How to Make Good Flashcards

Flashcards are the most misused study tool in existence. Students copy whole paragraphs onto cards, flip through them while half-watching a show, feel a warm sense of familiarity, and then blank on exam day. The tool gets blamed. The cards were the problem.

A flashcard works through one mechanism: retrieval. When you pull an answer out of memory, that memory gets stronger. Roediger and Karpicke (2006) showed that practicing retrieval beats repeated rereading for long-term retention, and Dunlosky et al. (2013) rated practice testing among the most effective techniques a student can use. None of that helps if your cards don’t actually force retrieval. A card that lets you nod along is a card that teaches you nothing.

This guide covers how to write cards that do force it. Reviewing them well is its own skill, covered in how to study with flashcards, and the science behind the review schedule lives in spaced repetition, explained. Here, we focus on the writing.

One fact per card

The single most important rule comes from the SuperMemo tradition and is called the minimum information principle: each card should test the smallest possible unit of knowledge. One question, one fact, one retrieval.

Why? Because a card containing four facts is really four cards welded together, and you will learn them unevenly. Suppose your card asks “Describe the three types of muscle tissue and where each is found.” You’ll nail skeletal, half-remember cardiac, and forget smooth entirely. Now the card is impossible to grade. Did you pass? Sort of? Spaced repetition software can’t schedule “sort of,” so the card keeps coming back, you keep half-failing it, and frustration builds.

Split it:

Q: Which muscle tissue type is striated and voluntary? A: Skeletal muscle.

Q: Where is smooth muscle found? A: Walls of hollow organs (gut, blood vessels, bladder).

Q: Which muscle tissue type is striated but involuntary? A: Cardiac muscle.

Three small cards feel like more work. They aren’t. Each takes two seconds to answer, each can be graded cleanly, and each can be scheduled on its own forgetting curve. Big cards feel efficient and review terribly. Small cards feel trivial and review beautifully.

Test recall, not recognition

Recognition is knowing the answer when you see it. Recall is producing it from nothing. Exams demand recall; bad cards train recognition.

The classic offender is the card you can answer from its shape. If your card says “Mitochondria?” and the back says “powerhouse of the cell,” you aren’t retrieving anything. You’ve seen the pair so often that the answer arrives by reflex, like the second half of a song lyric. The same problem hides in true/false cards, in multiple-choice-style cards, and in any card where the question contains most of the answer.

A useful test: could a stranger who knows the subject answer your card from the front alone? If the front says “The thing from lecture 4?” the card only works through memorized association with its own back. That association evaporates under exam conditions, where questions are phrased by someone else.

Write fronts that are specific, self-contained, and answerable in one direction:

Weak (recognition)	Strong (recall)
“Photosynthesis: light reactions happen in the thylakoid. True?"	"Where in the chloroplast do the light reactions occur?"
"1789?"	"In what year did the French Revolution begin?"
"Define osmosis (water, membrane, concentration…)"	"What is osmosis?”

That third row matters. Hints on the front feel kind. They are sabotage. Every hint you add is a hint the exam will not give you.

Good vs bad cards, subject by subject

Abstract rules stick better with examples, so here are paired cards from four common subjects.

Biology

Bad: “Explain the cardiac cycle.” Too big. This is an essay prompt wearing a flashcard costume.

Good: “During which phase of the cardiac cycle do the AV valves close?” A: “At the start of ventricular systole.” One event, one phase, instantly gradable.

History

Bad: “World War I: causes, dates, key figures, outcome.” This is a unit, not a card.

Good: “What event directly triggered World War I in 1914?” A: “The assassination of Archduke Franz Ferdinand.”

Language vocabulary

Bad: “ubiquitous, ephemeral, gregarious = everywhere, short-lived, sociable” Three words on one card guarantees interference between them.

Good: “What does ephemeral mean?” A: “Lasting a very short time.” One word per card, and ideally a reverse card too (“What word means lasting a very short time?”). Vocabulary has its own quirks, covered in how to memorize vocabulary.

Math and chemistry

Bad: “Solve: ∫x·eˣ dx” with a five-line worked solution on the back. You can’t grade a five-line answer pass/fail in your head.

Good: “What integration technique applies when the integrand is a product like x·eˣ?” A: “Integration by parts.” Card the decision, then practice full problems on paper separately. Cards train the recognition-of-method step that students lose under time pressure.

Cloze deletion vs question-and-answer

Cloze cards hide a word inside a sentence: “The powerhouse of the cell is the {{mitochondrion}}.” Q&A cards ask outright. Both work, and each has a home.

Cloze shines when context matters. Anatomy sequences, steps in a mechanism, lines of a legal definition: the surrounding sentence anchors the missing piece. Cloze cards are also fast to make from existing notes, since you delete rather than rewrite.

Q&A shines when you need flexible recall. A cloze card always shows you the same sentence frame, and your memory can hook onto that frame instead of the fact. If you only ever retrieve “insulin” inside one specific sentence, an exam question phrased differently may find nothing. For your most important facts, write a Q&A card even if a cloze version exists.

