Teach-back works only when correction follows

When a learner explains from memory, hidden relationships become audible. Sequence, cause, contrast and qualification can be examined. The performance is valuable because it exposes a mental model; it does not automatically certify that the model is accurate. Confidence, speed and polish can coexist with a missing condition.

Give the listener a concrete job: mark claims that need a source, transitions that are unclear, and places where a term changes meaning. “That was good” produces little information. A small observation tied to a sentence gives the learner something they can repair.

An open book can support a beautiful paraphrase while hiding what the learner can reconstruct. Begin with the source closed. Ask for a short explanation aimed at a specific listener: a child, a peer, a sceptic, or tomorrow’s self. A defined audience forces choices about what matters and which terms need unpacking.

Keep the first attempt brief. Ninety seconds is enough to expose structure without turning the task into a speech competition. The aim is desirable difficulty: enough effort to retrieve, not so much demand that performance anxiety replaces thought.

Now reopen the source. The learner, not only the teacher, compares the explanation with the evidence. Mark one accurate relationship, one omission and one overstatement. Correction is not a postscript to retrieval; it is the moment that prevents the retrieved error from becoming more familiar.

Ask for a second explanation that changes only what the comparison justified. This keeps revision connected to evidence. If the learner rewrites everything, the useful contrast between attempts disappears. A visible before-and-after record makes progress inspectable and gives the next retrieval a clear target.

Immediate success can reflect the freshness of the correction. Schedule another explanation after a delay and vary the prompt. Ask for a diagram, an example, a contrast with a nearby idea, or an answer to a deep “why” question. Variation tests whether the relationship travels beyond the original wording.

Keep the cycle humane. The record is learner-controlled evidence, not a grade or public ranking. Its purpose is to guide the next move: revisit the source, practise a weak relationship, discuss a genuine question, or teach again. Learning becomes durable through repeated reconstruction and repair, not through an impressive single performance.