Testing effect
Retrieving something from memory strengthens it more than reading it again.
Being asked to produce an answer, and succeeding, does more for later retention than restudying the same material for the same time. The act of retrieval is the mechanism, so the benefit survives even when no feedback is given, and grows with feedback. Learners consistently predict the opposite, rating restudy as more effective.
How it shows up in software
Learning products use quizzes, flashcards and recall prompts directly. Elsewhere the same mechanism appears whenever a product asks someone to produce rather than recognise: typing a command instead of picking it from a palette, or confirming a destructive action by naming the resource. The retrieval is doing memory work either way.
Using it well
- Replace a review screen with a recall prompt. Ask what the setting does before restating it.
- Give feedback right after the attempt, including the correct answer, since feedback enlarges the effect and prevents an error being cemented.
- Keep the difficulty where success is likely but not automatic. Retrieval that always fails teaches nothing.
- Design for the retrieval users will actually need later. Testing recognition when the real task needs recall trains the wrong thing.
Where it turns manipulative
- Gating access behind a quiz that exists to lengthen a session rather than to check understanding.
- Testing content the person has no way to have learned yet, which produces failure and drop-off instead of retention.
- Dressing up a lead capture form as a knowledge check, so the retrieval framing is a pretext for collecting answers.
Where you have seen it
Anki
Every card shows the prompt alone and requires an attempt before the answer is revealed.
GitHub
Deleting a repository requires typing its full name, which forces retrieval of what is about to be destroyed.
Duolingo
Translation exercises require producing the sentence rather than selecting the right one from a list.
What the research says
- Roediger and Karpicke, 2006Well evidenced
Students who studied a passage once then took recall tests outperformed students who restudied repeatedly, when tested a week later. The restudy group predicted they would do better.
- Rowland, 2014Well evidenced
A meta-analysis of 159 comparisons found a reliable retrieval practice advantage over restudy, larger when feedback followed the test.
Grades are a judgement about the evidence, not about how useful the idea is. Plenty of contested effects are still worth knowing, as long as you do not cite them as settled.