Learning Two Languages: Dual Language Learning Patterns, Predictors, and Outcomes

ABSTRACT

This secondary analysis study examined patterns, predictors, and outcomes of dual language learners’ (DLLs’) development in Spanish and English. Research Findings: Latent profile analysis identified three groups of DLLs: Balanced Bilinguals (54%), English-Dominant Bilinguals (25%), and Spanish-Dominant Bilinguals (21%). These groups had fairly comparable Spanish skills at Head Start (HS) entry but showed distinct learning patterns in English and Spanish during the 2.5 years in HS and kindergarten (K), which resulted in group differences in dual-language skills at K. When their home and classroom environments were compared, Balanced Bilinguals had more Spanish exposure than English-Dominant Bilinguals and more English exposure than Spanish-Dominant Bilinguals. Spanish-Dominant Bilinguals generally had more Spanish exposure at home and in the classroom, less English exposure at home, and had parents with lower English proficiency. English-Dominant Bilinguals had fewer home literacy materials in Spanish and had fewer opportunities for Spanish exposure in their first HS classrooms. Balanced Bilinguals presented the highest K achievement; the other two groups showed comparable skills. Practice or Policy: This study demonstrated variations in dual-language learning among DLLs in HS and indicated the need for tailored strategies that address the unique needs of diverse DLLs presenting varying proficiency in their two languages.

Approximately one-third of children five years old and younger in the United States are dual language learners (DLLs) who are acquiring a language other than English at home (Migration Policy Institute Data Hub, n.d.). Although children are capable of learning multiple languages (Espinosa, 2015), many DLLs living in the U.S. do not fully actualize their developmental potential to become bilinguals who are competent in both English and their home language (Fillmore, 1991, Hoff, 2013, Paradis, 2011). Bilingualism literature has drawn attention to the cognitive advantages of acquiring proficiency in multiple languages (Bialystok, 2001, National Academies of Sciences, Engineering, and Medicine [NASEM], 2017). Much of the relevant literature further indicates that bilinguals have additional academic and social advantages, including a strong foundation of school readiness; the establishment of strong cultural connections, identity, and family ties; along with the economic advantage of being bilingual into adulthood (Espinosa, 2006, NASEM, 2017, U.S. Department of Health and Human Services [US DHHS] & U.S. Department of Education, 2016). These bilingual advantages indicate the importance of understanding learning environments where DLLs are most likely become competent in both languages. In this study, we explored the patterns, contextual predictors, and developmental outcomes of dual-language learning over 2.5 years of Head Start (HS) and kindergarten (K). We used a nationally representative dataset for low-income DLLs who attended HS at age 3 for two years.

Dual-Language Learning

A consensus study report from NASEM (2017) acknowledged that “It is impossible to obtain an accurate assessment of a DLL’s developmental status and instructional needs without examining the child’s skills in both languages” (pp. 405–406). Such a notion highlights a holistic understanding of DLLs’ language skillsets across multiple languages they are developing. Cummins’s (1979, 1981) threshold hypothesis similarly provides that DLLs’ developmental status and needs can be precisely understood only when their skillsets in both languages are simultaneously considered. The hypothesis posits that bilingual advantage in cognition is evident for DLLs evincing a high degree of competency in both languages but not for DLLs whose competence in one language is dominant over the other. The hypothesis further presumed that DLLs have cognitive disadvantages if they fail to achieve proficiency in both languages, although such a concept of semilingualism has received criticism because of the vagueness of its threshold specification for distinguishing proficiency status, its deficit connotations for bilingual language practices, and the lack of reputable evidence supporting the concept (MacSwan, 2000).

In early childhood education (ECE), most research attention has been focused on understanding DLLs’ language skills in each language separately. Recently, however, there have been increasing efforts to holistically understand DLLs’ dual-language skills, acknowledging that children learn multiple languages simultaneously, rather than independently. These emerging studies have taken a person-centered approach to identify sub-groups of individuals who share similar characteristics based on a set of variables (Howard & Hoffman, 2018). A person-centered approach allows for holistic examination of both English and Spanish skills; thus, this approach can help to identify the unique variations in dual-language skillsets and development within DLL groups. Most often, DLLs’ language skills are studied in each language separately; by contrast, using a person-centered approach helps to extend our understanding of the developmental strengths and needs of a diverse group of DLLs with varied dual-language skills.

Recent studies using a person-centered approach have identified subgroups of DLLs based on their proficiency and use of English and their home language (mostly Spanish) (e.g., Kim et al., 2018, Halpin et al., 2021, López & Foster, 2021). For example, López and Foster’s (2021) study of Spanish-English DLLs attending HS identified four profile groups of DLLs based on their school readiness (cognition, language, literacy, and math) assessed in both English and Spanish upon exiting the HS program. They found that 30.63% of DLLs showed a high degree of school readiness in both languages (they were labeled Balanced Average), 38.75% of DLLs showed a higher degree of school readiness in English than in Spanish (English-Dominant), 17.81% of DLLs presented greater school readiness in Spanish than in English (Spanish-Dominant), and 13.12% of DLLs presented low school readiness in both languages (Balanced Low). It was also found that the Balanced Average group showed the most advanced school readiness skills in both languages among the four groups.

In another study, Kim et al. (2018) identified three subgroups of DLLs based on their Spanish use and exposure at home and in preschool classrooms (i.e., the language heard at home, the language used at home, the language used when speaking to class peers, and the language used to teachers). The groups were labeled Emergent Bilingual Children, who speak mostly Spanish and some English at home and in the classroom; Bilingual Children, who speak both at home but only English in the classroom; and Heritage Language Speakers, who speak only Spanish at home and in the classroom. The results showed more positive learning outcomes from Bilingual children in multiple developmental domains (i.e., language, literacy, math, cognitive, social-emotional, and physical) than other DLL groups. Furthermore, compared to non-DLL peers, DLLs, especially Bilingual Children, presented more advanced cognitive, social-emotional, and physical skills over the year. Research evidence clearly indicates that DLLs have very different proficiencies in each separate language they develop (Choi, Rouse, et al., 2018, Park et al., 2018, U.S. DHHS & U.S. DoE, 2016) as well as in the level of bilingualism (Kim et al., 2018, Halpin et al., 2021, López & Foster, 2021). This line of research raises the important question of why and how such differences emerge.

Contextual Factors Explaining Dual-Language Development

The field of ECE has long been interested in identifying contextual factors that promote early language development. Social interactionist theories, based largely on sociocultural theory (Lantolf & Thorne, 2006; Vygotsky and Cole, 1978), posit that language development occurs in culturally constructed settings (e.g., homes and schools) through interactions that motivate individuals to communicate. In line with sociocultural theory, studies have shown that the amount of quality language exposure is a significant predictor of children’s fluency and speed of language acquisition (Castro et al., 2020; Espinosa, 2015; Halpin et al., 2021; Hammer et al., 2014; Hoff et al., 2014; Thordardottir, 2019). For example, a study of school-aged bilingual children attending French schools in Montreal showed that the amount of exposure to each language, but not timing, was a stronger predictor of the children’s performance in each language (Thordardottir, 2019). A consensus study report from NASEM (2017) further noted that DLLs’ developmental differences in each language were due to language learners’ divergent experiences.

