On this page
A Class 5 classroom in India often holds children who cannot yet read letters sitting next to children who read fluently. One teacher, one textbook. Teaching at the Right Level, an approach developed by the NGO Pratham, tries something simple: for part of the day, group children by what they can actually do and teach that. This post looks at the research paper that traced how the idea was tested, failed, was changed and finally worked at scale.
What question did the research ask?
The paper is From Proof of Concept to Scalable Policies: Challenges and Solutions, with an Application by Abhijit Banerjee, Rukmini Banerji, James Berry, Esther Duflo, Harini Kannan, Shobhini Mukerji, Marc Shotland and Michael Walton, published in the Journal of Economic Perspectives in 2017. Its question was practical: a method can work in a small trial run by a committed NGO, but can it work when ordinary government schools run it?
The figures below are from the NBER working-paper version, revised in October 2017.
What is Teaching at the Right Level?
The basic idea, in the authors' words, is to group children "not according to their age, but according to what they know", and match teaching to that level. A short one-on-one test sorts children into groups such as "letters", "words", "paragraph" and "story" for reading. Each group then works with activities and materials pitched at its level.
What did the trials find?
The paper describes a chain of randomised trials over several years.
The proof of concept (Mumbai and Vadodara). In the Balsakhi programme, children who were behind were taught basic skills outside class for two hours a day by young women from the community. Their test scores rose by 0.28 standard deviations. We cover this trial in detail in our post on the Balsakhi study.
Bihar and Uttarakhand (2008-10). Giving schools Pratham's materials alone, or materials plus teacher training, had no effect in either state. Volunteers placed inside schools in Uttarakhand had no detectable impact either; they were absorbed into regular teaching or did not show up. But a one-month summer camp run by government teachers in Bihar improved reading by about 0.5 levels for children who attended.
Haryana (2012-13). Teachers were given an extra hour a day in which all children in grades 3 to 5 were regrouped by level, with monitoring by trained government staff. The trial covered 400 schools, 200 of which got the programme. Hindi scores rose by 0.2 levels. Over 90% of schools actually grouped children by level, unlike in the earlier trials.
Uttar Pradesh (2013-14). Short in-school "learning camps" of 10 or 20 days at a time, about 50 days a year, were run by trained volunteers.
The Uttar Pradesh results are the most striking. At baseline, 39% of children could not recognise letters and only 15% could read a paragraph or story. At endline, in control schools, 24% still could not recognise letters and 24% could read a paragraph. In camp schools, only 8% could not recognise letters and 49% could read a paragraph.
By the time of the paper, the teacher-led "Haryana" model had been implemented in 107,921 schools across 13 states, reaching almost 5 million children.
What does the study not show?
Most outcomes are basic reading and arithmetic. The trials measured early-grade skills such as reading a paragraph, not Class 9 science or board exam marks.
Gains were measured over one school year or less. The paper does not track these children into secondary school.
Context mattered. The authors note differences in state leadership and staff turnover between Bihar and Uttarakhand. A model that works in one state needs testing in another.
Some results are in "levels". A gain of 0.2 levels is not the same as 0.2 standard deviations, so do not compare those numbers directly with other studies.
What does this mean for Indian students, parents and teachers?
The lesson travels beyond primary school. Students who struggle in Class 11 physics are often missing something from earlier: fractions, rearranging an equation, reading a graph. The trials suggest that teaching at the right level for a short, focused period can close gaps faster than pushing on with the syllabus.
For teachers, the practical recipe is to test quickly, group by level for part of the time, use materials designed for each level, and check that the regrouping really happens. For parents, the message is to find out what your child can actually do, not just which class they are in. The same principle sits behind the Delhi Mindspark study, where software did the levelling.
How to read a study like this
Notice that this paper reports the trials that did not work as well as the ones that did. That makes it more trustworthy, not less. When a programme is described only by its best result, ask what other versions were tried and what happened to them.
If you do not know where your own gaps are, Fermi's AI tutor works out what you already know and builds a plan for the session, so you can start at your level rather than at the chapter you are supposed to be on.
This week, take one chapter you find hard and write down the earliest step in it you cannot do without help.
