The Near End of Bilingual Education

In the wake of California’s Prop 227

Bilingual education’s 26-year reign in California was supposed to end with the voters’ passage of Proposition 227 in June 1998. The proposition declared:

All children in California public schools shall be taught English by being taught in English. In particular, this shall require that all children be placed in English-language classrooms. Children who are English learners shall be educated through sheltered English immersion during a temporary transition period not normally intended to exceed one year.

Of course, there is often a disconnect between a law and its implementation, especially with an issue as controversial as bilingual education. State officials can subvert the law through interpretations that don’t conform to its intent; school districts can change their policies without making genuine changes in curriculum; or teachers can ignore the mandates, closing their classroom doors and doing as they please.

What happened in the wake of Prop 227? The answer should be of interest in Massachusetts, which is currently implementing a similar proposition, and in other states contemplating ending bilingual education or otherwise considering how best to educate students whose native tongue is not English. My research reveals that resistance to the new law was, in many schools and districts, quite intense, indicating the depth of support for bilingual education among teachers and principals. Of course, such opposition was to be expected after state officials and interest groups spent the past few decades aggressively promoting bilingual education. Yet gradually the intent of the legislation has prevailed in most places, apparently to the benefit of English Learners, at least judging by test scores. To explain these findings, however, I need to begin with some fundamentals about a much misunderstood topic.

Spanish speakers were virtually the only English Learners receiving authentic bilingual education because they were typically the only ones who fulfilled all the conditions for providing it efficiently.

Bilingual Education Before Proposition 227

During the past 25 years, essentially three different kinds of instructional programs have existed for students with limited English proficiency, also called English Learners. The first is English as a Second Language (ESL) tutoring mixed with regular classroom instruction, wherein both English Learners and English-speaking students are taught in English in the same classroom for most of the day. English Learners receive their supplementary ESL tutoring in a pullout setting for anywhere from an hour a week to several hours a day. The second program is sheltered English immersion, which involves teaching in English to a classroom filled only with English Learners. If all the children speak one language, the teacher may also speak in that language occasionally to clarify or explain a concept, but the children learn to read and write in English and they receive math, science, social studies, and other subjects in English. Teachers of children who function poorly in English will initially spend most of the day teaching them to read and write in English. Gradually, however, other subjects are introduced. For children in 1st grade or higher, it is usually just a matter of months before much of the day is devoted to these other subjects.

In the third instructional program, the only one that meets the definition of bilingual education in the theoretical literature, students are taught initial literacy and subjects like math and science in their native tongue as they progress toward fluency in English. English is taught as a separate subject for about an hour a day initially, although there may be almost no English at all in kindergarten. The amount of English is typically increased over time, but students are not taught entirely in English until they are literate in their native tongue.

The facilitation theory underlying bilingual education as just defined has two parts. The “threshold” hypothesis states that there is a threshold level of linguistic competence in the native language that all children must attain in order to avoid cognitive disadvantages, while the “developmental interdependence” hypothesis holds that the development of skills in a second language is facilitated by skills already developed in learning the first language. The implication is that children must first learn to read and write in their native tongue and should begin training in English literacy only after they have mastered their first language. Programs that deviate from this sequence violate the fundamental theory of bilingual education.

Yet observations I have conducted in more than 300 classrooms in California, Minnesota, New York City, and Massachusetts over the past 15 years indicate that local school systems have commonly used bilingual education as a generic term referring to all three types of language-instruction programs. Recall that sheltered English immersion consists of English instruction in a self-contained classroom of English Learners. However, school systems often call such programs bilingual education as long as the teacher is bilingual, the students are ethnically or linguistically similar, and the classes are formed with the stated intent of providing native-tongue instruction. For example, the program titles and descriptions for the Vietnamese, Russian, Khmer, and Chinese bilingual education programs in Minnesota, New York, California, and Massachusetts (before these programs were eliminated in the latter two states) often state that the children will be receiving native-tongue instruction. This is either completely false or a huge exaggeration. Children in these programs are always taught to read and write in English and receive subject-matter instruction in English.

In some Chinese bilingual-education classrooms, there may be some teaching in a non-English language, but it is neither a means of receiving subject-matter instruction nor of acquiring literacy. In some Chinese “bilingual” education programs, for example, the English Learners, all of whom are of Chinese origin, receive some instruction in Mandarin as a foreign language. (In fact, Mandarin could only be taught as a foreign language since it is only one of the many dialects spoken in China and is rarely the native tongue of Chinese immigrants to the U.S.) But these programs do not fit the theoretical model of bilingual education since the children learn to read and write first in English and the Mandarin is only a small part of their instruction.

