A flashcard is a small machine with one job: force a single piece of knowledge out of your memory. When a card does that, it works. When it does not, you can spend an hour with it and learn nothing, which is how most people conclude that flashcards do not work for them.
The cards are usually the problem, not the method. Here is what makes a card good, what makes it useless, and how to fix the ones you already have.
A good card asks exactly one thing
The single most common mistake is the two-in-one card. It looks like this: "What are the three types of X?" If you can only recall two of the three, the card marks you wrong, and you never learn which one you missed.
Split it. Three cards, one per type, each with its own answer. You will now know exactly which one is shaky, instead of flagging the whole set as failed.
The one-answer rule
A card should have one answer you can check. Not "roughly this, and also that," but one thing you either say or do not say. If marking your own answer requires judgement, the card is too vague to be useful.
This is why "What is stress?" is a bad card and "What happens to the body during the alarm stage of stress?" is a better one. The first has no checkable answer. The second does.
Cards that test recognition instead of recall
Some cards can be answered from the shape of the answer on the page. A multiple-choice card where you recognise the right option is testing recognition, and recognition is much easier than the recall an exam asks for.
The fix is to phrase the card so you have to produce the answer, not pick it. "Which of these is the powerhouse of the cell?" is weak. "Which organelle produces most of the cell's ATP?" makes you say "mitochondria" yourself.
The card that only makes sense in context
A card that reads "What did the court hold?" is fine while you are holding the chapter and meaningless in a review session three weeks later. Cards have to stand alone, because you will meet them alone.
Put the context in the question. "In the case about advertising to children, what did the court hold about commercial speech?" is longer, but it is a card you can still answer in December.
Three fixes that rescue most bad cards
- Split it. One card, one fact. If a card has two answers, make two cards.
- Make it checkable. If you cannot say your answer out loud and know whether it is right, rewrite the question.
- Add the missing context. If the card only works inside the chapter, put the chapter's clue into the question.
Run these three over a deck and you will usually delete a quarter of it and improve the rest. That is not a loss. A smaller, sharper deck is faster to review and harder to fool yourself with.
Cloze deletion, and when to use it
Cloze cards hide a word in a sentence and ask you to fill it in: "The mitochondria produces most of the cell's ___." They are fast to write and they work well for definitions and formulas. They are also easy to overuse.
The failure mode is a cloze where the missing word is guessable from grammar or context, so you fill the blank without understanding anything. If you can complete the sentence without knowing the concept, the cloze is decoration. Say the idea out loud instead and make a real question.
Good cards from a bad source
Most people do not write their cards from a blank page. They convert a lecture, a textbook chapter, or their own notes. That is fine, with one adjustment: the first pass from any source is a draft, not a final deck.
Read every generated card and apply the three fixes. The generator handles the typing; you handle the judgement, which is the part that actually decides whether the deck teaches you. If your source is a PDF, this is the same step the PDF workflow describes, and it is not optional.
One more thing worth doing: for the cards you keep getting wrong, look at why. It is almost never the fact you are missing. It is usually a card that asks two things, or a question so vague that you were never sure what a correct answer sounded like.
The test of a good deck
Run the deck with the source closed and notice how often you say "I know this but I can't phrase it." That phrase is the smell of a badly written card, not a gap in your knowledge. Fix the card, and the knowledge you already had becomes usable.
A ten-minute audit of a deck you already have
You do not have to rebuild a deck to fix it. Set a timer for ten minutes, open it cold, and delete anything that fails one of these tests.
The out-loud test. Read the question and answer it aloud before you look. If your answer was "sort of," the problem is the card, not your memory. It is either asking two things at once or phrased too vaguely to check.
The blank-stare test. Some cards make you pause not because you forgot, but because you cannot tell what is being asked. A card that confused you today will confuse you every time you meet it.
The too-easy test. If you answered without thinking, the card has already done its job. Retire it so it stops consuming review time you need elsewhere.
The duplicate test. Two cards that ask the same thing in different words are one card and one distraction. Keep the clearer one and delete the other.
The orphan test. A card that only makes sense inside a chapter you have since closed is an orphan. Add the missing context to the question or delete it, because you will meet it without the chapter in front of you.
Ten minutes of this usually removes a fifth of a deck and improves another fifth, and the deck gets noticeably faster to run. That speed matters more than people expect, because the review loop is the part you have to keep doing. A deck that takes twenty minutes gets reviewed; one that takes fifty gets postponed, then abandoned.
Run the audit once a month, or whenever a session starts to feel like a chore. The chore is usually the first sign that the deck has filled up with cards that stopped earning their place.
Why this matters more than the app
People spend hours choosing a flashcard tool and minutes writing the cards, which is the wrong way round. A good app running vague cards produces a frustrating hour and no learning, while a plain text file of sharp cards beats it every time.
The card is the unit of work, and the card is the thing you control. Get that right and the tool barely matters.
Frequently asked questions
How long should a flashcard question be?
As long as it takes to be unambiguous, and no longer. A card that is too terse loses context and becomes unanswerable later, so clarity beats brevity.
Is it bad to put two facts on one card?
Yes, split it. If you can only recall one of the two, the card marks you wrong and you never learn which half you missed.
When should I use cloze deletion?
For definitions and formulas where one word is the answer. Avoid it when the missing word is guessable from context, because then you complete the sentence without understanding anything.
Do cards need to make sense on their own?
Yes. You will meet them without the chapter in front of you, so the question has to carry whatever context it needs.
Is a smaller deck always better?
A sharper deck is better. If three cards ask the same thing, they are one card and two distractions, and deleting them makes the real ones faster to review.