Why Malherbe (1946) Is NOT Evidence
Against Bilingual Education
by Stephen Krashen
University of Southern California
In their review of bilingual education research, Rossell and Baker (1996)
categorize Malherbe (1946) as an instance of "structured immersion"
students showing more progress in second language reading than students
in transitional bilingual education. It is important to reexamine this
conclusion in detail, because it is one of ten studies interpreted as showing
"immersion" to be superior to bilingual education. Thus, it constitutes
10% of the case against bilingual education and for immersion. In addition,
Malherbe's study is not widely available and has not been described in
the professional literature by other scholars.
In this note I argue that Malherbe's study violates criteria for inclusion
in research reviews suggested by those on both sides of the issue. There
is no control for pre-existing differences and no statistical tests are
performed, violations of Rossell and Baker's criteria. In addition, a variety
of "bilingual" programs are included, none of which appears offer
much in the way of comprehensible input in English.
Malherbe studied the acquisition of English by Afrikaans-speaking children
and the acquisition of Afrikaans by English-speaking children under different
conditions, the result of a survey done in 1938. The published version
was considered preliminary, as additional analysis was cut short because
of World War II.
I focus here only on language acquisition. The report also covers the
impact of bilingual education on "social attitudes" and "discrimination."
The analysis is restricted to English-Afrikaans relationships. Black South
Africans are not considered in this or in the language analysis, and are
only mentioned once, in a footnote (p. 77).
The sample consisted of a total of 18,773 children, but only a tiny
subset of this sample is relevant to the question of bilingual education.
The only comparison that appears to speak to the issue involves the small
number of Afrikaans students coming from monolingual Afrikaans-speaking
homes who were enrolled in either English-only medium schools (n = 246
students), equivalent to submersion, and those enrolled in "bilingual"
schools (n = 440). Most Afrikaans-speaking children attended Afrikaans-medium
schools (85.5% of the unilingual Afrikaans-speaking children, or 4049).
It was not possible to study the progress of monolingual English speaking
children in Afrikaans-only medium schools because the sample size was too
Malherbe notes that "economic environment" and IQ data were
collected but it appears that this information did not enter into the analysis:
" ... before comparing different types of schools ... one had the
make sure that the other important factors (e.g. intelligence level and
economic environment) were also comparable. The investigation was therefore
a complex one. It was only by dealing with large numbers, such as were
covered in this survey, that anything like the semblance of a scientific
basis for our conclusions could be arrived at. Though thousands of tables
and graphs have already been worked out, this investigation has not yet
been completed owing to interruption by the war. It is hoped, however,
to complete it when the war is over and publish a full report" (p.
To my knowledge, the complete report was never published. Note that
this lack of control violates one of Rossell and Baker's conditions for
including studies in their analysis: Rossell and Baker insist that there
be some control for possible pre-existing differences in the experimental
and comparison groups, through randomization or statistical means. There
is clearly no control in this study: We have only raw data for fairly small
groups of subjects.
In my analyses of the research, I relaxed Rossell and Baker's requirement
somewhat (Krashen, 1996), requiring only that there be no reason to suspect
differences existed between the comparison and experimental groups, but
in this case there are very good reasons to suspect that important differences
existed. Malherbe notes that the Afrikaans monolinguals typically live
in "remote rural areas" (p. 41), while monolingual English speakers
are generally found in urban areas. It is unlikely that there were many
English-only medium schools in the remote rural areas where monolingual
Afrikaans speakers live, which suggests that there may have been important
environmental differences between those who went to English-only and bilingual
schools. Malherbe presents data showing strong differences in English language
attainment between Afrikaans speaking students in rural, town, and city
schools (p. 104).
As noted earlier, English-only schools for monolingual speakers of
Afrikaans must be considered a form of "submersion," as no special
effort was made to make English input comprehensible. But the "bilingual"
schools did not appear to be much better. Three models were used:
(1) Parallel class: Afrikaans and English speakers are taught in their
own language, in separate classes. Instructionally, there is thus no real
difference between this model and an Afrikaans-only school, even though
speakers of both languages are under the same roof, but Afrikaans speakers
heard much more English outside of class than those in Afrikaans-only schools
(2) Dual medium/repetition method: In this model, English and Afrikaans
speakers are in the same classroom, and the two languages are alternated.
a form of concurrent translation.
(3) Dual medium/separate subjects: Some classes are taught in English,
others in Afrikaans. Speakers of both languages are in the same classes.
The English medium classes could thus be considered submersion,
All three of these "bilingual" options are less than optimal
in supplying comprehensible input in the second language. In (1) students
are exposed to little English in school, in (2) they have concurrent translation,
a method known to produce poor results in English language development,
and in (3) there is no mention of whether any accommodation is made to
non-native speakers in the class.
We may simply be comparing "submersion" to three kinds of
inefficient bilingual programs. In addition, Malherbe does not tell us
how many students of the 440 were in each model. Also contrary to Rossell
and Baker's categorization, there is nothing "transitional" about
the bilingual programs, as the first language is continued, and expected
to continue development.
Results and conclusions
English language tests (vocabulary, reading comprehension, writing
and spelling combined) for Standards 4-10 (grades 5 to 11) were combined
and averaged and converted to standard scores (mean = 500 for the entire
sample). Malherbe reported no statistical tests and no standard deviations,
another serious violation of Rossell and Baker's criteria for including
studies in their survey.
Afrikaans speaking children in English-only schools clearly did better
(mean = 519) than those in bilingual schools (mean = 432). In fact, the
bilingual children's mean was not much higher than the mean score for children
in Afrikaans-only schools, where little English was available in school
or in the informal environment (mean = 412).
We cannot, however, score this as a victory for submersion over bilingual
education. Malherbe (1946), in fact, teaches us very little.
- It is likely that comparison students differed from bilingual students
in important ways. Those who attended English-only schools may have come
from a different and more privileged population, a conjecture consistent
with the finding that this group scored well above the population mean
on Afrikaans language (533) without any significant exposure to it in school.
- The three "bilingual" programs appeared to deliver little
comprehensible input in English. One version, in fact, was identical to
Afrikaans-only schooling. Also, we have no idea how many of the subjects
were in which version.
- No statistical tests were reported.
What is clear is that Rossell and Baker should not have included this
study as evidence for or against anything.
Malherbe, E. G. 1946. The Bilingual School: A Study of Bilingualism
in South Africa. New York: The Arno Press (reprinted in 1978, first
printed in 1946).
Krashen, S. 1996. Under Attack: The Case Against Bilingual Education.
Culver City, CA: Language Education Associates.
Rossell, C. and Baker, K. 1996. The educational effectiveness of bilingual
education. Research in the Teaching of English 30 (1): 7-74.
COPYRIGHT NOTICE: Copyright © 1999 by
Stephen D. Krashen. All rights reserved.