Connecting the Present to the Past

Terminology

An issue that has contributed to the controversies surrounding bilingual education has been the terminology used to address a growing population of bilingual students worldwide. As Table 1 illustrates, this group includes students who are labeled English language learners, dual language learners, English learners, students with limited English proficiency (LEP), and emergent bilinguals.

OPEN IN VIEWER

Table 1

Labels Referring to Bilingual Students Used in the Chapter or in the Reviewed Studies

Term	Definition
Dual language learner (DLL)	Students who are still developing language skills in Spanish at the same time that they are learning how to speak, read, and write in English (Center for Applied Linguistics, 2015)
Emergent bilingual (EB)	Children (ages 3–5) who speak a native language other than English and are in the dynamic process of developing bilingual and biliterate competencies with the support of their communities (García, Kleifgen, & Falchi, 2008)
English language learner (ELL)/English learner (EL)	Students whose first language is not English and who are in the process of learning English (Parrish et al., 2006); in this chapter, we use the term EL
English-only (EO)/non-LEP/non-ELL	Students who do not speak a language other than English at home (Guo & Koretz, 2013)
Limited English proficient (LEP)	Term used to identify students who have insufficient English to succeed in English-only classrooms (Goldenberg, Reese, & Rezaei, 2011)

Although these terms, in general, refer to students who speak a language other than English as their L1, each term reflects a somewhat different perspective on how the needs of bilingual students should be addressed and/or the priorities a state or school district has established to improve academic outcomes in an environment where the societal language is English. For example, the term emergent bilingual refers to children who have the possibility of becoming bilingual if they are instructed in both languages, whether they speak English or another language as their L1 (Garcia, Kleifgen, & Falchi, 2008). The term limited English proficient (LEP) has been used to refer to students who speak a language other than English as their L1 and who are limited in their English skills and therefore cannot benefit adequately from mainstream English instruction (Goldenberg, Reese, & Rezaei, 2011). The term has been criticized by researchers and practitioners who argue that it emphasizes the deficiencies of students who speak an L1 other than English rather than acknowledging that being bilingual likely increases their cognitive skills and provides them with better educational and economic opportunities (Agirdag, 2014; Calvo & Bialystok, 2013).

It is also difficult to determine the most appropriate term to refer to students who speak only English at home and are attending bilingual programs. These students are most commonly known as English-only (EO), non-LEP, or non–English language learners (non-ELLs; Guo & Koretz, 2013). In this chapter, for consistency and clarity, we use the term English learner (EL) to refer to U.S. students who do not speak English as their only language at home, and English-only (EO) to refer to U.S. students who speak only English at home. For studies that include participants outside the United States, we use the terminology used by the authors of those studies.

The terms that define bilingual programs are equally ambiguous. For example, programs where English and an L1 are used have been called bilingual, dual language, two-way immersion, transitional bilingual (early and late exit), and developmental, among other things. Although these terms describe very different bilingual programs (see Table 2), they are often used in similar ways, which makes it difficult for parents and others to understand the differences among programs and their potential advantages or disadvantages. For example, a transitional bilingual program (TBP) can be defined as a program where students are (a) taught entirely in their L1 for 1 or 2 years, (b) taught in their L1 throughout elementary school with some English language development (ELD) instruction (Slavin & Cheung, 2005), or (c) taught in their L1 for a certain percentage of the time each day (e.g., 30%) and in English for the rest of the time. In other words, in bilingual programs, the amount of time taught in each language in each grade can vary by school, classroom, or even time of year (D. L. Baker, Burns, Kame’enui, Smolkowski, & Baker, & 2015; Umansky et al., 2015). For the purposes of this chapter, we will use the following terminology to discuss the three most common kinds of programs currently in place: English immersion (EI) programs, where students receive mainly instruction in English; TBPs, in which students receive L1 instruction for a few years with some English instruction (e.g., ELD) before they are immersed in EO instruction; and dual language programs (DLPs), where students (ELs, or EOs and ELs) are taught in L1 and L2 from kindergarten with the goal of becoming balanced bilinguals (equally proficient in both languages). In our review of studies on bilingual programs we provide the specific amount of instructional time allocated to each language.

OPEN IN VIEWER

Table 2

Labels Referring to Bilingual Programs Used in the Chapter or in the Reviewed Studies

Program Label	Definition
50/50 Content model	Mathematics instruction is provided in all grades in English while social studies and science instruction is in Spanish. Language arts instruction is provided in the native language (L1) in pre-K and K and then in the second language (L2) in Grades 1–5. Language of instruction for non–content area–related instruction (announcements, calendar, physical education, etc.) alternates daily from English to Spanish. Students work in bilingual pairs in which English-dominant students are paired with Spanish-dominant students (Gómez, Freeman, & Freeman, 2005).
English as a second language (ESL)	Instruction aims to develop the English skills of students with limited English proficiency. ESL may or may not be content based (Guo & Koretz, 2013).
English for speakers of other languages (ESOL)/ESL	This bilingual model involves individualized instruction focusing on students’ acquisition of English language skills (Guo & Koretz, 2013).
ESL pullout	ELs are removed from the mainstream classroom for a portion of the day to receive individualized instruction with an ESL teacher. This method is more common in elementary schools (Guo & Koretz, 2013).
ESL push-in	An ESL teacher works with ELs in the larger mainstream classroom setting (Guo & Koretz, 2013).
Late exit/developmental/maintenance	ELs receive a substantial portion of educational instruction in their native language for several years of schooling. The goal is to develop both languages equally so as not to lose the L1 but rather use it to support L2 (Tong, Irby, et al., 2008).
Paired bilingual	Children learn to read in both English and their native language during different time periods each day or on alternating days. Within a few years, the native language reading instruction may be discontinued as children develop the skills to succeed in English (Slavin & Cheung, 2005).
Sheltered English immersion (SEI)	Nearly all classroom instruction is in English but with the curriculum and presentation designed for children who are learning the language. Although teachers may use a minimal amount of the child’s native language when necessary, no subject matter is taught in any language other than English, and children in this program learn to read and write solely in English (Commonwealth of Massachusetts, 2002). In the case of structured English immersion models (a form of SEI), teachers have a bilingual education endorsement or ESL credential and understand the students’ first language. In the law, SEI and structured English immersion are used interchangeably (Parrish et al., 2006).
Submersion	ELs are not offered any special language services, and instruction is entirely in English. This is often referred to as the “sink or swim” approach (Francis, Lesaux, & August, 2006).
Transitional bilingual program (TBP)	Children are taught to read primarily in their native language through the early elementary grades. Then they transition to instruction in English reading. The hallmark of TBPs is teaching reading in the native language for a period of time. Children in these programs are initially taught mostly in their native language and later are shifted gradually to English instruction (Guo & Koretz, 2013).
TBP early exit	Students transition to English instruction by second or third grade.
TBP late exit	Students receive native language instruction throughout elementary school (Slavin & Cheung, 2005). Late exit programs may use home language instruction 40% of the time or more throughout the elementary school years, even for students who have been reclassified as fluent English proficient (Parrish et al., 2006).
Academic language acquisition (ALA)	Instruction is divided into language blocks throughout the school day. The goal is for students to become proficient in English and in the general curriculum by using the L1. In general, there are three achievement goals of ALA: academic, linguistic, and psychosocial (Hoffstetter, 2004).
Extended foreign language (EFL)	Instruction is about 70% in English and 30% in Spanish. In the study reviewed here, students received 2 hours of English language arts, 30 minutes of independent reading, 30 minutes of social studies in English, 1 hour of mathematics in English, 1 hour of Spanish language arts, and 30 minutes of science in Spanish (M. G. López & Tashakkori, 2004b).
Two-way immersion	Both native English speakers and native speakers of the partner language are enrolled in this dual language program. Neither group makes up more than two thirds of the class. Core academic instruction is provided in both languages (Center for Applied Linguistics, 2015).
Two-way bilingual education	The goal is to facilitate the development of L2 skills while maintaining and enhancing L1 skills and the cultural integrity of students from both the minority and majority language groups (M. G. López & Tashakkori, 2006).

Historically, the question of whether bilingual instruction increases or reduces student learning opportunities has been controversial, and the most serious concerns about this type of instruction were not raised until 50 years ago with the immigration to the United States of large numbers of individuals whose first language is not English (D. L. Baker, Richards-Tutor, Gersten, Baker, & Smith, in press). According to the U.S. Census Bureau (2015), currently more than 400 languages are spoken in the United States; Spanish is by far the most widely spoken other than English (i.e., more than 37 million people speak Spanish), followed by Chinese (2.9 million), Tagalog (the official language of the Philippines, 1.5 million), Vietnamese (1.4 million), Korean (1.1 million), and German (1 million). The main concern in schools responsible for educating large numbers of ELs is the challenge of teaching and learning English in addition to the usual academic content.

To illustrate the problem, examination of recent data from the 2013 National Assessment of Educational Progress (NAEP; Kena et al., 2015) indicates that ELs scored significantly below EOs on the fourth- and eighth-grade assessments of reading (by 38 and 45 points, respectively) and mathematics (by 25 and 41 points, respectively). It is important to point out, however, that language proficiency can be confounded with low socioeconomic status (SES) and immigration status (Umansky & Reardon, 2014). Hispanics, for example, are more likely to live in poverty compared to Whites (33% of Hispanics are poor, compared to 13% of Whites; Kena et al., 2015). Hispanics also constitute the majority of ELs at 76.5% (Excelencia in Education, 2015). What is striking, however, is that most ELs who struggle academically are second- and third-generation students who were born in the United States (Hamilton, Martin, & Ventura, 2010). Thus, immigration cannot always fully explain low academic achievement, or English language proficiency. Moreover, NAEP data can also be complex to interpret because they do not adjust for ELs who have been reclassified as EOs (i.e., students who were ELs and are now judged to be fluent English proficient and sufficiently prepared for EO classrooms without EL instructional supports). Once reclassification occurs, it is more difficult to track students and determine, when they struggle, whether their low academic performance is due to continued difficulties in L2 acquisition or to learning difficulties, especially with academic content. Thus, it is possible that more ELs are struggling academically than the numbers indicated by the NAEP assessments (Umansky et al., 2015).

In summary, the diversity of the EL population and the lack of clarity in describing the bilingual programs and the students attending the programs (ELs and EOs) are important to address in future research, particularly considering the rapid increase of the EL population and the increase in the demand for bilingual programs in the country (K. López, 2014; Maxwell, 2012; Reljic, Ferrig, & Martin, 2015). Next, we summarize briefly the history of bilingual education in the United States to anchor our findings from the reviews cited, and to provide guidance for future research.

Historical and Sociocultural Background of Bilingual Education

Historically, the United States has been a bilingual country (August, Goldenberg, & Rueda, 2010), and bilingual education policies have shifted over time according to changing events (Crawford, 2004). Ovando (2003) argues that the shift in support or rejection of bilingualism through history has occurred not because of a shift in systematic ideas about language, but rather as a consequence of changing localized politics and social and economic forces. In the middle of the 19th century, bilingual education was supported and provided in more than 25 states. For example, schools could teach in Czech in Texas; in Spanish in the Southwest; in German in Colorado, Illinois, Indiana, Maryland, Missouri, Nebraska, Ohio, Oregon, and Pennsylvania; and in Swedish, Norwegian, and Danish in Illinois, Iowa, Minnesota, Nebraska, North Dakota, South Dakota, Washington, and Wisconsin. German was, by far, the most widely spoken language; by 1900, more than 600,000 children were being instructed in German (Kloss, 1977; Ovando, 2003). Stricter enforcement of linguistic assimilation started around 1880 with concerns about linguistic, cultural, and ideological competition between English and other languages (Salomone, 2010).

