Comment by Catherine E. Snow
[The Federal Bilingual Education Program]
In these comments on Christine H. Rossell's paper about the federal bilingual education program, I will respond specifically to her principal conclusions and her four recommendations. I acknowledge the line of reasoning that underlies those recommendations, and in some cases I agree with both the reasoning and the conclusions. Who could disagree, after all, that the language of federal legislation is often characterized by poor writing, redundancy, and irrelevancies? However, Rossell and I differ in basic orientation to the issues and in the knowledge bases we bring to this topic. Furthermore, Rossell makes some basic misstatements in her paper, misstatements that are in some cases so egregious that it would be remiss of any discussant to let them pass unchallenged.
Political Science versus Developmental Psychology
Rossell is a political scientist whose major focus in her discussions of bilingual education seems to be to identify obstacles to the rational implementation of policies and unpredictabilities in the process by which policies have been decided upon. I am a developmental psychologist, interested in the conditions under which children develop to their full capacities, and in particular the conditions under which the largest possible percentage of children will successfully learn to read. Teaching children to read is, in my view, the most important task schools face, because failure in reading will doom children in most other school-related domains. Furthermore, learning to read is the first serious task children face in school, and success or failure in this task has a long-term impact on children's views of themselves as students and learners.
Thus, my primary interest is how to ensure success in reading--or, to put the issue in the public health perspective adopted by the Committee on the Prevention of Reading Difficulties in Young Children, which I chaired, how to minimize risk of reading failure. Learning to read is, in a sense, a natural act for a normally developing child in a literate culture with good schooling. Factors that enhance the risk that children will not follow this normal developmental pathway, as reviewed in the committee's report, include living in poverty, arriving at school not speaking English, attending schools with generally poor performance, and having [End Page 244] poorer than average language and preliteracy skills on school entry.36 The superordinate prevention strategies identified by the committee were: (1) ensuring that all children have access to excellent, language- and literacy-rich, preschool environments, and (2) ensuring that all children have access to excellent reading instruction.
What constitutes excellent reading instruction? The answer the committee gave to that question derives from its definition of reading: Reading is the act of constructing meaning from print. In other words, good readers of English use the alphabetic principle--that letters map in regular ways onto sounds--fluently and automatically to access word meaning and construct textual meaning. Excellent reading instruction must give children opportunities to grasp the alphabetic principle, in the context of a continual focus on reading for meaning, and must ensure that children have enough practice in reading to become automatic at the identification of words and to become skilled users of strategies for comprehension.
So what is the relevance of this view of reading to policies for bilingual education? The committee report concluded that the relevance was great--that this view of reading dictated that instruction that involves teaching children to read in a language they do not understand cannot constitute good instruction. Getting to the meaning is the whole point of reading, and children (particularly children from homes with low levels of parental education and little literacy exposure) can be expected to become confused and discouraged if expected to learn to read meaningless words. Furthermore, recognizing the words one starts out reading is an enormous support to the acquisition of the details of the alphabetic principle for most children. Children who cannot use meaning to support their reading during the initial stages of mastering the alphabetic principle are, inevitably, facing a much harder task than children being taught to read meaningful words.
At slightly more advanced stages, as well, knowing the language in which one is reading is crucial to developing sophisticated comprehension strategies and to spending enough time on task to develop fluency and automaticity with word recognition and the implementation of comprehension strategies. Enthusiasm for the practice of reading is a major input to developing fluency and success. Young children are unlikely to be enthusiastic or to engage in the practice they need if they are reading mostly incomprehensible texts. [End Page 245]
Thus, the report on preventing reading difficulties made the following recommendation concerning reading instruction for children who arrive at school not knowing English and not knowing how to read: that such children be taught to read in their native language if that is feasible (that is, if instructional materials and qualified teachers are available, and if such children are sufficient in number to justify native language reading instruction), while also receiving instruction focused on building oral proficiency in English, and that transition to English reading instruction take place only after a reasonable level of oral proficiency in English had been achieved. For children for whom these conditions do not hold, the committee recommended that formal reading instruction be postponed until some reasonable level of proficiency in English has been achieved and that the schools design programs to provide rich English language and preliteracy environments for such children, to ensure the most rapid possible acquisition. Furthermore, research should be undertaken to address the question of what constitutes "sufficient oral proficiency in English" to serve as a safe basis for reading instruction.