The specific type and amount of exposure to a language that effectively supports the development of two languages is not well understood (NASEM, 2017); however, emerging evidence points to the importance of home and ECE classroom experiences for DLLs’ dual-language acquisition. Environmental factors relating to DLLs’ acquisition in each language (e.g., vocabulary size) include the relative and absolute amount of exposure to each language at home and in ECE settings, the language used during learning activities, the type of language exposure experienced from TV, proficiency in each language among the adults in the child’s life (e.g., mothers and teachers), parents’ attitudes toward their children’s home language learning, and parents’ beliefs/values surrounding bilingual development (Castro et al., 2020; Halpin et al., 2021; Hindman & Wasik, 2015; Hoff et al., 2012; Hurtado et al., 2014; Hwang et al., 2022; Li, 1999; McCabe, 2013; Partika et al., 2021; Thordardottir, 2019; Wong et al., 2021). Research has further indicated that DLLs benefit from consistent exposure to each language (both the home language and English) in ECE settings (NASEM, 2017; Raikes et al., 2019), yet non-native English-speaking parents’ use of English has limited benefit for DLLs’ English learning (Barnett et al., 2007; Hoff et al., 2014; McCabe et al., 2013).

We know little about whether and how language experiences differ among DLLs who present different levels of bilingualism (NASEM, 2017). Given the bulk of the literature highlighting the cognitive, academic, social, and economic benefits of bilingualism, the social contexts in which DLLs grow up and become bilingual warrant scholarly attention. Gaining insight into the social context of DLLs’ language acquisition has implications for DLL education.

Current Study

This study first aimed to identify subgroups of DLLs based on their Spanish and English receptive vocabulary skills assessed at K and their gain (or growth) scores during the time between entry to HS to K (approximately 2.5 years). We then explored learning environment factors in the home (at the time of HS entry) and classroom (during HS) to study whether and how DLLs with different levels of dual-language skillsets were exposed to different learning environments. Finally, we examined whether DLLs presenting different dual-language skills show different levels of achievement at K.

DLLs’ vocabulary learning in each language is important, given its predictive power in reading comprehension (Hemphill & Tivnan, 2008; Proctor et al., 2005). Taking advantage of a longitudinal dataset, this study considered both K skills and gain scores during HS to K years for each language. The skillsets that children demonstrate at K have received substantial research attention due to the national emphasis on school readiness. Empirical studies have also shown that children’s K vocabulary skills and their gains in vocabulary (or the rate of development) are significant predictors of subsequent vocabulary skills and academic achievement (Grimm et al., 2018; Ramsook et al., 2020; Rowe et al., 2012).

The context of this study is HS, a federally funded comprehensive early education and family support program for children and families from low-income households. HS is an important ECE setting for DLL education: more than one-third of all children in HS are from homes where a language other than English is spoken (US DHHS, Administration for Children and Families, Office of Planning Research and Evaluation [ACF OPRE], n.d.a). Given its emphasis on cultural and linguistic competency, HS is regarded as a model program for DLL education (McNamara, 2016). DLLs in the U.S. are disproportionately from low-income households (Child Trend, n.d.; Migration Policy Institute Data Hub, n.d.); this fact further underscores the importance of understanding DLLs’ development within the HS context.

Perhaps one of the main reasons for the scarcity of literature on dual-language development is the dearth of data regarding DLLs’ home language skills. Traditionally, U.S. ECE research has focused on DLLs’ English acquisition, and much less attention has been paid to their home language acquisition. This approach often yields only partial information about DLLs’ language development. Fortunately, the Head Start Family and Child Experiences Survey 2009 (Child and Family Data Archive, n.d.; Malone et al., 2013) provides data on nationally representative Spanish-English DLLs’ vocabulary skills in both English and Spanish during the 2.5 years between entry into HS and K. This unique feature made it possible to study the longitudinal pattern of dual-language development.

Using FACES 2009 data, we studied DLLs who had two years of HS participation starting at the age of three. This study set out to investigate the following research questions (RQs):

1. What patterns of dual-language learning do Spanish-English DLLs present during the 2.5 years between HS and K?

2. How do home and classroom language exposure differ among DLLs presenting different dual-language learning patterns?

3. Do K academic skills differ among DLLs presenting different dual-language learning patterns?

Methods

Study Participants

This study used the Head Start Family and Child Experiences Survey 2009 (FACES, n.d.; US DHHS, ACF, OPRE, n.d.b). This survey is a longitudinal dataset on nationally representative children who entered HS in 2009. FACES 2009 contains data on both English and Spanish skills of low-income Spanish-English DLLs during 2.5 years between entry into HS and K and thus provides a unique resource for studying the longitudinal patterns of dual-language development.

Our study included 232 Spanish-English DLLs from 54 HS centers who attended HS for the first time at age three in 2009. The children were identified as Spanish-English DLLs if their parents reported that Spanish is their primary language when speaking to their children. Children were excluded from this study if they had no child assessment data at K, most often due to early program exit (n = 191); were not applicable to have Spanish assessments upon entry into HS (n = 6); or had identified speech or language impairment during their years in HS (n = 3) or K (n = 2). A kindergarten longitudinal sampling weight, “WESTATWT,” was applied across all analyses. The findings of the current study are, therefore, nationally representative of DLLs who entered HS in the fall of 2009 at the age of three, stayed in HS for two years and then went to K in 2012 (FACES/AIAN FACES Helpdesk, personal communication, February 14, 2022). All children in our sample had two years of HS experience, and their data were collected four times over 2.5 years, including HS entry (fall 2009), HS year 1 (spring 2010), HS year 2 (spring 2011), and K (spring 2012).

Measures

English and Spanish Language Skills

Children’s receptive vocabulary skills in English and Spanish were assessed using the Peabody Picture Vocabulary Test (PPVT-4; Dunn & Dunn, 2007) and the Test de Vocabulario en Imagenes Peabody (TVIP; Dunn et al., 1986), respectively. When administering the PPVT and TVIP, children are verbally instructed to point to the picture that best reflects the word that the assessor utters. The reliability of the PPVT and TVIP was over .91 and .89, respectively, for the total FACES sample across all data collection points (Malone et al., 2013). In the main analysis, we used raw scores obtained at K and gain scores over 2.5 years between HS entry to K (estimated by subtracting scores at HS entry from K scores) for each language. Thus, our analytical approach considered children’s skill levels at K and their longitudinal gains in vocabulary development across two languages.

Home Language Environment

Parents reported their home language environment at the time of their child’s entry to HS at age three. This study focused on five aspects of the home language environments relating to the quality and quantity of each language exposure: 1) parents’ primary language use when reading books to children; 2) the presence of home literacy materials in each language; 3) the parents’ proficiency in each language; 4) the extent to which parents value their children’s English or Spanish learning; and 5) the language spoken in the TV programs that the children watch. All the variables examined in this study are listed in Table 3.

Table 1. Model fit comparisons among 1-class to 4-class models.