Occasionally, even ESL pullout programs, where students spend most of the day learning in English in a mainstream classroom, are mistakenly characterized as bilingual education when the children in the ESL pullout class are of the same ethnicity. The fact that these classes are actually taught in English is ignored by administrators, policymakers, parents, and advocates of bilingual education. Indeed, the latter usually deny it, perhaps seeing a political advantage in categorizing many different types of programs as bilingual education.

In short, official statistics on bilingual-education enrollment consistently overestimate the number receiving native-tongue instruction. Nevertheless, California government figures indicate that in 1997-98, the year before Proposition 227 was implemented, only 410,000 students were enrolled in bilingual education statewide, while 1.14 million Hispanic English Learners were enrolled in California public schools. Even if the only children enrolled in programs labeled bilingual education were Spanish speakers, at most only 36 percent of Hispanic English Learners could have been enrolled in such programs. Thus critics of bilingual education most likely have exaggerated its aggregate harm and supporters most likely have exaggerated its aggregate benefits, since only a minority of English Learners were enrolled in programs that were even nominally bilingual. Moreover, the impact of bilingual education was concentrated almost exclusively on Hispanics.

Spanish speakers were virtually the only English Learners receiving authentic bilingual education because they were typically the only ones who fulfilled all the conditions for providing it efficiently. In order to provide authentic bilingual education, schools must have teachers who are fluent in the language and enough English Learners from the same language group to fill a classroom without combining students from more than two grade levels in one classroom. In addition, the students must all speak the same dialect (Spanish has no important dialects), and the native tongue must be a phonetic language with a Roman alphabet (otherwise few of the skills learned in the native tongue can be transferred to English). Finally, there must be published textbook materials in the native tongue that conform to the U.S. curriculum.

Predictably, then, I have not found any bilingual-education programs that actually teach initial literacy in native languages, such as Chinese and Japanese, that use ideographic characters. I also have not found any non-Roman alphabet bilingual-education programs that teach initial literacy in the native language, even if the alphabet is phonetic (as in the case of Hebrew, Arabic, the Indian dialects, Russian, Armenian, and Khmer). Teachers have told me that it is too difficult or confusing to teach initial literacy, particularly to young children, in a language with an alphabet different from English. These classes are therefore typically taught completely in English either as a pullout supplement to the mainstream classroom, in which case the emphasis is on teaching the English language itself, or as a substitute for the mainstream classroom, in which case all subjects, including math, social studies, and science, will be taught in English at a pace the children can understand. This does not prevent these programs from receiving official approval as “bilingual” education programs and whatever funding is associated with that label.

Critics of bilingual education most likely have exaggerated its aggregate harm and supporters most likely have exaggerated its aggregate benefits, since only a minority of English Learners were enrolled in programs that were even nominally bilingual.


Though most English Learners, in California and elsewhere, did not receive bilingual education, Prop 227 passed in June 1998 largely on the strength of the allegation that the low achievement and high dropout rates of immigrant children were caused by “costly experimental language programs.” As a remedy, Prop 227 required all English Learners to be educated in sheltered English-immersion classrooms during a temporary transition period not to exceed one year. Once English Learners acquired a good working knowledge of English, they were to be transferred to English-language mainstream classrooms. Parents could request a waiver from these requirements, but only after their child had spent 30 days in a sheltered English-immersion classroom and only if the parent personally visited the school.

However, in practice Prop 227 has been dramatically changed by school districts, as evidenced by guidelines for school principals issued by Los Angeles Unified, San Diego Unified, and San Francisco Unified, apparently without protest from the state board of education. For one thing, the school districts have redefined a sheltered English classroom to include not only self-contained classrooms of English Learners taught in English, but also mainstream classrooms with ESL pullout instruction and self-contained classrooms of English Learners receiving up to 30 percent of their instruction in Spanish. Teachers have been permitted to recruit children for bilingual classrooms, even though the initiative says parents must initiate this process. Parents have been allowed to mail in their requests for waivers, when Prop 227 requires a personal visit. The school districts have also failed to require detailed documentation of the need for a bilingual education classroom, as the initiative requires, and they have changed the requirement of a year in a sheltered English-immersion classroom from a maximum to a minimum. In addition, children are being required to spend 30 days in an English-language classroom only when they first enroll in school–something the initiative says must happen each year.

