What language should South African children be taught in? The ongoing debates about the language of instruction in schools evoke strong (and often emotional) responses, as has again been seen in the various contributions in the media during the past two weeks.
However, these responses are seldom backed by evidence. A more scientific approach is required to address such an explosive topic and this week we released a working paper that we believe offers this.
First, though, some background. In his recent Percy Baneshik Memorial Lecture to the English Academy of South Africa, Professor Jonathan Jansen recommends the introduction of English as the language of instruction as early as possible ("Wrest power from English tyranny", Mail & Guardian, October 4).
As Jansen observes, many black parents recognise that English proficiency is important for successful participation in the economy, and therefore conclude that their children should be instructed in English. But this conclusion rests on a key assumption: "Instruction in English from as early as possible is the best way to become fluent in English."
At first this seems like a fair assumption to make: Surely the earlier you start with English the better? Doesn't practice make perfect?
However, most linguistic theory doesn't agree. Many linguists argue that when it comes to learning a second language it is crucial to first have a solid foundation in one's first language.
These theories predict that several years of mother-tongue instruction will lead to better second-language acquisition than being instructed in that second language from the first day of school.
So education policymakers in countries with multiple home languages, such as South Africa, have a dilemma. The goal is to optimise educational outcomes, which include English fluency, mother-tongue fluency and skills in other subject areas, for pupils whose first language is not English but how does one best achieve this?
The choice is between English instruction from grade one or starting with mother-tongue instruction and transitioning to English as the language of instruction at a later grade. Although there is a lot of theory about this, there is very little empirical evidence on the causal impact of these alternative instructional models.
As researchers focused on education in South Africa, we recently conducted a study that we believe provides some preliminary evidence on one important part of the language in education debate. We will describe our findings below.
Because non-language subjects in the matric exam are available only in English or Afrikaans, schools with pupils who are neither Afrikaans- nor English-speaking must figure out when to begin using English (or in a minority of cases, Afrikaans) as the language of instruction.
The current language in education policy encourages the use of mother-tongue instruction in the first three years of primary school followed by a switch to English or Afrikaans in grade four, but allows schools to make the final decisions.
Some schools choose to commence immediately with English as the language of instruction from grade one (these are sometimes referred to as "straight-for-English" schools). On the other hand, most schools choose to teach in the home language of the majority of the children in the school during grades one, two and three while they take either English or Afrikaans as a language subject to help them to prepare for the switch (we refer to these schools as mother-tongue schools).
About 60% of children in grade three learn in a language other than English or Afrikaans. By grade four, this proportion is only about 5%. Most children experience a transition from mother-tongue instruction to English instruction in grade four, something that by all reports is a difficult process.
However, the suggestion that using English as the language of instruction from grade one would avoid this difficult transition is too good to be true. Learning, including language learning, begins well before children enter formal schooling. Therefore, a "straight-for-English" approach still involves a transition in a child's language of learning, and children may be less well prepared for such a transition in grade one than in grade four.
But these arguments only take us so far. What we need is rigorous evidence.
Comparing apples with apples
Measuring the impact of this language choice on the academic performance of children later in life is no simple task.
First, schools that decide to teach in English are, on average, more likely to charge school fees, have smaller classes and have access to more resources than mother-tongue schools. Even more importantly, the quality of teachers, their English background and other aspects of school quality are typically stacked in favour of these "straight-for-English" schools.
In addition, children who attend the "straight-for-English" schools are on average from richer households where they receive more academic support from better-educated parents or caregivers, are less likely to go to bed hungry, have fewer siblings with whom they need to share resources and are more frequently exposed to English on television and in the home.
Both sets of factors – school quality as well as the home environment – have been shown by research to impact strongly on the academic performance of children.
When we therefore observe that children in the "straight-for-English" schools perform much better than their peers in the home-language schools in both English and mathematics, it would be naive to conclude that this is primarily driven by the language of instruction. Rather, what is required is to separate out the effects of the various factors affecting academic performance.