Two important points must be made about these recommendations. First, they are designed to reduce risk of reading failure. The committee recognized that children can learn to read in a language in which they have very low oral proficiency. Many millions of children have demonstrated that this is possible, not just in U.S. schools but also in colonial school systems across the world. However, such an approach is inherently more risky--more children will fail under this approach than if taught to read under conditions where meaning can be used as a support. The level of school failure that is normal in third-world countries with universal second language reading instruction simply cannot be tolerated in the United States.
Second, this recommendation applies only to children who cannot read on arrival at school. The committee did not make any recommendation concerning the value of bilingual education for older arrivals who can already read in their native language, nor for precocious readers who enter first grade reading fluently in their native language. However, much research suggests that such children can learn to read in the second language relatively quickly and then use second language literacy as a resource for second language oral acquisition. [End Page 246]
Specific Claims That Deserve Response
In responding to quotations from Rossell's paper, I recognize that some are less than central to the arguments she is making but nonetheless feel it is imperative to note when they are misleading or unfounded. In each case, her statement is provided in italics.
The U.S. Department of Education has enthusiastically embraced bilingual education. This claim is based on the distribution of funds available through the Bilingual Education Act and its successors. The largest part of these funds has typically been reserved for bilingual programs. But the total funding available has never been sufficient to serve most of the children who would have qualified for and benefited from bilingual programs. Thus, one could with as much justification argue that the Department of Education has failed to support bilingual education enthusiastically.
There is no treatment called "bilingual education" that is implemented in the same way and understood to be the same thing by everyone. This claim is true. The three program types noted in Rossell's paper fail to reflect the full range of program types or of variation within types. A major problem with the evaluations of bilingual programs, which as she notes are often exemplars of the worst in educational research, is that the variation within program type, which is often as great as that between program types, has not typically been taken into account. Many of the studies Rossell cites in her research synthesis would be rejected as fatally flawed by others who demand at least some information about classroom practices before evaluating a program's outcomes.
A major issue in thinking about these program labels is to realize that program labels do not distinguish educational experiences of children efficiently. The program labels mostly refer to classroom configurations--what mix of languages do the children in the classroom speak as native languages, and what language capacities are expected of the teacher? Thus, a transitional bilingual program is defined by the presence of children dominant in one language and a teacher with at least some competence in that language (and, one hopes, in English as well). Details of pedagogy, of distribution of instructional activities over the two languages, and of curriculum are all left undefined by the program category. [End Page 247]
"Structured immersion" is described by Rossell as "all-English" instruction in a self-contained classroom containing (exclusively) English language learners. Here, again, classroom configuration rather than explicit models for what to teach and how to teach it defines the program type. No one knows what should be going on in a structured immersion classroom, only that it should be going on primarily in English. Structured immersion is the program type that has been mandated in California by Proposition 227 and that is being proposed for Massachusetts and other states under Proposition 227-like legislation. No one has been trained in how to teach children in structured immersion settings. All that is prescribed is to restrict the use of the children's native language to a minimum. In many parts of the country, places where all the English language learners come from the same language background, structured immersion may not differ much in practice from programs labeled as transitional bilingual in which, as observational research shows, the children's native language may also be used rather little.
What does this mean? At a minimum, it implies that evaluations that compare programs called structured immersion with programs called transitional bilingual, but that provide no data on pedagogical activities and language use in the classroom, are worthless. More depressingly, all the complaints about the level of academic and English language accomplishments of children in transitional bilingual programs almost certainly will be replicated for structured immersion programs.
Only the Spanish speakers ... are receiving bilingual education through native tongue instruction, according to the theory. This claim reflects Rossell's beliefs that children who arrive at school speaking languages other than Spanish do not receive reading instruction in their native languages and that native language reading instruction is the key defining element of bilingual education. Both these claims could be challenged--on factual and theoretical grounds, respectively.
The majority of children receiving "bilingual education according to the theory" in the United States are Spanish speaking. Also, the majority of language minority children eligible for bilingual education but not receiving it "according to the theory" are Spanish speaking. Because 85 percent of non-English-speaking children in the United States speak Spanish, bilingual education not surprisingly has been developed largely with their needs in mind. Furthermore, the proximity of Spanish-speaking countries and the availability of Spanish language materials facilitate the [End Page 248] introduction of initial Spanish reading instruction, which is undeniably more widespread than initial instruction in other languages. The force of this observation in Rossell's argument is not, however, entirely clear. Even if only Spanish speakers were benefiting from native language reading instruction within bilingual programs, this does not constitute any indictment of the policy from an educational perspective.