San Diego’s interpretation and practice come close to subverting the intent of the law. Spanish-speaking English Learners in many sheltered immersion programs in San Diego schools are being taught to read and write in Spanish. My visits to two San Diego schools in September 2001 revealed that kindergarteners who knew no English were being assigned to classrooms called “waivered bilingual” during the first 30 days of the school year and were being instructed almost entirely in Spanish.

Nevertheless, although implementation has been uneven, enrollment in bilingual education has dropped dramatically across the state (see Figure 1). The total share of English Learners in California enrolled in bilingual education plummeted from 29 percent in 1997-98 to 12 percent in 1998-99, with the implementation of Prop 227. By 2001-02, it had declined a bit further, to just under 10 percent. Among elementary schools, where bilingual education was most common, the decline was more dramatic, but again bilingual education was not eliminated entirely. The share of English Learners enrolled in bilingual education in elementary school dropped from 39 percent in 1997-98 to 13 percent in 2001-02. Among secondary students, 10 percent of English Learners were enrolled in bilingual education before Prop 227, and about 3 percent in 2001-02. In the post-Prop 227 world, bilingual education is essentially an elementary-school program.



In the Classroom

To further investigate the implementation of Prop 227, I observed 170 classrooms and interviewed teachers and administrators in 29 elementary and junior high schools in eight California school districts during the spring of 1999 and the fall of 2001. The districts included Oceanside, Los Angeles, San Diego, San Francisco, two small San Francisco Bay-area school districts, and two small school districts near Los Angeles. The schools in Oceanside, Los Angeles, San Diego, and San Francisco were selected randomly from among those with large numbers of Hispanic (in San Francisco, large numbers of Chinese and Hispanic) English Learners. Thus my observations are representative of the school districts where bilingual education once flourished and also of the few schools with nominally bilingual programs for Chinese English Learners. (As explained above, the latter are actually sheltered English-immersion programs for students of Chinese origin.)

Visiting the school to sign a waiver authorizing bilingual education is not an idea that typically originates with the parent. My interviews suggest that bilingual education is in this sense like medical care. Teachers, like doctors, create supply by the criteria they use to define a child as needing treatment and they create demand by telling the patient what treatment he or she needs. In every school that I visited in the spring of 1999, teachers explained that they had “worked very hard” to get parents to sign waivers. They held meetings during the first 30 days of school and called parents to persuade them that their child would be better off in the bilingual-education program.

Just as Hispanic students were the only ones receiving authentic bilingual education before Proposition 227, they are the only ones being waivered after 227. The number of waivered classrooms is not caused simply by the number of students whose parents initially seek waivers. It also depends on the total number of Hispanic English Learners in a school, the school’s definition of eligibility for bilingual education, and the school’s strategies for filling a bilingual-education classroom. The most important step a principal can take is to control classroom assignments so that students who had been recommended for bilingual education before Prop 227 are in the same classroom, making it easier to convert the entire classroom to bilingual education on the 31st day of school.

Then parents must be contacted to obtain approval for the waiver of their child. When a simple majority of waivered students is obtained for a given teacher and classroom, the other parents can be told by phone that their child would need to change teachers if they do not sign a waiver. Alternatively, telephone calls might not be made until a decisive majority of waivered students had been obtained. In either case, telephone calls are very effective in converting additional students; most parents simply do not want their child’s education to be disrupted by changing classrooms, and many of them care more about that than they do about the language of instruction.

Thus the number of bilingual waivered students and bilingual waivered classes is not necessarily indicative of parental support for bilingual education. Rather, it seems to reflect staff support for bilingual education and, to some extent, parental deference to staff. Parents in schools with small numbers of Spanish-speaking English Learners, or in schools where a district-wide decision had been made to adopt sheltered English immersion, may not even have been aware of their right to apply for a waiver. In these districts, there was little or no likelihood of having enough students to maintain a bilingual-education program and thus no motive for the school to recruit parents. When pressure from above is absent, parental demand for bilingual education is low.

Interestingly, I also discovered that even the teachers of students still receiving Spanish bilingual education are using more English than in the past. The teachers of these classes offered two reasons. First, Prop 227 expressed the preferences of the electorate for a greater emphasis on English. Many teachers stated they were being responsive to their clients by increasing the English in bilingual education. Second, because there is no guarantee that a waivered class can be assembled for the next grade in the following year, teachers in bilingual-education classes told me they were preparing their students for the possibility that they would have to go into an English-language classroom because there were no bilingual-education classrooms available.

