Nine groups, one teacher, ten minutes

Splitting them matters because they need different fixes, and teachers often over-correct the management side when the design is the problem.

Design faults:

• The information gap had too little to exchange, so groups finished in two minutes - a task that can be completed in 120 seconds was never a 10-minute task.
• No outcome or product was required, so there was nothing to be accountable for at the end.
• No success criterion (e.g. find three differences, agree on a ranking, produce a written list) and no reason to talk to each other rather than to me.
• Nine groups of four in fixed rows is an organisational problem built into the plan; I should have used pairs or row-based turn-taking instead of nine simultaneous groups.

Management faults:

• Instructions were given while materials were already distributed, so half the class was looking at paper, not at me.
• I did not check understanding of the task before release, and did not model the interaction with two students.
• No agreed norm about noise level or L1 use was established beforehand.
• No monitoring route: I drifted toward the nearest groups rather than sweeping all nine.
• The feedback stage is a stage, not a luxury - dropping it means the task had no payoff, which is probably why students concluded it was "not real English".

Not a fault: the use of L1 for task management and clarification is normal; the issue is whether the target language was needed to complete the task.

Setup (2 min), materials face-down:

• "You will work in pairs with the student next to you. Do not turn the paper over yet."
• Demonstrate with one student at the front: I ask, she answers, she asks back. Two exchanges only.
• Check instruction: "Marco, what do you ask first?" - one learner restates, not "do you understand?".
• Set the outcome before release: "At the end, three pairs will report one thing they found out. You have four minutes."

Run (4 min):

• Pairs, not groups - nine groups becomes eighteen pairs, and pairs mean everyone speaks.
• Turn the paper over only now. The task requires unequal information (A has the times, B has the prices) so they must talk.
• I monitor on a fixed route (left column, middle, right column) with a timer visible, noting two good utterances and two errors for later - not correcting mid-task.
• Phones face-down on the desk as a class norm, established before, not enforced as a punishment now.

Close (4 min):

• Two-minute timer for the report stage; three nominated pairs report.
• I write up the two good utterances and one recurring error, and we fix the error together.
• One sentence on what we will reuse next lesson.

The point of the close is that students see the activity had a product. Without it, group work looks like time off.

Off-task groups: proximity and a question, not a telling-off. I walk over and ask a task question ("Which two times do you still need?"), which re-engages them without a public confrontation and preserves the relationship. If a phone is the issue, I apply a pre-agreed routine - "phones face down while we work" - to the whole class, not to the two students, so it is a norm rather than an accusation. Escalation in a class of 36 costs me the lesson.

The "just do the book" request: I take it as information, not as defiance. It usually means the students cannot see what the group work is for, and that it feels less safe than choral work. My response:

• Acknowledge it: the book work is familiar and they know how to succeed at it.
• Make the purpose explicit: the exam and real life require them to speak to each other without me; pair work is the only place that happens, since 36 students cannot each speak to me.
• Show the evidence: I tell them what I heard in monitoring that I could not have heard in a whole-class drill.
• Keep a balance: I continue whole-class work and book work, and I say so, so pair work is not framed as a replacement for everything they trust.

And I check my own assumption. If a well-designed pair task still fails repeatedly, I look at whether the task genuinely needs interaction, whether pairing is stable enough to build confidence, and whether the room layout is defeating me. In fixed rows, pairs within a row work; four-way groups do not.