In fact, though, sizable numbers of children who speak languages other than Spanish are receiving traditional bilingual education in which literacy is first introduced in the native language. Rossell and I spoke in the fall of 1998 at a meeting at Simmons College where we heard half a dozen bilingual teachers stand up and say that they were teaching children initial reading in languages ranging from Haitian Creole to Chinese.
The second aspect of this claim is that teaching reading in the native language is the key defining element of bilingual education. While not contesting the importance of native language literacy instruction to effective bilingual programs, I would disagree that this has been the primary justification within the intellectual history of the bilingual education movement. Rossell identifies a single theory as justifying bilingual education--Jim Cummins's theory of facilitation, or a common underlying proficiency. This theory has been widely cited in documents explaining the value of bilingual education, in particular in California where it has been combined with another theory, S. Krashen's notion of "comprehensible input" in informing educational practice.37
But facilitation is only one take on the value of bilingual education and, as Rossell's exposition reveals, the theory postdates the movement. The major intellectual input to bilingual education was practical wisdom--observations within the Coral Way program that bilingual education was possible--combined with a certain level of desperation concerning the academic accomplishments of non-English-speaking children. Justifications for bilingual practice have ranged widely, for example:38
--Basic humanity. It is neither kind nor nurturant to little children to put them in strange situations where they cannot communicate.
--Cultural continuity. Building on the skills and capacities children bring from home presupposes creating greater connections between home and school in language and in other ways.
--Achievement motivation. Children who are academically successful and well adjusted during their first year of formal schooling continue to show better school outcomes. [End Page 249]
--Bilingualism. The children's home language skills constitute resources that should be exploited.
--Sociolinguistic realities. In many immigrant communities, both the ancestral language and English are used widely, and thus if schools are part of those communities, both languages will be used in school, though perhaps for different purposes.
I would argue, perhaps agreeing with Rossell, that the most urgent justification for bilingual programs is the value of native language literacy as an academic protective factor, but it is a misrepresentation to think that this is the only rationale for bilingual programs that has ever been proposed.
The procedures used by school districts to identify students as limited-English proficient are inherently illogical. Rossell has identified a number of cases in which school districts have shifted their judgment of who was limited-English proficient (LEP) and some practices of identification and of decisionmaking that are less than exemplary. The "technology" of language proficiency assessment is relatively new, and her worst cases (from twenty years ago) reflect that. It is wrong, though, to suggest that all of the variation in identification rates she notes reflect incoherent identification procedures. The rise in the LEP population between 1987 and 1995 noted in her figure 1 is what would be predicted from immigration rates during those years.
Rossell is right that determinations must be made of whom bilingual programs are meant to serve and how best to identify those children. She proposes, in effect, a secondary prevention strategy--cast a wide net that requires little in the way of individual testing or screening, and provide prevention services to all within those risk groups, because enhanced education cannot hurt those who do not need it. Much can be said for this model as a mechanism for distributing money. Less can be said for it if it becomes a mechanism for ignoring the specific educational needs of language minority children. Furthermore, within the language minority population, different children may need access to different prevention strategies. If teaching children to read occurs optimally in a language they speak well, then before assigning language minority children to English-medium classrooms where formal reading instruction will be offered, whether they speak English well enough must be ascertained. How can this be done? By judicious use of some of the tests that Rossell discusses. She complains about those tests on various grounds. However, [End Page 250] that some native English speakers perform poorly on the same tests does not indict their usefulness as predictors of reading outcomes. Those low-scoring native English speakers no doubt would also profit from special educational treatments--but different ones than the non-native speaker needs.
