The lecture's example lands perfectly: gym memberships and sunk cost, explained so clearly you can recite it. You do the practice problem, get it right, feel like you understand. Then the exam asks about sunk cost in a software license — same mechanism, different story — and the recognition evaporates. You knew the example; you didn't know the mechanism.

That's the recognition-transfer gap, and it's the most common way "I understood the lecture" fails the exam. Recognition is knowing the concept when you see it in the story you learned it in; transfer is producing it in a story you've never seen. The exam always uses a new story — that's what makes it a test instead of a recital.

After you can answer the lecture's own example, change the people, the numbers, or the domain. If performance collapses, you had recognition — the story was carrying the concept, and without the story there's nothing left. If the mechanism survives, you have transfer — the idea works wherever it applies.

Why drilling the original example fails

Practicing the lecture's own example again is the recognition trap in its purest form — same story, same characters, same numbers. You get faster at the example without getting better at the concept, because the example's surface is doing the work: the gym membership cues "sunk cost" automatically, no retrieval needed. The exam's different surface produces no such cue, and the student who only drilled the original has nothing to retrieve.

Karpicke & Blunt showed retrieval supports transfer — producing the concept, not just re-exposing to it, is what makes it portable. Recognition of the original story is a weak proxy: it tells you the example was processed, not that the mechanism was learned. The new-story test is the distinction — the concept that survives a domain change is the one that'll work on the exam.

The same-mechanism-new-story workflow

Start with the flagship example. The lecture's central demonstration — the one the professor spent the most time on. That's the mechanism the exam will test; the example's specific story is the costume it'll wear differently.

Change the surface, keep the mechanism. New characters, new numbers, new domain — but the same underlying move. Sunk cost stays sunk cost whether it's a gym membership or a software license; the mechanism (past payment doesn't justify future use) is what transfers, the story is what changes.

Two transfer distances. Near transfer: same domain, new surface — different gym scenario, different membership math. Far transfer: different domain, same mechanism — sunk cost in a project you should abandon, a relationship you should leave. The far version is the exam's version.

Name the fake-progress version. For each item, say what re-drilling would have missed: "I can recite the gym example but can't do it with a subscription I already paid for." The contrast makes the transfer visible — recognition of the original story isn't ownership of the mechanism.

Retest with a new story. Same mechanism, different domain — the 1/3/7-day spacing holds, and the new story prevents memorizing the variant instead of learning the transferable move.

A 40-minute session on the clock

The weak answer is the recognition reflex — "keep going because you paid" — which is the misconception sunk cost exists to correct, applied to the new story without the mechanism. The strong answer transfers the move: the payment is gone regardless, so the only question is whether the software still helps now. That's the mechanism surviving the domain change — the gym costume is gone, and the idea still works.

Why the story has to change

The original example's story is scaffolding — it makes the mechanism concrete so you can see it. But scaffolding that never comes down becomes the structure: the gym membership isn't just illustrating sunk cost, it's becoming sunk cost in your memory, and the concept can't leave the building. The exam's different story is the test of whether the mechanism is free-standing or story-dependent.

This is why "I got the example right" isn't evidence of understanding. The example is where you learned it; the transfer is where you prove you have it. Same mechanism, new story — and if the mechanism still works, the concept is yours.