As a result, attitudes and legislation throughout the United States began to focus more on supporting the development of English language skills in schools. This approach became more widespread during World War I, when teaching in German and other foreign languages during school hours was forbidden because it was considered a threat to the state. The antagonistic attitude toward the teaching in foreign languages (particularly German) led to the Supreme Court case Meyer v. Nebraska (1923), in which the Court ruled that the prohibition of teaching in a foreign language in schools prior to eighth grade was unconstitutional. Similar prohibitions were annulled in Iowa and Ohio (Kloss, 1977).

After World War II, with the beginning of the Cold War, the educational focus shifted toward teaching math and science skills because they were essential for military, business, and diplomatic reasons. The lack of emphasis on teaching foreign languages, however, led to the National Defense Education Act of 1958, which had as one of its primary goals increasing foreign language education (Crawford, 2004). This shift in perspective coincided with the Cuban Revolution in 1959 and the Civil Rights Movement in the 1960s. The immigration of Cubans fleeing the revolution led to the creation of a very successful bilingual program at Coral Way Elementary in Dade County, Florida. Key to its success was the strong support of professional Cuban parents, well-trained Cuban teachers, and federal assistance through the Cuban Refugee Act (Crawford, 2004).

The Civil Rights Movement emboldened Mexican Americans and Puerto Ricans to fight against segregation and lack of educational opportunities for their communities. Although the movement to improve the conditions and opportunities for Hispanics has received less media attention than the movement to end racial segregation and discrimination against African Americans (Dauphine, 2014; Nuñez, 2009), the activities were extensive. An example of Hispanic efforts to end discrimination was Aspira. The movement was started in New York by activists in the Puerto Rican community, and its aim was to improve the living conditions and educational opportunities of Hispanics. Aspira is considered to be the first national educational and leadership development institution in the Puerto Rican community in the United States (Enck-Wanzer, 2006; Nuñez, 2009).

Another important historical event in support of bilingual education was the 1965 Immigration Act, signed into law by President Lyndon B. Johnson. This law revoked previous immigration policies that were based on a quota system with allocations assigned to nationalities according to their representation in past U.S. Census figures. The purposes of the new law were to reunite families whose members were separated by social and political conflicts and to attract skilled labor to the United States. As a result, 18 million immigrants (3 times more than in the preceding 30 years) entered the country, most of them from Asia and Latin America. This large immigration led to a rebirth of bilingual education as most immigrants spoke very little, if any, English (C. Baker, 2011). The same year, Congress also passed the Elementary and Secondary Education Act (ESEA), which provided assistance to educational agencies, through Title I, to support the education of children of low-income families. Although ESEA did not directly support bilingual education programs, it did provide additional funds to schools that served a large number of ELs who were also low-income.

Each of these historical events contributed to the eventual passage of the Bilingual Education Act of 1968, which was intended to assist school districts with the establishment of educational programs for immigrant children who did not speak English (Lyons, 2013). The Bilingual Education Act implied an acknowledgment by the federal government that immigrant children who did not speak English at home experienced unique language-based challenges in meeting educational goals (Gándara, 2015) and that, therefore, federal funding was needed to establish innovative educational programs for them. School districts could apply for the funds through grants intended for (a) purchasing educational programs, (b) training teachers and teacher aides, (c) developing and disseminating educational materials, and (d) involving parents in the education of their children. Funding, however, did not explicitly require schools to provide bilingual instruction (Stewner-Manzanares, 1988).

Nonetheless, bilingual programs began to proliferate, more to reduce social and economic inequalities than to develop the linguistic skills of all children, particularly ELs (Gándara, 2015). As a result, several civil lawsuits were filed alleging that equal opportunities were being denied to immigrant students. One highly influential case, Lau vs. Nichols (1974), was brought against the San Francisco school district alleging that 1,800 Chinese students were being denied equal education because of their limited English skills. The Supreme Court ruled in favor of the children, stating that children who did not understand the language of instruction were being denied equal educational opportunity and that, therefore, the district had violated the civil right of equal treatment for non–English-speaking students. School districts were subsequently required to provide “appropriate action” as evidence that their programs met the academic, linguistic, and sociocultural needs of students. If school districts were not in compliance with the guidelines, they would be denied federal funds (Ovando & Collier, 1998).

The term appropriate action, however, was not clearly defined. This lack of specificity led to Castañeda v. Pickard (1981), wherein the Fifth Circuit Court of Appeals provided a flexible and practical definition, stating that (a) a school district must use a program informed by education theory, (b) programs and practices need to be implemented with adequate resources and personnel, and (c) if the program implemented fails, or produces English proficiency without overcoming academic deficits, then the program does not constitute appropriate action (Lyons, 2013). The term appropriate action left room for later lawsuits focused on issues regarding what adequate instructional programs for bilingual students should entail. For instance, Gómez v. Illinois Board of Education (1987) claimed that it was the responsibility of the state to provide language remediation programs. Horne v. Flores (2009) involved parents suing the state of Arizona, its state board of education, and its superintendent for not providing adequate services to ELs.

One of the issues regarding the general terminology from the Castañeda v. Pickard ruling is that, to a certain extent, many programs can be justified by theory, and that to determine whether a program is adequate requires passage of time for evaluation. It is sometimes difficult to determine why, exactly, a program has failed, and for whom it has failed. The ambiguity in terminology has made it easier for states to implement a wide range of programs, as long as there is a theoretically sound reason, even if there is little or no evidence to support it (Lyons, 2013).

In 2001, Congress passed the No Child Left Behind (NCLB) Act, designed to increase the accountability of states and districts for the education of their students. NCLB required states to develop a set of standards and measurable goals to ensure that public school funding resulted in student performance gains in English. For ELs, NCLB allowed local districts to decide on programs of instruction (e.g., bilingual or EO), but it required high-stakes testing in English for all ELs after they had been in the U.S. education system for 3 years. Schools also had to demonstrate that all of their students, including ELs, were making “adequate yearly progress” in English (Menken, 2010). As a consequence, the test-based accountability system placed stronger emphasis on English acquisition without promoting the L1 of ELs. Thus, many schools decided to eliminate their bilingual programs in the belief that EO instruction was the best way to ensure that ELs would make yearly progress (Menken, 2013).

In December 2015, President Barack Obama signed the Every Student Succeeds Act, which was designed to replace NCLB and the 50-year-old ESEA. The Every Student Succeeds Act does not include any explicit provisions related to bilingual education or the education of bilingual students, and it shifts the responsibility of identifying the interventions and supports for struggling students to the states and local education agencies. Consequently, states and districts need to decide whether to make bilingual education programs available to ELs and EOs (and evaluate the quality of any such programs). Although it is too soon to determine the effects of the law on school programs for bilingual students and on student academic performance, history suggests that bilingual programs may continue to proliferate with or without the support of federal policies.

Recent evidence indicates that ELs and EOs can obtain significantly higher scores in reading in fourth grade if they attend schools in states with less restrictive EO rules and more support for bilingual education, including mandated trainings for all teachers (F. López & McEneaney, 2012). Although this study included only eight states, a similar study by F. López, McEneaney, and Nieswandt (2015), which included all 50 states and mathematics and science outcomes, found that Hispanic ELs living in states with higher proportions of Hispanics tended to demonstrate higher academic outcomes when there was a longer history of Hispanic presence in the state and when state-level policies allowed bilingual programs to exist. In addition, studies that have examined the academic performance of ELs in states that passed EO laws (e.g., Proposition 227 in California and the 2002 Massachusetts Act Relative to the Teaching of English in Public Schools; for the latter, see Commonwealth of Massachusetts, 2002) found that academic outcomes of ELs did not improve after these laws were passed (see Gándara & Hopkins, 2010; Guo & Koretz, 2013; Parrish et al., 2006).

In summary, bilingual education in the United States has been a contentious topic in both political and educational arenas. The recent evidence suggests, however, that restrictive policies against bilingual programs, instead of improving outcomes for ELs, can significantly affect their opportunities for educational attainment by reducing the funding and the additional supports and trainings required to address their needs, particularly the needs of ELs at risk for academic failure.

Next, we summarize the results of reviews on bilingual education designed to improve the academic performance of L2 learners. We include one meta-analysis conducted in Europe.

Summary of Reviews on Bilingual Education

In the past three decades, five meta-analyses (Francis, Lesaux, & August, 2006; Greene, 1998; Reljic et al., 2015; Rolstad, Mahoney, & Glass, 2005; Willig, 1985) and four other systematic reviews (K. Baker & de Kanter, 1981; Genesee, Lindholm-Leary, Saunders, & Christian, 2006; Rossell & Baker, 1996; Slavin & Cheung, 2005) have been conducted to examine the effects of bilingual education on the academic achievement of bilingual students. According to Goldenberg (2013), more meta-analyses and reviews have been conducted in bilingual education than in any other field. We summarize findings from these studies in Tables 3 and 4. Although findings from two of the earliest reviews indicated that EO programs produced better English outcomes (K. Baker & de Kanter, 1981) or that bilingual programs did not demonstrate greater effectiveness than EO programs (Rossell & Baker, 1996), each of the other reviews and meta-analyses obtained findings in support of bilingual programs.

OPEN IN VIEWER

Table 3

Review Studies in Chronological Order

Review	No. of Studies Reviewed	Types of Program Reviewed	Outcome	Effect Size
Baker and de Kanter (1981)	28	TBPs	EO programs were more beneficial. There was not sufficient evidence to support TBP.	NA
Rossell and Baker (1996)	75	TBP, structured immersion	EO programs were more beneficial.	NA
Slavin and Cheung (2005)	17	English immersion, TBP, paired bilingual model, two-way bilingual	BPs were superior, particularly paired bilingual programs.	ES = 0.45
Genesee, Lindholm-Leary, Saunders, and Christian (2006)	200	Two-way bilingual, developmental, ESL, SEI	BPs were more beneficial in oral language, literacy, and academic outcomes.	NA

Note. BP = bilingual program; EO = English-only; ES = median effect size; ESL = English as a second language; NA = not applicable; SEI, sheltered English immersion; TBP = transitional bilingual program.

OPEN IN VIEWER

Table 4

Meta-Analyses in Chronological Order

Meta-Analysis	No. of Studies Reviewed	Types of Program Reviewed	Outcome	Effect Size
Willig (1985)	23	Transitional bilingual, submersion	BP were as beneficial or sometimes more than EO programs.	ES a = 0.63 between bilingual programs and submersion
Greene (1998)	11	Transitional bilingual, structured immersion	ELs who are taught using at least some of their native language perform significantly better than similar children who are taught in English.	g = 0.21 for reading and g = 0.12 for math
Rolstad, Mahoney, and Glass (2005)	17	Transitional bilingual program, developmental bilingual	BP programs are favored, particularly developmental bilingual programs in reading, writing, mathematics, and language proficiency.	ES b = 0.23 bilingual education
Francis, Lesaux, and August (2006)	20	English-only, transitional bilingual, paired bilingual, two-way bilingual	Bilingual education has positive small to moderate effects on English reading outcomes.	d = 0.20 to 0.50
Reljic, Ferrig, and Martin (2015)	5	Bilingual, submersion, dual bilingual, transitional bilingual	Evidence points toward support for BPs as defined by programs that include the home language for the language minority students.	g = 0.23

Note. BP = bilingual program; d = Cohen’s d; EO = English-only; ES = effect size; g = Hedges’s g.

a

Median effect size. ^bMean effect size.

In the first meta-analysis of bilingual education research, Willig (1985) analyzed the same studies reviewed by K. Baker and de Kanter (1981) and found that bilingual programs appeared to be as beneficial as EO programs, if not more so, for supporting ELs’ English literacy development. Almost 14 years later, Greene (1998) obtained findings that corroborated those obtained by Willig (1985). Results of both meta-analyses favored bilingual programs, with effect sizes of ES = 0.21 for reading and ES = 0.12 for math. This consistency occurred even though only a few studies were the same in both meta-analyses (Krashen & McField, 2005).

