Bilingual Education in Arizona:
What the Research Says

by Stephen Krashen and Grace K. Park
University of Southern California 
Dan Seldin
University of California, Riverside

In the fall of 2000, Arizona voters will vote on Proposition 203, a measure more severe than California’s Proposition 227.  If passed, it will dismantle bilingual education in Arizona (Crawford, 2000a). Proposition 203 claims that Arizona’s public schools are doing an “inadequate job” of educating immigrant children, that bilingual education has been a failure, and that the solution is “heavy” exposure to English. Is this correct? 

In this paper, we provide a very brief review of research on bilingual education in Arizona.  We restrict our survey to controlled scientific studies in which the progress of a group who experienced bilingual education is compared to a control group, focusing specifically on English language acquisition. We present a fairly complete description of each study as well as the actual test scores, as this data may not be easily available to all readers. 

The first two studies examine the immediate impact of bilingual education on limited English proficient children.

De la Garza and Medina (1985) compared children in a bilingual program to English- dominant children in an all-English program. This is a very stringent test; most studies compare limited-English-proficient (LEP) children in bilingual programs with equally LEP children in all-English programs. The study was done in Tucson. Eighty percent of the bilingual education children were classified as LEP, but 94% of the comparison students were English-dominant. In addition, the socio-economic class of the English speaking children may have been higher (37% received free lunches, versus 76% in the bilingual group). 

In the bilingual classes, the first language was used 75% of the time in grade 1, 70% in grade 2 and 50% in grade 3. Reading instruction was done in Spanish, and the two languages were alternated weekly or monthly in subject matter classes, with the instructional language “contingent upon the L2 [English] proficiency of the LEP student” (p. 251). Children in the bilingual program scored slightly higher than the English-dominant comparison students in every grade tested, although differences were not statistically significant.  The small sample size for the bilingual group is of concern, but the results are spectacular. Children in bilingual education did as well, and perhaps slightly better than, a group of children who were dominant in English, and performed at national norms.

Table 1: Bilingual education students vs. English-dominant students (de la Garza and Medina, 1985) 

  Bilingual education English mainstream
number of students 25 117
  mean standard deviation mean standard deviation
grade 1 50.3 7.3 49.6 7.3
grade 2 52.9 8.l 51.3 8.5
grade 3 51.9 8.2 50.4 7.3

A study by Saldate, Mishra and Medina (1985) compared programs in Douglas, Arizona, a city on the Mexican border in which a great deal of Spanish is spoken. Bilingual education students were all from one school whose enrollment was nearly entirely Mexican-American. Comparison subjects were from other schools, “whose enrollment ranged from 60 to 90 percent” Mexican-American (p. 26). While bilingual education students were described as “low socio-economic,” there was no description of comparison students, but Saldate et. al. note that in the study, pairs of students from each group were matched on the basis of the Peabody Picture Vocabulary Test. 

The description of the bilingual program was provided in Medina, Saldate and Mishra (1985), a follow up study discussed below.  For subject matter, the “preview-review” method was used in which material is presented in one language and then reviewed in the other, followed by another review in the first language. Spanish reading and ESL was also included. According the Medina et. al., the first language was used 90% of the time in grade 1, 80% in grade 2, and 50% in grade 3. Also, by grade 3, alternate-day instruction was introduced for subject-matter teaching. 

Children were tested at the end of grade 2 on the Metropolitan Achievement Tests and at the end of grade 3 on the Wide Range Achievement Tests. For the WRAT, scores for English reading are presented here, a test that emphasizes word recognition and decoding. As shown in Table 2, the comparison group did slightly better than the bilingual education students in grade 2, but differences were small and not statistically significant. In grade 3, the bilingual education students outperformed the comparison students; here the difference was statistically significant and quite large. These results should be interpreted cautiously, however, because of the small sample size and the fact that the grade 3 sample was reduced from grade 2. Still, they are impressive.

Table 2: English reading performance of students in bilingual vs. nonbilingual programs (Saldate, Mishra and Medina, 1985)

  bilingual group comparison group
  number of students mean standard deviation number of students mean standard deviation
grade 2 (MAT) 31 119.70 30.21 31 128.19 29.04
grade 3 (WRAT) 19 62.47 15.71 19 51.26 7.63

The results of these two studies are fully consistent with the results of studies done elsewhere. Greene (1997),  for example, concluded that the use of the native language in instructing limited English proficient children has “moderate beneficial effects” and that “efforts to eliminate the use of the native language in instruction ... harm children by denying them access to beneficial approaches.” 

The next two studies examine the progress of graduates of bilingual education.

Medina, Saldate and Mishra (1985)  is a  follow-up of Saldate et. al. (1985). Subjects were graduates of the bilingual program described in Saldate et. al. (but were not the same children examined in that study), and had received bilingual education up to grade six. After grade six, instruction was all in English. All subjects in both the experimental and control group had graduated high school "on schedule." In both the experimental and comparison groups, all who remained to graduate on schedule were tested: in the bilingual group, 19 out an original 136 (14%), and in the comparison group, 24 out of an original 136 (18%) nearly identical percentages. The results of tests administered in grades 6, 8 and 12 are presented in Table 3. There were no significant differences between the groups at any of the levels, although the comparisons did slightly better. Note that at grade 12, both groups were near national norms. Once again, sample sizes were quite small.

