Dual Language Skills and Socioemotional Development for Spanish‐English Dual Language Learners in Head Start

ABSTRACT

The current study explored whether the associations between language and socioemotional (SE) development differ across various facets of language and SE domains, as well as across dual language learners (DLLs) with and without baseline English proficiency at Head Start entry. Cross‐lagged panel analyses were conducted on 471 Hispanic Spanish‐English DLLs (M _age = 41 months old; 49% girls) in Head Start programs, who were followed for two years from the program entry to kindergarten. Results supported unidirectional effects of prior Spanish receptive vocabulary and conceptual expressive vocabulary on subsequent SE development for the full sample. When examining the two DLL groups separately based on English proficiency at program entry, we found that the full‐sample results were primarily driven by Emergent Bilinguals, that is, DLLs developing English proficiency. For Bilinguals, that is, DLLs with stronger English proficiency, a reciprocal, bidirectional association emerged between Spanish receptive vocabulary and SE skills. Our findings suggest the significant role of DLLs’ home language (Spanish) in their SE development, and vice versa, indicating possible benefits of intentionally integrating heritage language supports into SE‐focused curricula and classroom.

Summary

• Spanish receptive and conceptual expressive vocabulary predicted DLLs’ later SE growth

• Language‐SE relations differed by DLLs’ English proficiency level at Head Start entry

• Emergent Bilinguals showed unidirectional effects from language to SE skills

• Bilinguals showed reciprocal relations between Spanish receptive vocabulary and SE

1. Introduction

One third of young children aged 0–5 in the United States (U.S.) live in a household where a language other than English is spoken by a parent or caregiver, and among these, 59% speak Spanish (Park and Pompa, 2021). This large population of dual language learners (DLLs) includes young children who are acquiring proficiency in two or more languages, either simultaneously or sequentially. When provided with appropriate support, DLLs have a high potential for achieving bilingualism and biliteracy, which are associated with cognitive benefits that support their academic success (National Academies of Sciences, Engineering, and Medicine, 2017). Further, these children often bring unique and valuable linguistic and cultural assets to their schools and community (Migration Policy Institute, 2021).

Given the increasing linguistic diversity of early childhood settings, early childhood education (ECE) researchers have devoted great attention to understanding their development and supporting their learning. However, within the ECE literature on DLLs, research has emphasized language acquisition and academic achievement more than socioemotional (SE) development, despite its critical role in children's overall development, including academic achievement (Choi et al., 2016; Guo et al., 2023; Sorrenti et al., 2024). Children learning multiple languages need to navigate between two sets of linguistic and cultural expectations that may have different social and behavior goals (Halle et al., 2014). DLLs often experience unique SE development influenced by their linguistic backgrounds (Halle et al., 2014; Huang et al., 2024). For instance, Huang et al. found that societal language proficiency (i.e., English) predicted SE skill development among DLLs speaking both English and their home language at home but not for DLLs speaking only their home language at home. As such, it is important to further our understanding of the relations between DLLs’ SE development and their language competence as they are essentially intertwined with each other (Huang et al., 2024). Such understanding can facilitate better support to meet DLLs’ unique needs and serve as a way to make developmental science more representative and inclusive.

This study examined the cross‐lagged associations between DLLs' language development (both receptive and expressive) and their SE development over approximately two years in preschool. Of particular interest is whether and how the potential reciprocal associations between language and SE development differs among DLLs based on baseline English proficiency at program entry. The context of this study is Head Start (HS), a federally funded comprehensive ECE program for children from low‐income families. HS is designed to promote school readiness through support for children's cognitive, language, and SE development, as well as family well‐being. Given its commitment for culturally and linguistically responsive services, and the fact that it serves over one‐third of all enrolled children who are DLLs (Early Childhood Learning & Knowledge Center, 2024), HS continues to stand as a significant ECE setting for DLLs. Thus, HS provides a developmentally and policy‐relevant context for examining how linguistically diverse children's English and Spanish language skills are associated with SE development. Understanding these associations in HS can inform efforts to provide linguistically responsive support that address the unique needs of children served by the program.

1.1. Language and SE Development

Preschool years are a critical period for language development. Language skills provide children with means to understand people's intentions and needs, as well as allow children to express their needs and better communicate and solve social conflicts verbally (Schultz et al., 2001). Language skills involve receptive language, the ability to understand words, and expressive language, the ability to express or produce words to convey meaning. Both receptive and expressive language skills are essential for positive social interaction, enabling children to both understand and appropriately respond to others’ requests and comments (Longoria et al., 2009).

SE development involves many important aspects of social and emotional skills that help children control their emotions, navigate social situations, and engage in self‐regulated learning. During early childhood, SE skills focus on engaging positively with others and effectively regulating emotions and behaviors, especially as children begin navigating more complex peer interactions (Denham and Brown, 2010; Parker and Gottman, 1989). Within ECE classroom context, children's social skills serve important roles in interacting and establishing positive relationships with teachers and peers (Denham and Brown, 2010; Halle et al., 2014). On the other hand, problem behaviors, including internalizing (e.g., anxiety, worry, sadness, and shyness or social withdrawal) and externalizing problems (e.g., hostile/aggressive behavior and hyperactivity; Halle and Darling‐Churchill, 2016) often predict fewer opportunities for meaningful interactions in the classroom (Bulotsky‐Shearer et al., 2011; Sutherland et al., 2018).

In the broader developmental literature, decades of research have documented the interconnected nature of language and SE skills (Barnett et al., 2012; Huang et al., 2024; Petersen and LeBeau, 2021). While the current research base remains relatively limited to non‐DLL populations, a smaller but growing body of research has documented similar language‐SE connections among DLL children. For example, studies with school‐aged Chinese and Spanish‐English DLLs in the U.S. have shown that both home and English language skills relate to SE competence reported by teachers and parents, with stronger associations typically observed for English (Chen et al., 2014; Collins et al., 2011). Additionally, Sun et al. (2021) studied Singaporean bilingual preschoolers who spoke Mandarin, Malay, or Tamil and found that both English and home language vocabulary were linked to fewer SE difficulties, but only English vocabulary (the language of instruction) was related to better prosocial skills in classrooms. Studies separately examining receptive and expressive language skills in their associations with SE skills are limited; however, existing evidence shows that children's expressive language skills in their home language, but not in English, are indirectly linked to their parent‐reported SE skills, including social strengths and problem behaviors (Chung et al., 2019). Together, these studies suggest that understanding the interconnections between language and SE skills requires considering multiple factors, including language (e.g., home language versus English), the context in which SE skills are assessed (e.g., home by parents versus school by teachers), and the type of language skills (e.g., receptive versus expressive).