The basic logic here is simple if one avoids being misled by Rossell's confused psychometric presentation. A six-year-old child who knows only three hundred words in English is going to have trouble in an English-medium first grade where the teacher is focusing on teaching reading. The child with a three-hundred-word vocabulary in English does not have a stable set of phonological distinctions, knows too few words to develop the phonological analysis skills basic to learning letter-sound correspondences, and will encounter many words critical to understanding first-grade texts that he or she does not know the meaning of. If such a child is a native English speaker, I would recommend English language enrichment combined with emergent literacy activities to promote prereading skills. If this child also knows six thousand words in Spanish, then the child should be given formal reading instruction in Spanish so that he or she can master the alphabetic principle and automaticity in letter and syllable recognition and develop the habit of reading, while acquiring more oral skills in English. If this child has become an on-grade-level reader in Spanish by the end of first grade, introducing English reading material at that point is likely to promote oral English development as well.
Nationally it is only possible to have half the student population at grade level. This claim, which Rossell not only makes but also excoriates others for failing to understand, is patently wrong. Rossell fails to display an understanding of the not-very-subtle distinction between norm-referenced and criterion-referenced measures. On norm-referenced measures, some test takers will always score below average. On criterion-referenced tests, however, everyone in a class, a school, or a nation can "pass." Grade level for reading is a criterion--established as a function of curricula and consensual standards. Rossell likely would counter that the curricula and the consensus standards shift to reflect average performance. Standards do shift--in the current period of educational reform, standards are being shifted upward, with the result that more than half the children in the nation are failing to meet many of them. If grade-level standards were averages, as Rossell seems to think they are, such an outcome would also be impossible. [End Page 251]
Educators apparently believe that children who score below average ... are children who are in academic difficulty. Rossell accuses educators of being as confused as the general public as to what tests mean and of appearing to believe that a low score has some absolute meaning. She is right that low scores on norm-referenced tests have no absolute meaning. But they do have power nonetheless, as predictors of academic outcomes and of literacy achievement. One could construct a language test on which first-grade children with excellent language and literacy skills score as low as the 10th percentile (the Scholastic Assessment Test [SAT] for college-bound high school students, for example). But if language minority or English-speaking students are scoring at the 10th percentile on valued age appropriate tests of vocabulary, knowledge of grammar, and oral comprehension in English, they are going to have trouble learning to read and will need adapted programs. All of Rossell's smoke and mirrors about the relativity of norm-referenced tests do not change those basic facts. The shift she notes in New York state's criterion for classification as LEP from the 23rd to 40th percentile may have reflected the reaction of wise practitioners to this finding. Whether the 40th percentile on the Language Assessment Battery (LAB) represents a level of English that is high or low depends on the norming sample. A cutoff at the 60th percentile would be justifiable, if children scoring at the 59th percentile were "having difficulty in English." No absolute meaning is attributable to either high or low scores, but that does not imply that test scores are meaningless. Rossell's university no doubt imposes as an entrance requirement for non-native English speakers a score on the Test of English as a Foreign Language (TOEFL) that is well above the 70th percentile for that test--and then occasionally still admits students whose English proves not to be adequate for the academic demands of a U.S. university.
School districts and states can increase or decrease the number of LEP students simply by changing the standard. Moreover, they do this on a regular basis without any rationale other than the desire to help students. School districts can change incidence of students classified as LEP or any other category by changing cutoff points. Rossell acknowledges that they do this through a desire to help students. While being below average on a universally administered test is the fate of half its takers, scoring below average on a test normed on non-native English speakers is not a likely fate for native speakers of English. Being below grade level could, in principle, afflict all or none of any given population. Again, Rossell confuses [End Page 252] norms with criteria, and norms for the language minority population with norms for the entire population.
[Facilitation theory] ignores the issue of the great variation in written language. In particular, it is silent on how to teach Asian children to read and write in their native tongue and why doing so is desirable. Essentially, Rossell is claiming that learning to read Asian languages could hardly develop skills that transfer to English. Rossell is general in her reference to Asian languages, which fall into a number of different language groups and use a variety of orthographies. Vietnamese and Korean, for example, use alphabetic writing systems--in the first case the Roman alphabet; Khmer's writing system is described as "alpha syllabic"; Japanese uses a syllabic system that displays many of the principles of alphabet writing but at a slightly less abstract level; the languages of the Indian subcontinent mostly use consonant-centered alphabetic systems.39 Thus, Rossell seems to have generalized Chinese morpho-syllabic orthography (which is borrowed for use in sophisticated Japanese and Korean writing, but not for initial literacy instruction) to the entire continent. Nor do the data suggest that it is harder to learn to read in syllabic or morpho-syllabic orthographies. Initial reading in such systems is much easier, though acquiring full literacy (two thousand to twenty thousand Kanji) in morpho-syllabic systems does take some years.