More recent reviews and meta-analyses have obtained similar findings in favor of bilingual programs, and they have provided some additional insights about the benefits of specific bilingual programs in comparison with others (e.g., Francis et al., 2006; Genesee et al., 2006; Rolstad et al., 2005; Slavin & Cheung, 2005). Rolstad et al. (2005) found that developmental bilingual programs, where the goal is bilingualism and biliteracy, produced better results than TBPs, where the primary goal is for students to become proficient in English. Similarly, Genesee et al. (2006) found that bilingual programs with extended instruction in the two languages (e.g., dual language and late exit programs) appeared to be more effective than early exit bilingual programs. In addition, while much of the research in this area has focused on elementary-age students, results from the meta-analysis by Francis et al. (2006) indicated that elementary- and secondary-level students instructed in their L1 and English performed better on measures of English language and reading than students instructed only in English (d = 0.20 to 0.50). Most recently, Reljic et al. (2015) conducted a meta-analysis examining the effects of bilingual programs versus submersion programs in the societal language for language minority students living in Europe. The results indicated a small positive effect (g = 0.23) in favor of bilingual programs over submersion programs on reading outcomes. These findings, however, need to be interpreted cautiously because only five studies met the inclusion criteria and were included in the analysis.

In summary, this collection of evidence appears to indicate that bilingual programs may produce more positive English outcomes for ELs than EO programs. Two limitations, however, must be noted with this body of work as a whole. First, many of the studies included in the reviews and meta-analyses contained methodological problems, such as how students were assigned to experimental and comparison groups, lack of comparability between the two groups (e.g., variability in language proficiency or student background characteristics), and lack of information about the quality of the bilingual programs because fidelity of implementation data were not reported (Willig, 1985). Second, few of the studies provided readers with specific details about the bilingual program models that were implemented (e.g., amount of instruction in both languages, length of the bilingual program), and few provided information on how instruction was organized and delivered (Francis et al., 2006; Genesee et al., 2006). Consequently, the results of the meta-analyses and reviews have not allowed researchers and practitioners to determine exactly for whom such programs are the most beneficial and under what circumstances.

Next we review studies published since 2003, the last year of studies included in the Francis et al. (2006) meta-analysis and the Genesee et al. (2006) review. We include any studies that meet our criteria, and that examine the effects of bilingual programs on the academic performance of bilingual students who are learning an L2.

Selected Empirical Studies of Bilingual Programs Published Since 2003

Findings

In this section we review studies by domain. Effect sizes are included if they were reported by the authors. We decided to summarize the studies in the tables using the more general terminology to describe bilingual programs defined earlier for consistency and clarity. We did, however, include the amount of time spent on each language in bilingual programs and the length of bilingual interventions if applicable. We synthesize the reading studies first, followed by the studies in writing, science, and mathematics. Appendix B lists the measures used in the studies, with their abbreviations.

Reading Studies

Given the large number of studies with reading outcomes that met our inclusion criteria (n = 48), we divided them into five types: (a) studies that employed a randomized controlled trial (RCT) design, (b) studies that employed a quasi-experimental or descriptive design, (c) studies that examined the effectiveness of bilingual programs and/or interventions for at-risk students, (d) studies that examined the role of the quality of instruction within the context of bilingual programs and/or interventions, and (e) studies that examined the effectiveness of bilingual programs outside the United States.

Studies comparing bilingual programs using an RCT design

We found five studies published since 2003 that met our inclusion criteria and that compared the effectiveness of different bilingual programs (or interventions) using an RCT design, as indicated in Table 5. Of these, all but one (Slavin, Madden, Calderón, Chamberlain, & Hennessy, 2011) were conducted with students in the preschool/early elementary grades; two (Durán, Roseth, & Hoffman, 2015; Slavin et al., 2011) were longitudinal. The most common comparisons were between a mainly EO program and a DLP or a TBP. All findings from the preschool studies indicated no significant differences in English outcomes between students in the bilingual program (i.e., in either the DLP or the TBP) and students in the EO program. In the longitudinal study by Durán et al. (2015), ELs in the TBP had growth trajectories similar to those of ELs in the mainly EO program in English. In Spanish, however, ELs in the TBP maintained their Spanish advantage compared to ELs in the EO program, corroborating previous findings with ELs in the primary grades, suggesting that bilingual programs present an opportunity for ELs to maintain their L1 without a detriment to their English reading and language development.

OPEN IN VIEWER

Table 5

Reading Studies Comparing Bilingual Programs Using a Randomized Controlled Design

Study	Grade Level	N	Program Models Compared	Academic Measures	Outcomes
Barnett, Yarosz, Thomas, Jung, and Blanco (2007)	Preschool	131	Two-way immersion: Language of instruction alternates weekly; 8 hours/day for 200 school days	ENG: PPVT-3, WJ-R	ENG: no significant differences
				SP: TVIP, WM-R	SP: TWI performed significantly better on TVIP (ES = 0.61, 0.56) and rhyme recognition (ES = 0.45); significant gains on TVIP, phoneme deletion, and letter recognition for English-dominant in TWI
				Both languages: phoneme deletion, rhyme recognition
			English immersion: 6 hours/day for 180+ school days
Durán, Roseth, and Hoffman (2010)	Pre-K	31	Transitional bilingual: All instruction in SP during Year 1	ENG: PPVT-4, WMLS-R, EL-IGDIs	ENG: No significant differences between groups on ENG measures; both groups demonstrated little growth in letter-word identification
			EO: All instruction in English	SP: TVIP, WMLS-R	SP: TBP showed higher rates of growth on oral vocabulary, expressive vocabulary, and letter-word identification
Durán, Roseth, and Hoffman (2014)	Pre-K	28	Transitional bilingual: ENG introduced until rate of 70% SP/30% ENG	ENG: PPVT-3, WMLS-R, DIBELS, MAP	ENG: No significant differences
			PE: SP introduced until 30% SP/70% ENG	SP: TVIP, WMLS-R	SP: TBP had significantly higher scores on TVIP (η2 = 0.41), WMLS-R letter-word identification (η2 = 0.30), WMLS-R picture naming (η2 = 0.17), and demonstrated significantly greater growth on DIBELS LNF (p < .05)
Durán, Roseth, Hoffman, and Robertshaw (2013)	K	31	Transitional bilingual: ENG introduced until rate of 70% SP/30% ENG	ENG: PPVT-4, WMLS-R, DIBELS PSF and NWF	ENG: No differences between TBP and PE programs on any measure
			PE: SP introduced until 30% SP/70% ENG	SP: TVIP, WMLS-R, picture naming, letter word	SP: Students in TBP demonstrated significantly greater growth on all Spanish measures
Farver, Lonigan, and Eppe (2009)	Preschool	94	Transitional bilingual: High Scope + Literacy Express Preschool Curriculum in SP until Week 9, transitioning to ENG over 3- to 4-week periodEO: High Scope + Literacy Express Preschool Curriculum in ENG (20 minutes/day, 4× per week)	ENG: P-CTOPPP	ENG: Students in the TBP intervention and EO intervention had significantly higher vocabulary, blending, elision, and print knowledge scores than control group (p < .01); TBP intervention group had significantly higher vocabulary and print knowledge scores than EO intervention group (p < .01)
				SP: P-CTOPPP	SP: TBP had significantly higher vocabulary, blending, elision, and print knowledge scores than EO and control groups (p < .01)
			Control: High Scope Curriculum
Roberts (2008)	Preschool	33	Storybook reading intervention: Primary language storybook reading at home; English storybook reading at home, plus all instruction and classroom routines in ENG	ENG: PPVT-3, pre-IPT, researcher-developed storybook and weekly vocabulary tests	ENG: Students who received primary language storybooks at home identified significantly more ENG storybook words than students who received English storybooks at home and at school (d = 0.87 to 2.38)
				SP: TVIP	SP: TVIP scores increased over time, and there were no significant differences in PPVT and TVIP scores
Slavin, Madden, Calderón, Chamberlain, and Hennessy (2011)	Grades 1–4	482	Transitional bilingual: SP in K + ESL instruction; transition to ENG began, on average, in Grade 2; all ENG by Grade 4	ENG: PPVT, WLPB-R	ENG: Scores were higher in SEI group across Grades 1–4 (ES = 0.08 to 0.31)
			Structured English immersion: All instruction in ENG	SP: TVIP, WLPB-R	SP: Significantly higher TVIP and WLPB-R scores for TBP students in Grades 1–4 (ES = 0.04 to 0.65); differences were not significant in Grade 4 in English, and only significant in Spanish passage comprehension favoring students in TBP (ES = 0.39)

Note. DIBELS = Dynamic Indicators of Basic Early Literacy Skills; EL-IGDI = Early Literacy Individual Growth and Development Indicators; ENG = English; EO = English-only; ES = effect size; IPT = IDEA Oral Language Proficiency Test; MAP = Measures of Academic Progress; NWF = Nonsense Words Fluency; P-CTOPPP = Preschool Comprehensive Test of Phonological and Print Processing; PE = predominantly English instruction; PPVT = Peabody Picture Vocabulary Test; pre-IPT = preschool IDEA Oral Language Proficiency Test; PSF = Phoneme Segmentation Fluency; SEI = structured English immersion; SP = Spanish; TBP = transitional bilingual program; TVIP = Test de Vocabulario de Imágenes; TWI = two-way immersion; WJ-R = Woodcock Johnson–Revised; WMLS = Woodcock Munoz Language Survey; WLPB-R = Woodcock Language Proficiency Battery–Revised; WM-R = Woodcock Muñoz–Revised; WMLS-R = Woodcock Muñoz Language Survey–Revised.

The Farver, Lonigan, and Eppe (2009) study compared the effect of an intervention in Spanish for ELs in the TBP program to the effect of the same intervention in English for students in the EO program; ELs in a control group did not receive the intervention. The findings indicated that ELs in the intervention in Spanish or English outperformed ELs in the control group on English vocabulary, blending, elision, and print knowledge tasks, suggesting that even for students at risk, interventions in the L1 can improve English reading skills. Roberts (2008) also found that preschoolers who received storybook reading in English at school, and in the L1 (i.e., either Spanish or Hmong) at home, had significantly higher knowledge of the storybook words than did preschoolers who received storybook reading in only English at home and at school, confirming that providing L1 supports either in preschool or at home is effective in helping children improve their reading skills in English.

Slavin et al. (2011) found that at the end of Grade 1, ELs in the TBP outperformed ELs in the EO programs on all the Spanish measures, but ELs in the EO program outperformed ELs in the TBP on all the English measures. At the end of Grade 4, however, the two groups performed similarly on all the English measures, but ELs in the TBP outperformed ELs in the EO program on Spanish reading comprehension (ES = 0.39). In summary, these studies corroborate the findings from the previous meta-analyses suggesting that bilingual programs help maintain and potentially even improve reading skills in students’ L1, as well as support their learning of English. In addition, given that all these studies were RCTs, findings can be causally related to outcomes.

