bae & Joshi, 2018), and syntactic awareness (Proctor et al., 2017) to L2 reading comprehension, though the degree of transfer depends on multiple factors, including the specific language skills in question, the similarities and differences between a child’s two languages, and their relative proficiency in each (Chung et al., 2019).

As such, the concept of cross-language transfer may be particularly relevant for closely related languages such as Spanish and English, where commonalities in vocabulary, phonology, morphology, and syntax enable skills to transfer from one language to the other (Koda, 2007). Therefore, cross-language transfer may be shaped by orthographic transparency and grain size differences between Spanish and English. Because Spanish has more consistent phoneme–grapheme correspondences, phonological skills may transfer more readily in early stages, whereas English’s deeper orthography increases the importance of morphological processing over time (Ziegler & Goswami, 2005). Further, according to the interdependence continuum (Proctor et al., 2010), the strength of cross-language associations is dependent on both the languages and the linguistic skills in question. In their view, oral language skills are hypothesized to be more language specific than reading skills. Aligned with this hypothesis, studies that have examined the role of both English and Spanish language abilities as predictors for English reading comprehension suggest that English language ability plays a more dominant role in English reading comprehension than Spanish language skills (e.g., Proctor et al., 2017). Further, bilingual students vary in proficiency in their L1 and L2, which manifests differently in language ability, and consequently literacy performance (Proctor et al., 2006). Another factor that we should consider is the specific language skills in question. Some skills appear to be more language-specific (e.g., vocabulary) and others may be more language-universal (e.g., phonological awareness; Branum-Martin et al., 2015; Melby-Lervåg & Lervåg, 2011). As bilingual children’s exposure to English increases over time, the associations between specific language skills and English reading comprehension may shift over time as a function of language proficiency. Drawing on reading stage theories (Chall, 1983) and the lexical quality framework (Perfetti, 2007), early reading relies heavily on code-based processes such as phonological awareness, which enable children to map print to sound. As bilingual children’s oral proficiency and lexical representations in English strengthen, meaning-based processes, such as morphological awareness, become increasingly central for constructing word- and text-level meaning. This process of shifting dominance is underexplored and raises questions about how language skills in L1 and L2 develop to influence reading comprehension across developmental stages.

The role of language skills in reading comprehension

Language comprehension encompasses a range of language skills, including vocabulary knowledge, phonological awareness, as well as knowledge of morphology and syntax (Oakhill & Cain, 2007; Silverman et al., 2020). For instance, vocabulary knowledge directly supports comprehension by providing access to word meanings, while syntax helps readers understand the relations between words in a sentence. The ability to integrate these skills efficiently may ultimately lead to successful reading comprehension (Kieffer & Lesaux, 2012). For bilingual children, oral language proficiency in both languages has been found to be a robust predictor in their L2 reading comprehension, though L2 oral proficiency typically shows stronger predictive power (Huang et al., 2021; Proctor et al., 2017; Shen & Coker, 2023; Wagley et al., 2022). Beyond these linguistic components, reading comprehension is also influenced by non-linguistic cognitive factors, including metacognitive skills such as comprehension monitoring and strategic reading and inferential reasoning abilities that enable readers to construct meaning beyond what is explicitly stated in the text (Channa et al., 2014; Kendeou, 2015; Soto et al., 2019). In the current study, we focus on a range of language skills in both languages.

Phonological awareness (PA)

PA refers to the ability to recognize and manipulate the sounds in spoken language (Tunmer & Rohl, 1991). It is a key predictor of early reading success, particularly in the decoding process, as it enables children to map sounds onto letters (Ziegler & Goswami, 2005). In monolingual children, PA is consistently associated with word reading ability and later reading comprehension (Kirby et al., 2003; Rezaei & Jeddi, 2020). The role of PA in English reading comprehension has been found to be strongest in the first years of reading instruction and steadily decreases in later years (Kirby et al., 2003). The importance of PA in English reading development has also been established for bilingual children (Tong et al., 2023; Wawire & Kim, 2018).

Following the seminal work of Durgunoğlu et al. (1993), studies have shown that bilingual children who develop strong PA in their L1 may benefit from cross-language transfer, which supports their ability to read in the L2 (Dickinson et al., 2004; Melby-Lervåg & Lervåg, 2011). When bilinguals begin to learn and recognize words in L2, they tend to use visual and phonetic cues from their L1 to facilitate L2 reading. For example, PA in Spanish has been found to predict English word recognition skills in Spanish–English bilingual children (Durgunoğlu et al., 1993). Similarly, in a study of Chinese–English bilingual Grade 2 children, Tong et al. (2023) also found that L1 PA indirectly predicted L2 reading comprehension through L2 vocabulary and word reading. However, cross-sectional research suggests that the contribution of PA to reading comprehension may follow a developmental trajectory. For instance, among Spanish–English first graders, Quiroga et al. (2002) found that both L1 and L2 PA predicted their English word reading skills; by second grade, Manis et al. (2004) found that only Grade 1 English PA but not kindergarten Spanish PA, predicted Grade 2 English reading comprehension. Swanson et al. (2008) found that English and Spanish PA did not contribute significantly to English reading comprehension for Grade 3 Spanish–English bilinguals, after controlling for other bilingual language skills. As bilingual children progress in their reading development and transition to “reading to learn”, the relative importance of PA may similarly diminish, and other language skills such as morphological awareness and vocabulary take on a greater role in supporting comprehension (Lonigan et al., 2018). This developmental shift suggests an “early importance and later decline” pattern for PA’s contribution to reading comprehension in bilingual populations. The current study explores how the contribution of PA in both L1 and L2 to English reading comprehension, relative to other oral language skills, changes over time for bilingual children, investigating whether this hypothesized pattern exists longitudinally.

Morphological awareness (MA)

MA involves the ability to reflect upon and manipulate morphemes, the smallest units of meaning in a language, and to combine them to form words (Carlisle, 2003; Kuo & Anderson, 2006). For example, recognizing that the word “unhappiness” is composed of the prefix “un-“, the root “happy,” and the suffix “-ness” enables readers to decode and comprehend its meaning. MA is hypothesized to influence the reading process because of its fundamentally integrative nature, such that recognizing and understanding morphologically complex words involves the integration of semantic, phonological, and syntactic information (Kieffer & Lesaux, 2012; Kuo & Anderson, 2006). Indeed, research on both English monolingual and bilingual children has demonstrated that MA is a unique and strong contributor to reading comprehension, above and beyond robust predictors such as cognitive abilities, PA, and decoding skills (Deacon et al., 2009; Kieffer & Lesaux, 2012; Kirby et al., 2012). Moreover, limited MA may explain why many EBs struggle to acquire language comprehension skills needed to understand complex texts (Kieffer & Lesaux, 2008).

It is evident that such association between MA and reading comprehension becomes increasingly stronger as children encounter more complex texts in later grades (Ke et al., 2021). For instance, following a sample of Spanish–English EBs from Grade 4 to Grade 5, Kieffer and Lesaux (2008) showed that not only did MA significantly predict EBs’ English reading comprehension in Grade 5, even after controlling for word reading skills, vocabulary, and PA, but also the predictive relation between MA and reading comprehension increased between Grade 4 and Grade 5.