For DLLs, language proficiency often develops along diverse trajectories across their two languages, reflecting differences in exposure and use across settings such as home and school (Ribot et al., 2018). Children's language development within these contexts can shape their social relationships, with the school language more closely linked to interactions with teachers and peers, and the home language often supporting socialization within families and communities (Collins et al., 2011).

1.2. Reciprocal Development Between Language and SE skills

Children develop early language and SE skills concurrently, with each domain mutually influencing and supporting the growth of the other. Language skills serve as a crucial foundation for communication and emotional and behavioral regulations, which in turn serve as a tool for facilitating social interactions and effective conflict management (Barnett et al., 2012; Halle et al., 2014). On the other hand, SE skills, such as high social skills and low problem behaviors, are an important prerequisite for engaging in and maintaining social interactions, often providing meaningful opportunities that foster language acquisition (Bruner, 1985; Erdemir and Brutt‐Griffler, 2022; Huang et al., 2024; Rose‐Krasnor, 1997; Sussman et al., 2023).

This view of mutual development of language and SE skills aligns with both social‐cultural and socio‐pragmatic views on language development and social learning (Tomasello, 2003; Vygotsky, 1978). From these perspectives, language functions as a primary cultural tool through which children engage in social interaction, internalize social norms, and regulate their behavior, thereby shaping SE development. In this sense, language not only influences other areas of development through social interactions, and its acquisition itself is shaped by children's participation in these social contexts.

Although theoretical assumptions support a reciprocal relation between language and SE development, empirical studies have not reached a clear consensus regarding the direction of this relation for DLL children. Studies investigating the longitudinal, reciprocal development of DLLs’ language and SE skills have yielded mixed findings. Some highlighted the effect of prior SE skills on later language skills (Ertanir et al., 2021; Hagan‐Burke et al., 2016; Huang et al., 2024 [for DLLs only speaking heritage language at home]), others revealed the effect of prior language skills on subsequent SE development (Huang et al., 2024 [for DLLs speaking both English and heritage language at home]), and some supported a bidirectional relation between the two developmental domains (Bohlmann et al., 2015). These mixed findings may reflect differences in how language and SE skills are operationalized, including whether receptive versus expressive vocabulary and specific SE indicators (e.g., self‐regulation, social skills, behavior problems) are examined. In addition, variation in DLLs’ language exposure contexts, developmental timing, and the language of assessment may affect the direction and strength of language–SE associations, contributing to inconsistent results across studies.

For instance, Ertanir et al. (2021) conducted a study testing reciprocal relations between receptive and expressive vocabulary and SE skills across three years among 216 DLLs aged 3–6 years old in Germany. Children in this study were exposed to at least one language other than German at home. The authors found a unidirectional relation where SE skills, operationalized as a composite including both strengths and behavior concerns, predicted later growth in both German receptive and expressive vocabulary. However, Bohlmann et al. (2015) followed U.S. preschool‐aged Spanish‐English DLLs across three time points and revealed a bidirectional associations between English expressive vocabulary and self‐regulation, even after controlling for child age, gender, ethnicity, income‐to‐needs ratio, and Spanish expressive vocabulary. Including both English receptive and expressive vocabulary, Clayton et al. (2022) showed that, among U.S. preschool‐aged Spanish‐English DLLs, beginning‐of‐year English receptive vocabulary, but not expressive vocabulary, predicted higher overall social skills and fewer inattention problems in the middle of the year, even after controlling for initial English and Spanish language proficiency. Further, controlling for initial English receptive and expressive vocabulary, children with higher mid‐year social skills had higher English expressive vocabulary at the end of the year, whereas children with fewer mid‐year inattention problems had higher English receptive vocabulary.

Given the variation in the associations between receptive and expressive vocabulary and different indicators of SE skills, it is important to examine the language‐SE relations in a comprehensive way, such as including both receptive and expressive vocabulary and a variety of SE skills indicators. Furthermore, although some previous studies have controlled for Spanish language skills (e.g., Bohlmann et al., 2015; Clayton et al., 2022), they did not explicitly examine how skills in different languages may have different associations with SE competence, leaving room for further examination.

1.3. Reciprocal Development of Language and SE Skills: Differences Across DLL Groups

DLLs, or children from homes where a language other than English is spoken, are not a homogenous group with uniform language proficiency; rather, they exhibit a wide range of abilities in languages they are learning (Choi et al., 2024; Rumper et al., 2023; Su et al., 2024). Extant research has revealed that English language proficiency is closely linked to DLLs’ social behaviors in early childhood classroom, especially in context where English holds dominant institutional and social status (e.g., U.S.): DLLs who possess proficient English language skills are inclined to participate in effective communication, form meaningful peer relations, and cultivate close relationships with their teachers (Halle et al., 2014), while DLLs with low English proficiency tended to be perceived as shy and may encounter peer rejection (Sung, 2014). In contrast, in bilingual contexts such as Canada, where English and French are both institutionally supported and hold relatively high status, bilingual development often occurs within a more additive environment (MacLeod et al., 2024; Mady, 2014), potentially altering how language proficiency relates to SE outcomes. Empirical studies have further shown that only children with high English proficiency have a SE advantage compared to their less proficient peers (Frechette et al., 2023; Melzi et al., 2017; Winsler et al., 2014). For instance, Frechette and colleagues (2023) found that, in the U.S. context, compared to monolingual children, proficient bilinguals had higher social skills, whereas emergent bilinguals (or DLLs developing basic English proficiency) showed similar social skills. As such, depending on DLLs’ English language proficiency, there may be differential relations between their dual language skills and SE development. For example, Huang et al. (2024) examined reciprocal relations between language and SE skills in United Kingdom children from ages 3 to 5 and found that these relations depended on children's home language use. They categorized DLL preschoolers into two groups: (a) children who speak both English and a heritage language at home, and (b) children who speak only a heritage language at home. Their findings showed that English expressive vocabulary predicted later social skills (specifically, greater growth in prosocial behavior and fewer SE difficulties) for children speaking both languages. In contrast, early problem behaviors predicted later gains in English expressive vocabulary for children who primarily spoke a heritage language. These results suggest that the foundational skills supporting later development may differ according to children's home language use, which is closely related to their language proficiency.