	1	2	3	4
Loglikelihood	−4171.77	−4054.77	−4001.85	−3975.77
# of parameters	8	13	18	23
AIC	8359.54	8135.53	8039.70	7997.55
Δ AIC		−224.01	−95.83	−42.15
BIC	8387.99	8181.77	8103.72	8079.35
Δ BIC		−206.22	−78.05	−24.37
Adjusted BIC	8362.63	8140.56	8046.66	8006.44
Δ Adj. BIC		−222.07	−93.90	−40.22
Entropy	-	.87	.83	.82
VLMR p-value	-	<.01	<.01	.18
LMR p-value	-	<.01	<.01	.19
% Profile 1	100%	27.03%	25.48%	14.29%
% Profile 2		72.97%	20.85%	23.94%
% Profile 3			53.67%	9.65%
% Profile 4				52.12%

AIC = Akaike information criterion; BIC = Bayesian information criterion; ABIC = sample size adjusted BIC; VLMR = Vuong-Lo-Mendell-Rubin likelihood ratio test; LMR = Lo-Mendell-Rubin adjusted likelihood ratio test.

Table 2. Descriptive statistics of receptive vocabulary skills in English and Spanish (N = 232).

Vocabulary skills		Total			Balanced Bilinguals (54%)			Spanish-Dominant Bilinguals (21%)		English-Dominant Bilinguals (25%)
		#	M	SD	M	SD		M	SD	M	SD
English Receptive vocabulary skills (PPVT)
	HS entry Standard score	196	61.29	18.08	64.40	17.57	a	49.78	15.61	63.72	17.67	a
	HS entry Raw scores	196	14.46	12.52	16.51	12.96	a	7.11	7.53	15.87	12.77	a
	K Standard score	231	82.91	10.74	89.16	7.57	a	70.36	7.71	79.94	8.18	b
	K Raw scores	231	73.42	19.00	85.27	12.01	a	48.99	11.20	68.31	14.43	b
	Raw gain scores (HS entry – K)	196	60.02	16.93	69.57	13.21	a	41.94	9.82	54.22	13.99	b
Spanish Receptive vocabulary skills (TVIP)
	HS entry Standard score	207	88.17	10.38	89.32	10.92	a	88.63	8.60	85.02	10.06	b
	HS entry Raw scores	211	8.82	7.07	10.07	7.28	a	7.91	6.04	6.77	6.94	b
	K Standard score	217	89.82	14.58	95.59	9.93	a	94.93	9.16	67.80	8.40	b
	K Raw scores	231	34.69	14.07	41.76	8.32	a	40.07	8.07	15.88	9.39	b
	Raw gain scores (HS entry – K)	210	26.10	12.73	31.49	8.16	a	32.28	7.52	8.71	8.18	b

Unimputed data were used. To compare group differences in the children’s vocabulary skills, a series of simple regression was run with an analytical weight applied (WESTATWT) as a post hoc analysis. All significant group differences at p < .05 were denoted as a, b, and c, where the same alphabetic value reflects no group difference.

Table 3. Descriptive statistics of child/family characteristics, home language experiences, classroom characteristics, and kindergarten outcomes.

Characteristics		Total				Balanced Bilinguals (54%)		Spanish-Dominant DLLs (21%)		English-Dominant DLLs (25%)
		#		Mean/%	SD	Mean/%	SD	Mean/%	SD	Mean/%	SD
Demographics
	Child sex (1=male)	232		45.08%		43.30%		39.46%		53.19%
	Child race/ethnicity (1=Hispanic/Latino)	231		100%		100%		100%		100%
	Child country of birth (1= U.S.)	231		97.50%		99%		92%		98%
	Household poverty (1= yes)	232		74.67%		71.59%		87.26%		71.09%
	Household size	232		4.97	1.69	4.91	1.58	5.24	1.90	4.87	1.74
	Number of children in the household	232		2.37	.99	2.40	1.02	2.49	.99	2.23	.91
	Maternal education (1=high school or higher)	229		33.50%		35.76%		24.48%		36.16%
	Parents born outside the US (1= yes)	229		88.09%		89.03%		88.77%		85.59%
	Language often spoken to the child (1=non-English)	232		100%		100%		100%		100%
	Parents’ first language (1= Spanish)	231		98.62%		98.15%		100%		98.49%
Home Language experience
Parent language used in reading
	English (1= yes)	232		22.08%		26.59%		4.48%		26.67%
	Spanish (1= yes)	232		61.33%		58.03%		87.07%		47.78%
	Both languages equally (1= yes)	232		16.42%		15.38%		8.45%		24.87%
Home literacy materials
	# of material types in English	230		2.13	1.28	2.17	1.21	1.91	1.40	2.21	1.32
	# of material types in Spanish	228		2.22	1.01	2.32	0.97	2.34	1.11	1.92	0.94
	# of children’s books	232		14.30	12.57	15.21	13.69	10.70	9.25	15.29	12.08
	% of children’s books in Spanish	222		47.26	31.59	45.35	30.10	65.08	29.24	37.93	31.42
Parent language proficiency
Spanish proficiency (1= yes to “very well”)
	Read	228		61.44%		67.06%		56.58%		54.54%
	Understand	228		62.08%		67.51%		53.88%		58.09%
	Speak	228		62.15%		69.65%		49.87%		57.06%
English proficiency (0 = not at all ~ 3= very well)
	Understand	231		1.18	0.80	1.22	0.83	0.91	0.65	1.30	0.82
	Speak	231		1.05	0.80	1.10	0.80	0.76	0.64	1.16	0.85
	Read	231		1.02	0.88	1.07	0.87	0.72	0.81	1.16	0.92
Parent value (importance that …)
	Child communicates in English (1= essential)	232		29.85%		36.28%		22.72%		22.22%
	Child improves English (1= essential)	230		27.51%		33.91%		18.87%		21.42%
	Child communicates in Spanish (1=essential)	231		28.19%		31.48%		23.78%		24.84%
Language spoken on TV child watches
	English (1=yes)	232		84.84%		89.35%		67.57%		89.16%
	Spanish (1=yes)	232		53.36%		56.30%		60.10%		41.95%
Classroom characteristics
Year 1
Teacher Spanish proficiency (1= not at all ~ 4=very well)
	Speak	230		2.58	1.38	2.60	1.37	2.97	1.35	2.20	1.35
	Understand	230		2.63	1.39	2.68	1.39	2.97	1.34	2.24	1.35
Spanish is used for instruction (1=yes)		229		71.07%		70.52%		84.49%		61.20%
Year 2
Teacher Spanish proficiency (1= not at all ~ 4=very well)
	Speak	223		1.43	1.02	1.37	0.94	1.54	1.14	1.47	1.09
	Understand	223		1.44	1.02	1.37	0.94	1.56	1.14	1.47	1.09
Spanish is used for instruction (1=yes)		221		62.99%		78.34%		59.81%		57.07%
Kindergarten outcomes
	Conceptual expressive language (EOWPVT)		231	102.22	16.19	108.54	14.05	94.17	12.61	95.49	17.49
	Math (IRT based on WJ and ECLS-B)		232	36.75	6.92	39.45	5.42	33.01	8.29	34.14	6.15
	Literacy (IRT based on WJ and ECLS-B)		220	27.48	7.53	29.42	6.87	24.87	8.43	25.22	7.06

Unimputed data were used. A sampling weight was applied (WESTATWT). SD-DLL = Spanish-dominant DLL; ED-DLL = English-dominant DLL.

