Does explaining the wrong answers actually help?

Supported, with a precise boundary. Little et al. (2012) is real and has held up, but the advantage is specific to related, untested information. On directly tested material, short answer with feedback wins — Kang et al. (2007), and the synthesis in Üner, Tekin & Roediger (2021).

For an exam of novel passages testing adjacent knowledge, related-untested is arguably the right target. So the claim survives for the MCAT specifically, not in general.

This is the weak link. Three problems:

1. Butler, Karpicke & Roediger (2007) tested feedback type on MC directly and found no difference. Timing mattered; type did not.
2. Finn & Metcalfe (2010) found the winning elaboration was learner-generated — scaffolded hints. Provided elaboration tied with plain correct-answer feedback.
3. Alamri & Higham (2022) found corrective feedback actively impairs related-item performance — the exact outcome the Little benefit is measured on.

Revised: MC works for transfer when the learner is made to reason about the distractors — not when the explanations are handed to them.

No research supports this, because nobody has studied it. What exists is adjacent: LLM distractor generators still produce invalid distractors needing expert review, and automated explanatory-feedback pipelines report meaningful error rates.

The likelier explanation is structural. Anki ships no multiple-choice note type, no distractor field, and no way to record confidence. The format is absent from the tooling, not merely expensive — and a bad distractor is worse than none, since the entire effect is conditional on competitiveness.