In bilingual children, MA in their L1 can contribute to their L2 reading comprehension, with research suggesting that knowledge of morphology in the L1 can transfer to the L2 (Bae & Joshi, 2018; Ke et al., 2021). For example, Spanish-speaking bilingual children who have a strong understanding of morphological structures in Spanish may be better equipped to understand morphologically complex words in English due to the similarities between the two languages, particularly with regard to shared roots and affixes, which in turn facilitate their ability to comprehend increasingly complex texts (Ramírez et al., 2013). However, bilingual children’s reliance on MA in L1 and L2 to successful L2 reading comprehension may differ, with L2 MA appearing to be more critical (see Ke et al., 2021 for a meta-analysis). Adding to this complexity, Proctor et al. (2012) showed that Spanish MA was not associated with English reading comprehension for Spanish–English EBs and non-EBs. The current study explores how the dominance of MA in Spanish and English shifts over time and contributes to EBs’ reading comprehension.

Syntactic awareness (SA)

SA refers to the ability to understand and manipulate the structure of sentences, which is essential for making sense of how words relate to each other within sentences (Cain, 2007). Strong SA allows readers to comprehend complex sentences, infer meaning from context, and resolve ambiguities in text through parsing complex sentences into smaller, more manageable components (Graesser et al., 2011). The contribution of SA to reading comprehension has been supported by a growing body of studies, even after controlling for other factors including nonverbal ability, word reading, PA, and vocabulary (e.g., Deacon & Kieffer, 2018; Tong et al., 2014). For instance, Deacon and Kieffer (2018) observed an association between SA and reading comprehension among monolingual English-speaking children in Grades 3 and 4. Such effects were also evident in bilingual children (Sohail et al., 2022; Xie & Yeung, 2022). In a study of English-French bilinguals, Sohail et al. (2022) found that L2 SA contributed to L2 reading comprehension, although indirectly through L2 word reading. Bellocchi et al. (2017) even showed that SA may be a more important predictor in L2 reading comprehension, relative to L1.

There is also emerging evidence for the cross-language transfer of L1 SA to L2 reading comprehension, although the finidngs are mixed (Proctor et al., 2012, 2017). For example, Sohail et al.‘s (2022) study of English-French bilingual children revealed that L1 SA contributed to their L2 reading comprehension, both directly and indirectly through word reading. In a study of Grade 1 and Grade 3 Chinese–English bilinguals, Siu and Ho (2015) found that children’s L1 SA was significantly associated with L2 reading comprehension directly and indirectly through L2 SA, suggesting a cross-language transfer. Among Spanish–English bilingual children in primary school, Proctor et al. (2017) observed that early L1 SA contributed to later L2 reading comprehension. Including Spanish SA in a composite Spanish core academic language skills, Galloway et al. (2020) found that Spanish language skills significantly predicted English reading comprehension. However, some studies of Spanish–English bilinguals have failed to find such transfer effects. Proctor et al. (2012) did not find a significant effect of Spanish SA on English reading comprehension; Swanson et al. (2008) found that only English SA, but not Spanish SA, predicted English reading. This study examines how SA in both Spanish and English contributes to English reading comprehension over time, offering insights into the evolving role of syntax in English literacy development.

Vocabulary knowledge

Vocabulary knowledge is instrumental to reading comprehension because it provides readers with the ability to understand the meanings of words they encounter in texts (Anderson & Freebody, 1981; McKeown et al., 1983; Quinn et al., 2015). According to the lexical quality hypothesis (Perfetti, 2007), the number of words available to readers and the quality of their lexical representations have a direct impact on reading comprehension. In particular, receptive vocabulary is one of the best known predictors of reading comprehension, above and beyond other factors such as nonverbal intelligence and word reading (Ouellette, 2006). The relation between vocabulary knowledge and reading comprehension has been supported in studies with bilinguals (Zhang & Zhang, 2022). Studies have also shown that vocabulary knowledge is more important in reading comprehension for bilingual readers (Babayiğit & Shapiro, 2020). For instance, Shen and Coker (2023) found that L2 vocabulary knowledge predicted EBs’ L2 reading comprehension, but such association was not found for monolingual English speakers. As such, vocabulary knowledge may be of particular importance for bilingual readers.

When examining the cross-language transfer of L1 vocabulary to L2 reading comprehension, existing research showed mixed evidence, with most studies pointing to limited associations. For instance, some studies of Spanish–English bilingual students failed to find an association between L1 Spanish vocabulary and L2 English reading comprehension (Leider et al., 2013; Proctor et al., 2017). However, Proctor et al. (2006), in a study of Spanish–English bilingual 4th graders, found a main effect of Spanish vocabulary knowledge on their English reading comprehension, as well as an interaction between Spanish vocabulary and English fluency. They suggest that fluent English readers benefited more from Spanish vocabulary knowledge than their less fluent peers. Moreover, Goodrich et al. (2023) showed that among Spanish–English bilingual kindergarten and Grade 1 children, Spanish vocabulary made a significant contribution to English reading comprehension beyond the effects of English oral language. This study explores how vocabulary knowledge in bilingual children’s L1 and L2 contributes to English reading development over time.

The current study

Despite growing interest in the literacy development of bilingual children, there remains a significant gap in the literature regarding how the relative importance of different bilingual language skills shifts over time. Much of the existing research has relied on cross-sectional designs or examined L1 and L2 skills in isolation, without considering the dynamic interactions between these languages as children develop English reading skills. Additionally, traditional approaches to modeling predictors of reading comprehension, such as multiple regression or structural equation modeling, often face challenges related to multicollinearity, where predictors are highly correlated, making it difficult to disentangle their unique contributions (Budescu, 1993) and determine which skills are most influential at different developmental stages.

This study addresses these issues by using dominance analysis (DA), a robust statistical method specifically designed to compare the relative importance of multiple correlated predictors (Azen & Budescu, 2003). The use of DA aligns with the study’s primary aim of identifying and comparing the contributions of different bilingual language skills, rather than testing causal or directional pathways, as would be the focus of structural equation modeling. Moreover, given the modest sample size, structural equation modeling was not feasible. By applying DA to bilingual language predictors at each of two time points, this study provides a clearer and statistically robust understanding of how the relative contributions of vocabulary, PA, MA, and SA in both Spanish and English to English reading comprehension evolve over time. Such methodology contributes to our understanding of the development of reading comprehension in bilingual children and how bilingual language skills may develop to play different roles. The study also offers practical implications for educators, suggesting ways to tailor interventions based on the shifting linguistic needs of bilingual learners as they progress through different stages of literacy development. Building on the theoretical framework and literature review, this study seeks to answer the following research questions:

1. How do various bilingual language skills (vocabulary, PA, MA, and SA) in both Spanish and English contribute to English reading comprehension at two distinct time points?

2. How does the relative importance of these linguistic skills shift over time, particularly as bilingual children gain greater proficiency in English (L2)?