Studies comparing bilingual programs using quasi-experimental or descriptive research designs

As Table 6 indicates, we located 21 studies published since 2003 that met our inclusion criteria and used quasi-experimental research designs (i.e., with comparison groups and/or controlling for pretest performance) or correlational research designs (i.e., where hypothesized causal relations were identified, but research design elements that were needed to make causal inferences, such as pretests or comparison groups, were absent; Shadish, Cook, & Campbell, 2001). Nine of these studies focused on younger students (Grades K–2), eight focused on upper primary students (Grades 2–6), and four focused on longer trajectories (Grades K–5, K–8, or K–11; Hofstetter, 2004; Reese, Gallimore, & Guthrie, 2005; Umansky & Reardon, 2014; Valentino & Reardon, 2015). The majority of the studies conducted in the early grades used primarily measures of phonological awareness, word reading, and vocabulary to determine program effectiveness. Three of the studies in the lower grades found that students receiving bilingual instruction outperformed their peers receiving English instruction with or without minimal ELD supports (Anderberg & McSparran Ruby, 2013; Lindholm-Leary, 2014; M. G. López & Tashakkori, 2004b). Results of the other five studies indicated either no significant differences between bilingual programs and EO programs or higher scores in English for ELs who participated in the English program and higher scores in Spanish for ELs who participated in the bilingual program. Six of the eight studies in the upper primary grades (Grade 2 and above) found that students in bilingual programs significantly outperformed their peers receiving EO instruction on English outcomes. Studies that included Spanish outcomes found, as expected, that Spanish outcomes were higher for students who participated in the bilingual programs than for students who participated in the EO program. Two of the eight studies in the upper primary grades (F. López, Scanlan, & Gorman, 2015; M. G. López & Tashakkori, 2006) found no difference at posttest on English outcomes between the bilingual program and the EO program.

OPEN IN VIEWER

Table 6

Reading Studies Comparing Bilingual Programs Using a Quasi-Experimental or Descriptive Design

Study	Grade Level	N	Program Models Compared	Academic Measures	Outcomes
Anderberg and McSparran Ruby (2013)	Preschool	45	Transitional bilingual: 20-minute ENGDual language: 50% SP/50% ENG	ENG: PPVT-3	ENG: Students in all three programs made significant gains from pretest to posttest; 78% of students in TBP made gains from pretest to posttest, compared to 54% in DLP and 51% in EO
				SP: TVIP	SP: No significant gains for any group
			EO: All instruction in English
Berens, Kovelman, and Pettito (2013)	Grades 2–3	213	Dual language: 50% SP/50% ENGDual language: 90% SP/10% ENGEO: All instruction in English	ENG: Phonological awareness tasks, CORE regular and irregular word reading, WLPB-R, LCEP	ENG: Students in 90/10 DL significantly outperformed 50/50 DL on phonological awareness and regular word reading tasks (p < .05), but 50/50 DL obtained significantly higher scores on irregular word reading (p < .05)
				SP: Phonological awareness tasks, CORE regular and irregular word reading, WLPB-R, LCEP	SP: Students in 90/10 DL significantly performed 50/50 DL on all tasks ( η p 2 = 0.11)
Carlo, Barr, August, Calderón, and Artzi (2014)	Grades 3–5	102	Transitional bilingual (early exit): All instruction in SP then transitioned to ENG in Grade 3	ENG: LAS-O, CAAS, WLPB-R	ENG: In Grade 3, EO significantly outperformed TBP-E and TBP-L on reading comprehension (p < .05), but by end of Grade 5, EO and TBP-E were performing comparably (and significantly better than the TBP-L group, p < .05); EO also significantly outperformed TBP-E and TBP-L on the CAAS and LAS-O (while the bilingual programs did not differ significantly); TBP-E grew at a significantly faster rate than EO on reading comprehension (ES for growth were 1.00 for TBP-E when compared to EO and .47 when compared to TBP-L); initial level of SP reading comprehension was a significant predictor after controlling for home language use and SES, indicating that students who started with higher skills demonstrated greater growth (p = .01 for TBP-L and p = .05 for TBP-E)
			Transitional bilingual (late exit): All instruction in SP then transitioned to ENG in Grade 5	SP: WLPB-R	SP: In Grade 3, both bilingual groups outperformed EOs (p < .05), but did not differ significantly from each other
			EO: All instruction in English
Gómez, Freeman, and Freeman (2005)	Grades 3–5	240	50–50 content model: Mathematics instruction in ENG; science, and social studies instruction in SP; language arts instruction in native language in pre-K and K and then in students’ L2 in Grades 1–5	ENG: TAKS	ENG: Of 117 Spanish-dominant Grade 3 students tested, 103 (88%) met proficiency standard; of the 56 English-dominant Grade 3 students tested, 51 (91%) met proficiency standard; overall, 154/173 (89%) met Grade 3 proficiency standard on TAKS Reading
Hammer, Lawrence, and Miccio (2007)	Pre-K–K	88	EO: ENG primary language of instruction for all studentsHEC: Home ENG communicationSEC: Students with SP as home language who had no prior ENG experience placed in classrooms where teacher or assistant spoke SP	ENG: PPVT-III, TELD-3, TERA-2, WLPB-R	ENG: Although SEC group started significantly lower than HEC group on ENG vocabulary (PPVT-III and TELD-3 component score) with p < .05, growth rates were comparable to HEC group; larger growth rates in SP during Head Start resulted in higher reading TERA-2 and ENG and SP WLPB-R letter-word identification scores
				SP: TVIP, PLS-3, WLPB-R	SP: SEC group significantly outperformed HEC group on SP vocabulary tasks at the end of Head Start (p < .05); early reading abilities in end of pre-K significantly predicted SP letter-word identification in K
Hofstetter (2004)	K–5	441	ALA: Instruction in K–1, 70% in L1, 30% in L2; percentage of ENG instruction increases until Grades 4/5, where ratio of instruction is 85% ENG and 15% SP	ENG: CELDT, SAT-9SP: SABE/2	ENG: SEI students earned higher scores on all CELDT subtests, and SAT-9 reading, than ALA students; ALA students who participated in Success for All (SFA) scored higher on SAT-9 than non-SFASP: ALA students scored in upper percentiles on SABE/2 and scored higher on SABE/2 compared to ENG SAT-9; ALA students who participated in SFA scored similarly to non-SFA students in reading.
			SEI: Students receive instruction predominantly in ENG, with 10% to 15% of instruction in ELD, 75% using sheltered ENG strategies
Lindholm-Leary and Block (2010)	Grades 4–6	Study 1: 193	DLP: 90% in SP in K–1, 80% in Grade 2, 70% in Grade 3, 60% in Grade 4, and 50% from Grade 5 onEO: Control group	ENG: ELA	ENG: In Study 1, there were no significant differences between scale scores of EP and EL students by program type; in Study 2, both EP and EL students in a DLP achieved at significantly higher levels than their peers in EO on the ELA CST (d = 0.71 for the EP group, and d = 0.35 for the EL group).
		Study 2: 466
				SP: Aprenda	SP: EP and EL students in DLP scored above grade level (and the state average) on the Aprenda reading subtest
Lindholm-Leary (2014)	K–2	283	Bilingual: (pre-K–2) 20% to 50% instruction in ENGBilingual/ENG: Bilingual pre-K (20% to 50% instruction in ENG), ENG in Grades K–1ENG/bilingual: Pre-K, bilingual in Grades 1–2 (20% to 50% instruction in ENG)English-only: ENG instruction Pre-K–2	ENG: PreLAS/LAS, CELDT	ENG: Students who received bilingual instruction in K–3 did not perform significantly worse than students receiving EO instruction (rather, they showed more growth across the grades compared to EO students); the greatest rates of growth were for the bilingual/ENG group; students who were Mostly Proficient in SP scored significantly higher on the CELDT than students who were Mostly Limited in SP ( η p 2 = 0.14)
				SP: DRDP, PreLAS/LAS	SP: Students receiving ENG instruction who began pre-K with significantly higher SP PreLAS scores performed significantly better (p < .01) than students receiving bilingual instruction and maintained higher scores for K and 1 (similar scores by program type in Grade 2); students in bilingual preschool performed comparably (bilingual and bilingual/ENG) and significantly lower than EO but demonstrated significantly greater gains after 1 year of pre-K (compared to the control group)
F. López, Scanlan, and Gorman (2015)	Grades 3–5	995	Dual bilingual: 90% SP/10% ENG in K; ENG introduced incrementally through Grade 4 (until 50% SP/50% ENG)	ENG: State reading assessment (fall/spring)	ENG: No significant differences in slopes between dual bilingual and EO classrooms, indicating that the relation between language modeling and reading achievement was the same in these settings; the slope of this relationship was weaker, however, in dual language classrooms
			Dual language: 50% SP/50% ENG starting in K (1 week SP instruction followed by 1 week ENG instruction)
			EO: All instruction in English
M. G. López and Tashakkori (2004a)	K–1	215	TWI: 30% SP/70% ENG	ENG: KAG, ERSA, High Frequency Word List, two writing samples, SRI	ENG: In K, students in the ESOL (control) group performed significantly better than students in the TWI (treatment) group on alphabet knowledge (η2 = 0.31); in Grade 1, no significant differences were found on any of the measures at posttest.
			English for speakers of other languages: 10% SP/90% ENG
M. G. López and Tashakkori (2004b)	K–1	66	EFL: 30% SP/70% ENG	ENG: SRI, KAG, High Frequency Word List, OLPSI-R	ENG: In K and 1, significant differences were observed on measures of alphabet knowledge, high-frequency word reading, and writing ( η p 2 = 0.001 to 0.12), but all differences favored the comparison (EO) group; despite being at greater levels of risk at pretest in K, students in the EFL group made sufficient progress, so that there were no statistically significant differences by the end of Grade 1
			EO: All instruction in ENG except 2.5 hours/week of SP language arts
M. G. López and Tashakkori (2006)	Grade 5	344	Two-way bilingual: 40% SP/60% ENG	ENG: FCAT	ENG: No significant differences in performance on FCAT reading for students in the two groups, but results indicated that students who entered with lower levels of ENG proficiency in K or Grade 1 scored significantly lower than students with higher levels of ENG proficiency (η2 = 0.11); the main effect of program type on oral language development was significant, with students in TWBE requiring less time to learn ENG than students in TBP (η2 = .02); students with higher initial levels of English proficiency exited ESOL program faster (η2 = 0.22), and students with lower levels exited faster in the TWBE program than in the TBP program (η2 = 0.04)
			Transitional bilingual: SP and ENG until students attained a level 3 in ENG, then option of receiving 2.5 hours/week of SP instruction
				SP: EDL	SP: The effect of program type was statistically significant, with TWBE outperforming TBP students on the measure of SP reading fluency (η2 = 0.26)
Marian, Shook, and Schroeder (2013)	Grades 3–5	2,009	TWI SP: SP instruction in Grades K–2, ENG in Grade 3	ENG: SMAG; SSAT	ENG: TWI-SP students showed significant improvements across grade levels in reading (η2 = 0.24); similar improvements not observed in the TPI group
			TPI: All instruction in ENG with pull-out ESL instruction as needed
			TWI ENG: ENG instruction in Grades K–2; SP in Grades 3–5
			EO: All instruction in English
Proctor, August, Carlo, and Bar (2010)	Grades 2–5	101	SpO: All instruction in SPEO: All instruction in ENGTBP: Transitioned out of SP into ENG after Grade 2, 3, or 4 (had attained Grade 2 SP literacy proficiency)	ENG: WLPB-R	ENG: EO and TBP students significantly outperformed SpO students in reading comprehension (p < .05)
				SP: WLPB-R	SP: SpO and TBP students had higher reading comprehension and oral vocabulary scores than EO students (p < .01), and SpO had significantly higher scores than TBP (p < .01); average reading growth for SpO and EO was comparable (but EO was 2 SD below the mean, showing absence of SP proficiency); SpO showed greater growth than TBP (p < .01) but no differences in growth for TBP and EO
Reading (2009)	K–1	101	French immersion: No ENG instruction in K; 1 hour/day of ENG literacy instruction in Grade 1	ENG: DIBELS ISF, LNF, PSF, NWF, ORF	ENG: NWF scores were higher for SP immersion compared to EO (d = 0.37, p < .05); median subtest scores on all measures were higher for SP immersion than for French immersion at the beginning of Grade 1 (d values for LNF, PSF, and NWF are 0.47, 0.44, and 0.46, respectively, p < .05 for all); significantly fewer SP immersion students were identified as needing intervention compared to French immersion students
			Spanish immersion: No ENG instruction in K; 1 hour/day of ENG literacy instruction in Grade 1
			EO: ENG reading instruction using commercially available phonics program
Reese, Gallimore, and Guthrie (2005)	K–8	91	TBP: Move to EO when reading ability in SP is at the third-grade level and have conversational ability in ENG	ENG: CTBS, MAT, IDEA Proficiency Task, BSM, ITBS, TAAS, ITAS	ENG: Students who were identified as high achievers in SP transitioned earlier and, despite experiencing a sharp decline in Grades 2–3 (when the transition occurred) recovered earlier and more quickly; scores for students identified as average or low achievers in SP in K began to decline in SP prior to the transition to ENG and continued to decline after the transition; by Grade 7, grade of transition to ENG was negatively correlated with performance (r = .48, p < .0001).
				SP: SABE, CTBS-S, Prueba de Lectura Inicial	SP: Students who received intensive instruction in K outperformed their peers through Grade 3 (p < .01), after which their advantage decreased and was no longer significant
Reese, Goldenberg, and Saunders (2006)	K–2	83	Transitional bilingual (early): Reading/language arts and other content area instruction provided in SP for approximately 1–3 years, then transition to EO	ENG: WLPB-R	ENG: Students in programs that emphasized ENG instruction scored higher in ENG
			DB: SP instruction through Grades 2–3, when ENG is added	SP: WLPB-R	SP: Students in programs that emphasized SP instruction scored higher in SP
			DLP: SP instruction in Grades K–1, ENG introduced in Grade 2, then 50% ENG/50% SP in Grades 3–5
			EI: All instruction in English
Uchikoshi and Maniates (2010)	Grade 2	67	Transitional bilingual: 10% Cantonese/90% ENGTransitional bilingual: 40% SP/60% ENG	ENG: PPVT-3, WLPB-R	ENG: Students in Cantonese/ENG TBP performed, on average, similarly to monolingual ENG norms for Grade 1 expressive vocabulary, while SP-speaking students in SP/ENG TBP performed similarly to monolingual norms for ENG kindergarten; students in both TBPs performed about 1 SD below the mean on receptive vocabulary and listening comprehension
				SP: TVIP, WLPB-R	SP: Students scored 0.5 SD below monolingual SP norms for receptive vocabulary and listening comprehension and had expressive vocabulary scores comparable to those of monolingual SP students in K
				Cantonese: PPVT-3	Cantonese: Students scored about 1.5 SD lower than monolingual Chinese norms for receptive vocabulary
Umansky and Reardon (2014)	K–11	5,423	Transitional bilingual: 80% to 90% SP instruction in K, with amount of ENG instruction increasing each year until Grade 3, when all instruction is in ENG	ENG: CELDT, CST (ELA)	ENG reclassification: Significantly larger proportions of students in English immersion programs demonstrated ENG proficiency on the CELDT and were reclassified in elementary school than in other programs; however, the proportion of students in transitional bilingual and maintenance programs reclassified as ENG proficient was significantly greater than the proportion of students in EI programs in middle school, and DI surpassed EI by high school
			SP maintenance bilingual: 80% to 90% SP in K, transitioning to 50% to 60% ENG by Grade 5
			SP DI: Mainly SP instruction in K to 50% SP/50% ENG by end of elementary; some programs continue through middle/high school		ENG reading: Students in EI significantly outperformed students in other programs in elementary school, but DI students significantly outperformed EI by Grade 5 (continuing to show advantages throughout middle and high school) and maintenance bilingual students significantly outperformed EI students by Grade 6
			EI: All instruction in ENG
Valentino and Reardon (2015)	K–8	13,750	Transitional bilingual: SP instruction in K and transition to ENG by Grades 2–3DB: In K, 50% to 90% of instruction in target language, proportion of ENG instruction increasing each yearDI: 80% to 90% of instruction in target language in K–1, transitioning to 50% ENG and 50% target language by Grade 5EI: All instruction in ENG + at least 30 minutes ELD support	ENG: State standardized tests of ELA	ENG: ELA test scores of students in TBP and DI increased significantly faster (0.064 SD faster) than those of students in EI; rates of growth were comparable for DB and EI (conditional on covariates); scores of students in DI (lowest scores in Grade 2) increased at a rate as much as 0.092 SD faster per grade than scores of students in other programs