Taken together, prior studies provide important evidence that language and SE skills are linked among DLL children, but they also highlight several remaining questions. First, findings regarding the directionality of language–SE associations remain mixed, with some studies suggesting that earlier SE skills predict later language skills (e.g., Ertanir et al., 2021; Hagan‐Burke et al., 2016; Huang et al., 2024 [for DLLs only speaking heritage language at home]), others finding that earlier language skills predict later SE development (e.g., Huang et al., 2024 [for DLLs speaking both English and heritage language at home]), and still others documenting bidirectional associations (e.g., Bohlmann et al., 2015). Second, prior studies vary in how language has been operationalized, with many focusing primarily on English skills or on a single language modality. Less is known about whether associations with SE development differ across DLLs’ societal language skills, heritage language skills, and broader conceptual vocabulary. Third, relatively little is known about whether these longitudinal associations differ depending on children's English proficiency at entry into preschool program. This heterogeneity is important because DLLs with developing English proficiency and those with stronger English proficiency may experience different linguistic demands and opportunities in ECE settings. This study aims to address these gaps in the literature.

1.4. The Current Study

The current study examined longitudinal, cross‐lagged associations between Spanish‐English DLLs’ language skills and SE development across approximately two years in HS. Rather than focusing only on whether language and SE skills are associated, we examine the directionality of these associations across three waves. In addition, we consider multiple dimensions of DLLs’ language skills, including English receptive vocabulary, Spanish receptive vocabulary, and conceptual expressive vocabulary, as well as multiple indicators of SE development, including social skills and problem behaviors. Finally, we test whether these associations differ between children who entered HS with developing versus stronger English oral proficiency. HS offers a particularly meaningful context for studying these associations: the program serves many DLLs from low‐income families, and has an explicit mission to promote school readiness, including both language and SE development. For many Spanish‐English DLLs, HS classrooms also represent an early learning context in which children must navigate English‐dominant classroom expectations while continuing to develop and use their home language. Thus, studying language–SE associations in HS can inform linguistically responsive practices in a major ECE system serving DLLs.

In this study, we draw on a sample of Spanish‐English DLLs in HS programs in the U.S. to answer two research questions: (1) Are DLLs’ receptive and expressive language differentially associated with their SE development? and (2) Does the association between DLLs’ language and SE development differ between children with varying baseline English proficiency at HS entry? We first examined research questions across the full DLL sample. Then, we divided the DLL sample into two groups based on their baseline English proficiency to explore differences in comparison: DLLs developing basic English proficiency (i.e., Emergent Bilinguals) and DLLs with stronger English proficiency (i.e., Bilinguals). The study findings can contirubte to our understanding of the reciprocal relations between language and SE development by simultaneously considering language modality (receptive vs. expressive), language context (English vs. Spanish), and developmental heterogeneity in English proficiency. Findings may also and facilitate the development of language and SE interventions to meet the needs of young DLLs with different English proficiency during the early childhood, a crucial period for language and SE development (Halle et al., 2014; Marchman and Fernald, 2008).

2. Method

2.1. Data Source

The current study leveraged the data came from the 2009 cohort of the Head Start Family and Child Experiences Survey (FACES 2009; U.S. Department Of Health And Human Services. Administration For Children And Families. Office Of Planning, Research And Evaluation, 2013). FACES 2009 is the fifth cohort in the series of FACES. The FACES 2009 data provide rich information on nationally representative samples of children and families served by HS, such as demographic characteristics, home/family environment and experiences, and developmental progress of children, as well as charecteritics of HS programs and services, such as enrollment, program partnerships, staff qualifications and professional development, and classroom practices and quality. As of this paper's preparation, eight series of FACES have been released, with the most recent cohort from 2021–2022. However, FACES 2009 is the most recent series to follow children for up to two years and to collect both English and Spanish vocabulary data for all DLLs. The FACES 2009 collected data through multiple questionnaires and instruments, including parent interview, direct child assessments, teacher child reports, classroom observations, teacher interviews, and HS director and education coordinator interviews (Malone et al., 2013). Over 3,000 3‐ and 4‐year‐old children enrolled in HS for the first time in fall 2009 participated in FACES 2009, who were followed each year through kindergarten.

Given our focus on long‐term language and SE development across the preschool years, we first restricted the analytical sample to children who were 3 years old at HS entry (i.e., 3‐year‐old cohort) and who had assessment data across the first three waves (W) of their HS years: fall 2009 (HS Year 1; W1), spring 2010 (HS Year 1; W2), and spring 2011(HS Year 2; W3). Within this restricted 3‐year‐old longitudinal sample, a total of 471 children in 168 classrooms were identified as Spanish‐English DLLs, whose primary home language was Spanish. Children's baseline English oral proficiency was determined using a language screener administered in fall 2009 (see Measures section). Based on this screener, 308 children did not pass the English oral proficiency screener and were classified in the present study as Emergent Bilinguals, whereas 163 children passed the screener and were classified as Bilinguals. These subgroup labels refer to children's English oral proficiency status at HS entry and were used to examine whether longitudinal associations between language and SE development differed by children's initial English proficiency. Table 1 presents demographic information for the full sample, Emergent Bilinguals, and Bilinguals separately. Almost half of the children in the full sample were female (49.26%). The majority of the children had both parents born outside the U.S. (72.19%). Most parents had a high school diploma or less (74.95%) and reported household income of $25,000 or less (73.67%).

TABLE 1.

Demographic information (unweighted) and descriptive statistics (weighted).