We hypothesized that PA would show stronger relative importance at earlier stages of learning to read and this contribution would diminish at the later time point, when vocabulary, MA, and SA would show increased importance, reflecting the developmental shift from “learning to read” to “reading to learn” (Chall, 1983; Lesaux et al., 2010; Lonigan et al., 2018). We further hypothesized that L2 English language skills would show greater relative importance than L1 Spanish skills at both time points, with this pattern becoming more pronounced over time as children develop greater L2 proficiency.

Method

Participants

The current study is a secondary data analysis of 57 Spanish–English emergent bilingual children, selected from a multi-year study on Spanish–English bilingual children’s language and literacy development (Huang et al., 2022). They were recruited from Spanish–English Dual Language Immersion (DLI) programs in five elementary schools in a Southwestern city in the U.S. All five schools have a high concentration of students of Mexican heritage. Sampling criteria included (a) being Grades 1 or 3 Spanish–English bilingual children in DLI programs and (b) having no documented speech, hearing, or other developmental disorders. All children were followed for two years and assessed on language and literacy skills at two time points. Time 1 data were collected in Fall 2019 and Time 2 data were collected two years later in Fall 2021. All participating schools were closed in Spring 2020 and implemented remote instruction during the 2020–2021 academic year, before returning to in-person instruction in Fall 2021. At each time point, child participants completed the language and reading measures in two separate sessions (one with English measures and one with Spanish measures) that were at least four days apart to avoid practice effects. The orders of the two language sessions at both time points were counterbalanced.

In this study, we used a cohort design where children in Grade 1 and Grade 3 at Time 1 were followed longitudinally to Grade 3 and Grade 5, respectively, at Time 2. This design allows us to track the development of bilingual language skills and their role in English reading comprehension over time. At Time 1, children’s mean age was 7.62 years old (standard deviation [SD] = 1.03), with 25 children in Grade 1 and 32 children in Grade 3. At Time 2, all children moved up to either Grade 3 or Grade 5. Almost half (45.6%) of the children were girls. Most of them were born in the United States (87.7%). All children were categorized as English language learners by their schools at Time 1. In terms of parent education, 26.3% of the parents had received high school diploma; 3.5% had some college education; 14% had associate or college degree; 5.3% had graduate degree; and 50.9% did not report their educational level. All students were enrolled in Spanish–English DLI programs designed to promote bilingualism and biliteracy. For the instructional program, teachers reported that they followed the 50/50 model regardless of the grade level, such that instruction was evenly divided between Spanish and English. Literacy instruction was provided in both Spanish and English in their dedicated Spanish language arts and English language arts blocks. Math was taught in Spanish whereas science and social studies were taught in English. Instruction aimed to develop balanced proficiency in both languages rather than transition students to English-only instruction.

Measures

All children’s language skills (i.e., PA, MA, vocabulary, and SA) in both Spanish and English and English reading comprehension abilities were measured at two time points. The percentages of correct scores were used for all variables.

PA

The Comprehensive Test of Phonological Processing–Second Edition (CTOPP-2; (Wagner et al., 2013) and the Test of Phonological Processing in Spanish (TOPPS; Francis et al., 2001) were used to evaluate participants’ PA in English and Spanish, respectively. Following CTOPP-2’s guidelines, we used the Elision and Sound Matching subtests for Grade 1 and the Elision and Phoneme Isolation subtests for Grades 3 and 5. Phoneme Isolation and Sound Matching measured similar constructs, and they are appropriate for different age groups. For TOPPS, we used the Elision and Sound Matching subtests for Grade 1 and the Elision and Segmenting Words subtests for Grades 3 and 5. Participants produce words and sounds based on the examiner’s instruction. CTOPP-2 has an internal reliability index of 0.88–0.94 for children between ages 4 and 17 years. TOPPS reported a reliability index of 0.83 for the entire test.

MA

The Early Lexical Morphology Measure (ELMM) in English and its parallel in Spanish (ELMM-S; see Marks et al., 2022) were administered to assess participants’ implicit morphological awareness in both languages. For both tasks, participants were given a word (e.g., quickly) and are requested to complete a sentence using part of that word (e.g., That lion was really ___ [quick]). Both measures included items that targeted compound morphology and derivational morphology, but not inflectional morphology. The reliability coefficient α for ELMM and ELMM-S are 0.93 and 0.95, respectively.

SA

The Morphosyntax subtest of the Bilingual English Spanish Assessment-Middle Extension (BESA-ME; Peña et al., 2010) was used to assess participants’ SA in both English and Spanish. BESA-ME was normed with children aged 7;0–12;6 and is thus appropriate for the participants in the current study (Grades 1, 3, and 5). The test includes both cloze and sentence repetition items. Participants completed cloze items and repeated sentences including basic morphosyntactic constructions, such as possessives, regular and irregular past tense, and relative clauses, by selecting the correct picture from multiple options (e.g., Look, this is a doctor and this is the doctor’s watch. This is a lady, and this is the… Child: ___ [lady’s watch]). The coefficient alpha for this section is 0.877 in Spanish and 0.880 in English.

Vocabulary

The English vocabulary subtest (Test 1: Picture Vocabulary) and the Spanish equivalent (Test 10: Vocabulario sobre dibujos) in the Woodcock-Johnson IV Tests of Oral Language (WJ-IV OL; Schrank et al., 2014b) were used to evaluate receptive vocabulary knowledge. For both subtests, participants were shown a picture of an object and were requested to name the object in the picture. The reliability coefficient ranges from 0.77 to 0.94 for both the English and Spanish subtests.

English reading comprehension

The Passage Comprehension subtest of the Woodcock-Johnson IV Tests of Achievement (WJ-IV ACH; Schrank et al., 2014a) was used to evaluate English reading comprehension skills. The test includes several task types that measure children’s ability to use semantic and syntactic cues to supply missing words within text. The initial items ask the participant to match a printed word with a picture of the corresponding object. Subsequent items require the participant to read short phrases and select the picture that best represents each phrase. The final set of items requires the participant to read short passages and choose the word that completes the sentence based on the context. The reliability for the subtest ranges from 0.87 to 0.93 for the target grade levels.

Procedure

At each time point, children participated in two assessment sessions conducted approximately four or more days apart to reduce potential practice effects. The order of the English and Spanish sessions was counterbalanced across participants and time points. Each session was administered by a trained research assistant who was a native or heritage speaker of the target language and provided all instructions in that language. Children were given several short breaks throughout each session to maintain engagement and minimize fatigue. Research assistants attended a one-day training session led by two project leads. During this session, research assistants learned standardized administration procedures for each assessment and received detailed guidance on scoring and data recording. After the training, they were paired in small groups to practice administering the tests with one another until they reached satisfactory accuracy and consistency. Each research assistant then completed a supervised administration session with one of the project leads and received feedback and a competency score before collecting data independently. To ensure data quality, a dedicated quality-check team at the second author’s lab verified all raw data entries as well as double-scored 20% of assessments to confirm scoring accuracy. The study was approved by the Institutional Review Board at the University of Texas at San Antonio, and all participants provided informed consent before participating in the study.

