The question of mother-tongue, dual-medium or English-only instruction in grades one to three is starting to look more nuanced than many of us have Âsupposed. So too is the matter of the instruments — the tests — used to assess competence and achievement in these foundation-phase grades.
These are among our findings based on results we released this month of grade one children’s Âoverall cognitive ability, their language competence and their knowledge of mathematics. The results come from pilot studies we have designed in the Institute for Childhood Development at the University of Johannesburg.
We tested pupils at two Gauteng schools. One was in Ekurhuleni where English is the only medium of instruction.
The 30 grade ones tested here come from homes where nearly all the African languages in the Âcountry are spoken, but there is very little classroom support in their mother tongue.
The other was the Funda UJabule School on the Soweto campus of the University of Johannesburg, where the two grade one classes are taught in either Sesotho or isiZulu, with English translations. This is done to acquaint with the English terminology of Âmathematics and to link mother tongue and English systematically. As described recently in this newspaper, the 76 children tested here are thus taught largely in their mother tongue but their teachers use a structured dual-language translation approach (“Languages of excellence”, Mail & Guardian, October 21). The mathematics tests are designed and standardised in Germany but translated and Âlocalised — they are conducted in the children’s mother tongues and adapted to objects known to the Âchildren, for instance soccer balls and fruits.
Reading the results
The first notable result is that all the children have a reasonable understanding of number and size in a test designed for children in Europe. The average for the Âchildren in the Soweto school was 61% and for the children in the English-language school in Ekurhuleni it was 55%. We had expected that the children in this school would not cope well with the test, given what the literature says about mother-tongue instruction and also what is intuitive to South Africans: we are quite set in our ways and beliefs when it comes to mother-tongue instruction. Yet the Âchildren in this school seemed to have Âmastered some of the more challenging conceptual items well. They also did better on some of the more Âdifficult questions.
But there is little doubt that, based on results from Funda UJabule, a dual-language teaching approach is a good thing. By dual language we mean that the children’s mother tongue is used as a support tool while the teaching increasingly takes place in English.
But there is more to it than simply ad hoc support: the teachers in this school follow a systematic translation process in which they do not encourage random language-mixing or code-switching. While the spoken dialects, such as the urban varieties of isicamtho or tsotsitaal, are not ignored, the children are encouraged to use their mother tongue in the way it is used in their reading books in the early grades, for instance. They thus get to know the terminology of numeracy and mathematics in both English and the mother tongue and they explore meaning in both English and the home language. And so they build a repertoire of terms and phrases in English. In grade two they will take the test in English.
Introducing English as the medium of teaching mathematics at Funda UJabule was not merely a pragmatic decision arising from teachers’ discomfort with using only isiZulu or Sesotho in the two classes. It had to do with our concern about large-scale testing. It is general knowledge in this country that our children perform poorly in these tests. No one would disagree that language is an obstacle in these tests, but we do not know enough about just how this obstacle manifests.
We conducted one-hour individual interviews with each of the children during the testing and these strongly reinforced our belief that the language of abstract concepts, or academic language, is the necessary partner to understand and cope in school. We think children’s limited repertoire of this language may inhibit their understanding.
Thus, we reasoned, the results of mathematics tests are partly the results of linguistic representation, with all the problems embedded in this. We say this even though we know that some of the conceptual understanding of young children is pre-linguistic. In other words, they may understand something but they cannot verbalise it or recognise it in the form of verbal language. Many learners struggle with mathematical word problems.
So we are encouraged by the results of both schools. The Ekurhuleni results suggest English on its own does not deter grade one children from learning numbers solely in that language. We are also encouraged by the results from Funda UJabule. But we know the numeracy test translations will have to be refined. For instance, in some cases, including everyday language, a colloquial “kushota ngezingaki?” — “how many short?” — could succinctly express “how many less?” better than an ambiguous “zincane kangakanani?” — “how much smaller?”.
The plan
For the next four years we will follow the mathematics understanding of the 2012 cohort of grade ones in these two schools, as well as in two other schools where there is only mother-tongue — isiZulu and Sesotho — instruction in numeracy and mathematics in the first three grades. Between the four schools we should be able to draw some conclusions about the learning of about 200 children over four years, until they are in grade four.
The numeracy test is not the only instrument in our piloting. We also assess the children’s overall cognitive ability on what is generally regarded as a reliable South African “intelligence” test. This test was translated into isiZulu and Sesotho. In the pilot test there was a good correlation between the numeracy scores of this test and the one we have described more fully already.
In addition to these two instruments, we also designed an oral language competence test in English. These results were not unexpected. The school in Ekurhuleni performed substantially better than Funda UJabule, where the children had less exposure to English. We can now only speculate where these scores will lie in four years’ time and how this will correlate with the learners’ performance in school, in the annual national assessments and in the provincial tests.
For the next four years we will administer mother-tongue reading tests for children at Funda UJabule at the end of grade one and in English for the Ekurhuleni school. We plan to include a test on children’s conceptual development in science in this way.
By 2015 we will have large volumes of data in all these areas, showing how each individual child in the Funda UJabule school has Âdeveloped over time compared with data on the children in three control schools. With all these tests we wish to achieve only one aim, namely to give detailed analysis of individual children’s learning over time and to understand dual-language Âeducation better. We hope to show that Âlarge-scale surveys answer only some of the questions we need to ask about learning in the foundation phase.
Elbie Henning is professor of educational linguistics and lead researcher in the University of Johannesburg’s Institute for Childhood Education, Graham Dampier is a researcher at the university’s Centre for Education Practice Research and Nick Welch is a scholar of urban Bantu languages, a vernacular rap artist and a “Strictly Vernac” comedian who also works as a part-time researcher at the ÂUniversity of Johannesburg.