You've planned a speaking activity. You tell students to find a partner. Within seconds, the same thing happens: best friends pair up, the two quiet students end up together by default, and there's one person left over looking around the room.
You know this isn't ideal. But assigning pairs yourself takes time, feels arbitrary, and sometimes creates its own drama. So you let it go. Again.
What if there was a better way — faster, fairer, and backed by years of classroom research?
The problem with "find a partner"
When students choose their own pairs, a few things happen that quietly undermine the activity:
Friends coast. They're comfortable, which sounds good — but comfort often means switching to their first language when things get tricky, or settling for short answers because their partner won't push back. The same dynamics repeat. Dominant students pair with passive ones. Popular students cluster. Less confident students get whoever's left. The social hierarchy from break time walks straight into your classroom. Some students never mix. After a whole term of self-selected pairs, it's common to find students who have never spoken to half the class. That's a missed opportunity — every new partner is a fresh conversation with different vocabulary, different ideas, and different communication challenges.What happens when you randomise
A researcher called Peter Liljedahl spent six years studying what happens when you make groups visibly random — students see the shuffle happen, so they know it's not the teacher playing favourites.
The results were dramatic. When groups were self-selected, about 80% of students entered the activity planning to let someone else do the thinking. After just three to four weeks of consistent random grouping, that number flipped completely. Every student came in ready to contribute.
Why? Because randomisation removes the social baggage. There's nothing to negotiate, no one to blame, and no status games. It's just "here's your partner, let's go." Students reported feeling more willing to work with anyone, and teachers noticed that knowledge moved more freely around the room.
The problem with ability grouping too
Some teachers assign pairs by ability — putting strong speakers together and weaker ones together, or deliberately mixing levels. Both approaches have merit in theory, but both carry hidden costs.
Putting weaker students together can reinforce the idea that they're "the low group." Large-scale studies have found that children placed in lower groups show increased emotional and behavioural problems over time. The label sticks, even when it's informal. Mixing strong with weak can work brilliantly — but only if the stronger student isn't dominating the conversation. Research shows that when the more proficient partner controls the interaction, both students get less out of it. The weaker student stays passive and the stronger one doesn't get challenged.The real finding from the research? No single pairing is always best. What matters is variation. Sometimes mixed ability is ideal. Sometimes similar levels work better. The power is in rotating — not in finding the one perfect configuration and sticking with it.
Why random is the right default
Random grouping solves the biggest practical problems all at once:
It's fast. No agonising over who goes where. A shuffle takes five seconds. It's fair. No student can complain about favouritism — it's visibly random. Nothing to argue about. It builds community. After a few weeks, every student has worked with every other student. Cliques soften. The room feels more connected. It removes your bias. Teachers are human. We have unconscious preferences about which students "work well together." Randomisation takes that variable out of the equation. It stays fresh. The same pair doing the same activity every week gets stale. New partners bring new energy, new accents, and new ideas.When to add constraints
Random doesn't mean unthinking. There are legitimate reasons to add rules on top of the randomisation:
- Two students who genuinely can't work together? Set a conflict avoidance rule so the shuffle respects it.
- Want to mix proficiency levels deliberately for a particular activity? Use a stretch-pairing mode.
- Want similar levels for exam practice? Use a matched mode.
The key is that these are constraints on the randomisation, not fixed assignments. The shuffle still happens — it just respects a few boundaries.
Tools like YapYapGo offer exactly this: random, stretch, matched, and mixed pairing modes with optional conflict avoidance — all handled by the shuffle so you don't have to manage it manually. Worth trying if group formation eats into your lesson time.
Give it three weeks
The one thing the research is very clear about: the benefits of random grouping aren't instant, but they come fast. Students resist for about three weeks. After that, working with anyone becomes normal. The social dynamics of your classroom shift — and they stay shifted.
Three weeks of mild awkwardness for a year of better engagement. That's a trade worth making.
Interested in the research? Here are the key sources:
- Liljedahl, P. (2014). The Affordances of Using Visibly Random Groups. In Transforming Mathematics Instruction, Springer. — The six-year VRG study. Also expanded in Building Thinking Classrooms (Corwin Press, 2021).
- Long, M. (1996). The Role of the Linguistic Environment in Second Language Acquisition. In Handbook of Second Language Acquisition, Academic Press. — The Interaction Hypothesis: new partners create fresh acquisition opportunities.
- Storch, N. & Aldosari, A. (2013). Pairing Learners in Pair Work Activity. Language Teaching Research. — Collaborative dynamics matter more than proficiency pairing.
- Kieffer, M., Proctor, C., Weaver, J. et al. (2025). Heterogeneous vs Homogeneous Grouping. American Educational Research Journal. — No single grouping is best; it depends on the student.
- Papachristou, E. et al. (2022). Ability-Grouping and Problem Behaviour Trajectories. Child Development. — UK cohort study (7,000+ children) linking lower-group placement to increased problems.
- Yule, G. & Macdonald, D. (1990). Resolving Referential Conflicts in L2 Interaction. Language Learning. — Stretch pairing only works if the stronger student isn't in the dominant role.