				Full sample		Emergent Bilinguals		Bilinguals
				N = 471		n = 308		n = 163
Variable	Min.	Max.	Missing %	Mean/n	SD/%	Mean/n	SD/%	Mean/n	SD/%
Female (n/%)			2.76%	232	49.26%	149	48.38%	83	50.92%
Parent immigrants (n/%)			5.73%
Both parents born in US				30	6.37%	11	3.57%	19	11.66%
One parent born outside US				74	15.71%	40	12.99%	34	20.86%
Both parents born outside US				340	72.19%	239	77.60%	101	61.96%
Maternal education (n/%)			5.31%
Less than high school diploma				247	52.44%	165	53.57%	82	50.31%
High school diploma or GED				106	22.51%	71	23.05%	35	21.47%
Vocational/technical associate degree or some college				65	13.80%	37	12.01%	28	17.18%
Bachelor's degree or higher				28	5.94%	18	5.84%	10	6.13%
Income (n/%)			4.03%
$0–$5,000				22	4.67%	10	3.25%	12	7.36%
$5,001–$10,000				49	10.40%	37	12.01%	12	7.36%
$10,001–$15,000				105	22.29%	63	20.45%	42	25.77%
$15,001–$20,000				103	21.87%	70	22.73%	33	20.25%
$10,001–$25,000				68	14.44%	47	15.26%	21	12.88%
$25,001–$30,000				41	8.70%	28	9.09%	13	7.98%
$30,001–$35,000				22	4.67%	14	4.55%	8	4.91%
$35,001–$40,000				12	2.55%	7	2.27%	5	3.07%
$40,001–$50,000				11	2.34%	7	2.27%	4	2.45%
$50,001–$75,000				12	2.55%	8	2.60%	4	2.45%
$75,001 or higher				7	1.49%	4	1.30%	3	1.84%
Age in month	33	52	2.76%	41.14	3.67	40.58	3.76	42.25	3.21
PPVT (fall HS1)	0	80	10.19%	15.98	14.04	9.15	7.75	27.99	14.54
PPVT (spring HS1)	2	86	21.02%	28.37	16.63	20.72	10.33	41.67	17.13
PPVT (spring HS2)	8	115	35.03%	50.39	21.34	42.89	18.41	66.02	18.33
TVIP (fall HS1)	0	49	4.46%	7.88	7.19	7.73	6.70	8.15	8.04
TVIP (spring HS1)	0	44	14.65%	12.50	10.00	12.78	9.99	11.92	9.99
TVIP (spring HS2)	1	56	34.61%	23.91	13.72	24.56	14.35	22.48	12.13
EOWPVT (fall HS1)	0	44	0.00%	17.36	9.49	14.88	8.70	22.28	9.08
EOWPVT (spring HS1)	0	54	11.46%	24.48	9.42	22.27	8.99	28.89	8.69
EOWPVT (spring HS2)	11	62	33.97%	34.77	9.34	32.56	9.01	39.46	8.21
Social skills (fall HS1)	2	24	3.82%	14.13	4.81	13.59	4.85	15.19	4.53
Social skills (spring HS1)	0	24	12.10%	16.69	4.78	16.30	4.89	17.50	4.43
Social skills (spring HS2)	0	24	38.22%	19.36	4.11	19.12	4.05	19.88	4.18
PB (fall HS1)	0	23	3.82%	4.91	4.60	5.02	4.48	4.70	4.83
PB (spring HS1)	0	22	11.68%	4.40	4.59	4.46	4.62	4.28	4.52
PB (spring HS2)	0	19	38.00%	2.80	3.68	2.95	3.66	2.49	3.70

Abbreviation: EOWPVT, Expressive vocabulary; HS1, Head Start Year 1; HS2, Head Start Year 2; PB, Problem behaviors; PPVT, English receptive vocabulary; TVIP, Spanish receptive vocabulary.

2.2. Measures

2.2.1. English Proficiency

We identified children's baseline DLL status based on their performance on an English oral proficiency screener (the Preschool Language Assessment Survey 2000 [preLAS]; Duncan and De Avila, 1998). Children were assessed on two subtests of 10 items each. In the Simon Says task, children were assessed on their receptive language; they were asked to follow the instructions that Simon says. In the Art Show task, children were assessed on their expressive language skills; they were presented with pictures and asked to identify the objects in English for each picture. The Cronbach's alpha estimate was.93 in fall 2009 (Malone et al., 2013). FACES 2009 used children's performance on the language screener to determine the language assessment routes. However, regardless of the language assessment routes, all Spanish‐English DLLs’ English and Spanish vocabulary were both assessed. In the present study, we used the language screener to create baseline English proficiency groups among Spanish‐English DLLs: Children who did not pass the screener were classified as Emergent Bilinguals, and those who passed the screener were classified as Bilinguals.

2.2.2. Dual Language Skills

The present analyses used three longitudinal language measures administered to Spanish‐English DLLs across waves: English receptive vocabulary, Spanish receptive vocabulary, and conceptual expressive vocabulary. The availability of these measures for the full study sample allowed subgroup comparisons of the longitudinal associations between language and SE development among children with differing levels of English oral proficiency at HS entry.

2.2.3. English Receptive Vocabulary

Children's English receptive vocabulary knowledge was measured using the Peabody Picture Vocabulary Test‐4^th Edition (PPVT‐4; Dunn et al., 2006). For each item, the assessor orally presented a series of words in English, from easy to difficult, accompanied by four‐line drawings of each word. Children were asked to indicate, by pointing, which of four pictures best showed the meaning of a word said aloud by the assessor. Raw scores were used in the analysis. The reliability (Cronbach's alpha) estimates of the PPVT‐4 were 0.97, 0.95, and 0.93 in the fall of 2009, and springs of 2010 and 2011, respectively (Malone et al., 2013).

2.2.4. Spanish Receptive Vocabulary

The Test de Vocabulario en Imagenes Peabody (TVIP; Dunn et al., 1986) was used to assess Spanish‐English DLLs’ Spanish receptive vocabulary knowledge. During this test, children were asked to identify a picture corresponding to the meaning of a word said aloud in Spanish by the assessor. Raw scores were used in the analysis. The Cronbach's alpha reliability estimates were 0.93, 0.94, and 0.92 in fall 2009, spring 2010, and spring 2011, respectively (Malone et al., 2013).

2.2.5. Conceptual Expressive Vocabulary

The Expressive One‐Word Picture Vocabulary Test–Spanish‐Bilingual Edition (EOWPVT–SBE; Brownell, 2001) was used to measure the conceptual expressive vocabulary of Spanish‐English DLL children. The assessor asked children to accurately name an object shown in a picture. Conceptual scoring was used, such that responses are accepted in either English or Spanish and are scored correctly when children accurately identify an object regardless of whether they label it in English or Spanish. Raw scores were used in the analysis. The reliability (Cronbach's alpha) estimates were 0.86, 0.80, and 0.90 in fall 2009, spring 2010, and spring 2011, respectively (Malone et al., 2013).

2.2.6. SE Skills

2.2.6.1. Social Skills

Children's engagement in cooperative classroom behaviors was reported by teachers using 12 items from the Personal Maturity Scale (Entwisle et al., 1987) and the Social Skills Rating System (SSRS; Elliott et al., 1988; Gresham and Elliott, 1990). The items encompassed behaviors such as “following the teacher's directions,” “helping to put things away,” “complimenting classmates,” and “following rules when playing games.” Teachers were prompted to provide a response from “1 = never” to “3 = very often.” A total summary score was created by FACES 2009, with the Cronbach's alpha reliability estimates of 0.89, 0.89 and 0.88 in fall 2009, spring 2010, and spring 2011, respectively (Malone et al., 2013).