Primary Language Use in Reading

Parents reported the primary language they use when reading books to children by selecting one of three mutually exclusive responses: English, Spanish, or both languages equally. We examined each response separately using the yes vs. no indicator (1 = yes; 0 = no).

Home Literacy Materials

Parents reported whether their child had the following four materials in English or Spanish at home: (a) children’s comic books or magazines; (b) children’s computer programs or games; (c) adult books or magazines; and (d) CDs or tapes with singing. The FACES team created a total number of material types by each language. The composite score ranged between 0 (none) to 4 (all) for each language.

In addition, parents reported the number of children’s books at home in any language. The original variable ranged from 0 to 100. Because the variable was positively skewed, we truncated the variable so that it ranged from 0 (no books) to 50 (indicating more than 50 books).

Last, we used data on the percentage of children’s books in Spanish at home. This FACES-created variable indicates the relative number of children’s books in Spanish over in English at home.

Parent Language Proficiency

Parents reported their proficiencies in reading, understanding, and speaking in English and Spanish using a 4-point Likert-type scale from 0 (not at all) to 3 (very well). For Spanish, the data distribution of parent proficiency was highly skewed toward “very well” (64–65%); thus, these variables were re-coded into the following dichotomous variables: 1 (very well) or 0 (not very well). For parents’ English proficiency, we used the original Likert-type scale, including 0 (not at all), 1 (not well), 2 (well), and 3 (very well) to capture variations in the sample.

Parental Values

Parents reported how important it was for their children to communicate in English, improve their English skills, or communicate in their primary language (Spanish). Each variable was rated using a 4-point Likert-type scale from 0 (not at all) to 3 (essential). Because data distribution was highly negatively skewed, a dichotomous variable indicating 0 (not essential) and 1 (essential) was created for each variable.

The Language Spoken in TV Programs

Parents reported the language spoken in the TV shows their children watched. Parents reported 1 = yes or 0 = no for English and Spanish separately. These mutually inclusive responses were examined separately.

Classroom Language Environment

Teachers reported on their classroom environment in the spring of each academic year. All children in our study had two years of HS participation, and each year’s data were analyzed separately.

Teacher Proficiency in Spanish

Teachers reported their Spanish understanding and speaking proficiency using a 4-point Likert-type scale ranging from 1 (not at all) to 4 (very well).

Spanish Instruction

Teachers reported whether Spanish was used for classroom instruction (1 = yes; 0 = no).

Kindergarten Achievement

Conceptual, Expressive Vocabulary

EOWPVT Spanish-Bilingual Edition (EOWPVT-SBE; Brownell, 2001) was used to assess conceptual expressive vocabulary skills of children whose parents spoke Spanish to children at home (All children included in this study). Children were instructed to name objects, actions, and concepts in pictures in Spanish and English. A correct response in either language meant that the item was scored as “correct.” We used a bilingual-normed standard score (Std). The assessment’s internal reliability (Cronbach’s alpha) was .90 at K.

Math

We used IRT-based math scores estimated based on (a) the Applied Problems Subscale of the Woodcock – Johnson Tests of Achievement (AP WJ – III; Woodcock et al., 2001), which assesses numbers and operations, and (b) the math assessment from the Early Childhood Longitudinal Study, Birth Cohort (ECLS-B; Snow et al., 2007; U.S. Department of Education, 2002), which assesses knowledge of numbers and numeric operations such as addition and subtraction, size, pattern, and shape, as well as mathematical word problem-solving skills. The IRT score represents the number of items children would have correctly answered if both AP WJ-III and ECLS-B math assessments were given. The IRT score reflects a child’s true ability given “the ability estimate, along with the difficulty of the item, the discrimination power of the item, and the probability of guessing correctly” (Malone et al., 2013, p. 169). Even children who had not received all items due to basal or ceiling rules had IRT scores that presented their true ability. The internal reliability of the assessment was .95 at K.

Literacy

We used the IRT-based scores estimated based on WJ Letter-Word (Woodcock et al., 2001) and ECLS-B Letter-Sounds tasks to measure knowledge of letters, sounds, and words. The IRT scores represent the number of items children would have correctly answered if both WJ Letter-Word and ECLS-B Letter-Sound responses were given. The internal reliability of the assessment was .96 at K.

Analysis

Research Question 1

Latent profile analysis (LPA) was employed using Mplus 8 to identify the profile patterns of dual-language development (Asparouhov & Muthén, 2014; Nylund et al., 2007). Four variables were used in the analytical model: PPVT (English) scores at K, the PPVT gain scores between HS entry and K, the TVIP (Spanish) scores at K, and the TVIP gain scores between entry into HS and K. Separate LPA models were run with one to four profiles. The criteria used to determine the best model were: (a) Akaike Information Criterion (AIC; Bozdogan, 1987); (b) Bayesian Information Criterion (BIC; Schwarz, 1978); (c) Vuong – Lo–Mendell – Rubin likelihood ratio test (VLMR-LRT; Vuong, 1989); (d) entropy (Jedidi et al., 1993); and (e) that a reasonable proportion of children (no less than 1% of the sample) were assigned to each profile (Jung & Wickrama, 2008). To account for the nesting issue (i.e., that some children were enrolled in the same HS center), we used center IDs in the CLUSTER command in Mplus. Missing data were handled by full information maximum likelihood estimation. A sampling weight, “WESTATWT,” was applied.

Research Question 2

To explore whether children’s home and classroom environments differ across DLL profile subgroups, we ran a series of logistic regressions (when examining dichotomous variables) or simple linear regression (when examining continuous variables) using Stata 14. Each language environment was treated as an outcome variable, and dummy-coded DLL groups were included as independent variables. To test group differences across all DLL subgroups, we ran each model repeatedly with different DLL subgroups as a reference group. Across all models, we used “vce(cluster)” to adjust the standard error by allowing intragroup correlation at the HS center level. Given that this study is exploratory in nature, we did not adjust the p-value for multiple comparisons (Moran, 2003; Saville, 1990). To handle missing data, 50 imputed datasets were created using multiple imputations (MI). “WESTATWT” was applied as a sampling weight.

Research Question 3

To test whether children’s K achievement differed across the different DLL profile subgroups, we ran a series of simple linear regressions using Stata 14. Each K achievement was treated as an outcome variable. Dummy-coded DLL groups were included as independent variables. Each model was run twice against a different reference group to test differences across all DLL profile subgroups. Throughout all analyses, “vce(cluster)” was used to adjust standard errors by accounting for intragroup correlation at the HS center. The same imputed dataset used for RQ 2 was used for the analytical models examining RQ 3. “WESTATWT” was applied as a sampling weight.

