How to Make Flashcards That Actually Work: The Atomic Card Method

Most flashcards are made wrong. They contain three facts on the back. They are copied verbatim from lecture notes. They ask vague questions with multiple valid answers. They test recognition rather than recall.

The result: a deck that feels productive to review but fails to build reliable long-term memory. The problem is not spaced repetition — it is card quality. Spaced repetition is the scheduling mechanism; the cards are the content that mechanism acts on. Bad cards produce bad results regardless of how well-optimised your review schedule is.

This guide covers how to build cards that actually work, based on Piotr Wozniak's evidence-based principles for knowledge formulation.

The core principle: one fact per card

Wozniak's 1999 guide "20 rules for formulating knowledge" is the most cited practical resource on flashcard design. Rule 4 — the minimum information principle — is the most important: formulate cards to contain the minimum amount of information necessary to answer a single question.

A card with five facts on the back does not test five memories. It tests whether you can retrieve all five simultaneously. When you fail to recall one, spaced repetition software cannot determine which of the five is weak — the whole card resets. If you mark the card "Good" after recalling four of the five, the forgotten fact stays on too long an interval.

Split complex cards ruthlessly:

Before:

Front: What are the five stages of the Leitner system? Back: Box 1 = daily, Box 2 = every 2 days, Box 3 = weekly, Box 4 = fortnightly, Box 5 = monthly.

After (five separate cards):

Front: How often is a card in Leitner Box 1 reviewed? Back: Daily.

Front: How often is a card in Leitner Box 3 reviewed? Back: Weekly.

This feels extreme. It is correct. Five simple cards, each testing one retrieval pathway, are more useful than one complex card testing five at once.

What to put on the front: question structure matters

The front of a card is a question. It must be specific enough that exactly one correct answer exists. Compare:

The weak fronts could generate dozens of valid responses. The strong fronts have one correct answer. This is what makes a card reviewable: you either retrieve the correct answer or you do not.

Use these question starters:

• "What is..." — definition
• "Why does..." — mechanism
• "What happens when..." — cause-and-effect
• "Define:" — terminology
• "What is the [year/name/number] of..." — specific facts

Avoid "Tell me about..." or "What do you know about..." — these are essays, not flashcards.

What to put on the back: minimum viable answer

The back contains the minimum information that correctly answers the front. A good back is 1–3 sentences or a short phrase.

For definitions: one clear sentence.

Front: "What is the spacing effect?" Back: "The finding that distributing practice over time produces better long-term retention than the same total time spent in a single massed session."

For mechanisms: cause + effect in one sentence.

Front: "Why does spaced practice outperform massed practice for long-term retention?" Back: "Because retrieval after partial forgetting requires more effort, and that effort strengthens the memory trace in a way that easy retrieval from a fresh memory cannot."

For lists of 3+ items: use a bullet format, but only if those items belong together as a unit.

Front: "What are the three components of a minimum viable SRS system?" Back: "1. Card creation workflow. 2. Daily review schedule. 3. Triage process for weak items."

Note: if you can make three separate cards from this, you should.

Cloze deletion: when it beats Q&A

Cloze deletion fills in a blank rather than answering an open question. It is more effective than Q&A for:

• Specific facts, dates, and names (the context cues recall)
• Formulas (the surrounding terms cue the missing part)
• Vocabulary in context

"Ebbinghaus published Über das Gedächtnis in [1885]." "The SM-2 default ease factor is [2.5]." "Cepeda et al.'s meta-analysis covered [254] spaced practice studies."

Q&A is better for conceptual understanding, mechanisms, and anything where you want to generate an explanation rather than fill a blank.

The three mistakes that ruin decks

1. Copying notes verbatim. Notes are written to inform, not to be tested. "Spaced repetition is a learning technique that schedules review at increasing intervals" is a note. "What scheduling principle does spaced repetition exploit to maximise long-term retention?" is a card. If you cannot convert a note into a testable question, it contains passive information — and that information is unlikely to be retrieved reliably anyway.

2. Cards you do not understand. Wozniak's rule 1: do not learn what you do not understand. A card for a fact you do not understand at all will produce endless failed retrievals and will never promote to longer intervals. Understand first; encode second.

3. Interference cards. If two cards ask for similar facts that are easy to confuse (two drugs with similar names, two historical events from the same year), they will interfere with each other. Add discriminating context to each: "the drug ending in -pril is the ACE inhibitor, not the ARB" on one of the backs.

A practical workflow: build while you learn

The most effective time to create cards is immediately after reading or studying, when the material is fresh enough to phrase questions precisely.

1. Read a section
2. Do a brief free-recall summary (5 minutes — what can you retrieve without looking?)
3. Open your card deck and create cards for:
   • Key definitions and terms
   • Mechanisms and cause-effect relationships
   • Specific numbers, dates, or names that matter
   • Items you got wrong in the free recall
4. Review the deck the next day before reading the next section

This workflow integrates active recall (the free recall step) with spaced repetition (the review schedule). Each step reinforces the other.

Build your deck now

The WarpRead Flashcard Tool lets you build atomic cards directly in-browser — with a paper index-card aesthetic (ruled lines, red margin) for each card's front and back. You can also import a deck from an AI-generated text file using the built-in prompt template, and export as a standalone HTML file for offline focus-mode review.

For the full science behind card design and scheduling, take the free Spaced Repetition course — Lesson 5 covers Wozniak's complete 20 rules with worked examples.

References

• Wozniak, P.A. (1999). Twenty rules for formulating knowledge in learning. SuperMemo website.
• Kornell, N., & Bjork, R.A. (2007). The promise and perils of self-regulated study. Psychonomic Bulletin & Review, 14(2), 219–224.
• Dunlosky, J., et al. (2013). Improving students' learning with effective study techniques. Psychological Science in the Public Interest, 14(1), 4–58.