Analytical approach

In this study, we employ DA to rigorously assess the relative importance of predictor variables within our regression models. DA provides a robust methodological framework for evaluating the contribution of each predictor by systematically examining all possible subset models (Azen & Budescu, 2003). This approach enables us to quantify not only the individual impact of each variable but also their combined effects when interacting with other predictors, thereby offering a comprehensive understanding of their hierarchical structure and interdependencies (Budescu, 1993). DA is well-suited for small to moderate samples, as it focuses on decomposing explained variance rather than estimating regression coefficients directly, thereby reducing sensitivity to multicollinearity and sample fluctuations (Azen & Budescu, 2003; Budescu, 1993). DA evaluates the relative importance of predictors by assessing how much each predictor contributes to the variance explained in the prediction of an outcome (i.e., R²). However, it also accounts for the interdependencies between predictors, considering their unique contribution while controlling for shared variance. A predictor is considered more important if its unique contribution to the explained variance is greater compared to other predictors in the model. There are three levels of dominance that can be achieved within each pair of predictors in DA. One predictor has complete dominance over another predictor if dominance holds across all possible subset models. One predictor is considered conditionally dominant over a competitor predictor if the predictor’s averaged additional contribution within each model size (i.e., the number of predictors in a given model) is greater than that of another predictor. General dominance is achieved when the overall averaged additional contribution of one predictor is greater than another predictor (Azen & Budescu, 2003). These three levels of dominance are associated with each other in a hierarchical fashion, such that complete dominance implies conditional dominance, which, in turn, implies general dominance. By calculating the incremental explanatory power of each variable, DA enhances the robustness and interpretability of our findings, facilitating a nuanced understanding of the underlying data dynamics (Johnson & Lebreton, 2004).

To answer our research questions, we used the same concurrent predictors (vocabulary, PA, MA, and SA) to predict English reading comprehension at the same time point. Grade level was included as a control variable to control for developmental differences between children in different grades. The DA models for two time points were specified separately. The DA models were specified in R using the DominanceAnalysis package (Bustos & Soares, 2024). Given our modest sample size, we conducted a bootstrapped dominance analysis with 1,000 replications to assess the robustness of the DA results. For each replication, the dominance ordering among predictors was recalculated to examine the stability of predictor importance across samples.

Results

Preliminary analysis

Table 1 presents descriptive statistics for all study variables. There were no missing data in the current study. Table 2 displays the correlation matrix at Time 1 and Time 2 separately. At Time 1, all language skills in English and Spanish were positively correlated with English reading comprehension (r ranged from 0.45 to 0.83, all ps < 0.001). In addition, bilingual language measures tended to positively correlate with each other, except that English vocabulary was not correlated with Spanish vocabulary or Spanish SA. Similarly, at Time 2, all language skills in English and Spanish were positively correlated with English reading comprehension (r ranged from 0.34 to 0.75, all ps < 0.01). Moreover, English vocabulary was not correlated with Spanish vocabulary, MA, or SA, with all other bilingual language skills positively correlating with each other.

Table 1.

Descriptive statistics of study variables

Variable	Min.	Max.	M	SD	Skewness	Kurtosis
Time 1
English vocabulary	0.11	0.54	0.37	0.08	−0.40	0.74
English phonological awareness	0	0.97	0.56	0.25	−0.22	−0.73
English morphological awareness	0	0.78	0.28	0.20	0.98	0.51
English syntactic awareness	0.13	0.98	0.68	0.22	−0.81	0.05
Spanish vocabulary	0.23	0.60	0.43	0.10	−0.48	−0.61
Spanish phonological awareness	0	1	0.54	0.34	−0.05	−1.42
Spanish morphological awareness	0.03	0.93	0.51	0.28	−0.06	−1.28
Spanish syntactic awareness	0.05	0.83	0.53	0.24	−0.50	−1.06
English reading comprehension	0.14	0.62	0.36	0.13	−0.12	−1.20
Time 2
English vocabulary	0.28	0.59	0.44	0.07	−0.04	−0.50
English phonological awareness	0.21	0.97	0.72	0.21	−0.71	−0.70
English morphological awareness	0.12	0.88	0.54	0.22	−0.10	−1.40
English syntactic awareness	0.13	1	0.78	0.19	−1.70	3.31
Spanish vocabulary	0.19	0.64	0.45	0.12	−0.36	−0.87
Spanish phonological awareness	0	1	0.74	0.31	−1.05	−0.44
Spanish morphological awareness	0.13	0.98	0.68	0.24	−0.74	−0.36
Spanish syntactic awareness	0.04	0.94	0.61	0.25	−0.59	−0.77
English reading comprehension	0.17	0.75	0.50	0.11	−0.76	0.87

Percentage of correct scores were used for all study variables

Table 2.

Correlation matrix with time 1 below the diagonal and time 2 above the diagonal

	1	2	3	4	5	6	7	8	9
1. English vocabulary	1	.40**	.67***	.66***	.08	.35**	.20	.09	.57***
2. English phonological awareness	.44***	1	.66***	.54***	.30*	.70***	.45***	.37**	.62***
3. English morphological awareness	.52***	.73***	1	.61***	.40**	.62***	.63***	.49***	.75***
4. English syntactic awareness	.66***	.70***	.64***	1	.27*	.59***	.40**	.42**	.62***
5. Spanish vocabulary	.25	.47***	.50***	.40**	1	.53***	.80***	.74***	.34**
6. Spanish phonological awareness	.40**	.82***	.71***	.56***	.49***	1	.65***	.54***	.54***
7. Spanish morphological awareness	.42**	.75***	.75***	.56***	.64***	.78***	1	.80***	.52***
8. Spanish syntactic awareness	.24	.52***	.49***	.43***	.80***	.54***	.71***	1	.38**
9. English reading comprehension	.57***	.83***	.74***	.67***	.46***	.66***	.70***	.45***	1

*p < .05, **p < .01, ***p < .001

Dominance analyses

Column 2 of Table 3 presents the overall average unique contributions of each predictor for English reading comprehension models, from greatest to least predictive contribution. Figure 1 presents the average variance explained in English reading comprehension at Time 1 and Time 2. At Time 1, which includes students in Grades 1 and 3, English PA made the largest average unique contribution (R² = 0.201), followed by English MA (R² = 0.117). Spanish MA made the third largest average unique contribution (R² = 0.091), followed by grade level (R² = 0.090). English SA (R² = 0.087), Spanish PA (R² = 0.077), and English vocabulary (R² = 0.070) made modest contributions. Spanish vocabulary and SA made the least contributions to Time 1 English reading comprehension (both R²s < 0.040). At Time 2, which includes students in Grades 3 and 5, English MA made the largest average unique contribution (R² = 0.176). English PA made the second highest average unique contribution (R² = 0.098), followed by English SA (R² = 0.097). English vocabulary, Spanish MA, and Spanish PA ranked next in the contributions to Time 2 English reading comprehension (all R²s > 0.050). Grade level, and Spanish SA and vocabulary made the least average contributions (all R²s < 0.030). Taken together, these findings suggest a developmental progression of linguistic skills, such that English PA remains a strong contributor in the early stages of reading development, while English MA becomes increasingly dominant as children progress into later elementary grades.