Results

The Profile Patterns of Dual-Language Development

Table 1 summarizes model fit comparisons across the one- to four-profile models. AIC and BIC values were lower for the models with a greater number of profiles, indicating that the n +1 model fits the data better than the n-model (Kline, 2016). All models showed an entropy value over .80, which is acceptable (Lubke & Muthén, 2007). Both LMR and VLMR were statistically significant at p < .01 for the three-profile model, indicating that the three-profile model fits the data better than the two-profile model, and the model fit does not improve by adding one more profile (i.e., four-profile model) to the three-profile model (Nylund et al., 2007). The three-profile model included approximately 21%, 25%, and 54% of children in each respective profile group. Based on these model fit indices, we concluded that the three-profile model best fits the data. The three profile groups were labeled as Balanced Bilinguals, English-Dominant Bilinguals, and Spanish-Dominant Bilinguals based on their significant characteristics (i.e., the language of dominance at K, the language that the group developed more over the 2.5 years). All groups were similar in that their average standardized (std) scores in Spanish (TVIP) were higher than in English (PPVT) upon entry to HS (Table 2).

To better understand how the three groups of DLLs differ in their language skillsets at HS entry, K, and the gain scores (between HS entry and K), a series of simple regressions were performed with an analytical weight applied (WESTATWT) as post hoc analyses. All three groups were compared by changing the reference group. The group differences in English and Spanish skills are described below and presented in Table 2. Figure 1 visually presents mean scores of children’s English and Spanish receptive vocabulary skills presented in Table 2.

Figure 1. Comparison of English (PPVT) and Spanish (TVIP) receptive vocabulary skills (raw scores) across three groups of dual language learners (DLLs).

Note. The graph is to visually present mean scores of children’s English and Spanish receptive vocabulary skills presented in Table 2.

Balanced Bilinguals

The Balanced Bilingual group comprised approximately 54% of the children in the sample. As shown in Table 2, average Balanced Bilinguals had PPVT scores comparable to those of the English-Dominant group at HS entry (M _{Std score} = 64.40) but showed the highest gain scores between HS entry and K (M _{gain score} = 69.57) and eventually presented the highest skills at K (M _{Std score} = 89.16) among the three groups. Balanced Bilinguals’ TVIP scores were comparable to those of Spanish-Dominant Bilinguals across all three-time points: at entry into HS (M _{Std score} = 89.32), at K (M _{Std score} = 95.59), and the gains scores from HS entry to K (M _{gain score} = 31.49).

Spanish-Dominant Bilinguals

The Spanish-Dominant group comprised approximately 21% of the children in the sample. As shown in Table 2, the average Spanish-Dominant group had the lowest PPVT scores at HS entry (M _{Std score} = 49.78) and K (M _{Std score} = 70.36) as well as the lowest gain scores from HS entry to K (M _{gain score} = 41.94). Spanish-Dominant Bilinguals had TVIP scores comparable to those of English-Dominant Bilinguals at HS entry (M _{Std score} = 88.63) but had higher gains (M _{gain score} = 32.28) and showed higher skills at K (M _{Std score} = 94.93).

English-Dominant Bilinguals

The English-Dominant group comprised approximately 25% of the children in the sample. As shown in Table 2, the English-Dominant group’s PPVT scores were comparable to those of Bilinguals at HS entry (M _{Std score} = 63.72), but they showed lower gains (M _{raw gain score} = 54.22) and lower scores at K (M _{Std score} = 79.94) than Balanced Bilinguals. The English-Dominant group showed TVIP scores comparable to those of the Spanish-Dominant Bilinguals at HS entry (M _{Std score} = 85.02) but achieved the lowest gains (M _{gain score} = 8.71) and K achievement (M _{Std score} = 67.80) in TVIP among the three groups.

Group Differences in Home Language Environment

Table 3 presents descriptive statistics of the language environments at home (when entering HS) and in the classroom (during HS years). Table 4 shows whether and how the three groups had statistically different language environments. All parents of DLLs in our sample primarily spoke Spanish to their children.

Table 4. Language group differences in home and classroom language environment.

	Balanced bilinguals vs. English-Dominant DLL (ref.)						Balanced bilinguals vs. Spanish-Dominant DLL (ref.)					Spanish-Dominant DLL vs. English-Dominant DLL (ref.)
	b		OR	p-value	95% CI		b	OR	p-value	95% CI		b	OR	p-value	95% CI
Home Language experience
Parent language use in reading
English	0.00		1.00		-0.82	0.81	2.04	7.72	**	0.61	3.48	-2.05	0.13	*	-3.65	-0.45
Spanish	0.41		1.51		-0.42	1.24	-1.58	0.21	***	-2.39	-0.78	-2.00	7.36	***	-3.09	-0.90
Both languages equally	-0.60		0.55		-1.95	0.75	0.68	1.97		-0.39	1.75	1.28	0.28		-0.16	2.71
Home literacy materials
# of material types in English	-0.05				-0.41	0.31	0.25			-0.26	0.75	-0.29			-0.80	0.21
# of material types in Spanish	0.38			*	0.05	0.71	-0.04			-0.43	0.35	0.42		*	0.04	0.80
% of children’s books in Spanish	7.29				-3.05	17.63	-.19.77		***	-30.82	-8.72	27.07		***	12.99	41.14
# of children’s books	-.80				-4.09	3.93	4.51		*	.52	8.50	-4.59		*	-8.51	-.67
Parent language proficiency
Parent Spanish proficiency (1= very well; 0 = not very well)
Understand	0.41		1.51		-0.27	1.09	0.57	1.76		-0.24	1.37	-0.16	0.85		-1.12	0.81
Speak	0.53		1.71		-0.21	1.28	0.82	2.28		-0.04	1.68	-0.29	0.75		-1.25	0.68
Read	0.50		1.66		-0.24	1.25	0.43	1.53		-0.30	1.15	0.08	1.08		-0.90	1.06
Parent English proficiency (0 = not at all ~ 3= very well)
Understand	-0.08				-0.29	0.14	0.31		**	0.08	0.55	-0.39		**	-0.68	-0.10
Speak	-0.05				-0.26	0.15	0.35		***	0.14	0.55	-0.40		**	-0.66	-0.15
Read	-0.09				-0.35	0.17	0.35		*	0.05	0.64	-0.44		*	-0.80	-0.08
Parent value (importance that child …)
improves English skills (1= essential)		0.62	1.85		-0.22	1.45	0.77	2.17	*	0.04	1.51	-0.16	0.85		-1.11	0.79
communicates in English (1= essential)		0.69	1.99		-0.05	1.43	0.66	1.94		-0.02	1.34	0.03	1.03		-0.92	0.98
communicates in Spanish (1= essential)		.33	1.39		.76	2.53	.39	1.42		.68	3.17	-.06	.94		.44	2.03
Language spoken on TV child watches
English (1=Yes)	0.02		1.02		-1.37	1.41	1.39	4.03	**	0.38	2.41	-1.37	0.25	*	-2.59	-0.16
Spanish (1=Yes)	0.58		1.78		-0.19	1.35	-0.16	0.86		-1.08	0.76	0.73	2.08		-0.34	1.81
Classroom characteristics
Year 1
Teacher Spanish proficiency
Speak	0.40			*	0.00	0.81	-0.37			-0.99	0.25	0.78		*	0.08	1.47
Understand	0.43			*	0.01	0.85	-0.29			-0.92	0.34	0.72		*	0.03	1.42
Spanish used for instruction (1= yes)	0.44		1.55		-0.28	1.15	-0.82	0.44		-1.89	0.25	1.25	3.50	*	0.28	2.23
Year 2
Teacher Spanish proficiency
Speak	-0.09				-0.39	0.22	-0.17			-0.60	0.25	0.08			-0.34	0.50
Understand	-0.09				-0.39	0.22	-0.19			-0.62	0.24	0.10			-0.33	0.52
Spanish use for instruction (1= yes)	0.13		1.14		-0.74	1.01	-0.84	0.43		-2.14	0.46	0.97	2.65		-0.29	2.24

Imputed data were used. Group comparisons were performed for each home and classroom factor separately. When simple logistic regression was performed, Odd Ratio (OR) was reported. A sampling weight was applied (WESTATWT). * p < .05; ** p < .01; *** p < .001.