Note. ALA = academic language acquisition; BSM = Bilingual Syntax Measure; CAAS = Computer-Based Academic Assessment System; CELDT = California English Language Development Test; CTBS = Comprehensive Test of Basic Skills; CTBS-S = Comprehensive Test of Basic Skills Spanish; CST = California Standards Test; DB = Developmental Bilingual; DI = Dual Immersion; DIBELS = Dynamic Indicators of Basic Early Literacy Skills; DLP = Dual Language Program; DRDP = Desired Results Developmental Profile; EDL = Español Desarrollo de la Lectura; EFL = Extended Foreign Language; EI = English Immersion; ELA = English language arts; EL = English learner; ELD = English language development; ENG = English; EP = English proficient; EO = English-only; ERSA = Emergent Reader Screening Assessment; ESOL = English for Speakers of Other Languages; FCAT = Florida Comprehensive Assessment Test; HEC = Home English Communication; ISF = Initial Sound Fluency; ITAS = Individual Test of Academic Skills; ITBS = Iowa Test of Basic Skills; KAG = Kindergarten Assessment Guide; LAS = Language Assessment Scales; LAS-O = Language Assessment Scales–Oral; LCEP = Language Competence/Expressive Proficiency Task; LNF = Letter Naming Fluency; MAT = Metropolitan Achievement Test; NWF = Nonsense Word Fluency; OLPSI = Oral Language Proficiency Scale Interview–Revised; ORF = Oral Reading Fluency; PLS = Preschool Language Scale; PPVT = Peabody Picture Vocabulary Test; PreLAS = Preschool Language Assessment Scale; PSF = Phoneme Segmentation Fluency; SABE = Spanish Assessment of Basic Skills; SAT-9 = Stanford Achievement Test; SEC = students with Spanish at home without prior ENG experience; SEI = Structured English Immersion; SFA = Success for All; SMAG = State Measure of Annual Growth in English; SP = Spanish; SpO = Spanish only; SRI = Scholastic Reading Inventory; SSAT = State Standards Achievement Test; TAAS = Texas Assessment of Academic Skills; TAKS = Texas Assessment of Knowledge and Skills; TBP = transitional bilingual program; TBP-E = Transitional bilingual program–early exit; TBP-L = Transitional bilingual program–late exit; TELD-3 = Test of Early Language Development; TERA-2 = Test of Early Reading Ability; TPI = Transitional Program of Instruction; TVIP = Test de Vocabulario en Imágenes; TWI = two-way immersion; TWBI = two-way bilingual education; WLPB = Woodcock Language Proficiency Battery; WLPB-R = Woodcock Language Proficiency Battery–Revised Spanish and English Forms.

The four longitudinal studies (Hofstetter, 2004; Reese et al., 2005; Umansky & Reardon, 2014; Valentino & Reardon, 2015) examining the effectiveness of bilingual programs over time produced three important findings:

Students’ initial proficiency levels in Spanish were important for determining their rate of growth in English reading skills when participating in a TBP (Reese et al., 2005).

Although students in EI programs may have started with higher skills initially, students in bilingual programs often exhibited faster rates of growth and were able to close the gap and even outperform students in EI programs over time (Umansky & Reardon, 2014; Valentino & Reardon, 2015).

The timing of ELs’ reclassification as English proficient was often related to the bilingual program models (i.e., students in EI programs were reclassified at greater rates in elementary school, while students in dual immersion, transitional bilingual, and language maintenance programs were more often reclassified in middle school).

The delay in reclassification might have important positive and negative implications for the academic performance of bilingual students. On one hand, reclassification might allow students to participate fully in all course offerings, including those with rigorous content. On the other, delays in reclassification might allow students to receive additional support services, such as ELD, that they would not receive if they were reclassified. The drawback of delayed classification, however, is that students might miss rigorous content because of having to attend ELD classes, which are usually more focused on developing student English language proficiency than building their content knowledge (Umansky & Reardon, 2014).

Studies comparing bilingual programs for at-risk students

We found five studies that disaggregated the data by student risk status or that included students with disabilities. Two studies defined risk status based on academic performance (D. L. Baker, Park, Baker, Basaraba, et al., 2012; Gerber et al., 2004), and two defined it based on the presence of autism spectrum disorders. One study (Kovelman, Baker, & Petito, 2008) examined whether age of acquisition (students are termed early bilingual, late bilingual, or monolingual) was related to English and Spanish literacy skills. Although this study did not screen students for potential reading disabilities, we included it in Table 7 because bilingual students were disaggregated by age of exposure to Spanish and English, and age of acquisition was found to be a risk factor for later language development.

OPEN IN VIEWER

Table 7

Reading Studies Comparing Bilingual Programs for At-Risk Students

Study	Grade Level	N	Program Models Compared	Academic Measures	Outcomes
D. L. Baker, Park, Baker, Basaraba, et al. (2012)	Grades 1–3	210	Paired bilingual: 90 minutes SP + 30–45 minutes ENG in Grades 1–2; 60 minutes SP + 60 minutes ENG in Grade 3	ENG: DIBELS ORF, SAT-10	ENG: ELs in the paired bilingual program had higher rates of ORF compared to ELs in EO (d = 0.33 to 0.53); ELs at risk in the paired bilingual program significantly outperformed ELs at risk in the EO program in reading comprehension in Grade 2 (d = 0.51)
			EO: 120 minutes ENG instruction
Gerber et al. (2004)	K–1	43	CIM: SP small group direct instruction in K, transitioned to ENG for all but 14 students who continued to receive SP instruction	ENG: Rime/Onset Detection, Phoneme Segmentation, WIAT	ENG: ELs in the SP intervention demonstrated greater gains and closed the achievement gap with comparison students on all ENG measures except Onset Detection (p < .01); no significant interactions between group and time
				SP: Rime/Onset Detection, Phoneme Segmentation, WIAT	SP: ELs in the SP intervention made significantly greater gains on SP Rime Detection compared to comparison group (p < .01)
Hambly and Fombonne (2012)	Pre-K	75	MON: Exposed to one language only (e.g., ENG, French, or SP)	ENG: ADI-R, MCDI, VABS	ENG: No significant differences between the bilingual and monolingual groups on measures of expressive vocabulary, receptive language, or early language milestones (e.g., age of first words and phrases; p > .05); SIM and MON consistently scored higher than SEQ
			SIM: Bilingual exposure before 12 months of age
			SEQ: Bilingual exposure after 12 months of age
Kovelman, Baker, and Petito (2008)	Grades 2–3	150	In California: Instruction for EOs—ENG in K–2, then SP introduced in Grade 3; for SP-speaking ELs instruction in SP in K–2 and then ENG in Grade 3	ENG: CORE Assessing Reading, LCEP, WLPB-R	ENG: Age of first exposure to ENG had significant impact on student’s performance in ENG on phonological awareness, reading, and language (p < 0.01); students from monolingual ENG homes and early bilinguals significantly outperformed late bilinguals; significant group difference (early vs. late bilingual) for all tasks favoring the early exposure group (p < .001); EOs in bilingual program outperformed EOs from monolingual ENG homes on phonemic awareness skills
			In New Hampshire: English only
				SP: CORE Assessing Reading, LCEP, WLPB-R	SP: Children from monolingual ENG homes performed as well as children from SP-speaking homes on phoneme awareness tasks, but SP early bilinguals significantly outperformed students from monolingual ENG homes on reading tasks (p <.001)
			In Connecticut: Formal reading instruction was introduced in student’s dominant language (SP) and then in the other language (ENG); print material was available in both languages
Ohashi et al. (2012) a	Pre-K	60	Simultaneous bilingual: Ongoing exposure at home in two languages (one of which was ENG or FR)	ENG: ADOS, ADI-R, PLS-4, VABS	ENG: No significant differences between simultaneous bilingual and monolingual exposure groups for age of first words or phrases, or for ADOS Total Communication scores
			Monolingual exposure: Exposed to only one language (ENG or FR) from birth to age 3 or 4

Note. ADI-R = Austim Diagnostic Interview–Revised; ADOS = Autism Diagnostic Observation Schedule; CIM = Core Model Intervention; DIBELS = Dynamic Indicators of Basic Early Literacy Skills; EL = English learner; ENG = English; EO = English-only; FR = French; LCEP = Language Competence/Expressive Proficiency Task; MCDI = MacArthur Communicative Development Inventory; MON = Monolingual; ORF = Oral Reading Fluency; PLS = Preschool Language Scale; SAT = Stanford Achievement Test; SEQ = sequential bilingual exposure; SIM = simultaneous bilingual exposure; SP = Spanish; VABS = Vineland Adaptive Behavior Scales; WIAT = Word Identification/Word Attack; WLPB-R = Woodcock Language Proficiency Battery–Revised.

a

We included the Ohashi et al. (2012) study here although it was conducted in Canada because we wanted to compare findings from the two studies that included students with autism.

