Almost everything on this site tells you to practise retrieval. The testing effect is one of the sturdiest findings in learning research, and active recall is the single best use of study time available. This page is the asterisk.
In 1994 Michael Anderson, working with Robert and Elizabeth Bjork, showed that retrieving one memory can actively suppress related memories you did not retrieve. Not merely fail to strengthen them — impair them, relative to items from categories you never practised at all. They called it retrieval-induced forgetting.
The experiment
The design is neat enough to hold in your head.
- Study a list of category-item pairs: fruit–orange, fruit–banana, drink–whisky, drink–gin, and so on.
- Practise retrieving only some items from some categories: fruit–or___, repeatedly. You never practise banana, and you never touch the drinks at all.
- Later, take a full test on everything.
Three outcomes, and the third is the interesting one:
- Practised items (orange) — much better recalled. That is the testing effect doing its job.
- Unpractised items from unpractised categories (whisky, gin) — the baseline.
- Unpractised items from practised categories (banana) — worse than baseline.
Banana was studied exactly as much as whisky. The only thing that happened to it is that a competitor from its own category got practised. Anderson's interpretation is that retrieval works partly by inhibiting competitors: to get orange out cleanly, your mind has to suppress the other fruit crowding the cue, and that suppression lingers.
How much should you care?
Less than the name suggests, for three reasons.
It is small next to the benefit. The practised items gain far more than the neighbours lose. A meta-analysis of the paradigm found the effect is real and replicable, but modest — nobody has ever shown that self-testing produces a net loss. If you skip retrieval practice to protect your unpractised material, you trade a large gain for a small one.
It fades. The suppression is a temporary state of the retrieval system, not damage to the memory. Later restudy or retrieval of the suppressed item recovers it.
The mechanism is contested. The inhibitory account is the leading explanation but not the only one; interference-based accounts explain much of the same data, and the debate is live. What is not in dispute is the empirical pattern.
Where it genuinely bites
The effect matters most where recall of one item and recall of its neighbours compete for the same cue, and where the neighbours are the ones you will actually need.
Studying only your flagged cards. If you drill the twelve facts you marked important and never touch the other thirty in the same topic, you are not just neglecting them — you are actively making them harder to reach. This is the strongest practical argument for reviewing a whole set rather than a favourite subset, and for the queueing logic in the Leitner system and spaced repetition, which force the neglected material back into rotation.
Rehearsing one version of an event. Tell the same story about a meeting four times and the details you left out get harder to retrieve, not just less rehearsed. This is well documented in eyewitness research and is worth knowing about your own memory of events. It is also a good argument for taking notes at the time rather than trusting the retelling.
Exam revision by topic. Hammering one sub-topic in a subject can transiently depress the rest of it. Mixing them — interleaving — is already the better strategy for other reasons, and this is one more.
The practical rules
- Test yourself on the whole set, not the highlights. Rotate through everything, even the parts you think you know. Coverage is the defence.
- If you self-test on part of a topic, revisit the rest soon after. The suppression is temporary; a quick pass over the unpractised siblings undoes it.
- Do not let this become a reason to stop testing. It is a footnote to retrieval practice, not a rebuttal. The direction of the overall evidence has not moved.
- Write things down. The most reliable protection against a quirk of biological retrieval is not needing to rely on it. That is the whole argument for external memory, and one of the honest reasons a note beats a rehearsed anecdote.
Related: how memory works for the encoding-storage-retrieval split, recall vs recognition for why the easy kind of remembering misleads you, and the wider memory and focus section.