Flashcards

A prompt-and-answer card produces a generative retrieval effect — but only when it tests one thing at a time and stays restricted to the outer, arbitrary layers of knowledge. Used past those limits, it turns into an expensive, overwhelming maintenance task that crowds out deeper work.

What makes a usable card

• A single prompt per card. The prompt can be a question, a fill-in-the-blank, or image occlusion (blanking part of an image).
• One thing tested at a time. Bundling several facts into one prompt makes it impossible to score cleanly — a mix of right and wrong leaves you unable to mark the card correct or incorrect.
• Outer-layer content only. Reserve cards for the arbitrary detail (the upper layers), not the conceptual relationships that should already be encoded.

Failure modes

• Too many cards. Hundreds or thousands of cards is the signature of a system compensating for shallow encoding — it demands constant repeating time and creates overwhelm. Proper layered encoding shrinks the deck because most information becomes relevant and deeply held without a card.
• Confusing flashcards with summary cards. A summary card states information; it has no generative effect. A flashcard forces retrieval. Only the second one trains memory.

Links into the system

One of the two techniques under Rote Learning and Memorisation, alongside Method of Loci. Card load drops as Layers of Learning and Higher Order Learning do their work first; schedule the reviews through Spaced Interleaved Retrieval.