The study by D. L. Baker, Park, Baker, Basaraba, et al. (2012) compared the effects of two programs on reading outcomes: a paired bilingual program (where students received English and Spanish reading instruction simultaneously in Grades 1–3 at different times of day) and an EO reading program. The results indicated no significant differences in the effects of the two programs on English comprehension outcomes for students not at risk for reading failure in Grades 1 to 3. However, students at risk for reading difficulties appeared to benefit more from the paired bilingual program than from the EO program across the three grades. These findings were significant in Grade 2 (d = 0.51). Gerber et al. (2004) also found bilingual instruction favorable for at-risk kindergarten students who received a 30-minute Spanish intervention, in comparison with students who did not receive the intervention on any English measures except Onset Detection (p < .01).

Finally, two studies (Hambly & Fombonne, 2012; Ohashi et al., 2012) examined whether the social impairments associated with autism spectrum disorder could cause language delays and/or communication impairments in preschool bilingual children exposed to (a) one language (e.g., English, Spanish, or French), (b) a bilingual language environment prior to 12 months of age (i.e., simultaneous bilingual exposure), or (c) a bilingual language environment after 12 months of age (i.e., sequential bilingual exposure). Results indicated that although there were no statistically significant differences in the expressive vocabulary, receptive vocabulary, or language skills of students in the three groups, students who were exposed to only one language or to two languages prior to 12 months of age performed significantly better than students exposed to two languages after 12 months of age.

Ohashi et al. (2012) took a slightly different approach to answering the research question above by examining the age at which students with autism from monolingual (English or French) or simultaneous bilingual language environments (English and French) spoke their first words and phrases and were able to engage in functional communication. Results indicated no statistically significant differences between the two groups, suggesting that bilingual language environments do not cause language delays or disadvantage young children with autism spectrum disorder. In summary, the findings of the studies examining the effects of bilingual programs on students at risk suggest that bilingual programs do not hinder or otherwise impair the development of literacy skills; rather, the results indicate that the programs may support L1 and L2 literacy development for this population of students.

Studies examining the role of quality of instruction within bilingual programs

We found nine studies that examined quality of instruction in the context of bilingual programs and that met our inclusion criteria as indicated in Table 8. All studies were conducted in Grades preK to 2, with the exception of Vaughn et al. (2008), which followed students through Grades 4 and 5 after they had received a bilingual intervention or had been part of a control group in Grade 1. Seven of the eight studies included a detailed description of the high-quality instruction provided during the interventions. High-quality instruction was defined by evidence-based principles of instruction, such as teacher modeling and verbalizations, guided practice, brisk lesson pacing to maintain student engagement, multiple practice opportunities, specific academic feedback, and lessons structured to focus on the core components in reading (Coyne, Kame’enui, & Carnine, 2011). Overall, results of these studies indicated that students who received the high-quality instruction outperformed students in the comparison groups. For example, in the Cena et al. (2013) experimental study, Grade 1 students in a TBP were randomly assigned to receive either a scripted vocabulary intervention or general vocabulary instruction using the standard program and vocabulary routines designed to increase the explicitness of the instruction. Findings indicated that students who received the scripted vocabulary intervention significantly outperformed students who received the general vocabulary instruction, on overall Spanish depth of word knowledge, Spanish vocabulary usage, and Spanish vocabulary definition tasks (η² = 0.17, 0.10, and 0.20, respectively). This advantage, however, did not have an effect on English reading and vocabulary outcomes.

OPEN IN VIEWER

Table 8

Reading Studies Examining the Quality of Instruction in Bilingual Programs

Study	Grade Level	N	Program Models Compared	Academic Measures	Outcome
Buysse, Castro, and Peisner-Feimberg (2010)	Pre-K	193	English-only PD with evidence-based accommodations to support ELs (visual cues, prereading activities, observing and documenting second language learning)	ENG: WLPB-R, PPVT-III, PAT, NL	ENG: No significant differences between treatment/control on overall PAT score, but treatment students demonstrated significantly greater gains on onset segmentation task (ES = 0.61)
			Treatment: Nuestros Niños PD for 8 weeks in SP	SP: WLPB-RS, TVIP, PAT, NL, WMTS-PC	SP: Treatment students demonstrated significantly greater gains on PAT (ES = 0.69) and specifically on the rhyme-matching subtask (ES = 0.68).
Calhoon, Al Otaiba, Cihak, King, and Avalos (2008)	Grade 1	76	Two-way immersion with 60 PALS sessions (30–35 minutes, 3×/week)	ENG: DIBELS LNF, PSF, NWF, ORF	ENG: Significant time × condition interaction favoring PALS for PSF, NWF, and ORF (ES = 0.53, 0.50, 0.51); ELs receiving PALS instruction outperformed control group (ES = 1.29, 1.15, and .38 for NWF, LNF, and ORF)
			Two-way immersion, no PALS
Cena et al. (2013)	Grade 1	50	Traditional bilingual program (TBP): Reading in SP with 30 minutes ELD	SP: DOK SP Vocabulary, TVIP, IDEL FLO, BVAT	SP: Students who received vocabulary intervention significantly outperformed control group (TBP only) on DOK definitions (η2 = 0.20), DOK usage (η2 = 0.10), and DOK total scores (η2 = 0.17); no effects on bilingual verbal ability
			Vocabulary intervention: Taught 32 vocabulary words over 8 weeks using scripted lessons
Tong, Irby, Lara-Alecio, and Mathes (2008)	K–2	262	Traditional bilingual program–Enhanced (TBP-E): 70/30 SP/ENG (70/30 in K, 60/40 in Grade 1, 50/50; all SP instruction in content areas until spring of Grade 2) + tiered intervention (75 minutes K; 90 minutes Grade 1)	ENG: Letter naming and letter sound, CTOPP, WLBP-R, DIBELS	ENG: Students in TBP-E significantly outperformed students in the “typical” program on Rapid Object Naming (ES = 0.12), Blending Phonemes (ES = 0.44), Segmenting Phonemes (ES = 0.71), Oral Language (ES = 0.48), and Reading (ES = 0.16 for DIBELS and greater than 0.40 for WLPB-R; after adjusting for initial skill)
			Traditional transitional bilingual program (TB): 80% SP/20% ENG implementing typical practice	SP: TOPP-S, WLPB-RS, IDEL	SP: After controlling for initial skill, students in the TBP-E significantly outperformed students in the “typical” program on SP letter naming (ES = 0.25), Blending Phonemes (ES = 0.38), Segmenting Phonemes (ES = 0.21), Listening Comprehension (ES = 0.19), and Reading Comprehension (ES = 0.38)
Tong, Lara-Alecio, Irby, and Mathes (2011)	K–1	140	Same as above (Tong, Irby, et al., 2008)	ENG: WLPB-R, IDEA Oral Proficiency Test	ENG: ENG oral language skills developed faster from Time 2 to Time 3; students in TB had higher reading comprehension scores than students in TBP-E
				SP: WLPB-R, IDEA Oral Proficiency Test	SP: Students demonstrated higher Spanish overall language scores and students in TBP-E had significantly higher scores than students in TB ( η p 2 = 0.41), with SP developing faster from Time 1 to Time 2; students in TBP-E demonstrated stronger reading skills in SP than ENG and significantly higher scores than students in TB ( η p 2 = 0.08)
Tong, Lara-Alecio, Irby, Mathes, and Kwok (2008)	K–1	1,234	Same as above (Tong, Irby, et al., 2008)	ENG: WLPB-R, IDEA Oral Proficiency Test	ENG: Oral ENG growth was linear regardless of program type, and the rate of growth was significant for each group (.07 through .09); initial levels of performance varied significantly by group, and all but SEI-T showed heterogeneous growth; SEI-T students had significantly higher initial skills (ES = 0.40), but SEI-E students demonstrated larger growth (ES = 0.71); there were no differences in initial skill between TB and TBP-E, but TBP-E students demonstrated significantly greater growth (ES = 1.61); SEI-T students had significantly higher initial skills than TB (ES = 1.96), but the two groups grew at comparable rates
Vaughn et al. (2008)	Grade 1	86	Transitional bilingual program and intervention in SP. Length of intervention: 32 weeks, 5 days per week, 50 minutes/day in Grade 1. After first grade, core reading instruction was conducted in SP and in ENG, and by Grade 4 all instruction was in ENG.	ENG: WLPB-R, TOWRE, Spelling, Aimsweb Mazes, and TOSRE in Grades 4–5.	ENG: There were nonsignificant transfer effects favoring the SP treatment group in ENG letter word ID (0.39), ENG word reading (0.34), and ENG sentence reading (0.30).
				SP: WLPB-R, TOWRE-SP, IDEL FLO, spelling	SP: Results were mixed and nonsignificant. However, effects sizes were moderate in FLO (0.32), passage comprehension (0.33), word reading (0.39), and spelling (0.43).
Vaughn, Linan-Thompson, et al. (2006)	Grade 1	69	Transitional bilingual program and intervention in SP: 32 weeks; 5 days per week; 50 minutes/day	ENG: WLPB-R, CTOPP, DIBELS ORF, letter naming, sound identificationSP: WLPB-R, TOPP-S, IDEL FLO, letter naming, sound identification	ENG: No significant differences on reading outcomes between treatment and control groups with the exception of listening comprehension and verbal analogies favoring the control group (ES = −0.46 and −0.86, respectively). SP: The treatment group significantly outperformed the comparison group on TOPP-S (d = 0.73), WLPB-WA (d = 0.85), IDEL (d = 0.58), WLPB-Reading Comprehension (d = 0.55), WLPB-Listening Comprehension (d = 0.43)

Note. BVAT = Bilingual Verbal Ability Test; CTOPP = Comprehensive Test of Phonological Processing; DIBELS = Dynamic Indicators of Basic Early Literacy Skills; DOK = Depth of Knowledge; EL = English learner; ELD = English language development; ENG = English; ES = effect size; FLO = Fluidez en la Lectura Oral; IDEL = Indicadores Dinámicos del Éxito de la Lectura; LNF = Letter Naming Fluency; NL = Naming Letters; NWF = Nonsense Word Fluency; ORF = Oral Reading Fluency; PALS = Peer-assisted learning strategies; PAT = Phonemic Awareness Task; PD = professional development; PPVT = Peabody Picture Vocabulary Test; PSF = Phoneme Segmentation Fluency; SEI = structured English immersion; SP = Spanish; TOPP-S = Test of Phonological Processing–Spanish; TOSRE = Test of Sentence Reading Efficiency; TOWRE-SP = Test of Word Reading Efficiency FORM B adapted into Spanish; TVIP = Test de Vocabuario de Imágenes; WLPB-R = Woodcock Language Proficiency Battery–Revised; WLPB-RS = Woodcock Language Proficiency Battery–Revised Spanish Form; WMTS-PC = Where’s My Teddy Story?–Print Concepts.