2.2.6.2. Problem Behaviors

Teachers reported children's negative behaviors associated with learning problems and later grade retentions, using an abbreviated version of the Personal Maturity Scale (Entwisle et al., 1987) and the Behavior Problems Index (BPI; Peterson and Zill, 1986). A total of 14 items in the measure were designed to capture the frequency of aggressive (e.g., “hits/fights with others”), hyperactive (e.g., “is very restless”), and anxious or depressed and withdrawn (e.g., “is unhappy”) behavior. A scale from “1 = never” to “3 = very often” was used. FACES 2009 derived a total summary score with higher scores representing more frequent or severe negative behavior, with the Cronbach's alpha reliability coefficients of 0.88, 0.87, and 0.86 in fall 2009, spring 2010, and spring 2011, respectively (Malone et al., 2013).

2.2.7. Control Variables

We also included several control variables from fall 2009 (i.e., at the entry of HS; W1), including child age and sex, parent immigrant status, maternal education, household income. Age in month and child sex were indicated by parents or caregivers. Parent immigrant status was derived from parent interview and identifies whether one or both of the child's parents was born outside of the U.S., with responses including “1 = Both parents born in the U.S.,” “2 = One parent born outside the U.S.,” and “3 = Both parents born outside the U.S.” Parents also indicated mother's highest education from following categories: “1 = less than high school diploma,” “2 = high school diploma or GED,” “3 = vocational/technical diploma, associate's degree, or some college,” and “4 = bachelor's degree or higher.” Household income variable was derived from parent interview, during which parents or caregivers indicated the category their household income fell into (see Table 1 for full response categories).

2.3. Analytic Approach

To examine the longitudinal reciprocal relations among DLLs’ language (i.e., Spanish receptive vocabulary, English receptive vocabulary, and conceptual expressive vocabulary) and SE development (i.e., social skills and behavioral problems), we specified a series of three‐wave cross‐lagged panel models using Mplus 8.8 (Muthén & Muthén, 2017). Cross‐lagged panel analysis is an approach used to describe reciprocal relations or directional associations across different variables over time (Kenny, 1975; Newsom, 2015). In the current study, the cross‐lagged panel analysis was used to examine (a) how language or SE skills predicts themselves over time (i.e., autoregressive paths) and (b) how each developmental domain subsequently predicts the other (i.e., cross‐lagged paths) across three waves. We specified the cross‐lagged paths from each language and SE skills at time t‐1 to SE and language skills at the subsequent time point t, respectively. Reciprocal associations between language skills and SE skills were examined in separate models for each of three language measures, Spanish receptive vocabulary, English receptive vocabulary, and conceptual expressive vocabulary, resulting in three sets of models. Language and SE skills were also set to be correlated at each time point. We first modeled cross‐lagged relations with SE skills among the full sample. Then, separate models were constructed for Emergent Bilinguals and Bilinguals. All control variables mentioned above were used to predict language and SE skills in the fall of HS Year 1 (W1). Because the data are nonexperimental, cross‐lagged paths are interpreted as longitudinal associations or statistical prediction from earlier to later measures, rather than evidence of causal effects.

Because children were clustered within classrooms (i.e., several children sharing one classroom), clustered‐robust standard error was used to account for the classroom‐level effect. Because we used a nationally representative dataset with complex survey design, in all models, we employed the Taylor Series linearization method, which is a design‐based variance estimation method, to estimate proper variance for complex survey data (Williams, 2000). In addition, since we focused on the sample of Spanish‐English DLLs only, we also used domain analysis in Mplus by specifying the SUBPOPULATION command to include specific subsample only.

To handle missing data, we used multiple imputation. The multiple imputation model included all study variables as well as language and SE variables assessed in kindergarten to get unbiased estimation. Twenty datasets were imputed and were used in subsequent analyses. Percentage of the missing data for each variable ranged from 0% to 38.22%, with an overall average of 14.68% missing.

3. Results

3.1. Preliminary Analyses

3.1.1. Descriptive Statistics

Table 1 presents the descriptive statistics on all study variables for the full sample and for Emergent Bilinguals and Bilinguals separately. To examine whether the two subgroups differed in terms of demographics and language and SE skills, we ran a series of linear regression models using DLL status as a predictor (Emergent Bilinguals = 1, Bilinguals = 0) while taking into account of the complex survey design using Taylor Series linearization method.

We found that compared to Bilinguals, Emergent Bilinguals were younger (B = −1.66, p < 0.001), were less likely to have both parents born in the U.S. (B = −0.11, p = 0.002) but were more likely to have both parents born outside of U.S. (B = 0.20, p < 0.001). No significant differences were found for child sex, maternal education, and household income (all ps > 0.05). In addition, Emergent Bilinguals scored lower on English receptive vocabulary (B = −18.85, B = −20.95, B = −23.13, W1‐3, respectively, all at p < 0.001) and conceptual expressive vocabulary (B = −7.40, B = −6.63, B = −6.90, W1‐3, respectively, all at p < 0.001) at all three time points than Bilinguals. Teachers reported that Emergent Bilinguals had less advanced social skills in the fall of HS Year1 (W1: B = −1.60, p = 0.012) than Bilinguals, but such differences in social skills were not found in later time points (i.e., W2 and W3). No significant group differences were observed on Spanish receptive vocabulary and teacher‐reported problem behaviors at any time points.

3.1.2. Correlational Analyses

Correlational results are presented separately for the full sample, Emergent Bilinguals, and Bilinguals in Tables S1 and S2 in the online supplementary materials. Language and SE skills were positively correlated across three time points. That is, children with more advanced language and SE skills at one time point also exhibited higher skills at the other two time points (r ranged from 0.34 to 0.80). This pattern stayed consistent for the full sample, Emergent Bilinguals, and Bilinguals. In general, Spanish receptive vocabulary and conceptual expressive vocabulary appear to be positively correlated with social skills and negatively correlated with problem behaviors, while English receptive vocabulary was generally not correlated with social skills and problem behaviors, with some exceptions.

