One atomic fact per spaced repetition card, answer under one sentence — multi-component cards produce vague retrieval the algorithm cannot schedule
Create spaced repetition cards testing exactly one atomic fact per card with answers under one sentence, because cards testing multiple components produce vague retrieval that the algorithm cannot accurately schedule.
Why This Is a Rule
Spaced repetition algorithms (SM-2, FSRS) schedule card reviews based on a binary signal: you remembered or you didn't. This signal is only meaningful when the card tests exactly one thing. A card asking "What are the three branches of government and their functions?" tests six facts simultaneously. If you remember the legislative and judicial branches but not the executive, what do you report? "Partially remembered" isn't a valid input — the algorithm needs remembered/forgotten. Either answer misrepresents your knowledge state, causing the algorithm to schedule the card incorrectly.
Atomic cards — one fact per card — produce clean retrieval signals. "What is the function of the executive branch?" has one answer. You either know it or you don't. The algorithm receives an accurate signal and schedules optimally. Over thousands of reviews, this precision compounds: atomic cards converge on each fact's true forgetting curve, while multi-fact cards produce noisy scheduling that wastes review time on facts you know and under-reviews facts you don't.
The "answer under one sentence" constraint prevents a subtler form of non-atomicity: cards with atomic questions but paragraph-length answers. If the answer is a paragraph, the card is actually testing multiple related facts bundled into one card. Split it.
When This Fires
- When creating flashcards for any spaced repetition system (Anki, RemNote, Mochi)
- When existing cards feel vague or hard to grade ("Did I really remember that?")
- When review sessions take too long relative to the number of cards
- Complements Create SRS cards only for information you already understand — memorization without comprehension produces brittle, unusable memory (understand before memorizing) with the card design standard for what you do memorize
Common Failure Mode
"Explain X" cards: "Explain the process of photosynthesis." This is an essay question, not a flashcard. You produce a partial, vague answer, grade yourself "Good" because it felt close, and the algorithm schedules it as known. The specific components you forgot never surface for targeted review.
The Protocol
(1) For each fact you want to memorize, create a card testing exactly that fact: one question, one answer. (2) The answer should be one sentence or less. If you can't express the answer in one sentence, the question is testing multiple facts — split it. (3) Use specific, closed questions: "What neurotransmitter is primarily involved in reward pathways?" (answer: dopamine) not "What does dopamine do?" (answer: many things). (4) When reviewing, grade strictly: if you couldn't produce the exact answer before flipping the card, grade it as forgotten. Partial credit prevents the algorithm from targeting your actual gaps. (5) For complex topics, create many atomic cards that collectively cover the topic rather than a few comprehensive cards. 10 atomic cards about photosynthesis produce better retention than 2 "explain photosynthesis" cards.