Primary Language Use in Reading

English

Approximately 4.48% of Spanish-Dominant Bilinguals were read to primarily in English, a significantly lower percentage than that of English-Dominant Bilinguals (26.67%; OR _{ref. English-Dominant} = .13; p < .05) and Balanced Bilinguals (26.59%; OR _{ref. Spanish-Dominant} = 7.72; p ≤ .01). Balanced Bilinguals and English-Dominant Bilinguals did not differ in the rate.

Spanish

Over 87% of the Spanish-Dominant Bilinguals were read to primarily in Spanish at home, a higher percentage than that of the English-Dominant Bilinguals (47.78%; OR _{ref. English-Dominant} = 7.36; p ≤ .001) and Balanced Bilinguals (58.03%; OR _{ref. Spanish-Dominant} = .21; p ≤ .001). Balanced Bilinguals and English-Dominant Bilinguals did not differ in the rate.

English and Spanish Equally

No statistically significant difference was found in the percentage of children who were read to equally in Spanish and English among the three groups (M _Range = 8% [Spanish-Dominant Bilinguals] to 25% [English-Dominant Bilinguals]).

Home Literacy Materials

Literacy Materials in English

No group difference was found in the number of different types of literacy materials in English at home (M _Range = 1.91 [Spanish-Dominant Bilinguals] to 2.21 [English-Dominant Bilinguals]).

Literacy Materials in Spanish

The English-Dominant Bilinguals had approximately two different types of home literacy material in Spanish (M = 1.92; SD = .94), fewer than the Spanish-Dominant Bilinguals (b = −.42; p < .05) and Balanced Bilinguals (b = −.38; p < .05) had.

Percentage of Children’s Books in Spanish

On average, less than 50% of children’s books at home were in Spanish (M = 47.26%; SD = 31.59). Spanish-Dominant Bilinguals had relatively more books in Spanish over English at home (M = 65.08%; SD = 29.24) than Balanced Bilinguals (b = 19.77; p ≤ .001) and English-Dominant Bilinguals (b = 27.07; p ≤ .001).

A Total Number of Children’s Books

On average, children had about 14 books at home (M = 14.30; SD = 12.57). Spanish-Dominant Bilinguals had fewer books (M = 10.70; SD = 9.25) than Balanced Bilinguals (b = −4.51; p ≤ .05) and English-Dominant Bilinguals (b = −4.59; p ≤ .05).

Parent Language Proficiency

Spanish

About 62% of parents reported they read, understand, or speak Spanish “very well.” The parents’ Spanish proficiency levels did not differ across the three groups.

English

On average, parents of Spanish-Dominant Bilinguals reported their English proficiency in comprehension (or understanding) (M = .91; SD = .65), speaking (M = .76; SD = .64), and reading (M = .72; SD = .81) to be “not well” (below 1). Parents of Spanish-Dominant Bilinguals rated their comprehension (understanding) lower than did the parents of Balanced Bilinguals (b = −.31; p ≤ .01) and the English-Dominant group (b = −.39; p ≤ .01). Their rates in speaking proficiency were also lower than those of the parents of Balanced Bilinguals (b = −.35; p ≤ .001) and English-Dominant Bilinguals (b = −.40; p ≤ .01). Similarly, their rates in reading proficiency were lower than those of the parents of Balanced Bilinguals (b = −.35; p < .05) and English-Dominant Bilinguals (b = −.44; p < .05). No difference in English proficiencies was reported between parents of English-Dominant Bilinguals and Balanced Bilinguals.

Parental Values

Nearly 34% of the parents of Balanced Bilinguals reported that their children’s English improvement was “essential ( = 1).” This percentage was about two times higher than the Spanish-Dominant Bilinguals’ parent reports (18.87%; b = .77; OR _{ref. Spanish-Dominant} = 2.17; p < .05). When parents asked whether it was important for their children to communicate in English or Spanish, no group difference was found.

The Language Used on TV

Spanish

No group differences were found in whether children watched TV in Spanish.

English

About 68% of Spanish-Dominant Bilinguals watched TV in English. This rate was lower than that for the Balanced Bilinguals (89.35%; b = 1.39; OR _{ref. Spanish-Dominant} = 4.03; p ≤ .01) and English-Dominant DLLs (89.16%; b = −1.37; OR _{ref. English-Dominant} = .25; p < .05).

When performing post hoc analyses, there was no group difference in time spent watching TV (M = 1.10 hours/day; SD = .76).

Group Differences and Classroom Language Environment

Teachers’ Spanish Proficiency

Year 1

On average, English-Dominant Bilinguals had teachers who rated their Spanish-speaking proficiency as “not well” ( = 2) (M = 2.20; SD = 1.35) in their first HS year. This rating was lower than those for teachers of Spanish-Dominant Bilinguals (b = −.78; p < .05) and Balanced Bilinguals (b = −.40; p < .05). Similarly, teachers of the English-Dominant group rated their Spanish comprehension (understanding) to be “not well” ( = 2) (M = 2.24; SD = 1.35), lower than teachers of Spanish-Dominant Bilinguals (b = −.72; p < .05) and Balanced Bilinguals (b = −.43; p < .05). Teachers of the Spanish-Dominant group and Balanced Bilinguals did not report different levels of Spanish proficiency.

Year 2

No difference was found in teachers’ ratings of their Spanish proficiency among the three DLL groups.

Spanish Use for Instruction

Year 1

In the first year of HS, nearly 85% of Spanish-Dominant Bilinguals were in classrooms where Spanish was used for instruction. This percentage was higher than that of English-Dominant Bilinguals (61.20%; b = 1.25; OR _{ref. English-Dominant} = 3.50; p < .05) but not different from that of Balanced Bilinguals (70.52%; p > .05). No group differences were found between English-Dominant Bilinguals and Balanced Bilinguals.

Year 2

About 63% of DLLs in our sample were in classrooms where Spanish was used for instruction in their second year of HS. No DLL group difference was found in the percentage of children in classrooms where Spanish was used for instruction in their second year of HS.

