A good flashcard tests one thing, takes under five seconds to read, and makes you pull the answer out of your head from scratch. Most student-made flashcards break all three rules, which is why decks that feel productive don't transfer to the exam. This guide is how to make ones that do, based on retrieval practice research.
Why is it so hard to make good flashcards?
Anyone can write a question on one side and an answer on the other. Deciding what goes on the card is where it falls apart.
Turning a page of notes into a deck means picking which facts matter, pulling them out of the surrounding context, and writing questions that test real understanding instead of "have I seen these words before." It's a compression problem. You're taking dense, connected information and breaking it into small independent pieces that can each be reviewed on their own schedule.
Most students skip this part entirely. They copy sentences from their notes onto cards and end up with a deck that tests whether they can recognise their own handwriting. The act of writing a good flashcard is itself the learning. Craik and Lockhart's levels of processing framework (1972) explains why. The deeper you process something while encoding it, the better you remember it. Deciding what's important enough to become a card is exactly the kind of deep processing that rereading never triggers.
How do you make good flashcards?
Start with material you've already read and understood. Don't make cards on your first pass through new content. You'll end up with cards about things that turn out not to matter, or worse, cards that lock in your misunderstandings.
One fact per card. Ask yourself: if I get this wrong, is there only one possible reason? If yes, the card is atomic. If no, split it. "What are the four chambers of the heart?" is a bad card. It tests a list. "Which chamber receives deoxygenated blood from the body?" is a good card. One fact, one answer.
Keep the front short. The question side should take under five seconds to read. Long wordy question stems waste your mental effort on parsing the question instead of retrieving the answer. If your question is a paragraph, rewrite it.
Make the answer specific. A back side that says "it's complicated" or "there are several factors" teaches you nothing. If you can't write a clean concise answer, you don't understand the concept well enough to make a card about it yet. Go back to the source.
Use images when possible. Paivio's dual coding theory (1971) and the picture superiority effect both show that people remember images roughly twice as well as words alone. A card with a labelled diagram of the heart will outperform a text-only card about heart anatomy. Not every card needs a picture, but for anything visual or spatial (anatomy, geography, diagrams, processes), images change the math.
Add context, not clutter. "Q: What is apoptosis? A: Programmed cell death" is a dictionary entry. Adding why it matters turns it into something worth remembering: "Q: Why is defective apoptosis central to cancer? A: Cancer cells evade programmed death signals, so they keep dividing." The context should clarify, not pad.
Watch for interference. Similar cards confuse each other in your memory. "What does the mitral valve do?" and "What does the tricuspid valve do?" will bleed together unless you give them distinguishing visual or contextual cues. Wozniak names interference as probably the single biggest cause of forgetting in long-term spaced repetition users. The fix is making similar cards look or feel different from each other.
What does a 45-minute flashcard creation session look like?
15 minutes
Read one section of your textbook or lecture notes. Don't make cards yet. Just read. Highlight or underline key facts, but only after you've read the full section. If you can't explain the main ideas without looking, reread.
15 minutes
Go back through your highlights and write the cards. For each key fact, write a question that needs recall, not recognition. One fact per card. Keep the front under one sentence. If the material is visual (a diagram, a process, a map), use image occlusion or include the image on the card. Aim for 10 to 20 cards from a typical lecture. More than that and you're either testing trivia or bundling facts.
10 minutes
Quiz yourself on the cards you just made. This first review, even with the material fresh, kicks off the spaced repetition cycle. Pay attention to cards that feel ambiguous or confusing. Those need rewriting, not just another review. If a card has two possible correct answers, the question is broken. Fix it now.
5 minutes
Review any older cards that your spaced repetition system has scheduled for today. These are the ones approaching the edge of your forgetting curve. The effort of pulling them back out when it feels hard is exactly what makes the memory stick.
The whole loop takes 45 minutes for one lecture's worth of material. Doing this for every subject every week is the part most students can't sustain, which is the gap Lexie is built for. Snap a photo of your notes and the app generates flashcards that follow these rules: one fact per card, short front, specific back, image occlusion for diagrams. The review discipline is the same either way. The card creation is the part that gets automated.
Frequently asked questions
Making your own has a real learning advantage. Deciding what matters and writing a good question forces deep processing before you start reviewing. Senzaki et al. (2017) found that students who made their own flashcards scored significantly higher on exams. That said, well-built pre-made decks (AnKing for medical students, for example) save a huge amount of time. The best approach is to treat pre-made decks as a starting point and then edit, delete, and rewrite based on your own understanding. A deck you've curated will outperform one you've imported and never touched.
10 to 20 well-built cards per hour of content. More than that usually means you're going too granular, testing trivia that won't show up on any exam. Fewer than that usually means you're bundling multiple facts into one card. Stick to the minimum information principle. Each card tests one atomic fact. If you're writing 50 cards for one lecture, some of them aren't worth memorising. Focus on the concepts that connect to other concepts, the things the instructor emphasised, and the material you'd need under exam pressure.
Both work, for different things. Basic Q&A cards are better for conceptual understanding because you have to generate the answer from nothing. Cloze deletions, where a word or phrase is blanked out of a sentence, are faster to build and work well for vocabulary, definitions, and factual recall. The risk with cloze is pattern matching. You learn the shape of the sentence instead of the meaning. For language learning and terminology, cloze is great. For mechanisms, relationships, and processes, Q&A cards build deeper memory.
Yes, you just have to change the format. For problem-solving subjects like math and physics, write cards that test the setup or the first step instead of the full solution. "What method would you use to solve this type of integral?" is a better card than "Solve this integral." For essay-based subjects, write cards that test key arguments, evidence, and connections. "What is the strongest counterargument to utilitarianism?" forces you to pull and organise ideas you can later stitch into an essay under exam pressure. The flashcard becomes a building block, not the finished thing.
Five big ones. Cramming several facts onto one card, so when you miss one part the scheduler reschedules everything, including the bits you knew. Writing cards you don't understand yet, against Wozniak's first rule. Testing recognition instead of recall by nodding along at answers you merely recognize. Using someone else's deck without rewriting the cards in your own words. And dropping cards after one correct answer instead of letting the algorithm decide when you're done (Kornell & Bjork, 2008).