In the fall of 2001, I asked several former bilingual-education teachers who were now teaching in sheltered English-immersion classrooms whether they would ever go back to bilingual education. Not a single teacher said yes. All preferred sheltered English immersion, even though they thought it was harder work for them as teachers. A recurring theme was that “bilingual education was a good theory, but in practice it just didn’t work very well.” One practical problem facing bilingual education was the fact that many students change their residence from year to year, and even within a year. Thus they could find themselves in bilingual education in one school, all-English instruction in the next, and back to bilingual education in a third school. Another problem was the discontinuity between the bilingual-education curriculum and the curriculum in mainstream classrooms.

In general, my interviews indicate that, despite some uneasiness about the future and an unwillingness to renounce the theory of bilingual education, former bilingual-education teachers teaching in sheltered English-immersion programs now strongly support sheltered English immersion. They perceive themselves as giving their students the nurturing environment that they previously believed only a bilingual-education program could provide, while at the same time providing the exposure to English that they worried was lacking in the bilingual-education programs they used to teach in. As we will see below, early evidence of Prop 227’s effect on the achievement of English learners seems to support this view.

Prop 227 and Achievement

One of the biases that evaluations of programs for English Learners must overcome is that a much smaller percentage of students are actually tested in bilingual education than in English-immersion programs. One reason given by advocates and administrators is that it is unreasonable to administer English-language tests to students who are learning literacy in their native tongue. This may be true, but it gives the bilingual-education programs an unfair advantage because schools and teachers tend to exclude the lowest-scoring students from testing.

This problem exists in California with English Learners as a group and with bilingual education in particular. According to state regulations, all English Learners must be tested on the statewide Stanford 9 tests first administered in 1997-98, the year before Prop 227. However, only 68 percent of English Learners were tested in 1997-98 in reading; the share increased to only 84 percent in 2000-01. Although math is less language-based than reading, the testing rates for English Learners in math are only a few points higher: 72 percent of English Learners were tested in math in 1997-98; 86 percent in 2000-01.

The cause of this in California is threefold. First, a loophole in the state law gives parents the right to remove their child from testing. Second, since special-education students may be excused from testing, an English Learner can be classified as special education and excused on that basis. Third, English Learners tend to have lower socio-economic status, making them more likely to be absent from school on the day tests are administered. These factors bring about considerable variation in testing rates among schools and school districts.

Valentina Bali of Michigan State University found that in 1997-98 the Pasadena school district in southern California tested only 50 percent of its bilingual-education students, versus 89 percent of those who were in ESL programs. A 1998 Los Angeles school district report showed that bilingual-education students scored higher than students in English-immersion programs after five years. But only 61 percent of the bilingual-education students were tested, versus 97 percent of the students in the all-English program. Under these circumstances, the kind of casual comparisons made by the media of achievement before and after Proposition 227 and across school districts are risky. Moreover, any trends in aggregate achievement can be obscured by increases in the testing rates of the target population, as has happened in the wake of Prop 227 for English Learners. Evaluating the effect of Prop 227 on achievement is also complicated by the lack of data on student achievement broken down by which program they participated in.

It is possible, however, to analyze the effect of Prop 227 indirectly by examining the relationship between the percentage of students enrolled in bilingual education and the achievement of English Learners across the more than 9,000 schools in California. A simple comparison, examining only those elementary schools with significant bilingual- education programs (more than 120 students enrolled before Prop 227), reveals that the schools that eliminated their bilingual education programs had a 10-point gain in reading and a 13-point gain in math, but those that maintained some form of bilingual-education program had only a 6-point gain in reading and a 14-point gain in math.

This comparison may underestimate the impact of eliminating bilingual education, since even the schools that kept more than 120 students in bilingual education still had a large reduction in bilingual-education enrollment. Moreover, even if a school maintained a scaled-down bilingual-education program, my interviews suggest that in many schools it is no longer the same program–more English is being used and students are being transitioned faster since there are fewer bilingual-education programs in the upper grades. Trying to isolate the true effect of a program that is no longer the same is difficult even at the individual level; it is even more difficult at the school level. Differences in testing rates by program introduce further bias. The percentage of English Learners tested in reading was four points lower in the schools that eliminated bilingual education, while the percentage of English Learners tested in math was three points lower. All else being equal, these lower testing rates should inflate the test scores of the schools that retained bilingual education.