The key issue here, though, is whether transfer of literacy skills from a first to a second language occurs only when the two languages share a writing system. From my own infinitely greater difficulty trying to learn to read Arabic and Hebrew as compared with Spanish and Dutch, I can sympathize with Rossell's intuition-based claim, that transfer to new alphabetic systems is less extensive than transfer within alphabetic systems. Unfortunately a systematic research base does not exist from which to argue either side of this case. However, there clearly are things one learns from knowing how to read in Chinese that could be useful in learning to read English--the analyzability of print, the need to access a phonological representation of words being read, something about the difference between spoken and written language, the value of literacy skills, and so on. And knowing any alphabetic system prepares one for the phonological analysis of words that constitutes the basis for any other alphabetic system--and often is a great barrier to young readers of English because English has a deeper orthography than most alphabetic languages. I must agree with Rossell that it may not be worth the time and [End Page 253] effort to teach children to read first in a language in which they will not have ongoing opportunities to read. Transfer of skills from first to second language literacy may require more than initial literacy accomplishments in the first language. In this case, given the absence of realistic opportunities to become an advanced or fluent reader in some languages because of absence of reading materials or ongoing support and instruction, some children perhaps should postpone formal literacy instruction until after the development of sufficient English proficiency.
Although Title VII favors bilingual education, not an ounce of evidence in 1968, and no consistent evidence in the thirty years since then, proves its superiority. The claim that research fails to confirm the value of bilingual education, central to Rossell's paper and to her work in this area, is demonstrably false. She and I agree that much of the research on this topic is execrable. She has included many of those studies in the overview presented in table 1. A. C. Willig and J. P. Greene reviewed far fewer studies in their meta-analyses because they identified crucial design flaws in many of the seventy-two studies Rossell allows to "vote" in her method.40 Such flaws ranged from including graduates of transitional bilingual programs in the putative control group to failing to distinguish between Canadian (elite, elective) and U.S. immersion programs. The flaws in Rossell's approach to synthesizing research on bilingual education have been repeatedly discussed in numerous venues. Despite her willingness to stretch her methods so as to disfavor bilingual education as a model, her results do not demonstrate that bilingual programs are bad for children. The scores come out about even for the various programs across the various assessments offered. A fairer presentation of data and a stricter selection limited to well-designed studies make clear that bilingual education is much better for children entering school without literacy skills in English.
Four Recommendations
First, eliminate reference to limited proficiency in English and substitute "language minority" as the criterion for distributing Title VII funding. This proposal would increase the population base for distributing Title VII funding and would streamline the process of estimating the numbers of children eligible. Though it would not solve all problems of identification, it would simplify identification and could reduce error. If [End Page 254] elimination of consideration of children's language skills did not extend to the decisions made about their educational program, but was limited to determining a basis for distribution of funding, it might be sensible and deserves a trial.
Second, eliminate reference to specific educational programs in Title VII. This recommendation is based on a claim with which I categorically disagree, that research cannot be used to justify bilingual approaches to educating language minority children with limited skills in English. Thus, while I endorse the underlying notion that states and districts should be required to use solid research evidence in justifying their decisions about education, I cannot endorse this recommendation. Bilingual programs offer the best education to many children. Their quality needs to be improved, as Rossell points out. But the United States is not responding to the evidence that math and science programs are often ineffective by eliminating them. Instead, efforts are made to improve them. The same approach should be adopted for bilingual education.
Third, rewrite Title VII to eliminate redundancies and irrelevancies. Rossell and I agree entirely that Title VII is badly written. The impact of this on schools and children is not, however, so negative that major efforts to call in the language police need to be launched.
Fourth, Title VII legislation should require school districts to collect and make available outcome data, not to carry out evaluation studies. Like Rossell, I am appalled at the money and time wasted in carrying out evaluations of educational programs, generally with inadequate designs and insufficient resources to do a good job. However, requiring school districts to provide outcome data to outside evaluators will hardly solve the problem of generating good evaluations. A major issue in evaluating educational programs (those for language minority children and all others) is knowing what is going on inside the classroom. Program labels are not enough. Thus, I agree that evaluation efforts need to be rethought but do not agree that the solution offered by Rossell is adequate.