Table 3.

Dominance analysis results

Predictor	R 2
	Average	1 predictor	2 predictors	3 predictors	4 predictors	5 predictors	6 predictors	7 predictors	8 predictors	9 predictors
Predicting Time 1 English reading comprehension
English phonological awareness	0.201	0.689	0.326	0.201	0.145	0.115	0.097	0.086	0.078	0.071
English morphological awareness	0.117	0.554	0.216	0.107	0.061	0.039	0.027	0.020	0.015	0.012
Spanish morphological awareness	0.091	0.496	0.171	0.073	0.035	0.019	0.012	0.008	0.006	0.004
Grade level	0.090	0.378	0.146	0.082	0.055	0.042	0.034	0.028	0.024	0.019
English syntactic awareness	0.087	0.454	0.161	0.076	0.042	0.025	0.015	0.008	0.004	0.000
Spanish phonological awareness	0.077	0.437	0.134	0.049	0.020	0.011	0.009	0.009	0.012	0.014
English vocabulary	0.070	0.320	0.113	0.059	0.038	0.028	0.022	0.018	0.016	0.014
Spanish vocabulary	0.034	0.213	0.042	0.013	0.007	0.006	0.006	0.006	0.007	0.008
Spanish syntactic awareness	0.030	0.201	0.030	0.006	0.003	0.004	0.005	0.006	0.007	0.008
Predicting Time 2 English reading comprehension
English morphological awareness	0.176	0.568	0.324	0.210	0.146	0.106	0.079	0.061	0.049	0.041
English phonological awareness	0.098	0.385	0.185	0.103	0.064	0.043	0.032	0.026	0.022	0.019
English syntactic awareness	0.097	0.381	0.182	0.102	0.064	0.043	0.032	0.026	0.023	0.023
English vocabulary	0.082	0.321	0.170	0.103	0.064	0.039	0.023	0.012	0.005	0.000
Spanish morphological awareness	0.058	0.269	0.108	0.054	0.031	0.021	0.015	0.011	0.009	0.007
Spanish phonological awareness	0.053	0.288	0.107	0.043	0.017	0.007	0.003	0.002	0.003	0.005
Grade level	0.029	0.134	0.058	0.032	0.019	0.010	0.005	0.003	0.001	0.001
Spanish syntactic awareness	0.024	0.143	0.036	0.011	0.005	0.004	0.004	0.005	0.005	0.006
Spanish vocabulary	0.020	0.118	0.034	0.013	0.006	0.003	0.002	0.002	0.001	0.001

Fig. 1.

Average unique variance explained by bilingual language skills in English reading comprehension at Time 1 and Time 2. Note. PA = phonological awareness, MA = morphological awareness, SA = syntactic awareness

Table 4 reports the complete, conditional, and general dominance relations among all pairs of predictors when predicting English reading comprehension at Time 1 and Time 2. Complete dominance was first established by comparing the contributions of each predictor for all subset models (Fig. S1 in supplementary material). At Time 1, in every subset model, the variances explained by English PA were larger than those for all other eight predictors; therefore, English PA completely dominated other eight predictors in the prediction of English reading comprehension. Further, English MA and grade level also completely dominated Spanish vocabulary and SA at Time 1. At Time 2, English MA showed complete dominance over all other eight predictors in the prediction of English reading comprehension. English PA completely dominated Spanish vocabulary, PA, SA, and grade level. Further, Spanish MA also showed complete dominance over Spanish vocabulary.

Table 4.

Dominance relations in the prediction of English reading comprehension at Time 1 and Time 2

Dominant variable	Complete dominance	Conditional dominance	General dominance
Predicting Time 1 English reading comprehension
English phonological awareness (EPA)	EV, EMA, ESA, SV, SPA, SMA, SSA, G
English morphological awareness (EMA)	SV, SSA	ESA, SMA	EV, SPA, G
Spanish morphological awareness (SMA)			EV, ESA, SV, SPA, SSA, G
Grade level (G)	SV, SSA	EV	ESA, SPA
English syntactic awareness (ESA)			EV, SV, SPA, SSA
Spanish phonological awareness (SPA)		SV, SSA	EV
English vocabulary (EV)		SV, SSA
Spanish vocabulary (SV)			SSA
Spanish syntactic awareness (SSA)
Predicting Time 2 English reading comprehension
English morphological awareness (EMA)	EV, EPA, ESA, SV, SPA, SMA, SSA, G
English phonological awareness (EPA)	SV, SPA, SSA, G	SMA	EV, ESA
English syntactic awareness (ESA)		SV, SPA, SMA, SSA, G	EV
English vocabulary (EV)			SV, SPA, SMA, SSA, G
Spanish morphological awareness (SMA)	SV	SSA, G	SPA
Spanish phonological awareness (SPA)		SV	SSA, G
Grade level (G)		SV	SSA
Spanish syntactic awareness (SSA)			SV
Spanish vocabulary (SV)

Second, we established conditional dominance for those predictors that did not establish complete dominance over other predictors. We compared the average unique contributions of each predictor at each model size (e.g., 5-predictor models) in the prediction of English reading comprehension at Time 1 and Time 2. Table 3 (columns 3–9) and Fig. S2 in supplementary material presents conditional dominance relations among predictors. The figure shows that the average unique contributions of all predictors decreased as a function of increasing the number of predictors in the models at both Time 1 and Time 2. At Time 1, English MA had larger average unique contributions at each model size relative to English SA and Spanish MA, indicating that English MA conditionally dominated these two predictors. Further, grade level showed conditional dominance over English vocabulary; Spanish PA and English vocabulary both showed conditional dominance over Spanish vocabulary and Spanish SA. At Time 2, English PA showed conditional dominance over Spanish MA. English SA conditionally dominated Spanish vocabulary, PA, MA, SA, and grade level. Spanish MA further showed conditional dominance over Spanish SA and grade level. Finally, Spanish PA and grade level showed conditional dominance over Spanish vocabulary.

Lastly, we examined general dominance among predictors that did not establish conditional dominance relations; we compared the overall average unique contributions of each predictor across all subset models without considering the model size (Table 3, column 2; Fig. 1). At Time 1, English MA generally dominated English vocabulary, Spanish PA, and grade level. Spanish MA generally dominated English vocabulary, English SA, Spanish vocabulary, Spanish PA, Spanish SA, and grade level. Grade level further generally dominated English SA and Spanish PA. English SA showed general dominance over English and Spanish vocabulary, and Spanish PA and SA. Spanish PA further generally dominated English vocabulary. Finally, Spanish vocabulary generally dominated Spanish SA. At Time 2, English PA generally dominated English vocabulary and SA. English SA generally dominated English vocabulary, which further showed general dominance over all four Spanish language skills and grade level. Spanish MA generally dominated Spanish PA. Spanish PA showed general dominance over Spanish SA and grade level. Grade level showed general dominance over Spanish SA, which further generally dominated Spanish vocabulary.