Group differences in the correlations between language and SE skills were observed. For Bilinguals, their English receptive vocabulary across all three waves showed stronger positive correlations with W1 social skills (r ranged from 0.23 to 0.24), and Bilinguals’ Spanish receptive vocabulary at W3 showed stronger positive correlations with social skills and stronger negative correlations with problem behaviors across all waves. For Emergent Bilinguals, Spanish receptive vocabulary at W3 was correlated with concurrent social skills and problem behaviors only. Their conceptual expressive vocabulary at W1 and W2 showed stronger negative correlations with problem behaviors across all waves. Further, problem behaviors were negatively correlated with English receptive vocabulary (r = −0.18) and conceptual expressive vocabulary (r = −0.21) at W3 for Emergent Bilinguals but not for Bilinguals.

3.2. Cross‐Lagged Panel Analysis

We first report the results of the models in the full sample (RQ1), followed by the two DLL subgroups with different English proficiency (RQ2).

3.2.1. English Receptive Vocabulary and SE Development

Figure 1 displays the results of the cross‐lagged models of English receptive vocabulary and SE development separately for the full sample, Emergent Bilinguals, and Bilinguals. The path standardized coefficients are presented in Table S3.

FIGURE 1.

Cross‐lagged models of english receptive vocabulary and SE development. Note. Coefficients presented were standardized. Significant coefficients are presented in bold on solid lines. Nonsignificant coefficients are presented on dashed lines. PPVT, English receptive vocabulary, Social, Social skills, PB, Problem behaviors, HS1, Head Start Year 1, HS2, Head Start Year 2. *p < 0.05, **p < 0.01, ***p < 0.001.

3.2.1.1. Full Sample

No significant cross‐lagged effects were observed between English receptive vocabulary and SE skills, indicating that English receptive vocabulary did not significantly predict later SE development, nor did SE skills predict later English receptive vocabulary. Concurrent associations between English receptive vocabulary and SE skill were observed at W1: English receptive vocabulary was positively associated with social skills and negatively associated with problem behaviors. In addition, social skills and problem behaviors were negatively correlated at each wave, indicating consistent co‐occurrence of SE dimensions over time. Autoregressive paths were significant for all developmental domains across waves (all ps < 0.001). Children with higher English receptive vocabulary and more adaptive SE skills at earlier waves tended to maintain relatively higher skills in the same domains at later waves.

3.2.1.2. Group Differences by English Proficiency (Emergent Bilinguals vs. Bilinguals)

Consistent with the full‐sample results, no cross‐lagged effects were found in either Emergent Bilinguals or Bilinguals. However, group differences emerged in concurrent associations at W1: (a) for Emergent Bilinguals, English receptive vocabulary was negatively associated with problem behaviors but not social skills; but (b) for Bilinguals, English receptive vocabulary was positively associated with social skills but not problem behaviors. Social skills and problem behaviors were negatively correlated at each wave. Autoregressive paths were significant for all developmental domains across waves (all ps < 0.001) for both groups.

3.2.2. Spanish Receptive Vocabulary and SE Development

Figure 2 displays the results of the cross‐lagged models of Spanish receptive vocabulary and SE development separately for the full sample, Emergent Bilinguals, and Bilinguals. The path standardized coefficients are presented in Table S4.

FIGURE 2.

Cross‐lagged models of spanish receptive vocabulary and SE development. Note. Coefficients presented were standardized. Significant coefficients are presented in bold on solid lines. Nonsignificant coefficients are presented on dashed lines. TVIP, Spanish receptive vocabulary, Social, Social skills, PB, Problem behaviors, HS1, Head Start Year 1, HS2, Head Start Year 2. *p < 0.05, **p < 0.01, ***p < 0.001.

3.2.2.1. Full Sample

When we examined cross‐lagged associations, we found unidirectional relations between Spanish receptive vocabulary and both aspects of SE development. That is, higher Spanish receptive vocabulary at W1 and W2 was associated with greater development in social skills and fewer problem behaviors at W2 and W3, respectively.

3.2.2.2. Group Differences by English Proficiency (Emergent Bilinguals vs. Bilinguals)

Results from cross‐lagged associations showed that like the full sample, Emergent Bilinguals showed a unidirectional relation between Spanish receptive vocabulary and SE development, such that Spanish receptive vocabulary at W2 predicted children's problem behaviors development (i.e., fewer problem behaviors) at W3. For Bilinguals, we found bidirectional associations between Spanish receptive vocabulary and their SE development. These children with more advanced Spanish receptive vocabulary at W2 showed more positive skills in both social skills and problem behaviors at W3. In addition, Bilinguals with higher social skills at W1 had higher gains in Spanish receptive vocabulary at W2. Similar to the full sample, Emergent Bilinguals showed significant concurrent associations between Spanish receptive vocabulary and both social skills and problem behaviors at W1 and W3. Such associations were not found in Bilinguals. Autoregressive paths were significant for all skills for both groups as shown in the full sample (p < 0.001), indicating that children with higher performance in Spanish receptive vocabulary knowledge, social skills, and fewer problem behaviors at the previous time point tended to maintain advantages in the same domains at subsequent time points. The two SE skills were concurrently associated for both groups as they were in full sample (all ps < 0.001).

3.2.3. Conceptual Expressive Vocabulary and SE Development

Figure 3 displays the results of the cross‐lagged models of conceptual expressive vocabulary and SE development separately for the full sample, Emergent Bilinguals, and Bilinguals. The path standardized coefficients are presented in Table S5.

FIGURE 3.

Cross‐lagged models of expressive vocabulary and SE development. Note. Coefficients presented were standardized. Significant coefficients are presented in bold on solid lines. Nonsignificant coefficients are presented on dashed lines. EOWPVT, Expressive vocabulary, Social, Social skills, PB, Problem behaviors, HS1, Head Start Year 1, HS2, Head Start Year 2. *p < 0.05, **p < 0.01, ***p < 0.001.

3.2.3.1. Full Sample

In the full sample, unidirectional cross‐lagged effects were observed from conceptual expressive vocabulary to SE development. Specifically, higher conceptual expressive vocabulary at W1 and W2 was associated with greater subsequent social skills at W2 and W3, respectively. No cross‐lagged effects were found from SE skills to later conceptual expressive vocabulary. Concurrent associations were observed at W1. Conceptual expressive vocabulary was positively associated with social skills and negatively associated with problem behaviors. In addition, social skills and problem behaviors were negatively correlated at each wave, indicating consistent co‐occurrence of SE dimensions over time. Autoregressive paths were significant for all developmental domains across waves (all ps < 0.001), indicating that children who showed more advanced conceptual expressive vocabulary knowledge and social skills and fewer problem behaviors at a previous time point tended to maintain advantages in the same domains at subsequent time points.

