Differentiation for mixed-ability classrooms

Three tiers, one lesson

The differentiation pattern that actually fits a 40-minute Nepali class period is the three-tier rubric: scaffolded, standard, stretch. Same concept, same lesson objective, three parallel versions of the material — not three different lessons, three different entry points to one lesson.

Scaffolded. Extra structure. Simpler language. Fewer steps per problem. Worked example given first. Vocabulary glossed in the margin. For a fractions worksheet, the scaffolded version shows the first item solved in full, breaks each subsequent item into named sub-steps (step 1: find common denominator), and uses smaller numbers so the arithmetic does not get in the way of the concept.

Standard. Textbook level. The version the curriculum assumes. This is the worksheet you would have made anyway. It is the centre of gravity; the other two are perturbations from it.

Stretch. An extension question or two beyond textbook depth — a problem that requires combining the concept with something from the previous chapter, a real-world application, or a why is this true? prompt that asks for reasoning rather than computation. Not harder arithmetic. Deeper thinking.

The same four core questions appear on all three versions. The student does not see the version of the worksheet they were not given; they see a worksheet on fractions with their name on it. The teacher sees three piles.

The prompt pattern

The prompt that produces all three versions in one pass is one of the highest-value templates a teacher can save. It is also one of the prompts where the source-only and named-segment discipline from Section 1 pays off most:

Using only the attached NEB Class 6 maths textbook page on adding and subtracting fractions, produce three parallel versions of the same worksheet, sharing the same 4 core questions but adjusting structure and language. Scaffolded version: worked example shown first, each question broken into named steps, vocabulary glossed in margin, smaller numbers where possible without changing the concept. Standard version: textbook level, no scaffolding, same 4 questions. Stretch version: same 4 questions plus 2 extension questions — one real-world application using a Nepali context (a Tihar sel preparation recipe, a Janakpur market vendor splitting stock), one reasoning question (why does the rule work, in 3-4 sentences). End each version with a different reflection question: scaffolded — which step was the hardest and why; standard — give one example from outside the textbook where you would use this; stretch — can you state the rule for adding fractions in your own words, and one case where it would not apply. Provide a single combined answer key, marked by version. Use only facts and methods from the attached source. Do not introduce techniques not in the page (e.g. cross-multiplication if the source uses LCM).

Several things in that prompt are doing structural work. The same four core questions across all three versions means the teacher can mark in one pass — the answers are the same; only the path to them differs. The different reflection question per tier does something subtler: it signals to each student that the work is theirs, not a watered-down or souped-up version of someone else’s. The Nepali context in the stretch version anchors the extension in a world the student knows, which is where extension questions tend to fail when imported from foreign textbooks. And the source-only instruction prevents the model from quietly importing cross-multiplication into a chapter that builds the concept through LCM.

The teacher’s judgement layer

The model produces the materials. The teacher decides who gets which version. This is the part no algorithm can do, and the part where Nepali teachers — who know their students better than any test score reveals — have an advantage that does not get talked about.

You know that the boy in row three solved the standard items quickly last week but stopped engaging when the stretch question was framed in English. You know that the girl who looks scaffolded on paper is actually advanced — she just lost her father last term and has been quiet. You know that the new transfer student from the Madhesh school district is reading two grades behind in Nepali but is the strongest mathematician in the room. The model knows none of this. The model will, if asked, propose a heuristic — give scaffolded to students who scored under 40%, standard to 40-75%, stretch above 75% — and that heuristic will be wrong for a third of the class, in ways that matter.

The right division of labour is: the model produces three versions in fifteen minutes; the teacher spends two minutes at the start of the period sorting forty-five papers into three piles, based on what they know about each child and the previous week’s work. The piles change week to week. A student in the scaffolded pile this week may be in the standard pile next week, because the topic changed and her relationship to it changed. Differentiation should be a moving target, not a permanent label.