Kindergarten Achievement

Conceptual Expressive Vocabulary Skills

As shown in Table 5, Balanced Bilinguals presented higher conceptual, expressive vocabulary skills at K than English-Dominant Bilinguals (β = .81; p ≤ .001) and Spanish-Dominant Bilinguals (β = .90; p ≤ .001). English-Dominant Bilinguals and Spanish-Dominant Bilinguals did not differ in their conceptual expressive vocabulary skills assessed in both languages at K (β = −.09; p > .05).

Table 5. Language group differences in kindergarten outcomes (N = 232).

	Conceptual Expressive Vocabulary		Math		Literacy
	β	SE	β	SE	β	SE
Ref. Balanced bilinguals
English-dominant DLLs	−0.81***	0.16	−0.72***	0.13	−0.53**	0.16
Spanish-dominant DLLs	−0.90***	0.20	−0.87***	0.18	−0.66***	0.18
Intercept	0.33***	0.10	0.34***	0.07	0.26**	0.08
Ref. English-dominant DLLs
Spanish-dominant DLLs	−0.09	0.23	−0.15	0.21	−0.13	0.22
Intercept	−0.48**	0.16	−0.38***	0.10	−0.27	0.14

Imputed data were used. All outcome variables were z-scored (M = 0; SD = 1). A sampling weight was applied (WESTATWT).

Math

As shown in Table 5, Balanced Bilinguals demonstrated higher math skills at K compared to English-Dominant Bilinguals (β = .72; p ≤ .01) and Spanish-Dominant Bilinguals (β = .87; p ≤ .001). No difference was found between English-Dominant Bilinguals and Spanish-Dominant Bilinguals in their K math skills assessed in English (β = −.15; p > .05).

Literacy

As shown in Table 5, Bilinguals had higher literacy skills at K than English-Dominant Bilinguals (β = .53; p ≤ .01) and Spanish-Dominant Bilinguals (β = .66; p ≤ .001). English-Dominant Bilinguals and Spanish-Dominant Bilinguals did not differ in their literacy skills assessed in English at K (β = −.13; p > .05).

Discussion

Using FACES 2009, this study identified three groups of Spanish-English DLLs who showed different levels of K skills in English (PPVT) and Spanish (TVIP) and their gain scores during the 2.5 years in HS and K. The three groups were given the following labels: Spanish-Dominant Bilinguals, English-Dominant Bilinguals, and Balanced Bilinguals.

Spanish-Dominant Bilinguals

Spanish-Dominant Bilinguals presented gains scores and K achievement in Spanish comparable to those of Balanced Bilinguals and showed the lowest gains and the lowest K skills in English among the three groups. Spanish-Dominant Bilinguals are disadvantaged in the traditional and most common educational settings in the U.S., where instruction and interactions are predominantly in English. Indeed, research points to the importance of early English acquisition for DLLs’ later academic achievement (Choi, Jeon, et al., 2018; Choi, Rouse, et al., 2018; Han, 2012; Halle et al., 2012), indicating that DLLs may suffer disadvantages when their learning environment does not match their primary language.

However, it should be noted that the field acknowledges the benefit of a strong home language foundation for DLLs. Beyond the social and economic advantages of developing and maintaining a home language, strong home language skills benefit second language acquisition (Cummins, 1981; NASEM, 2017). Cummins’ linguistic interdependence hypothesis posits that a common underlying proficiency (or language-independent knowledge) facilitates the transfer of knowledge across languages, and thus, having a strong home language skill aids in the acquisition of a second language. Such a premise indicates that the Spanish-Dominant group, whose home language skills at K were comparable to those of monolingual Spanish speakers (on which the TVIP was normed), will acquire English more efficiently in the future using their home language skills as a foundation. When home language is continuously supported, the Spanish-Dominant group may eventually become balanced bilinguals as they receive more exposure to English in formal schooling. In this sense, Spanish-Dominant Bilinguals should be considered a group of young learners with developmental potential to excel, rather than an at-risk group for development. However, given the concern over home language loss often observed from DLLs in the U.S. school system (Oller & Eilers, 2002), systematic efforts supporting continuous home language development should be paralleled with purposeful English support even for DLLs who already achieved solid home language skills during early years.

It is worth noting that Spanish-Dominant DLLs’ K achievement did not differ from that of English-Dominant Bilinguals. When interpreting these findings, it is important to consider that Spanish-Dominant Bilinguals had significantly lower English proficiency than English-Dominant Bilinguals at K, and their K math and literacy achievement were assessed in the English version of the assessment. This points to the possibility that our Spanish-Dominant Bilinguals would have performed better if their academic skills had been assessed conceptually in both languages or in their dominant language (Spanish). Perhaps Spanish-Dominant Bilinguals eventually outperform their English-Dominant DLL peers in academics as they become more fluent in English through schooling. This speculation warrants further investigation.

English-Dominant Bilinguals

On average, English-Dominant Bilinguals did not show a pattern of stagnation or decline in home language development over the 2.5 years in HS and K. This may reflect the unique characteristics of the study setting: HS has a long history of committing to providing “developmentally, culturally, and linguistically appropriate” services (US HHS, ACF, Office of Head Start, 2016). However, English-Dominant Bilinguals gained the least Spanish skills among the three groups. English-Dominant Bilinguals were the only group whose TVIP standard scores were lower at K than at HS entry, indicating that English-Dominant Bilinguals’ development in Spanish was slower relative to their peers of similar age. Our findings further showed that although English-Dominant Bilinguals became more proficient in English over Spanish at K, their gain in English was lower than that of Balanced Bilinguals.

Our English-Dominant Bilinguals are somewhat comparable to López and Foster’s (2021) Balanced Low group, which presented low school readiness in both languages. As discussed earlier, English-Dominant Bilinguals performed similarly to Spanish-Dominant Bilinguals in math and literacy assessed in English, even though their K PPVT scores were significantly higher. These observations suggest that DLLs whose developmental characteristics resemble those of our English-Dominant Bilinguals may need the most support in their development. It is interesting that English-Dominant Bilinguals’ initial English skills were comparable to those of Balanced Bilinguals, and their Spanish skills were comparable to those of Spanish-Dominant Bilinguals. Perhaps, this indicates that English-Dominant Bilinguals’ lower functioning in both languages may be attributed to the limited opportunities they had for language exposure and use during their preschool years.

Balanced Bilinguals

On average, the Spanish skills of the Balanced Bilinguals in our sample were superior to their English skills upon entry to HS, and they grew to become Balanced Bilinguals by improving English skills while also developing Spanish skills over the 2.5 years in HS and K. They presented Spanish skills comparable to those of Spanish-Dominant Bilinguals and had the highest gains in English among the three groups. Balanced Bilinguals also had the highest achievement at K. Balanced Bilinguals presented above average age-normed standard scores in the conceptual expressive vocabulary skills (M _{standard score} = 103.21), and their math and literacy skills were comparable to those of the overall sample of children who entered HS at age 3 (Malone et al., 2013). These findings reinforce the scholarly consensus that learning two languages neither imposes a cognitive burden on children nor impedes their learning (Espinosa, 2013, 2015; NASEM, 2017).