3.2.3.2. Group Differences by English Proficiency (Emergent Bilinguals vs. Bilinguals)

Group differences emerged in the cross‐lagged associations. For Emergent Bilinguals, unidirectional effects from conceptual expressive vocabulary to SE development were observed. Specifically, higher conceptual expressive vocabulary at W1 was associated with higher social skills and fewer problem behaviors at W2, as well as higher social skills at W3. In contrast, for Bilinguals, no significant cross‐lagged relations were observed between conceptual expressive vocabulary and SE development, indicating a lack of longitudinal prediction in either direction for this group. Across both groups, no cross‐lagged effects were found from SE skills to later conceptual expressive vocabulary, suggesting that SE functioning did not predict subsequent expressive vocabulary growth in either group. Concurrent associations varied by group. For Emergent Bilinguals, conceptual expressive vocabulary was negatively associated with problem behaviors at W1, whereas for Bilinguals, it was positively associated with social skills at W1. Social skills and problem behaviors were negatively correlated at each wave in both groups. Autoregressive paths were significant for all developmental domains across both groups (all ps < 0.001). This indicates that children who showed more advanced conceptual expressive vocabulary knowledge and social skills, and fewer problem behaviors in previous time point tended to do so in the same domains at subsequent time points.

4. Discussion

Among a sample of Spanish–English DLLs in HS, we found that language–SE associations varied by language modality and children's baseline English proficiency. Spanish receptive vocabulary and conceptual expressive vocabulary showed more consistent longitudinal associations with later SE development, whereas English receptive vocabulary showed no cross‐lagged associations. Importantly, patterns differed by DLL subgroup: Emergent Bilinguals primarily showed unidirectional effects from Spanish receptive and conceptual expressive vocabulary to SE development, whereas Bilinguals showed more reciprocal associations between Spanish receptive vocabulary and SE skills. These findings highlight the importance of distinguishing between heritage and societal language skills, as well as between different dimensions of SE development, when examining associations between DLLs’ language and SE development. They also suggest that language–SE associations are not uniform across DLLs but instead vary depending on children's baseline English proficiency at HS entry.

4.1. DLLs’ Language Skills Were Differentially Associated With SE Development

Three main findings stand out as particularly noteworthy. First, unlike previous studies demonstrating a significant relation between receptive vocabulary and SE skills in DLLs (Clayton et al., 2022; Ertanir et al., 2021; Sun et al., 2021), we did not find longitudinal associations between English receptive vocabulary and SE development (for both social skills and problem behaviors). Our findings suggest that, for DLLs whose primary home language was Spanish, English receptive vocabulary may play a limited role in the prediction of subsequent SE outcomes once prior skills are accounted for; likewise, SE skills are also less likely to predict the development in English receptive vocabulary. The inconsistency with previous literature may lie in differences in methodology. Much of the prior work has examined concurrent associations, whereas our study investigated how prior English receptive vocabulary skills predict subsequent growth in SE development, controlling for autoregressive effects. It is therefore possible that English receptive vocabulary may be linked to SE skills concurrently but not longitudinally. In addition, differences in SE measurement may also contribute to discrepant findings. Prior studies often relied on composite SE indices capturing multiple behavioral domains, whereas we examined social skills and problem behaviors separately. Composite SE measure may capture shared variance across multiple, interrelated social and regulatory behaviors, which may increase the sensitivity to detect language‐SE associations that may not emerge for more narrowly defined SE constructs.

Second, Spanish receptive vocabulary skills were linked to DLLs’ positive later social skills and reductions in problem behaviors reported by teachers. This highlights the importance of DLLs’ home language skills for their SE development. For Spanish‐English DLLs, stronger Spanish receptive vocabulary may support children's ability to understand and participate in interactions with Spanish‐speaking peers, teachers, family members, or other adults. These interactions may, in turn, provide opportunities to practice social behaviors, understand expectations, and regulate behavior. Although the current study cannot identify the specific contexts in which children used Spanish, the findings suggest that heritage language skills may serve as an important developmental resource for DLLs’ SE development.

Third, our findings showed unidirectional effects of conceptual expressive vocabulary on social skills, but not problem behaviors. Children with stronger conceptual expressive vocabulary skills may be better able to initiate and sustain peer interactions, communicate needs, respond to others, and participate in classroom routines, all of which are central to teacher‐rated social skills (Denham and Brown, 2010; Longoria et al., 2009). However, due to the conceptual nature of the expressive language measure used in this study, we were unable to disentangle these skills across the two languages. Thus, this finding may reflect the role of children's broader expressive vocabulary repertoire across languages rather than expressive ability in one specific language. The lack of association with problem behaviors may indicate that conceptual expressive vocabulary is more directly tied to children's positive social engagement than to reductions in behavior difficulties. Our null finding contrasts with Huang et al. (2024), who found an association between DLLs’ expressive language and reduction in problem behaviors. The difference may be explained by measurement variation: Huang et al. measured children's expressive vocabulary in English, whereas we used a conceptual expressive vocabulary, which capture expressive vocabulary repertoire in both languages. It is possible that expressive skills in the societal language may be more directly linked to managing classroom‐related behavioral demands, whereas broader conceptual expressive knowledge across languages may be less sensitive to context‐specific behavioral regulation in ECE settings.

4.2. Associations Between Language and SE Skills Differ Between Emergent Bilinguals and Bilinguals

Subgroup analyses showed that associations between language and SE development differed by children's baseline English proficiency at HS entry. For Spanish receptive vocabulary, Emergent Bilinguals showed a unidirectional effect of Spanish receptive vocabulary predicting later reductions in problem behaviors. In contrast, Bilinguals showed bidirectional associations, such that social skills predicted later Spanish receptive vocabulary growth, and early Spanish receptive vocabulary predicted later social skills and problem behaviors. Our findings suggest that for DLLs who have developed basic English proficiency (i.e., Bilinguals), social interaction processes may become more mutually reinforcing with heritage language development. Bilinguals’ stronger social skills may have enabled more frequent engagement in meaningful social interactions that provide opportunities for language learning. In fact, previous research has shown that SE skills predict more frequent classroom verbal interactions among Bilinguals (referred to as DLL‐proficient) but not Emergent Bilinguals (referred to as DLL‐early) in a HS setting (Choi et al., 2026). Importantly, we found that both DLL groups showed an effect of Spanish receptive vocabulary on later problem behaviors, indicating that stronger receptive vocabulary knowledge in the heritage language (Spanish) may provide a stable linguistic resource that supports behavioral regulation regardless of their baseline English proficiency at HS entry.

