MCQ banks are the default practice format because they're available and scoreable — but they train only half the skill. Choosing the right option among four is discrimination: telling apart confusable answers. Producing the answer yourself is generation: retrieving it without cues. An MCQ-only student can recognize the right answer when it's listed; they can't always produce it when it's not — and the final exam often asks for both.

The fix is to pair the formats deliberately. For each concept, write a short-answer stem first — generate the distinction, the mechanism, the application. Then an MCQ with a trap — the foil that swaps the confusable, the plausible wrong answer. Finally, one sentence on why the foil was tempting — the discrimination made explicit. Generation then discrimination, in that order — because the short answer builds the knowledge the MCQ then tests.

Why MCQ-only practice fails

Drilling only multiple-choice banks trains a specific skill: picking the right answer from a list. That's discrimination, and it's real — but it's not generation, and exams that require producing an answer expose the gap. A 90% on MCQ practice means you can recognize the right option; it doesn't mean you could have written it — and the gap shows up when the exam asks for a sentence instead of a selection.

The pair covers both skills. The short answer is generation — the testing effect's core: produce the distinction without cues, and the retrieval strengthens it. The MCQ is discrimination — a desirable difficulty: choose among lookalikes, and the choice sharpens the boundaries. Generation is the testing effect; choosing among lookalikes is a desirable difficulty — and one format leaves a hole the other fills.

The generate-then-discriminate workflow

Short answer first, always. The order is the mechanism: generate the answer before the options can cue it. Writing the distinction, the mechanism, the application — that's the retrieval, the hard rep that builds the knowledge. Do the MCQ first and the options do the thinking; the generation never happens.

Then the MCQ with a real trap. Not "pick the definition" — a foil that swaps the confusable: Type I for Type II, mitosis for meiosis, the near-miss that's almost right. The trap is the discrimination test — the detail the short answer let you gloss over, now forced into a choice.

One sentence on why the foil was tempting. "It swaps the errors" or "it uses the wrong sign convention" — the discrimination made explicit. The tempting foil is the learning: it marks the boundary you almost crossed, and naming it sharpens where the concepts actually split.

Both misses are re-tests. A short answer you couldn't write is a generation gap; an MCQ trap you fell for is a discrimination gap. They're different failures — the first is missing knowledge, the second is fuzzy boundaries — and both get scheduled for the next session.

The MCQ checks, doesn't teach. With the short answer done, the MCQ is a verification: can you spot the right answer among traps, given that you already produced it? It confirms and sharpens; it doesn't substitute for the generation that came first.

A 40-minute session on the clock

The weak pair is definition-recall — restating the labels, then an MCQ that asks for the same label back. That's recognition twice, not generation plus discrimination. The strong pair generates the distinction in terms of what the errors do (rejecting true, missing false), then the MCQ trap swaps which error is the false alarm — the confusable detail the definition glossed over. The pair trains both the production and the precision.

Why the order matters

Short answer then MCQ is the sequence that makes each format do its job. The short answer builds the knowledge — you produce the distinction, which is the retrieval that consolidates it. The MCQ then tests the discrimination — with the knowledge built, the trap measures whether you can apply it precisely. Reverse the order and the MCQ substitutes for the generation: the options provide the answer, you pick it, and the building never happened.

This is why MCQ banks alone undertrain — they skip the generation step that makes the discrimination meaningful. A student who only drills MCQs can pick the right answer; a student who also writes short answers can produce it — and the production is what survives when the exam format changes.