The ECE field acknowledges that sole emphasis on English acquisition is not the optimal instructional approach for DLLs. The current scholarly consensus emphasizes that DLLs reap many benefits when continuously developing and maintaining their home language(s) while learning English (Espinosa, 2015; MacSwan et al., 2017; McCabe et al., 2013; Raikes et al., 2019; NASEM, 2017). Our findings corroborate such knowledge.

Language Environment

The wide variability in dual-language proficiency among DLLs who participated in HS raises the question of how their early language experiences differ. Our results uncovered several differences in the home and classroom environments. These findings contribute to the emerging evidence that DLLs benefit from consistent exposure to both languages (Barnett et al., 2007; Raikes et al., 2019). In line with the current literature, we found that if DLLs had more exposure to a language, they became more fluent in the language by K. In general, Spanish-Dominant Bilinguals had the most Spanish exposure at home (e.g., more of them were read to primarily in Spanish, and they had relatively more children’s books in Spanish over English), whereas their home environment provided less English exposure (e.g., fewer children were read to primarily in English, and they had less exposure to TV in English) and their parents had lower English proficiency than those of other groups. Furthermore, Spanish-Dominant Bilinguals, relative to English-Dominant Bilinguals, had more Spanish exposure in their classrooms during their first year in HS (e.g., Spanish was used for class instruction), and their teachers had higher Spanish proficiency. On the other hand, English-Dominant Bilinguals had more English exposure at home than Spanish-Dominant Bilinguals, but they had less Spanish exposure at home (e.g., fewer types of home literacy materials in Spanish) and in their first-year HS classrooms (e.g., Spanish was used for class instruction) compared to Spanish-Dominant Bilinguals.

Balanced Bilinguals, relative to Spanish-Dominant Bilinguals, had more English exposure (e.g., more children were read to primarily in English, their parents’ English proficiency was higher, and more children watched TV in English), and a higher percentage of Balanced Bilinguals’ parents reported that their children’s English improvement is essential to them. Compared to English-Dominant Bilinguals, Balanced Bilinguals had more Spanish exposure at home (e.g., they had more types of home literacy materials in Spanish) and were with teachers who were more proficient in Spanish during their first HS year. It is worth noting that although Balanced Bilinguals had more Spanish exposure in their homes and classrooms, they showed higher gains in English (PPVT) over the 2.5 years than did the English-Dominant Bilinguals. These findings may indicate that exposure to home-language-friendly environments does not delay DLLs’ acquisition of English skills but may even facilitate English acquisition when English development is simultaneously supported. It is also worth noting that English-Dominant Bilinguals, who had lower proficiency in both languages than Balanced Bilinguals, had similar English exposure but less Spanish exposure in their environment than Balanced Bilinguals. Such findings may indicate that home English exposure at the expense of home language exposure could contribute to slower growth in each language.

This study presents the importance of identifying and creating linguistically stimulating environments for DLLs’ learning in both languages. Although more research is needed to advance our understanding of effective environments for bilingual development (NASEM, 2017), existing literature offers many promising approaches to support DLLs. These recommendations highlight opportunities for proficient speakers’ frequent use of and exposure to both languages that do not diminish other language exposure and use (e.g., Hoff, 2018). Specific to ECE programs, the field has recommended providing frequent opportunities to converse with peers and adults, creating an environment where DLLs can be exposed to and use their home language (e.g., inviting adults who can speak DLLs’ language for book reading and conversation), providing intentional English and home language support (e.g., reviewing keywords in a child’s home language before reading a book), providing learning opportunities for English acquisition (e.g., pairing DLLs with English-speaking children; repeating key vocabulary words), and creating a welcoming environment (e.g., books, posters, visual aids representing different languages and culture; Castro et al., 2011; National Center on Early Childhood Development, Teaching and Learning, n.d.). An increasing volume of research has further shown the promise of two-way immersion programs, which offer academic instruction in a combination of English and the primary language spoken by children (e.g., Serafini et al., 2022).

These approaches highlight the importance of frequent high-quality language exposure to and use of English as well as DLLs’ home languages. However, we know little about whether such approaches are universally effective for all DLLs or whether such effectiveness differs depending on DLLs’ skillsets in each developing language. Given their unique skillsets and needs, Spanish-Dominant DLLs may need additional deliberate English support in the classroom to facilitate their learning and understanding of both languages, while English-Dominant DLLs may benefit more from frequent exposure to and use of their home languages in their classrooms. These speculations merit future research attention.

Limitations and Future Study

Several limitations should be considered when interpreting the findings of this study. First, criticisms of the TVIP should be noted. As reviewed by Wood et al. (2021), researchers have pointed out that TVIP’s normative sample, who were Spanish-speaking children living in Mexico and Puerto Rico, may not adequately represent DLLs growing up in the U.S. Wood et al. found that DLLs responded more accurately on items that were more likely to be used at home, indicating differential item functioning of items on the TVIP. Furthermore, TVIP has not been updated since 1980 and thus uses older features (e.g., black and white images and outdated technology depicted) that might depress students’ raw scores compared to PPVT, which has more modern features (e.g., color print). TVIP is widely used to assess the Spanish receptive vocabulary skills of DLLs living in the U.S., yet these concerns should be considered when interpreting the study findings.

Second, we used home language environment data collected at the time of children’s entry to HS. We acknowledge that home language practices and experiences can change as children’s skillsets develop. In FACES 2009, parents were asked to report their home language environment once, and the HS entry was the primary time point when the data were collected.

Third, our language environment variables broadly reflect the quality and quantity of the exposure and use of each language, but we cannot provide further details about actual experiences at home and in the classroom. Also, this study relies on parent and teacher reports on language environment variables, but we cannot verify that parents’ and teachers’ reports are reliable. We hope this exploratory study will encourage more scholarly research on the optimal environments for dual-language learning.

Fourth, our study focused only on DLLs whose home language was Spanish. Although Spanish-English DLLs are the largest DLL group in the U.S., we acknowledge that non-Spanish-speaking DLLs living in the U.S. are largely excluded from the discussion, even within DLL literature. These non-Spanish-speaking DLLs have their own difficulties and strengths, and they deserve more research attention.

Finally, given the correlational nature of this study, this study does not have the capability to draw causal inferences. For example, this study cannot provide causal evidence of bilingual benefits in children’s school readiness because children with more advanced skills across school readiness may also acquire vocabulary in both languages more rapidly than children with less advanced skills.

Conclusion

This study utilized nationally representative datasets from FACES 2009 that are unique in that they paid particular attention to the DLLs’ learning environments, experiences, and development in HS. This study contributes to the emerging body of research that aims to understand dual-language skillsets. Corroborating the view that DLLs are not a homogeneous group of children, the results showed that the pattern of dual-language acquisition differed widely during HS and K years. Such findings indicate the importance of continuous assessment in both languages that DLLs develop to understand their accompanying instructional needs and provide optimal support for their learning. Further, these findings indicate the need to better understand DLLs’ language experiences contributing to the different dual-language learning patterns. Our preliminary evidence suggests that both home and classroom language environments explain such differences in their dual-language learning during HS and K years. This study calls for additional research illuminating contextual factors promoting balanced bilingualism.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This study was supported by the American Educational Research Association’s Division G-Social Context of Education.