We then performed bootstrapped dominance analyses (1,000 replications) to examine the stability of the dominance ordering across resamples. We report the proportion of replications in which each predictor maintained its dominance rank, providing a measure of robustness and reliability of the findings (see Table 5). Results showed that on average, the replication consistency of general dominance (i.e., overall relative importance) was 83.2% for Time 1 and 83.6% for Time 2 DA models. For models at each time point, the first-ranking predictor was the most stable predictor across predictors, with bootstrapped dominance stability proportions ranging from 95.2% to 100% and 86.6% to 100% for Time 1 and Time 2 models, respectively. Overall, the bootstrapped dominance values indicate that the rank ordering of predictors was generally robust, despite the modest sample size (n = 57).

Table 5.

Replication consistency of bootstrapped dominance analysis with 1000 replications

Focal predictor	Comparison predictor	Replication consistency (%)
Time 1
English phonological awareness	English morphological awareness	95.2
	Spanish morphological awareness	99.1
	Grade level	95.7
	English syntactic awareness	97.0
	Spanish phonological awareness	100.0
	English vocabulary	99.5
	Spanish vocabulary	100.0
	Spanish syntactic awareness	100.0
English morphological awareness	Spanish morphological awareness	69.7
	Grade level	69.1
	English syntactic awareness	72.5
	Spanish phonological awareness	88.8
	English vocabulary	85.5
	Spanish vocabulary	98.9
	Spanish syntactic awareness	99.8
Spanish morphological awareness	Grade level	53.9
	English syntactic awareness	54.3
	Spanish phonological awareness	69.8
	English vocabulary	70.5
	Spanish vocabulary	96.0
	Spanish syntactic awareness	99.8
Grade level	English syntactic awareness	47.9
	Spanish phonological awareness	58.6
	English vocabulary	66.9
	Spanish vocabulary	89.2
	Spanish syntactic awareness	93.9
English syntactic awareness	Spanish phonological awareness	59.6
	English vocabulary	79.4
	Spanish vocabulary	95.4
	Spanish syntactic awareness	98.9
Spanish phonological awareness	English vocabulary	60.7
	Spanish vocabulary	97.0
	Spanish syntactic awareness	99.6
English vocabulary	Spanish vocabulary	85.7
	Spanish syntactic awareness	89.6
Spanish vocabulary	Spanish syntactic awareness	56.0
Time 2
English morphological awareness	English phonological awareness	90.9
	English syntactic awareness	86.6
	English vocabulary	94.0
	Spanish morphological awareness	99.0
	Spanish phonological awareness	98.6
	Grade level	100.0
	Spanish syntactic awareness	100.0
	Spanish vocabulary	100.0
English phonological awareness	English syntactic awareness	50.3
	English vocabulary	57.1
	Spanish morphological awareness	80.0
	Spanish phonological awareness	81.7
	Grade level	93.3
	Spanish syntactic awareness	98.2
	Spanish vocabulary	98.5
English syntactic awareness	English vocabulary	62.7
	Spanish morphological awareness	69.4
	Spanish phonological awareness	77.4
	Grade level	91.9
	Spanish syntactic awareness	94.6
	Spanish vocabulary	93.3
English vocabulary	Spanish morphological awareness	64.3
	Spanish phonological awareness	70.0
	Grade level	94.3
	Spanish syntactic awareness	93.2
	Spanish vocabulary	93.4
Spanish morphological awareness	Spanish phonological awareness	51.1
	Grade level	79.6
	Spanish syntactic awareness	93.5
	Spanish vocabulary	97.6
Spanish phonological awareness	Grade level	79.6
	Spanish syntactic awareness	93.6
	Spanish vocabulary	92.3
Grade level	Spanish syntactic awareness	57.5
	Spanish vocabulary	63.8
Spanish syntactic awareness	Spanish vocabulary	67.2

Discussion

In the current study, we investigated the relative importance of bilingual language skills in Spanish–English EB children’s English reading comprehension at two time points. Using a DA approach, our findings revealed a dynamic interplay between various linguistic skills, both within and across languages, as children develop their English reading abilities over time. Specifically, at Time 1, among the included bilingual language skills, English PA was the strongest predictor of English reading comprehension, followed by English MA. However, by Time 2, English MA emerged as the most robust predictor, surpassing the contribution of English PA. Spanish language skills, particularly MA and PA, showed notable contributions at Time 1 but had decreasing importance by Time 2. The results highlight the changing nature of the relations between these various linguistic components and English reading comprehension over time and provide insights into the complex process of reading development for Spanish–English EBs.

While with the emergence of developmental patterns, it is important to note that other factors that were not included in the current study, such as socioeconomic status, home environment, school instructions, and cognitive development, could also play significant roles. In addition, the two-year interval between Time 1 and Time 2 coincides with the COVID-19 pandemic, a period that brought major changes to children’s educational experiences. During the pandemic (2020–2021 academic year), our participating schools implemented remote instruction. These shifts likely affected children’s learning experiences and introduced variability in their developmental trajectories. Nevertheless, understanding such developmental relations has significant implications for assessing risk for reading difficulties among Spanish–English EB students, as well as for developing tailored instruction and targeted interventions to meet their specific needs.

English language skills as predictors of English reading comprehension over time

The most striking finding of our study is the shifting pattern of relative dominance of language skills predictors of English reading comprehension over time. At Time 1, when children were in Grades 1 or 3, English PA was the strongest predictor of English reading comprehension, followed by English MA. Such initial dominance of English PA aligns with extensive research highlighting its fundamental role in early reading development (Kirby et al., 2003). The strong predictive power of PA at this stage likely reflects its critical role in mapping sounds onto letters and supporting early decoding skills (Ziegler & Goswami, 2005), as well as children’s heavy reliance on phonological processing in early reading stages for successful comprehension of texts. Our finding also indicates that PA may be particularly important as Spanish–English bilingual children begin to develop reading proficiency in their second language. However, by Time 2, although English PA’s contribution was still substantial, it decreased to become the second strongest predictor. This shift suggests a developmental progression where the relative importance of PA diminishes as children engage with more complex texts, which may require higher-level language skills. This developmental pattern aligns with research suggesting that PA may become less central as children develop automaticity in word recognition (Ehri, 2005; Kirby et al., 2003).

By Time 2, when children were in Grades 3 or 5, we observed a notable shift in the dominance pattern, with English MA surpassing PA as the strongest predictor of English reading comprehension. This shift represents a crucial developmental transition in reading comprehension processes, such that readers tend to rely heavily on their ability to reflect upon and manipulate morphemes with increased demand for understanding morphologically complex words as texts becomes more complex (Ke et al., 2021). As such, children’s ability to analyze and manipulate morphological structures become increasingly important for reading comprehension, particularly in English, a language with a rich morphological structure (Carlisle, 2003; Kuo & Anderson, 2006). Our finding also extends previous research and demonstrates the heightened importance of MA in bilingual contexts (Kieffer & Lesaux, 2008). These results also align with theories of reading development, including Chall’s (1983) stages of reading, the SVR (Gough & Tunmer, 1986), and the Lexical Quality Hypothesis (Perfetti, 2007), highlighting how the relative importance of different language skills shifts as children build reading competence.