Table 3. Long-term performance of graduates of bilingual vs. nonbilingual programs (Medina, Saldate, and Mishra, 1985)

  bilingual group comparison group
  number of students mean standard deviation number of students mean standard deviation
grade 6 (MAT) 19 44.78 8.20 24 46,23 5.51
grade 8 (MAT) 19 46.82 8.69 24 47.83 6.95
grade 12 (CAT) 19 49.57 7.95 25 52.82 5.18

MAT = Metropolitan Achievement Test
CAT =California Achievement Test

Powers (1978)  studied the progress of graduates of bilingual education in Nogales. Students were tested at grade 5 on the Stanford Reading Test and at grade 7 on the Gates-MacGinitie Reading Test. We have little description of the bilingual program; we know only that students in the study had been in bilingual education for three to five years, and that in grade 1 85% of instruction was in Spanish, and in grades 4 and 5, 30% was in Spanish. Comparison students were Mexican-American students who did not receive any bilingual education.

Table 4 presents mean scores for bilingual education graduates and comparison graduates, adjusted for socio-economic factors. Comparison studies did better than students in bilingual education in grade 5, but this difference was reduced by grade 7, and was not statistically significant.

Table 4:  Adjusted means for tests of reading comprehension, bilingual vs. nonbilingual graduates (Powers, 1978)

  number of students bilingual group scores number of students comparison group scores
grade 5 40 4.6 41 5.09
grade 7 44 21.36 43 23.82

Powers also presented evidence that students who reported more English use at home on entry to school and at grade seven had higher reading scores. Of crucial importance is the fact that comparison students reported more English use. This factor, however, was not controlled in the analysis. 

The results of this study are only suggestive: An important control, English use outside of school, was not included and we have no description of the bilingual program. These factors could act to underestimate the effect of bilingual education.

In studies done outside of Arizona, graduates of bilingual education have been shown to do very well, sometimes outperforming native speakers of English in their districts (Burnham- Massey and Pina, 1985; San Francisco USD, 1998).  In the two Arizona studies involving graduates reviewed here, bilingual education graduates also did well, achieving close to national norms in one case, and doing only slightly worse than comparisons in another case, despite using less English outside of school.

When children are tested during bilingual programs or immediately after they are exited,  bilingual education shows a clear advantage. Results for the highest grade for Saldate et. al (1985). are quite positive, and de la Garza and Medina (1985) show that bilingual education students do as well as English-dominant students. The results of studies done with graduates of bilingual education are also encouraging, with graduates performing at national norms in one case.

These results do not support the view that bilingual education in Arizona is a failure. In fact, Arizona studies strongly suggest that bilingual education is beneficial, a conclusion that is consistent with the results of studies done in other states. What is clear is that calls for the elimination of bilingual education are completely unjustified.

The results presented here probably underestimate the positive effects of bilingual education. We have little information about the details of the bilingual education programs, and what details are provided suggest that the programs were not set up in an optimal way (Krashen, 1996). One would expect that bilingual programs that are better implemented would produce even stronger results. We have learned a great deal since the 1970s and 1980s. 

In recent years, students in bilingual education in Arizona have also consistently outperformed limited English proficient children in English-only programs on Stanford 9 tests of reading (Crawford, 2000b). As Crawford notes, these comparisons are very crude, as there is no control for any possible confounding variables, including pretest levels of English literacy, socioeconomic class, the type of program, or the duration of the treatment. We present here the 1999 SAT9 results, reported in January, 2000. Note that bilingual education students are better at every grade level. Bilingual education students did better than English-only students for the previous two administrations of the Stanford 9 as well.

Table 5. LEP student performance on Stanford 9 reading test, bilingual vs. English-only programs in Arizona, 1998-99

grade bilingual education English-only
  number mean score number mean score
2 1029 37.2 3594 37.0
3 1213 35.9 4150 32.3
4 1247 38.4 4021 33.8
5 2026 32.4 4775 31.6
6 1506 34.7 4327 34.3
7 1257 33.4 3891 31.8
8 1016 35.4 3798 32.6
9 768 33.4 2613 27.7
10 708 33.6 2196 25.5
11 564 33.4 1681 26.9

Mean scores are given in NCE’s (normal curve equivalents).

Acknowledgment: We thank David McField for his encouragement and assistance.

Burnham-Massey, L. and Pina, M. 1990. Effects of bilingual instruction on English academic achievement of LEP students. Reading Improvement 27(2): 129-132. 
Crawford, J. 2000a. English-only vs. English-only: A tale of two initiatives
Crawford, J. 2000b. Stanford 9 scores show a consistent edge for bilingual education.
de la Garza, J. and Medina, M. 1985. Academic achievement as influenced by bilingual instruction for Spanish-dominant Mexican-American children. Hispanic Journal of Behavioral Sciences. 7(3): 247-259.
Greene, J. 1997. A meta-analysis of the Rossell and Baker review of bilingual education research. Bilingual Research Journal 21(2,3): 103-122. 
Medina, M., Saldate, M., and Mishra, S. 1985.The sustaining effects of bilingual education: A follow-up study. Journal of Instructional Psychology 12(3): 132-139.
Powers, S. 1978. The influence of bilingual instruction on academic achievement and self-esteem of selected Mexican-American junior high school students. PhD Dissertation, University of Arizona. 
Rosenthal, R. 1984. Meta-Analytic Procedures for Social Research. Beverly Hills: Sage.
Saldate, M., Mishra, S., and Medina, M. 1985. Bilingual instruction and academic achievement: A longitudinal study. Journal of Instructional Psychology 12(1): 24-30.
San Francisco Unified School District. 1998. The Lau Report, 1997-98. Language Academy & Other Bilingual Programs, San Francisco Unified School District.