Two Languages In One Classroom

AWhen a child arrives at school speaking a language the school does not teach in, somebody must decide how long that language is allowed to stay in the room. Two answers have organised policy since the 1970s. The transitional model treats the home language as a bridge: it is used to keep mathematics and science comprehensible for two or three years while the majority language is built up, and is then dismantled once that language is secure. The dual-language model treats it as a destination: instruction is divided, often ninety to ten at first and fifty-fifty later, between the two languages for the whole of primary school, and classes are deliberately mixed so that speakers of each language are learners of the other. The models are not merely technical variants of one another. Each rests on a claim about what a school year is for, and the evidence assembled for and against them has proved unusually difficult to read.

BThe most cited evidence for the dual-language model comes from Ingrid Halvorsen of the Ravenshall Centre for Educational Policy, who followed 3,240 pupils in eleven districts from their entry in 1998 until they left primary school. Her finding was a crossing of lines. At the end of Year 3, children in dual-language classrooms were on average four months behind their transitional peers on the majority-language reading test; by the end of Year 6 they were seven months ahead, and the gap widened further where the local secondary school offered any subject in the minority language. Halvorsen's explanation borrows from an older hypothesis: that literacy is learned once rather than twice, so that a child who has learned to interrogate a written text in one language transfers the habit rather than building it again. The early deficit, on this reading, is not a cost of the model but an artefact of measuring it too soon.

CPeter Ashgrove, an economist of education at Dunmore, does not dispute Halvorsen's figures; he disputes what produced them. Dual-language places are almost never assigned. Parents apply for them, and the parents who apply are, in his phrase, unusual in ways that a questionnaire does not record. Ashgrove therefore reanalysed the twenty-four districts in which places had been rationed by lottery, so that families who had applied and lost could serve as a comparison. The Year 6 advantage survived, but shrank from 0.31 to 0.11 of a standard deviation — real, he concedes, though no longer large enough to justify the reorganisation of a school system on its own. His second objection concerns supply. Across those districts the benefit appeared only where the stream was staffed by teachers qualified in both languages; where schools had filled the posts with assistants translating on the spot, dual-language pupils did no better than transitional ones, and in three districts slightly worse.

DUnderneath the dispute lie two accounts of school time. The first, time-on-task, is arithmetical: a language is learned in proportion to the hours spent inside it, so hours given to one are hours taken from the other. The second, interdependence, holds that the two languages draw on a single underlying store of academic skill, so that time in either can feed both. Halvorsen and Ashgrove agree, unusually, that the evidence favours interdependence for reading comprehension and time-on-task for spelling and pronunciation, skills which are surface-level and do not transfer. Where they separate is over the balance point. Ashgrove reads the same data as showing that the transfer is genuine but slow, and that a system judged on examinations taken at eleven will be punished for waiting; Halvorsen answers that this is an argument about when children are examined, not about how they learn.

EWhatever the mechanism, the constraint that decides most cases is neither theory nor test scores but staffing. A survey in 2016 of 380 schools that had opened dual-language streams found that 38 per cent had closed them again within six years; only one closure in nine was attributed to disappointing results, while more than half were attributed to the departure of a single teacher whom the school could not replace. Transitional programmes are cheaper to run for the reason that makes them less ambitious: a teacher needs the home language in order to explain, not in order to teach a curriculum in it. Ashgrove takes this as decisive for policy, arguing that a model which works only under conditions most districts cannot meet is not a model but a hope. Halvorsen answers that the scarcity of such teachers is itself a policy choice, and that the two districts which began funding a two-year bilingual qualification saw applications for these posts roughly triple within four years.

FThe last difficulty is the one least often stated. Almost every study in the field, Halvorsen's included, ranks the models by results on tests written in the majority language, because those are the tests that exist and the ones on which a child's later prospects turn. A model designed to produce two literacies is thus assessed on one of them, and the second is recorded, if at all, by asking families whether the language is still spoken at home. When Ravenshall added such a question in 2019, the replies turned out to be a poor guide to anything measurable, and the researchers dropped them from the following report. Both camps accept that this leaves the central promise of dual-language education, that a child keeps one language while gaining another, largely untested. What each draws from the admission differs: for Ashgrove it is a reason to fund only what can be shown to work, and for Halvorsen a reason to build the tests that are missing.