Reading gains from first-language instruction in multilingual primary classrooms: a two-year study in 46 Karnataka schools

The objection to first-language instruction in Indian primary schooling is rarely made on learning grounds. It is made on the grounds that time spent in the home language is time taken from English, and that English is what parents are sending their children to school to acquire.

This is an empirical claim with an established answer in the international literature — literacy skills transfer across languages, so early first-language reading instruction typically improves later second-language reading — and almost no Indian evidence. This study supplies some.

Forty-six government primary schools across three Karnataka districts were assigned to a structured bridging programme or to business-as-usual instruction. Assignment was at school level to avoid contamination within staffrooms. The bridging programme allocated 40 minutes daily to structured reading instruction in the child's stronger language for Grades 1 and 2, with a planned transition to Kannada-medium literacy work in Grade 3.

Outcomes were oral reading fluency in Kannada and a decoding assessment in English, administered individually at baseline, twelve months and twenty-four months by assessors blind to condition. The analysis sample is 2,310 children; attrition was 11 per cent and balanced across arms.

At twenty-four months, children in bridging classrooms read Kannada connected text at 0.41 SD above control (95% CI: 0.28 to 0.54). English decoding was 0.23 SD above control (95% CI: 0.11 to 0.35), which is the result that matters for the policy argument: the bridging children had received less English instructional time and decoded English better.

Effects were heterogeneous in the direction theory predicts. Children whose home language was neither Kannada nor English — largely Urdu, Tulu and Lambadi speakers in our sample — gained 0.58 SD in Kannada, against 0.29 SD for children who already spoke Kannada at home.

The programme required no additional teaching posts and no reduction in the school day. Its real cost was materials in four languages and eleven days of teacher preparation, which came to ₹1,840 per child per year at the scale we ran it, and would fall substantially at district scale.

We are explicit about one limitation: the transition to Kannada-medium literacy in Grade 3 is the point at which bridging programmes elsewhere have most often failed, and our data stop before we can observe it. A four-year follow-up is underway and we will report it whichever way it comes out.