Every teacher faces the same dilemma when setting up pair work: should you pair strong students with weak students, or match similar levels together?
This is one of the most debated questions in ESL classroom management, and the honest answer is that neither approach is universally better. The research is clear on one thing: the pairing strategy you use should match the activity you're running. Using the same pairing approach for every activity is the mistake.
YapYapGo is a classroom speaking practice tool that offers four pairing modes - random, stretch, matched, and mixed - because different activities genuinely need different approaches. Here's how to decide which to use.What "stretch pairing" actually means
Stretch pairing means deliberately pairing a stronger speaker with a weaker one. The idea is that the stronger student provides a model, scaffolds the interaction, and pulls the weaker student's language to a higher level.
The concept comes from Vygotsky's Zone of Proximal Development - the idea that learners can achieve more with a more capable partner than they can alone. In theory, stretch pairing puts every weaker student in their ZPD.
In practice, it depends entirely on the interaction dynamics.
When stretch pairing works
Research on mixed-level pairs shows clear benefits when the interaction is collaborative - both students are genuinely exchanging information, negotiating meaning, and contributing equally to a shared task.
Information gap activities are the classic example. Student A has a description that Student B needs to draw. Student B has information that Student A needs to ask questions to discover. Both students must contribute for the task to work. In this format, stretch pairing tends to produce more complex language from the weaker student because they're hearing natural, level-appropriate models and being pushed to respond.
Discussion activities with open-ended questions also work well in stretch pairs, provided the question is genuinely open. "What would you do if you had a million pounds?" can be answered at A2 ("I would travel") or C1 ("I'd probably invest most of it and use the remainder to..."). Both students engage at their own level without either being excluded.The key condition: the activity must require both students to actively contribute. If the task has a "right answer" that the stronger student knows and the weaker one doesn't, you'll get a one-sided interaction that benefits neither.
When stretch pairing fails
The most common failure mode is dominance. When the stronger student controls the conversation - asking all the questions, giving all the long answers, correcting the weaker student's errors - the interaction becomes a mini-lecture. The weaker student gets less speaking time than they'd get in a matched pair, and the stronger student gets no real challenge.
Research by Yule and Macdonald found that when a more proficient student was placed in the information-holder role (they had information the other student needed), the weaker student asked more questions and produced more complex language. When the roles were reversed, the interaction collapsed into the dominant student providing everything.
This is the key practical insight: stretch pairing only works if the activity structure forces both students to contribute equally. Without that structural protection, the stronger student will default to doing most of the talking.
Tool tip: YapYapGo's stretch mode pairs students by proficiency level - higher with lower. You can set proficiency levels per student in class settings. The classroom group maker and random team maker also let you manually create mixed groups if you prefer to manage this yourself.
When matched pairing works
Matched pairing - similar proficiency levels together - works best when the activity is about fluency development under pressure rather than level-stretching.
Timed activities: If you're running the 4/3/2 fluency technique or any time-pressured speaking task, matched pairs work better because both students are working at the same pace. A B2 student paired with an A2 student in a two-minute talk will either dominate or wait. A B2 student paired with another B2 student will push each other. Exam practice: IELTS and Cambridge speaking practice should almost always use matched pairs. The exam is scored against band-level criteria - A2 students practising Part 3 abstract discussion with a C1 partner are not being assessed under exam-like conditions. They're either being carried or being overwhelmed. Debate activities: When both students have a position to argue and approximately equal language resources, the debate becomes a genuine contest. Significant level mismatches in debates produce frustrating outcomes for both students.The case for random pairing as the default
Both stretch and matched pairing require you to know your students' levels accurately and manage the logistics carefully. The research is also clear that neither approach is consistently better across all activities - what matters most is variation over time.
Random pairing solves the logistics problem and provides variation automatically. Over a term of random pairing, every student will end up with a range of partners at different levels. Some rounds will feel like stretch pairs. Some will feel like matched pairs. The average outcome tends to be better than any fixed strategy applied consistently.
For timed pair activities, a classroom timer keeps all pairs working at the same pace. A six-year study by Peter Liljedahl on visibly random groups found that after three weeks of consistent random pairing, 100% of students arrived ready to work with whoever they were paired with. That readiness - the willingness to engage fully regardless of partner - may be the most important outcome of all.
A practical framework
Use this as a starting guide:
- Default: Random pairing. Variation over time produces the best average outcomes.
- Information gap activities: Stretch pairing. Task structure protects equal contribution.
- Timed fluency practice: Matched pairing. Pace matching matters.
- Exam practice (IELTS, Cambridge): Matched pairing. Level-appropriate conditions matter.
- Open discussion: Random or stretch. Both work if the question is genuinely open.
- Debate: Matched pairing. Equal resources make for genuine contest.
Sources:
- Storch, N. & Aldosari, A. (2013). Pairing Learners in Pair Work Activity. Language Teaching Research. - Collaborative dynamics matter more than proficiency matching.
- Yule, G. & Macdonald, D. (1990). Resolving Referential Conflicts in L2 Interaction. Language Learning. - Role assignment in stretch pairs determines interaction quality.
- Vygotsky, L.S. (1978). Mind in Society. Harvard University Press. - Zone of Proximal Development: the theoretical basis for stretch pairing.
- Liljedahl, P. (2014). The Affordances of Using Visibly Random Groups. In Transforming Mathematics Instruction, Springer. - Random grouping produces better long-term outcomes than fixed strategies.
- Kieffer, M. et al. (2025). Heterogeneous vs Homogeneous Grouping. American Educational Research Journal. - No single grouping is universally best; variation matters most.