A decent default: cloze for facts embedded in sequences and definitions, Q&A for anything you must produce cold. And never cloze two blanks in one sentence on one card. That’s the minimum information principle again.

Use images where words are slow

Some knowledge is spatial. Anatomy, geography, chart patterns, molecular structures. Describing the position of the brachial artery in words is painful; an image with a label hidden is a two-second card. If your tool supports image occlusion, use it for anything you’d naturally point at in a textbook figure.

Images also help as memory anchors on text cards. A vocabulary card for gregarious with a photo of a noisy dinner party gives the memory a second handle. Keep it to one image with a clear purpose. Decorative clutter slows reviews and adds nothing.

What to do with lists

Lists are where the one-fact rule seems to break down. The five steps of glycolysis regulation, the eight cranial bones, the four causes of a recession: you can’t split them into independent cards, because the exam wants the set.

The trick is overlapping cards that each demand one retrieval. For an ordered sequence, write step-to-step cards: “In the cardiac conduction pathway, what comes after the AV node?” gets one card, “What comes after the bundle of His?” gets the next. Each card tests one link in the chain, and together they reconstruct the whole sequence. For unordered sets, write one card asking “How many cranial bones are there?” and then individual cards anchoring each member: “Which cranial bone forms the forehead?” A final card asking for the full list is fine as a capstone, graded pass only when you produce every member.

What you should not do is put the bare list on one card and recite it daily. You’ll learn the first and last items and permanently fumble the middle, which is exactly how serial recall fails under pressure.

The mistakes that quietly ruin decks

A few patterns account for most failed decks.

Copying sentences from the textbook. Verbatim text tests whether you recognize the author’s phrasing, which the exam will not reuse. Rewrite every fact in your own words. The rewriting is half the learning; if you can’t rephrase it, you don’t understand it yet.

Orphan cards. A card like “Why is this enzyme important?” with no context is unanswerable a month later when you’ve forgotten which lecture it came from. Add the minimum context: “In glycolysis, why is phosphofructokinase the key regulatory enzyme?”

Yes/no cards. “Is the femur the longest bone?” lets you pass with a coin flip. Convert to open recall: “What is the longest bone in the body?”

Carding what you already know. Every card costs review time forever. If you’ve known since childhood that Paris is the capital of France, that card is pure tax. Be ruthless at creation time; deck size determines your daily workload, a tradeoff explored in how many flashcards per day.

Making cards before understanding. Cards maintain memories; they don’t build understanding. If a mechanism confuses you, untangle it first with your notes, a diagram, or a method like the Cornell cue-column review. Carding confusion just schedules your confusion for daily delivery.

Hoarding instead of deleting. A deck is a working tool, not an archive. When a card turns out to be ambiguous, trivial, or a duplicate, delete it on the spot. Students treat deletion as wasted effort; in reality every deleted dud refunds review time to the cards that matter. The best decks I’ve seen shrink almost as often as they grow.

From raw notes to a working deck

The honest objection to all of the above: writing good cards takes time, and during a heavy semester that time is scarce. Two answers.

First, you don’t need cards for everything. Card the load-bearing facts, the ones whose absence would sink an exam answer. A typical lecture might justify 30 to 60 cards, not 200.

Second, drafting no longer has to start from a blank card. StudyDone reads your own notes, slides, or PDFs and drafts one-fact-per-card flashcards from them, which you then prune and rephrase. You can try the same idea in the browser with the flashcard generator, pasting in a section of notes to get a draft deck. Treat any generated draft the way you’d treat a classmate’s deck: edit aggressively, delete the obvious, and rewrite fronts until they force genuine retrieval.

However the cards get made, the standard stays the same. One fact. Recall, not recognition. Your own words. A deck built on those three rules turns review time into one of the highest-return hours in your week, and the next step is learning to run those reviews properly with active recall at the center.

FAQ

How many flashcards should one topic produce?

More than you expect. A single dense lecture slide often breaks into five or ten cards, because each card should hold exactly one retrievable fact. A 50-minute lecture commonly yields 30 to 60 cards. If a topic gives you only three cards, you are probably stuffing several facts into each one.

Should I make flashcards for everything in the course?

No. Cards are for material you must produce from memory: definitions, mechanisms, dates, formulas, vocabulary. Skills like essay writing or multi-step problem solving need practice in their full form. Card the facts that feed those skills, then practice the skill separately.

Are pre-made decks from other students worth using?

They are a weak substitute. Writing the card is itself a study act, and someone else's phrasing rarely matches how you learned the material. A shared deck can fill gaps near an exam, but treat it as a supplement. Cards built from your own notes almost always perform better.

Is it better to make flashcards on paper or digitally?

Digital wins for most students because software can schedule each card with spaced repetition, which paper cannot do at scale. Paper still works for small decks under about a hundred cards. Past that point, tracking what is due by hand becomes the bottleneck.

How long should writing flashcards take?

Roughly 20 to 30 minutes per lecture once you are practiced, less if you generate a first draft from your notes and then edit. If card writing eats whole evenings, your cards are too elaborate. Short question, short answer, move on.