The Tong, Irby, Lara-Alecio, and Mathes (2008) experimental study compared two different TBPs (an enhanced TBP and a typical TBP) for Grades K to 2. The enhanced TBP included PD that focused on helping teachers (a) reduce code-switching in the classroom in order to allow ELs to develop their language proficiency in each language without interference from the other language and (b) provide differentiated instruction for struggling ELs in a response to intervention model. The findings indicated that ELs in the enhanced TBP outperformed ELs in the typical TBP on almost all the measures in Spanish and English. English oral language proficiency, however, was still below the norm after 3 years, suggesting that students need time to develop their language proficiency in L2 independently of the type of bilingual program they attend (Genesee et al., 2006).

Vaughn, Linan-Thompson, et al. (2006) examined the effectiveness of a researcher-developed Spanish reading intervention that accompanied core instruction in Spanish in a TBP, first when it was implemented with two cohorts of Grade 1 students and then when the same cohorts were assessed again as part of a follow-up study 3 and 4 years later (Vaughn et al., 2008). Findings from the longitudinal study were mixed, and effects of the treatment in first grade in Spanish were mostly nonsignificant three and four years later. Effect sizes, however, ranged from 0.32 in Spanish connected text fluency to 0.43 in Spanish spelling favoring intervention students compared to comparison students in the Spanish study. Moreover, students in the Spanish intervention group appeared to have outscored students in the Spanish comparison group on letter word identification (ES = 0.39), on word efficiency (ES = 0.34) and on spelling (ES = 0.30). These findings are important (even if they were not significant) because they are apparent across decoding, fluency, and comprehension, and they suggest that interventions might have long-lasting effects, particularly if they could be provided across grades.

We could only locate one study that examined the direct effect of a PD program for preschool teachers called Nuestros Niños on Spanish and English student outcomes (Buysse, Castro, & Peisner-Feinberg, 2010). The Nuestros Niños program consists of 3 days of PD followed by classroom support from consultants and biweekly community-of-practice meetings that offer teachers a structured process to develop and refine instructional strategies. ELs in classrooms where teachers received the Nuestros Niños PD demonstrated significantly greater gains on the English onset segmentation task (ES = 0.61) and on Spanish phonological awareness tasks (ES = 0.69) compared to ELs in classrooms where teachers did not receive the PD.

Altogether, results from these studies confirm what Goldenberg (2013) and other researchers have suggested regarding the importance of providing ELs and EOs with explicit instruction and additional interventions, if necessary, that incorporate evidence-based principles of instructional design (Coyne et al., 2011). Although in Table 8 we described studies that measured quality of instruction, only the Buysse et al. (2010) study in preschool examined the effect of teacher PD in Spanish on EL outcomes in English and in Spanish. Thus, more research is needed to determine whether PD in the primary and upper grades in the language of instruction has a direct effect on student academic outcomes. Moreover, investigating exactly what factors of the PD are directly related to student outcomes (e.g., teacher knowledge of the linguistic systems, pedagogical knowledge, teacher bilingual language proficiency) would greatly benefit the field.

Reading studies comparing bilingual programs outside the United States

We found eight studies that were conducted outside the United States, as indicated in Table 9. The three studies conducted in the Netherlands (Admiraal, Westhoff, & de Bot, 2006; van der Leij, Bekebrede, & Kotterink, 2010; Verspoor, de Bot, & Xu, 2015) examined the effectiveness of different bilingual program models with students of different ages. In all three studies, students in bilingual programs in English and Dutch significantly outperformed their peers on English and Dutch measures in schools where Dutch was the language of instruction and English was taught for 2 hours a week or less.

OPEN IN VIEWER

Table 9

Reading Studies Comparing Bilingual Programs Outside the United States

Study	Grade Level	N	Program Models Compared	Academic Measures	Outcomes
Admiraal, Westhoff, and de Bot (2006)	Lower secondary (Grades 1–4)	1,305	Dual immersion: 50% English; 50% Dutch; Dutch only	ENG: EFLVT, CITO-Oral, MAVO, content exams (geography and history)	ENG: Main effect of bilingual program was large in terms of proportion of variance explained at school level in Vocabulary (R2 = .48–.97), Reading Comprehension (R2 = .94), Oral Language (R2 < .30); bilingual group earned higher scores on content exams in Dutch and ENG
Goldenberg et al. (2014)	Grades 1–2	571	United States: English immersion; transitional bilingual education; developmental bilingual; dual languageMexico: SP only	SP: TOPP-S, WLPB-R	SP language: ELs in the United States who received reading instruction in either ENG (ENG ELs) or SP (SP ELs) outperformed students in Mexico on Phonological Awareness tasks (p < .001); students from Mexico outperformed U.S. students on Picture Vocabulary and Listening Comprehension tasks (p < .001)
					SP reading: SP ELs outperformed ENG ELs and students from Mexico on all measures in Grade 1, but students from Mexico closed the gap and outperformed ENG ELs and performed comparably to SP ELs in Grade 2
Lorenzo, Casal, and Moore (2009)	Ages 9–14	1,320	Three content area classes taught in L2 (predetermined by teacher); during bilingual instruction students regroup into bilingual and monolingual groups	Diagnostic tests employed to assess language competencies (reading, listening, writing, speaking)	ENG, FR, and German: There were statistically significant differences between bilingual and control groups on reading, listening, writing, and speaking, with means favoring the bilingual group (actual t-test values not reported)
Lyster, Quiroga, and Ballinger (2013)	Grade 2	65	80% FR/20% ENG; 50% FR/50% ENG; 20% FR/80% ENG	ENG: PPVT, AAT, MAT	ENG: No significant effects by program type, but significant effects on MAT by group (treatment/control) and language dominance
				French: EVIP, AAT, MAT	FR: No significant effects by program type, but significant effects on MAT by group (treatment/control)
Parsons and Lyddy (2009)	Grade 2	82	English Reading First: ENG in 2nd year of schooling and Irish reading during the 3rd year	Translational equivalents of single-word reading tasks (50 words, ordered in increasing difficulty)	ENG: Accuracy scores of Gaeltacht children were significantly lower than those in the other groups (p < .01)
					Irish: ENG medium sample performed significantly lower than other groups (p < .01)
			Irish Reading First: Irish reading between 1st and 2nd year of schooling and ENG reading toward the end of the 2nd year
			Gaeltacht: Irish reading in Grades 1–2, then ENG in Grades 2–3; Irish reading in Grade 4
			English medium
Sailors, Hoffman, Pearson, Beretvas, and Matthee (2010)	Grades 1–2	801	Read Intervention Grade 1 and in Grade 2: 10 culturally relevant storybooks written in ENG and translated into home languages (used as read-alouds) + 8 picture books to be used during reading instruction	ENG: Object Identification, Letter/Sound Identification, Word Identification; Sentence Comprehension, Word Production; Sentence Production, Spelling (All assessments translated into home languages: Afrikaans, Xhosa, Sotho, Zulu, and others)	ENG: Students in the READ intervention significantly outperformed students in comparison schools on all language assessment subtests except Sentence Production (Grade 1, η2 = 0.01 to 0.03; Grade 2, η2 = 0.01 to 0.03); schools that were identified as strong implementers of the intervention performed significantly better on Object Identification, Sentence Comprehension, Word Production, and SpellingIn Grade 1, students in strong-implementing schools significantly outperformed “regular” intervention and comparison schools in Coding and Meaning subscales
			Comparison: Instruction provided using traditional phonics-based, culturally irrelevant textbooks		Home language: Students in the READ intervention significantly outperformed students in comparison schools on all language assessment subtests except Sentence Production in Grade 1 (Grade 1 η2 = 0.03 to .09; Grade 2, 0.03 to 0.15); “regular” intervention schools significantly outperformed strong implementers and control group on Word Identification task; “regular” intervention and control group scored significantly higher than strong implementers on Word Production and Spelling tasks; in Grade 1, intervention schools significantly outperformed comparison schools on Coding and Meaning subscales; in Grade 2, intervention schools significantly outperformed on Meaning subscale
Van der Leij, Bekebrede, and Kotterink (2010)	Grades 2–3	57	Bilingual: Concurrent bilingual instruction from the age of 4 and reading instruction in ENG and Dutch in Grade 1	ENG: PPVT,OMT, Loanword Reading Fluency, Orthographic Knowledge and Awareness	ENG: Bilingual group performed significantly better on Vocabulary (Grade 2, η p 2 = 0.12; Grade 3, η p 2 = 0.53), Word Reading Fluency (OMT; Grade 2, η p 2 = 0.23; Grade 3, η p 2 = 0.22), Loanword Reading Fluency ( η p 2 = 0.14), and Orthographic Awareness ( η p 2 = 0.24)
			Comparison: Did not receive any instruction or practice in ENG at school	Dutch: Similar to English measures	Dutch: Bilingual group performed significantly better on Reading Comprehension ( η p 2 = 0.13) in Grade 2, Multisyllabic Word Fluency (DMT; Grade 2, η p 2 = 0.08; Grade 3, η p 2 = 0.13), and on Orthographic Knowledge ( η p 2 = 0.10) in Grade 3
Verspoor, de Bot, and Xu (2015)	Ages 12–14	399	Bilingual: 50% (~15 hours/week) of instruction in ENG, 10 hours of which were in content subjects (history, math, art, physical education, etc.) and 5 hours of ENG	ENG: Receptive vocabulary measure, productive informal writing task	ENG: Using sum of vocabulary and writing scores, bilinguals had significantly higher scores than regulars and controls in winter and spring (p < .001); Year 1 bilinguals had significantly higher ENG proficiency scores than regulars and controls (η2 = 0.20) after controlling for initial proficiency, CITO score, out-of-school contact, and motivation); Year 3 bilinguals had significantly higher ENG proficiency scores than regulars (η2 = 0.16) but not controls
			Regular: 2 hours/week ENG instructionControl: 2 hours/week ENG instruction
				Dutch: CITO (scholastic aptitude), Questionnaire (ENG use, motivation, attitudes)	Dutch: Bilinguals were significantly more motivated than regulars and controls (p < .05), and bilinguals and controls had significantly higher CITO scores than regulars (p < .001)

Note. AAT = Auditory Analysis Test; CITO = Dutch National Institute for Test Development; EFLVT = English as a Foreign Language Vocabulary Test; EL = English learner; ENG = English; EVIP = Échelle de vocabulaire en images Peabody; FR = French; MAVO = National Final Examinations for English; MAT = Morphological Awareness Test; OMT = One Minute Test; PPVT = Peabody Picture Vocabulary Test; TOPP-S = Test of Phonological Processing–Spanish; WLPB-R = Woodcock Language Proficiency Battery–Revised.

The Goldenberg et al. (2014) study compared the Spanish reading and language performance of Spanish-speaking ELs in the United States who participated in one of five bilingual programs and Spanish-speaking students in Mexico who received Spanish-only instruction. The results indicated that while U.S. Hispanics who received instruction in either English or Spanish significantly outperformed students in Mexico on phonological awareness tasks (p < .001), students from Mexico outperformed students from the United States on picture vocabulary and listening comprehension tasks (p < .001). With respect to Spanish reading, Hispanic students receiving primarily Spanish instruction in the United States outperformed Hispanics receiving English instruction and students from Mexico on the Spanish reading tasks in Grade 1. By Grade 2, however, students from Mexico closed the gap and were performing better than Hispanics receiving primarily Spanish instruction in the United States. This result suggests that if the language of instruction is also the societal language, students perform better compared to when students are exposed to two languages in an environment where the societal language is English.