For conceptual expressive vocabulary, we observed a unidirectional effect of expressive vocabulary on Emergent Bilinguals’ SE development, whereas no such relation was found for Bilinguals. Previously, a study conducted in a HS program found that conceptual expressive vocabulary skills predicted more frequent classroom verbal interactions among Emergent Bilinguals (referred to as DLL‐early), but not among Bilinguals (referred to as DLL‐proficient; Choi et al., 2026). Conceptual expressive skills among Emergent Bilinguals may facilitate greater engagement in verbal interactions, which, in turn, may provide opportunities for developing SE skills through practicing social communication and interacting with peers and teachers. The absence of a significant association between conceptual expressive skills and classroom verbal interactions reported in Choi et al.’s study of bilingual children may offer one potential explanation for the lack of an association between conceptual expressive skills and SE skills observed in the present study. In this study, Emergent Bilinguals’ conceptual expressive language skills are likely to be representative of their Spanish expressive vocabulary, whereas Bilinguals’ conceptual expressive vocabulary is representative of their expressive language repertoire in both languages. These findings are consistent with prior work showing positive associations between heritage language expressive skills and SE outcomes (e.g., Chung et al., 2019), and extend this literature by demonstrating longitudinal effects within classroom settings.

4.3. Limitations and Future Directions

This study has several strengths, including its longitudinal design and the use of multiple language and SE measures; however, study limitations should also be considered when interpreting the findings. First, expressive vocabulary was measured conceptually, which captures DLLs’ conceptual language skills in both English and Spanish. Conceptual score is considered as a more equitable approach to assess multilingual children's language abilities (Mancilla‐Martinez et al., 2021) since multilinguals process language through a shared conceptual system (Kroll and Stewart, 1994). However, this also makes it difficult to identify whether differences exist between the two languages. Future studies may differentiate children's expressive language skills in English (societal) and Spanish (home languages) to gain a nuanced understanding.

Second, the data used in this study were collected between 2009 and 2011. The demographics and composition of children in HS may be different from what we witness nowadays. However, FACES 2009 is the FACES series that placed an enhanced focus on children who are DLLs and remains the most recent data having two years of HS data and included both English and Spanish vocabulary measures for Spanish‐English DLLs. This makes FACES 2009 the ideal data source available to examine HS DLLs' development. Future research using more current longitudinal data that assess DLLs’ language skills in both English and their home language, document children's language exposure and use across home and ECE settings, and include repeated measures of SE development is needed to further advance our understanding of language–SE associations.

Third, children's SE skills were reported by their teachers in the current study, as is typically done in early childhood research (e.g., Clayton et al., 2022; Ertanir et al., 2021). Given that SE functioning differs based on context and across teacher and parent reports, it is important for future studies to consider incorporating multiple informants and methods, such as parent reports and direct observations, to capture a more comprehensive picture of DLLs’ SE development across contexts.

Fourth, we classified children as Emergent Bilinguals or Bilinguals based on their English proficiency status at the beginning of HS, as determined by screening assessments. This approach does not capture changes in English proficiency over time; thus, some children classified as Emergent Bilinguals have developed higher levels of English proficiency during the HS period that are not reflected in our grouping variable. Future research should incorporate time‐varying or growth‐based measures of English proficiency to better capture dynamic changes in bilingual development.

Fifth, although the present study draws on a nationally representative sample of HS children in the U.S., the findings should be interpreted within the context of HS classrooms, which primarily serve children from low‐income families and often include a high proportion of linguistically diverse students. As such, we cannot make generalization to other ECE settings, where differences in socioeconomic composition, language policies, or levels of support for heritage language exist. Relatedly, the sample consisted exclusively of Spanish‐English DLLs; therefore, the findings may not generalize to DLLs who speak home languages other than Spanish.

There are other factors, such as home language environment, that may affect the language‐SE association (e.g., Huang et al., 2024). Although it is out of scope of the current study, future studies should incorporate more comprehensive measures of home language proficiency and home language context to clarify how children's bilingual experiences across home and ECE settings contribute to language–SE development. Finally, although we used cross‐lagged panel models controlling for previous autoregressive effects, the associations between language and SE skills cannot be interpreted to be causal.

5. Conclusion

This study contributes to the emerging literature on HS DLLs’ language and SE development by comprehensively investigating such relations across children's receptive and expressive language, and societal (English) and heritage (Spanish) language, as well as across children with and without basic English proficiency at HS entry. Our findings supported unidirectional effects of DLLs’ Spanish receptive vocabulary and conceptual expressive vocabulary on their SE development over a 2‐year period in HS, indicating the significant role of DLLs’ heritage language in their SE development. Analyses comparing the two DLL groups based on baseline English proficiency at HS entry indicated that the patterns observed in the full sample were primarily driven by Emergent Bilinguals. In contrast, for Bilinguals, we found a reciprocal, bidirectional association between Spanish receptive vocabulary and SE skills.

Together, these findings suggest that heritage language skills function as a critical SE resource for DLLs, although the direction and nature of these associations differ by children's baseline English proficiency. From an applied perspective, ECE programs serving DLLs, particularly those in linguistically diverse, low‐income contexts such as HS, may benefit from intentionally integrating heritage language supports within SE‐focused curricula and classroom practices. Such approaches may not only foster SE development but also promote more inclusive learning environments that leverage DLLs’ existing linguistic strengths rather than privileging societal language (English) proficiency alone. More broadly, these findings challenge deficit‐oriented assumptions that prioritize societal language (English) proficiency as the primary driver of DLLs’ SE development and highlight the importance of recognizing heritage languages as valuable linguistic resources for a means of supporting DLLs’ holistic development.

Author Contributions

Ye Shen: conceptualization, methodology, data curation, formal analysis, visualization, writing – original draft, writing – review and editing, software. Ji–Young Choi: writing – review and editing, conceptualization.

Funding

The current study was supported in part by (1) Johns Hopkins University School of Education Faculty Start‐Up Funding, (2) the Language Learning Early Career Research Grant, and (3) a grant from the American Educational Research Association which receives funds for its “AERA Grants Program” from the National Science Foundation under NSF and NSF‐DRL #1749275. Opinions reflect those of the authors and do not necessarily reflect those of AERA or NSF.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting Information: desc70275‐supp‐0001‐SuppMat.docx

Data Availability Statement

The data that support the findings of this study are available from ICPSR. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from https://doi.org/10.3886/ICPSR34558.v4 with the permission of ICPSR.