How to Make Effective Flashcards (Cognitive Science Edition)

You've probably seen a deck of someone else's flashcards and thought "I should make those." Two weeks later, you have 400 cards, none of them have been reviewed in five days, and the ones you have reviewed feel weirdly easy in the moment and weirdly impossible on the exam. The deck isn't bad because of the schedule. It's bad because of the cards.

Most flashcard advice online is about which app to use. The app matters less than people think — Anki, Queazy, Quizlet, paper, all work fine. What determines whether a deck actually moves your retention is the design of each individual card. Below: the three rules that matter, what they look like in practice across different subjects, and the common failure modes.

Rule 1: One fact per card

If you can split the answer into two unrelated sentences and they'd both still be true alone, it's two cards.

A card that violates this: "What are the four causes of metabolic acidosis with anion gap and which two are most common in alcoholics?" The answer contains two pieces of information that don't logically depend on each other in the moment of retrieval. When you review it, you'll get one half right and the other wrong, and the card scores as a miss — but you've actually mastered half of it. The schedule wastes your time re-testing the part you knew.

Split: "What are the four common causes of high anion-gap metabolic acidosis?" and "In an alcoholic patient with anion-gap metabolic acidosis, what are the two most likely causes?" Each card now has a single retrieval target. The schedule can promote each one independently.

The pragmatic version of this rule: if the answer is longer than three sentences, you wrote two cards.

Rule 2: The prompt is a question, not a topic

"Hyperkalemia" is a topic. "What ECG changes appear in hyperkalemia, in order of severity?" is a question.

Topic-fronted cards encourage a vague "tell me everything you know" response. That's not retrieval — that's free association. The card grades itself on whether the response contains the correct material, which means you can ramble for 90 seconds, mention the right thing in passing, and call it a hit.

Question-fronted cards force a specific answer. You either produced "peaked T waves → prolonged PR → widened QRS → sine wave" or you didn't. The grade is binary and the retrieval is real.

A test for whether your prompt is a question: read it out loud. Does someone hearing it for the first time know exactly what answer you want? If they'd respond "what about it?", you have a topic.

Rule 3: The answer is sayable

Out loud. In one breath. Without slides.

This is the rule most students break. They put a five-bullet definition on the back of the card, the front asks "what is X?", and they review by reading both sides and feeling like they learned something. They haven't, because:

• The answer was too long to retrieve from memory in a single pass — they had to read it instead of recalling it.
• "Read both sides" is the failure mode where the card becomes a re-reading exercise.

Operational rule: the back of the card should be answerable as a sentence or a short list (≤5 items). If your concept genuinely needs a five-bullet answer, split it: one card per bullet, with the prompt narrowed enough to elicit that single bullet.

What good cards look like — by subject

Three subjects, three styles. The rules are the same; the texture differs.

Anatomy / pharmacology (fact-dense)

Bad: Front: "Cranial nerve VII" / Back: "Facial nerve. Innervates muscles of facial expression. Exits at stylomastoid foramen. Branches: temporal, zygomatic, buccal, mandibular, cervical."

This is six facts on one card. Test for recognition, not recall. You'll feel familiar with all of them and remember none.

Good (split into 3 cards):

• "Which cranial nerve innervates the muscles of facial expression?" → "VII (facial)"
• "At what foramen does the facial nerve exit the skull?" → "Stylomastoid foramen"
• "Five terminal branches of the facial nerve, in order from superior to inferior?" → "Temporal, zygomatic, buccal, mandibular, cervical (To Zanzibar By Motor Car)"

The third one bundles five items because the mnemonic makes them a unit — and the prompt explicitly asks for "five, in order," which forces production.

Conceptual subjects (mechanisms, causation)

Bad: Front: "Mechanism of cardiac action potential" / Back: a 200-word paragraph describing all 5 phases.

Good (split):

• "Phase 0 of the cardiac action potential — which channel opens, what happens to membrane potential?" → "Fast Na+ channels open; rapid depolarisation from -90 to +20 mV"
• "Phase 2 of the cardiac action potential — what causes the plateau?" → "Balance between Ca²⁺ influx (slow L-type channels) and K+ efflux"
• "In phase 3, which channel is responsible for repolarisation?" → "Delayed rectifier K+ channel (IKr)"

Each card tests one transition. You can promote phase 2 to monthly review while keeping phase 4 in daily review if phase 4 is the one giving you trouble.

Languages (vocabulary, sentence patterns)

Bad: Front: "der Tisch" / Back: "the table"

This is fine as a starting card but it's testing recognition. You'll see "der Tisch" and feel familiar with it. You won't be able to produce "der Tisch" when you need to say "the table" in German.

Better: Front: "How would you say 'the table is in the kitchen' in German?" / Back: "Der Tisch ist in der Küche."

Now you're producing a full sentence with article agreement, preposition case, and word order — three lessons in one card with a single retrieval cue. Cloze-deletion (fill-in-the-blank) is even better for vocabulary acquisition: "Der ___ ist in der Küche." (Answer: Tisch.)

Quantitative subjects (formulas, problem patterns)

Bad: Front: "Quadratic formula" / Back: "x = (-b ± √(b² - 4ac)) / 2a"

This tests memorising a formula. Memorising the formula is fine but it doesn't help you apply it.

Good (paired):

• "State the quadratic formula." → "x = (-b ± √(b² - 4ac)) / 2a"
• "x² - 5x + 6 = 0 — solve using the quadratic formula. Show the discriminant first." → "Discriminant: 25 - 24 = 1. x = (5 ± 1) / 2 = 3 or 2."

The first card builds the recall. The second card is the actual exam skill — applying it to a specific problem. You need both.