Two Nepali scenarios

Class 6 maths, mixed community school. Forty-five students, half from families where the parents finished Class 5 themselves, half whose parents send them to evening tuition that already covered this chapter last week. A standard worksheet bores the second group and crushes the first. The teacher produces a scaffolded version with the first question worked through, smaller numerators, and a Nepali-medium glossary box (न्यून, अंश, हर); a standard version that matches the textbook; and a stretch version that asks the tuition kids to invent a real-world fractions problem from their own week and solve it. The tuition kids do not realise they have been given the harder paper; they think they got the interesting paper. The first-generation learners do not see “easy” on their paper; they see worked steps and a place to write. Both groups work for the full period.

Class 9 English, monolingual Maithili speakers next to bilingual students. A teacher in a Janakpur-region school has a class where roughly a quarter of the students speak Maithili at home and have weaker English, while another quarter come from families that switched to English at home and read English novels for fun. The standard textbook lesson on present perfect — I have eaten, she has gone — moves too fast for the first group and too slow for the second. The scaffolded version gives the conjugation table at the top, three Nepali-context example sentences, and fill-in-the-blank items with the verb provided. The standard version follows the textbook exercises directly. The stretch version asks students to write a short paragraph about a recent festival using five present-perfect verbs and one contrast with simple past. Same period, same concept, three rooms inside one room.

Avoiding the labelling trap

There is a failure mode of differentiation that is worth naming: the moment students realise they have been sorted by ability, the lesson stops being about the lesson and starts being about the label. A child who is told they are in the “easy group” can carry that for years. A child whose friends notice they have a different worksheet at lunch can decide they are stupid, in a way that is hard to undo.

The model produces the materials. The teacher protects the dignity. Four practical disciplines:

• Same header on all three versions. Class 6 Maths — Fractions Worksheet — [date]. Not Scaffolded / Standard / Stretch in the header. Mark the version in a small code in the bottom corner if you must — a single letter, not a word.
• Same core questions, same numbering. A student comparing papers with a friend sees the same four questions. The differences are in the structure around them.
• Vary the piles each week. A student who is always in the same pile knows it. Move the boundaries; let last week’s stretch student get a standard paper this week on a different topic.
• Never explain the system to the class. We are doing different worksheets so I can help everyone is for the staff room and the headteacher. In front of the students, the worksheet is simply the worksheet.

What this connects to

Tiered worksheets from Section 2 feed directly into the scaffolded / standard / stretch pattern here. A worksheet with three difficulty bands on one page is differentiation visible to the student; three parallel versions on three pieces of paper is differentiation invisible to the student. Both are useful in different periods, and both come from the same fact-first prompt discipline you have been building since Section 1.

The next chapter is on the back half of the teaching cycle — assessment, feedback, and grading. Differentiation in materials produces tiered work; tiered work needs tiered feedback. That is where Chapter 3 starts.

Check your understanding

Quick check

—

A school administrator argues: real differentiation in a 45-student Nepali classroom only became feasible with AI — without it, no individual teacher could sustain the practice. True or false?

1. ATrue — differentiation was a Western-school luxury that NEB classrooms could not previously afford.
2. BFalse — differentiation has always been good practice, and many Nepali teachers have done it for decades by hand, on their own time. AI reduces the time cost of doing it well, but it did not invent the practice and it is not what makes it pedagogically valuable.
3. CTrue — without AI, the cognitive load of designing three parallel versions exceeds what any human teacher can manage.
4. DFalse — differentiation is unnecessary because the NEB curriculum is uniform across the country.

Check answer→

What comes next

Chapter 2 has been about the front half of the teaching cycle — planning the lesson, building the materials, sorting the room. Chapter 3 turns to the back half: assessment, feedback, and grading. A tiered worksheet produces tiered work, and tiered work needs feedback that meets each student where they actually are — without inflating the teacher’s marking time back to where it was before. That is the next chapter, and it is where the time savings from this one either compound or quietly disappear.