Retrieval-induced forgetting - why studying one fact can bury another

The testing effect says retrieving a fact strengthens it - true, and one of the most reliable findings in memory research. Less well known is its uncomfortable sibling: retrieving one fact can actively weaken a related, competing fact you haven't retrieved at all. This is retrieval-induced forgetting, first documented by Anderson, Bjork, and Bjork in 1994, and it's a real reason "similar" material sometimes gets more confusing the more you study it, not less.

The mechanism

When you're asked to recall one specific item from a category (say, one specific muscle that flexes the elbow, when several do), your brain doesn't just retrieve that item - it actively suppresses the competing candidates so the right one can surface cleanly. That suppression is useful in the moment, but it has a cost: the items you suppressed become measurably harder to recall later, even though you never studied them incorrectly and never actively forgot them through simple decay. You forgot them because you successfully remembered something else in the same category.

Where this actually bites people

This shows up hardest in dense, categorical material: drug classes with overlapping mechanisms, muscles with overlapping actions, biochemical pathways with shared intermediates - exactly the kind of content the Ultimate MCAT Deck and similar large decks are full of. If you drill one card in a cluster repeatedly while never touching its close relatives, you can end up more confident about the drilled item and quietly worse at the others than when you started - which is the opposite of what confidence is supposed to track.

What actually protects against it

The fix isn't avoiding related material - it's making sure every item in a cluster gets its own independent retrieval practice instead of over-indexing on whichever one you happen to find easiest to recall. Interleaving instead of blocking helps here specifically because it forces you to retrieve competing items back-to-back rather than letting one dominate a study session while its relatives sit untouched. This is also a reason cloze and multiple-choice cards that include the confusable near-neighbors as distractors - rather than testing each fact in total isolation - tend to surface this exact failure mode before an exam does, instead of after. It's the same reasoning behind breaking a table or list down by what's actually being tested instead of drilling whichever row is easiest.

The related phenomenon worth knowing next

Retrieval-induced forgetting is about suppressing a correct-but-untested competitor. Its cousin, the hypercorrection effect, is about what happens when you're wrong with high confidence and then corrected - a different mechanism, but the same underlying idea that memory for related material doesn't move in the direction you'd naively expect.