Examining the relationship between the percentage of English Learners enrolled in bilingual education in a school and test scores for English Learners, taking into account differences in schools’ pre-Prop 227 test scores and the percentage of schools’ students eligible for free lunch, reveals another indicator of Prop 227’s effect on achievement. This approach shows that the percentage of an elementary school’s students enrolled in bilingual education is significantly and negatively related to a school’s average test score for English Learners in both reading and math, even after accounting for the characteristics of its students. The results suggest that elementary schools with no bilingual-education enrollment score six points higher in reading and three points higher in math than schools with all their English Learners enrolled in bilingual education. The magnitude of the effect in reading is greater than one-half of a standard deviation–a large effect by the standards of education policy research. The effect on math is .21 of a standard deviation (see Figure 2). Moreover, as with the comparisons discussed above, these estimates may underestimate the true impact of English-language instruction, since Prop 227 has also changed bilingual education.

Data for individual students still suffer from the testing-rate bias favoring bilingual education, but where it is available one can at least determine which program the student is enrolled in. Valentina Bali also analyzed the achievement of individual English Learners in the Pasadena school district. In 1998, 53 percent of Pasadena’s English Learners were enrolled in bilingual education, compared with less than 2 percent after Prop 227. Adjusting statistically for the lower testing rate among students in bilingual education, Bali found that the effect of being in bilingual education in 1997-98 was negative and statistically significant, but the magnitude was only 2.4 points in reading and 0.5 in math.Using the same technique to examine the gains made by the two groups following the implementation of Prop 227, Bali found that putting these same students in a structured immersion classroom the next year eliminated the small gap between English Learners who had been in bilingual education and those not in bilingual education. The English Learners who transferred from bilingual education to structured English immersion made gains of four points in reading compared with gains of only two points for the students who had been taught in English previously. (There was no difference in the gains the two groups made in math.) In short, Bali’s analyses suggest that putting English Learners who had been in bilingual education into structured immersion increased their reading scores by about two points (.18 of a standard deviation) and their math scores by about a half point or less (.03 of a standard deviation). These effects are somewhat weaker than those produced by my school-level analysis, but still indicative of substantial benefits.Prop 227 appears to have had a positive effect on the achievement of English Learners, but it is not going to turn them into high-scoring students. Although bilingual education may be a relatively ineffective way of teaching English Learners, it was not the primary cause of their low achievement. The root problem is the way children are designated limited English proficient. An English Learner is not just a child from a non-English-speaking family. He or she is a child from a non-English-speaking family who scores low in English. Children from non-English-speaking families who score above a state-designated standard in English when they are initially tested are not designated limited English proficient. Therefore, English Learners must, by definition, be low scoring in English regardless of which program they are enrolled in.Practically speaking, the movement away from bilingual education will not dramatically improve the performance of students of non-English-speaking background, in part because bilingual education, in the pure form theorists advocated, was not as widely practiced as generally believed and in part because no program can dramatically improve the achievement of a group that is defined by its low achievement. There is a ceiling effect, not on individual children, but on the group as a whole because children who improve ultimately disappear from the category. Prop 227’s effects on student achievement have also been moderated by the fact that some schools and school districts are subverting the law’s intent and assigning Spanish-speaking English Learners to classrooms taught largely or almost entirely in Spanish in the first 30 days of school.Nevertheless, the law has had beneficial consequences. Test scores of English Learners appear to have risen considerably, if less dramatically than some of Prop 227’s proponents had hoped. Even more impressive is the fact that Prop 227 changed the direction of California policy, reversing 26 years of aggressive support for bilingual education by advocates within the state department of education. It seems that when voters speak clearly, policy does change, if somewhat more unevenly than the stark phrasing of the law would require.Christine H. Rossell is a professor of political science at Boston University. This research project was funded by the Public Policy Institute of California.

Last Updated


Notify Me When Education Next

Posts a Big Story

Business + Editorial Office

Program on Education Policy and Governance
Harvard Kennedy School
79 JFK Street, Cambridge, MA 02138
Phone (617) 496-5488
Fax (617) 496-4428

For subscription service to the printed journal
Phone (617) 496-5488

Copyright © 2024 President & Fellows of Harvard College