English SA demonstrated an increasing contribution to English reading comprehension from Time 1 to Time 2, moving from fourth to third position among language skills predictors in terms of predictive power. When comprehending connected text with multiple sentences, strong SA allows readers to comprehend text through parsing complex sentences into digestible components (Graesser et al., 2011). As such, it is not surprising that SA becomes increasingly important as texts become more complex and require greater syntactic parsing abilities (Cain, 2007). This also indicates that the ability to process complex sentence structures tends to be more crucial for text comprehension in later years for Spanish–English EB children. Our result is consistent with previous studies of bilingual children (Bellocchi et al., 2017; Sohail et al., 2022; Xie & Yeung, 2022), demonstrating that L2 SA contributed to bilingual children’s L2 reading comprehension. The enhanced role of SA over time also aligns with theories emphasizing its importance in reading comprehension (Cain, 2007), which is particularly relevant for bilingual learners who must navigate potentially different syntactic structures across languages.

Extant research highlights the importance of vocabulary in the development of children’s English reading comprehension (Ouellette, 2006; Quinn et al., 2015). For bilingual children, in particular, their English L2 vocabulary showed more critical influences in their English reading comprehension (Babayiğit & Shapiro, 2020; Shen & Coker, 2023). In our study, considering all other seven bilingual language skills, the contribution of English vocabulary to reading comprehension showed notable growth, advancing from sixth position at Time 1 to fourth position at Time 2 among language skill predictors. This pattern supports the lexical quality hypothesis (Perfetti, 2007) and suggests that as texts become more sophisticated, vocabulary knowledge plays an increasingly important role in reading comprehension. This strengthening relation between English vocabulary and English reading comprehension over time may reflect both the increasingly sophisticated texts encountered by readers and the increasing importance of lexical knowledge in supporting higher-level comprehension processes.

One major contribution of our study is the investigation of relative importance of different language skills while considering all combinations of predictors in our DA models. Across all four English language skills included in this study, the relative importance pattern in predicting English reading comprehension showed a shift between PA and MA but stayed consistent for SA and vocabulary. As children progress through elementary school years, English MA gained more importance over English PA, followed by the contributions of SA and vocabulary. This may indicate that across primary years, children’s reading comprehension development follows a predictable progression from reliance on sound-based decoding skills to increased utilization of morphological knowledge and that SA and vocabulary may serve as foundational supports that work in concert with the more dynamic PA-MA relation. This pattern aligns with theory proposing that SA and vocabulary provide the scaffolding necessary for both phonological and morphological processing in reading comprehension (Perfetti & Stafura, 2014). While prior studies have often examined these predictors in isolation or in limited combinations (Kirby et al., 2012), our DA models provide a more comprehensive understanding of how these skills compete in explaining variance in English reading comprehension. These findings also provide theoretical insight into how the developmental trajectory of bilingual language skills may differ in bilingual environments that actively promote balanced L1 and L2 development (such as the DLI programs in the current study), compared to English-dominant contexts.

Cross-language transfer: Spanish language skills as predictors of English reading comprehension

Our findings further revealed complex patterns in the cross-language contributions of Spanish language skills to English reading comprehension that changed substantially over time. Specifically, at Time 1, Spanish MA and PA showed notable contributions, ranking third and fifth among language skills predictors, respectively. This finding suggests a cross-linguistic relation between MA and PA and L2 reading, supporting theories positing the interdependent nature of bilingual language development (Cummins, 1979, 1991; Koda, 2007). This also indicates important roles of L1 skills in early L2 reading for Spanish–English EBs. One interesting finding is that at Time 1, English PA explained the largest variance in English reading comprehension, followed by English MA. However, the contribution of Spanish MA surpassed that of Spanish PA. Previous research has shown that PA is more likely to be language universal (Chung et al., 2019), such that PA in one language can easily transfer to another language to support comprehension (Tong et al., 2023). Here, we observed the more critical role of Spanish MA, relative to Spanish PA, in English reading comprehension; the reason may be twofold.

First, unlike English, which has a deep orthography that encodes both phonological and morphological information, Spanish has a shallow orthography with a nearly perfect correspondence between phonemes and letters (i.e., phonological transparency). However, Spanish is a language with a richer morphology. Children thus need to develop a sensitivity to morphemes in Spanish because of the complex morphological system. For Spanish–English bilingual children, understanding English texts, which often consist of “deep” English words, may involve relying more heavily on their morphological sensitivity developed in Spanish. This aligns with the interactive transfer framework (Chung et al., 2019), positing that transfer tends to occur from the more morphologically complex language to the less complex language. Indeed, Ramirez et al. (2010) also found that Spanish MA contributed to English word reading over and above English MA. Further, Ramírez et al. (2013) showed that Spanish derivational MA contributed significant variance to English reading comprehension, with a similar magnitude of the effect of English MA. Second, English and Spanish share many structural similarities in derivational morphology. The common morphological principles across English and Spanish, such as the semantic and syntactic properties of morphemes and the morphological structure of complex words may contribute to the cross-language associations between Spanish MA and English reading comprehension.

However, by Time 2, all Spanish skills showed diminished contributions, ranking after all English skills. This developmental shift toward greater language-specific skills in English reading comprehension processes may indicate a natural progression as learners develop stronger L2-specific skills and rely less on cross-linguistic transfer. This finding aligns with the fact that Spanish–English bilinguals in the current study became more English dominant at Time 2. Children’s Spanish proficiency did not seem to grow as much as their English proficiency from Time 1 to Time 2. For example, the mean score of Spanish vocabulary was 0.43 at Time 1 and 0.45 at Time 2. Research has shown that the strength of the existing L1 oral proficiency foundation moderates cross-language transfer, such that cross-language transfer is more likely to occur when bilinguals maintain high L1 proficiency (Goodrich et al., 2014). Since children in the current study appear to develop higher proficiency in their L2 by Time 2, it is not surprising that they leverage their abilities in L2 to carry out L2 comprehension tasks.

Spanish vocabulary and SA showed consistently lower contributions across both time points, suggesting that these aspects of language knowledge may be more language-specific than MA and PA. This finding aligns with research suggesting that certain linguistic skills may be more amenable to cross-linguistic transfer than others (Proctor et al., 2010) and highlights the complexity of the relations between L1 and L2 skills in bilingual reading development. Indeed, the interdependence continuum model (Proctor et al., 2010) suggests that the strength of the cross-language transfer is dependent on the language pairs as well as the specific linguistic skills in question. Our finding is aligned with previous studies suggesting limited cross-linguistic transfer from L1 vocabulary (Leider et al., 2013; Proctor et al., 2017) and L1 SA (Proctor et al., 2012) to L2 reading comprehension. Further, research showed that within-language skills are more important for predicting English reading comprehension, relative to readers’ cross-language skills (Goodrich et al., 2023; Nakamoto et al., 2008). In fact, among Spanish–English bilinguals, Proctor et al. (2017) found a significant cross-language association between Spanish SA and English reading comprehension, whereas Proctor et al. (2012) failed to find such an association. The inconsistency is largely due to the fact that the later study included English language controls, suggesting the more critical role of within-language skills.