The study by Sailors, Hoffman, Pearson, Beretvas, and Matthee (2010) was conducted in South Africa in rural schools where students were provided with an intervention in their L1 (e.g., Afrikaans, Xhosa, Sotho) and in English in Grades 1 and 2. The findings indicated that ELs in the treatment schools performed better in their L1 and in English than the ELs in control schools. Teachers in the treatment schools were provided with high-quality instructional materials and PD. All of these studies suggest that bilingual programs can also benefit students whose L1 varies significantly from English. Moreover, bilingual instruction appears to provide students the support they need in their L2, as well as help maintain their L1.

Writing Studies

We found only one quasi-experimental study that examined the effects of bilingual programs on writing. As described in Table 10, Bae (2007) compared the effects of an English-Korean two-way immersion program on the writing performance of a bilingual English-Korean group in Grades 1 and 2 versus an EO group in the same grades in the United States. The findings suggested that the English-Korean group had mean scores similar to those of EOs on grammar and coherence by the end of Grade 2. More studies, however, need to be conducted on the effects of bilingual programs on student writing in L1 and L2.

OPEN IN VIEWER

Table 10

Studies Examining the Effectiveness of Bilingual Programs in Supporting Writing Development

Study	Grade Level	N	Program Models Compared	Academic Measures	Outcomes
Bae (2007)	Grades 1–2	192	KETWIP: 30% of instruction in ENG in K, increasing to 50% ENG in Grade 2	Researcher-developed narrative story task scored with respect to coherence, content, grammar, and text length	ENG: Grade 2 KETWIP students performed comparably to students in EO program on indices of grammar and coherence/content; KETWIP students wrote significantly longer narratives than EO students
			EO: All instruction in English

Note. ENG = English; EO = English-only; KETWIP = Korean-English two-way immersion program.

Science Studies

We found only three quasi-experimental studies that examined science outcomes for ELs and EOs attending bilingual programs as summarized in Table 11. Findings from the Ciechanowski (2014) study indicated that ELs and EOs who attended a DLP that focused on science content made substantial gains in science outcomes in English (the language of science instruction in the DLP) from pretest to posttest. The science language gains were also significantly correlated with science content gains. The Martínez-Alvarez, Bannan, and Peters-Burton (2012) study found no differences in science knowledge in English (the language of science instruction) between students attending a DLP and students attending a TBP (η² = 0.07). However, ELs in the DLP performed significantly better on Spanish reading measures.

OPEN IN VIEWER

Table 11

Studies Examining the Effectiveness of Bilingual Programs in Supporting Student Science Understanding

Study	Grade Level	N	Program Models Compared	Academic Measures	Outcomes
Ciechanowski (2014)	Grade 3	33	Dual language program: 50/50 ENG/SP, half of the day taught in ENG, half of the day taught in SP; ELs received an additional 30 min ELD instruction focused on science	ENG: FOSS curriculum tests	ENG: All emergent bilinguals (ELs and ENG proficient) made significant science gains from pre- to posttest
					Science language gains were significantly correlated with content gains for both groups (r = .39–.56)
M. G. López and Tashakkori (2006)	Grade 5	344	Two-way bilingual: 40% SP/60% ENG	ENG: FCAT	ENG: No significant differences in performance on FCAT science for students in the two groups, but results indicated that students who entered with lower levels of ENG proficiency in K or Grade 1 scored significantly lower than students with higher levels of ENG proficiency (η2 = 0.09)
			Transitional bilingual: EO with the option of receiving 2.5 hours/week SP instruction	SP: EDL
Martinez-Alvarez, Bannan, and Peters-Burton (2012)	Grade 4	78	InSciRead: Instructional model based on use of inquiry-based science cognitive strategies and reading comprehension of science texts written in SP	ENG: Adapted Detecting Incongruities task	ENG: After adjusting for pretest SP reading levels, students in the treatment group had significantly higher adjusted mean scores on total number of sentences correctly identified ( η p 2 = 0.07)
			Bilingual model: 50/50 two-way dual language model in which math, science, and language arts instruction was in SP; social studies and language arts instruction was in ENG	SP: WMLP

Note. EL = English learner; EDL = Español Desarrollo de Lectura; ENG = English; EO = English-only; FCAT = Florida Comprehensive Assessment Test; FOSS = Full Option Science System; SP = Spanish; WMLP = Woodcock Muñoz Language Proficiency.

Mathematics Studies

Seven of the studies in our review focused on mathematics (see Table 12). One of them took place in Germany (Kempert, Saalbach, & Hardy, 2011). None of them were randomized controlled studies, and only one included pretest data (Valentino & Reardon, 2015). The study by Kempert et al. (2011) indicated that proficiency in the language of testing was more influential than student mathematical problem-solving abilities. In addition, more variance was explained by cognitive abilities, SES, and reading comprehension than by language proficiency or arithmetic skills. This study, however, included only posttest data, so the results should be interpreted with caution.

OPEN IN VIEWER

Table 12

Studies Examining the Effectiveness of Bilingual Programs in Supporting Student Mathematics Understanding

Study	Grade Level	N	Program Models Compared	Academic Measures	Outcomes
Gómez, Freeman, and Freeman (2005)	Grades 3–5	240	50–50 Content model: Mathematics instruction in ENG, science, and social studies instruction in SP; Language Arts instruction in native language in pre-K and K and then in students’ L2 in Grades 1–5	ENG: TAKS	ENG: Of 103 Spanish-dominant Grade 3 students tested, 89 (86%) met the proficiency standard; of the 58 English-dominant Grade 3 students tested, 53 (95%) met the proficiency standard; overall, 142/159 (89%) met the third-grade proficiency standard on TAKS Reading
Hofstetter (2004)	K–3	441	Academic language acquisition (ALA): Instruction in K–1, 70% in L1, 30% in L2; percentage of ENG instruction increases until Grades 4/5, where ratio of instruction is 85% ENG and 15% SP	ENG: CELDT, SAT-9 mathematicsSP: SABE	ENG: The percentages of students who achieved the 50th National Percentile Rank on the mathematics test of the SAT-9 were comparable in ALA (38% to 46%) and SEI (40% to 56%)
			Structured English immersion (SEI): Students receive instruction predominantly in ENG, with 10% to 15% of instruction in ELD, 75% using sheltered ENG strategies
Kempert, Saalbach, and Hardy (2011)	Grade 3	74	German only: Primary language of instruction was German	German: BVAT, Mathematical word problems, reading comprehension test, language proficiency test, mathematics test for Grade 2, cognitive ability test (CFT 20)	German: Math skills and German language proficiency were the strongest predictors of German mathematics problem solving, accounting for 43.7% of the observed variance
			German and Turkish: Children with Turkish backgrounds were offered Turkish language classes	Turkish: BVAT, mathematical word problems	Turkish: Proficiency in the language of testing (Turkish) was the strongest predictor of Turkish problem solving (R2 = .27) after controlling for SES, cognitive ability, and math skills; no significant change to variance explained after including German language proficiency in the model
Lindholm-Leary and Block (2010)	Grades 4–6	Study 1: 193	DLP: 90% in SP in K–1, 80% in Grade 2, 70% in Grade 3, 60% in Grade 4, and 50% from Grade 5 on	ENG: Mathematics California Standards Test	ENG: In Study 1, there were no significant differences between scale scores of English proficient (EP) and English learner (EL) students by program type; In Study 2, both EP and EL students in a DLP achieved at significantly higher levels than their peers in EO on the Mathematics CST (d = 0.77 for EP group, d = 0.58 for EL group)
		Study 2: 466	EO: Control group
Marian, Shook, and Schroeder (2013)	Grades 3–5	2,009	TWI-S: SP in Grades K–2, ENG in Grades 3–5 (math in SP K–3, ENG in Grades 4–5, social studies/science in SP)	ENG: State measure of annual growth in ENG, State Standards Achievement Test	ENG: TWI-S students significantly outperformed TPI in math (and showed significant improvements across grade levels); for all grades, TWI-E students outperformed EO students in math
			TPI: ENG with pullout ESL instruction as needed
			TWI-E: ENG in Grades K–2; SP instruction in Grades 3–5 (math in SP in K–3, ENG in Grades 4–5, social studies/science in SP in K–5)
Valentino and Reardon (2015)	Grades 2–5	13,750	Transitional bilingual: SP instruction in K and transition to ENG by Grades 2–3Developmental bilingual: In K, 50% to 90% of instruction in target language, proportion of ENG instruction increasing each yearDual immersion: 80% to 90% of instruction in target language in K–1, transitioning to 50% ENG and 50% target language by Grade 5English immersion: All instruction in ENG plus at least 30 minutes ELD support	ENG: State standardized tests in mathematics	ENG: Grade 2 math scores were significantly higher for students receiving TB instruction (0.27 SD higher), but this difference did not hold to middle school; students receiving DLP instruction grew faster than students receiving English immersion instruction

Note. BVAT = Bilingual Verbal Ability Test; CELDT = California English Language Development Test; CFT = Cognitive ability test; DLP = dual language program; EDL = Español Desarrollo de Lectura; ELD = English language development; ENG = English; EO = English-only; ESL = English as a second language; FCAT = Florida Comprehensive Assessment Test; SABE = Spanish Assessment of Basic Skills; SAT-9 = Stanford Achievement Test; SP = Spanish; TAKS = Texas Assessment of Knowledge and Skills; TPI = Transitional Program of Instruction; TWI-E = two-way immersion–English; TWI-S = two-way immersion–Spanish.

Valentino and Reardon’s (2015) longitudinal study also compared outcomes in mathematics. The only statistically significant outcome was in second grade, favoring the TBP (ES = 0.27), but this effect did not last until middle school. Marian, Shook, and Schroeder (2013) found that ELs attending a two-way immersion program outperformed ELs attending EO programs across grades. Hofstetter (2004), on the other hand, found that ELs performed equally well in a transition program and a structured EI program in Grades K to 3. In summary, ELs attending bilingual programs might not be at a disadvantage in terms of their mathematics performance, but more research is needed to determine the benefits of bilingual programs on mathematics, whether it is taught in L1 or L2.

Cross-Linguistic Transfer and Cognition

Future Research

As we reflect on the political and scientific activities that have occurred since the publication of Educational Researcher’s special issue on bilingual education in 1992, we believe that we may be at a turning point in which bilingual education is viewed increasingly as an asset for ELs and EOs alike. This hypothesis is supported by the rapid rise of bilingual programs in the United States, including in states where EO is mandated, as well as by the large number of publications examining the effects of bilingual policies (Gándara & Hopkins, 2010), bilingual instruction, and bilingualism on economic outcomes (Gándara & Callahan, 2014), academic outcomes, and cognition (Bialystok et al., 2009). However, in 1992, Cziko suggested that instead of asking whether bilingual education works, we should consider the question of what can actually be done to make bilingual education better and accessible to more children. Next, we provide some guidelines on how Cziko’s suggestions could be addressed.

Conclusion

Bilingual education in the United States has been a central part of the culture and identity of many different ethnic groups, particularly of Hispanics. Moreover, our search indicates that bilingual programs, and specifically DLPs, have gained increasing attention from middle-income parents who view bilingual education as an opportunity for their children to obtain a broader view of the world and be more competitive in the job market. Although evidence of the economical and cultural benefits of bilingualism is starting to emerge, additional research targeting the PD of high-quality bilingual teachers and bilingual programs is necessary to create a stronger knowledge base that is well grounded in scientific evidence.