The real question is how much of the differences in performance is explained by children's home circumstances, how much by a school's quality and, lastly, how much by the language in which children are taught.
In light of these challenges, and bearing in mind the need for evidence-based policy, we have released a working paper containing our research on this issue.
We used data from the department of basic education's "annual survey of schools" to identify the language of instruction in each grade in each school for the years 2007 to 2011.
We also used test score data from the annual national assessments of 2012 for all children in grades one to six.
Combining these data sets, we are able to separate the effects of overall school quality and home circumstances from the impact of language of instruction on the performance of children in a standardised English test written in grades four, five and six.
The motivation for this study was to gain insight into the situation at the schools where this policy matters most.
Therefore, our analysis was limited to the 9 000 or so primary schools that serve predominantly black children who come from the poorest households in South Africa.
What we found is quite striking. Among children in schools of a similar quality and coming from similar home backgrounds, those who were taught in their home language during the first three years of primary school performed better in the English test in grades four, five and six than children who were exposed to English as the language of instruction in grades one, two and three.
The size of the difference is not inconsequential: it is equivalent to about a third of a year of additional learning for children who were instructed in their home language during grades one, two and three compared with their peers who were instructed only in English during that same period.
This finding seems to be in line with the thinking of education specialists, who have for many years promoted the advantages of mother-tongue instruction in the early stages of children's education.
What we can and cannot say
As with any study, our findings have their limitations. This research tells us the average effect of language of instruction in South African schools as things are currently being implemented.
Advocates of both "straight-for-English" approaches and mother-tongue instruction envisage a carefully thought-through set of instructional practices implemented by high-quality teachers and supported by sufficient materials. However, we estimate the impact of the alternative models as they are being implemented currently, within specific contexts of schools, teachers and homes.
Therefore, we cannot make deductions about what the impact would be if all teachers were able to speak English fluently and household poverty and gaps in school quality were eliminated from the South African landscape. We are therefore not making any conclusions regarding the impact of English instruction in a utopian version of reality where all policies are perfectly implemented, but rather looking at the school system that we do have.
Although our study confirms that the language of instruction is an important contributor to the academic performance of children, it is not the main contributor.
Factors such as community- and home-level poverty, weak school functionality, weak instructional practices, inadequate teacher subject knowledge, and a need for greater accountability throughout the school system all represent much more severe constraints to achieving better education.
Our results are not directly informative for policy decisions regarding the extension of mother-tongue instruction beyond the grade three level.
This policy decision would require additional research. Similarly, our study has no direct application to the language policy debates at the university level.
It is important to recognise these limitations, because there are many policy issues and arguments that often get confused in South Africa's language debates.
Our study provides preliminary evidence on one key aspect of the matter: as things are currently being implemented, the choice to use mother-tongue instruction as opposed to English instruction in grades one, two and three generally leads to better English learning in the long run.
Finally, although our study points to the value of the mother tongue as the language of instruction, it may well also be important to strengthen the teaching of English as a subject in the early grades to help to facilitate the transition to English in grade four, as the curriculum and assessment policy statements recommend.
In conclusion, our research indicates that, as far as we can tell from empirical analysis, the current language policy is on the right track.
Given the practical realities of the South African education system, the policy to encourage the use of mother-tongue instruction in the foundation phase but still allow schools to make the final choice based on their specific circumstances seems to be beneficial.
We believe these research findings to be an important starting point from which to take the language debate forward. In order to make headway on this important issue, we need to move beyond a discourse fuelled by ideology and emotions towards one informed by rigorous research. After all, evidence-based policy needs exactly that – evidence.
Dr Stephen Taylor works in the department of basic education and Marisa Coetzee is a researcher in the economics department at the University of Stellenbosch. They write here in their personal capacities as academic researchers. Their full academic paper is available at mg.co.za/wp1810