Retrieve before you reread: turn familiarity into an observable attempt
A practical study loop that asks for an answer first, preserves the miss, and uses feedback to choose the next practice step.
Rereading can make a fact feel familiar while leaving the harder question unanswered: can you produce it when the cue arrives? Retrieval practice makes that state observable. The point is not to turn every study session into a high-stakes exam. It is to attempt a response before reopening the source, then use the outcome to choose a better next encounter.
Write a cue that requires production
Start from a concept, relationship, person, place, work, event, or category boundary—not a copied archive clue. Write a short cue that has a checkable target and enough context to avoid accidental ambiguity. Hide the target. Say or type the response before exposing notes. A recognition-only multiple-choice prompt can have a place, but it does not measure the same production demand.
- Record the cue and target source
- Attempt before consulting the note
- Capture the response you actually produced
- Mark correct, incorrect, partial, or ambiguous with a reason
- Read concise feedback and repair the representation
- Schedule another encounter based on the result and study horizon
Keep the miss useful
A miss is not a character judgment. It can reveal that the cue was weak, the target was ambiguous, a near neighbor interfered, a relationship was missing, or the fact has not been retrieved enough. Preserve a compact error tag before correcting the answer. If the evaluator cannot distinguish a knowledge miss from a bad item, route the item to review instead of training on an uncertain key.
Do not let an answer checker turn a disputed spelling, alias, or wording into silent truth. Keep the submitted response, accepted variants, source, and evaluator decision together.
Use feedback to change the next attempt
Feedback should be short enough to inspect and rich enough to repair the error. Show the target, one discriminating relationship, the source, and the reason a variant was accepted or rejected. Then ask for retrieval again later rather than rewarding immediate rereading fluency. Research findings about retrieval do not supply one universal session recipe, so compare the learner's delayed results against the chosen method.
Measure at the delayed horizon
An immediate correct answer after feedback mainly proves that the answer is accessible now. The learning question is whether it returns after a delay and under a useful cue. Track first-attempt accuracy, delayed accuracy, repeated error type, and time-to-response where appropriate. Keep the raw practice private; publish only aggregate product behavior after privacy and evaluation review.
Open the local Jeopardy Intelligence study surface; its private history must never enter a feeder or SubjectGraph export.