One interesting finding in the current study is that relative importance shifts over time between Spanish vocabulary and SA, such that Spanish vocabulary ranked more important than Spanish SA at Time 1, whereas by Time 2, Spanish SA ranked more important. At early stages, Spanish vocabulary may facilitate English reading comprehension through cognate awareness (Ramírez et al., 2013). Cognates are words in different languages that are often similar in pronunciation, spelling, and meaning, which are very common in English and Spanish. By Time 2, however, the increased importance of Spanish SA may reflect the growing complexity of academic text structures that demand syntactic processing in the English reading comprehension development. Additionally, this shift might reflect limitations in our Spanish vocabulary measure used in the current study, which primarily assesses children’s ability to identify objects based on presented pictures. As children progress in school, their need for sophisticated academic language processing might make SA relatively more important than the vocabulary knowledge of objects captured by this measure (Snow & Uccelli, 2009).

One important consideration in interpreting our findings is the potential influence of differences in age and prior exposure to English between participants at Time 1 and Time 2. Given that children were assessed in Grade 1 and Grade 3 at Time 1 and subsequently followed to Grade 3 and Grade 5, differences in developmental trajectories may have contributed to variations in the relative importance of language skills across time points. For instance, younger children may rely more on foundational oral language and PA skills, while older children with longer L2 exposure might benefit more from advanced literacy skills such as MA and text comprehension strategies. These developmental shifts could partially explain changes in predictor dominance observed in our analyses. Future research with a sample of similar age using a longitudinal approach could help disentangle the role of age-related developmental changes from the broader trends in language and reading development identified in this study.

Implications

The findings of the current study have several implications for both assessment and instruction for Spanish–English EB children. However, these implications should be interpreted cautiously since DA does not imply causal effects or statistical significance. A predictor can explain relatively little variance and still have a statistically significant conditional effect. The following implications address only relative importance as quantified by DA. First, when assessing risk for reading difficulty among Spanish–English bilinguals, our findings underscore the importance of using assessment frameworks that are both bilingual and developmentally sensitive. Rather than relying on a single linguistic indicator, screening practices should consider the relative importance of different linguistic skills (including EB children’s L1 skills) at different stages of development. Screening and assessment should also take into account EB children’s language and literacy experiences, such as their instructional program and home literacy environment. Importantly, although the developmental progression from PA to MA has been documented previously, our findings suggest that the diagnostic utility of these skills shifts in ways that may be especially relevant for bilingual children educated in DLI settings. Second, our results have implications for instructional decision-making in bilingual contexts. While early reading instruction (early elementary grades) often emphasizes PA-based decoding strategies (National Reading Panel, 2000), our findings suggest that a gradual and intentional increase in explicit morphological instruction may be particularly beneficial as EB children progress into middle and upper elementary grades. However, the maintained importance of all four English language skills indicates that an integrative approach to literacy instruction, rather than a complete shift from one skill to another, would be beneficial. Furthermore, Spanish language skills, particularly MA and PA, should be maintained and developed early and strategically leveraged in bilingual programs to support English reading comprehension.

Limitations and future directions

Several limitations need to be acknowledged. First, children in the current study were from mixed grade levels at each time point (i.e., Grades 1 & 3 at Time 1 and Grades 3 & 5 at Time 2). This may obscure the developmental pattern of the relative importance of various bilingual language skills. Although the same cohort of children was followed over time, future studies would benefit from grade-specific or cohort-stratified analyses to more precisely examine how the dominance of predictors such as PA and MA shifts across distinct developmental stages. Larger samples would allow dominance analyses to be conducted separately by grade, strengthening inferences about developmental change. Second, our sample size is relatively small and focused on Spanish–English bilinguals, which limits the generalizability of the findings to other bilingual populations. Expanding this research to include more bilinguals with different language pairs could help distinguish between language-specific and universal aspects of English reading development. Furthermore, our focus on English reading comprehension limits our understanding of the role of bilingual language skills in children’s heritage language (i.e., Spanish) reading development, and future research should explore a broader range of literacy outcomes in both languages. Fourth, there are limitations in terms of the instruments used in the study. The MA measure is a new researcher-developed assessment that has been validated for children aged 5–9 years old (Marks et al., 2023), but our sample at Time 2 consisted of some children older than the age range, even though we didn’t observe a ceiling effect among Grade 5 students and this measure has been used with children across elementary school age (Marks et al., 2024; Sun et al., 2023). In addition, the SA measure is an experimental test that consists of two subtests: Morphosyntax and Semantics (Peña et al., 2020). Because of time and resource constraints, we were only able to administer the Morphosyntax test. The SA results may not be representative of participants’ full SA competencies, particularly for the fifth graders at Time 2. Fifth, Time 1 data were collected in Fall 2019 before the pandemic and school shutdowns, whereas Time 2 data were collected in Fall 2021 post-pandemic. Although at both time points, we used the same tests and testing modality (i.e., in person), the effect of the pandemic may have impacted children’s bilingual language and English reading outcomes at Time 2. Additionally, although we leveraged two time points, the analyses were conducted at each time point, making them more cross-sectional, despite following the same cohort of children. Future studies may consider using a full longitudinal design to further expand our findings. Finally, while the present study employed dominance analysis to address multicollinearity among bilingual language skills, the correlational design limits causal interpretation. Future research could build on this methodological approach by using longitudinal cross-lagged models to examine bidirectional and time-ordered relations among dual language skills and reading comprehension. Experimental and quasi-experimental intervention studies are also needed to test whether instructional approaches targeting MA yield incremental benefits beyond PA at particular developmental stages, especially for bilingual learners in upper elementary grades. Embedding dominance analysis within intervention studies may further illuminate how instructional focus alters the relative contribution of correlated linguistic skills to reading comprehension outcomes.

Conclusion

This study provides valuable insights into the dynamic nature of English reading development predictors in Spanish–English EB children. Our findings reveal that the relative importance of various linguistic skills, both within and across languages, changes as children develop their English reading abilities. English PA and Spanish MA and PA skills showed initial importance; however, as children progress, English MA becomes the strongest predictor, while Spanish language skills showed generally decreasing contributions. This pattern suggests a complex interplay between L1 and L2 skills in the development of English reading comprehension, underscoring the potential benefits of leveraging children’s L1 skills in early L2 reading instruction. In addition, the increasing importance of English MA, SA, along with vocabulary, highlights the need for a comprehensive approach to literacy instruction that addresses multiple linguistic domains. These findings have important implications for educational practices, suggesting dynamic and adaptive approaches to bilingual literacy assessment, instruction, and interventions. By recognizing the dynamic nature of reading comprehension predictors, educators and researchers can work toward developing more effective strategies to support bilingual children’s reading success.

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Declarations

Conflict of interest

The authors declare that they have no conflict of interest.