In October 2019, during the second month of my master’s in cognitive neuroscience at la facultad, I sat down with a stack of printed papers on long-term potentiation and tried to memorize the names of about forty receptor subtypes and their canonical pathways. I was reading each page three or four times, highlighting in two colors, and feeling productive. Two weeks later, in a peer study session, a colleague asked me to explain the difference between AMPA and kainate receptor function in mossy fiber synapses and I produced a sentence that was, politely, mostly nonsense. I had spent nine hours rereading that material and could not retrieve the basic distinctions under any pressure at all.
That was the week I downloaded Anki. I built a deck of 312 cards from a single course — Neurobiología Molecular y Celular — and committed to a daily review schedule. Eleven weeks later I ran a self-test on the original material and scored 91% retention on cards I had not actively studied for between three and seven weeks. I had spent, by my Anki history log, an average of fourteen minutes per day. The time investment was roughly a third of what rereading had cost me, and the retention measurably better. I have not seriously studied any factual material without Anki since that semester, and I now recommend it to every PhD student I supervise.
What follows is everything I have learned about making the system actually work — both from my own use across Spanish neuroscience vocabulary, English academic terminology, and a French verb deck, and from watching dozens of estudiantes try and frequently fail to integrate it into their lives.
What spaced repetition actually does
The empirical foundation goes back to Hermann Ebbinghaus, who in 1885 spent months memorizing nonsense syllables on himself and plotting the rate at which he forgot them. The resulting forgetting curve is one of the most replicated findings in psychology: information you learn today will be roughly half forgotten by tomorrow if left untouched, and almost entirely gone within a week. What Ebbinghaus also discovered is that each successful re-encounter with the material flattens the curve. Review on day one, the second forgetting curve decays more slowly. Review on day three, slower still. The intervals grow geometrically, until a single annual reminder is enough.
Spaced repetition algorithms exploit this directly. Anki’s original scheduler was SM-2, derived from Piotr Wozniak’s work on SuperMemo in the late 1980s — Wozniak was experimenting with optimal interval calculations on himself for years before any of this became usable software. Newer versions of Anki ship with FSRS, a successor that uses your individual recall patterns to fit the curve more precisely. Both algorithms do the same thing: when you mark a card as forgotten, they schedule it again within minutes; when you mark it as recalled with some effort, they push the interval out to a few days; when you mark it as effortless, the interval stretches to weeks, then months, then years.
The literature is robust. The Cepeda et al. 2008 meta-analysis in Psychological Science is the one I cite most often — it pulls together 317 experiments and shows the spacing effect holds across age groups, materials, and retention intervals up to a year. Roediger and Karpicke’s 2008 paper in Science, on the testing effect, is the second pillar: retrieval itself, not just spacing, is what consolidates memory. Anki combines both mechanisms in a single tool, which is part of why it works so well for me and the alumnas I supervise.
The practical result is that with somewhere between five and fifteen minutes of review per day, a serious learner can hold tens of thousands of facts in long-term memory indefinitely. That sounds like marketing copy until you have lived it for two or three years.
What Anki is good for, and what it is not
I want to be honest about the limits, because I see too many enthusiastic newcomers try to Anki-fy their entire intellectual life and end up resenting the tool. The fit is excellent for vocabulary across any domain — foreign language, medical and anatomical terminology, legal vocabulary, the species names a biology student needs, the molecular structures a chemistry student wrestles with. It works beautifully for formulas in mathematics and physics provided you understand the derivation already and just need to keep the formula itself accessible. Anatomy, geography, historical dates and figures, translation pairs, definitions of key concepts — all of this is what flashcards were invented for, and Anki refines the format.
The fit is poor for procedural skills. You cannot Anki your way to better essay writing or cleaner code; those require practice in the actual medium. The fit is also poor for open-ended understanding — questions like why the Roman Empire fell, or how monetary policy interacts with labor markets, do not fit on a card without becoming so reductive they are useless. For that kind of material I read, take notes, and write summaries; Anki only comes in afterward, to anchor specific facts I want to retain from the synthesis.
There is a middle zone too. Equations you need to derive rather than recognize sit awkwardly: Anki will help you keep the form of the equation accessible, but the derivation itself is procedural and needs separate practice. I have a deck of about sixty integration techniques where the cards prompt me with the integral and the back shows the substitution to try — that works because what I am rehearsing is pattern recognition, not the derivation steps themselves.
Daily volume, which is where most people break
This is where most attempts collapse, and I have watched it happen often enough that I want to be very direct. People start with extravagant ambition, add a hundred new cards on day one, and within three weeks they are staring at a queue of 800 due reviews. The math is unforgiving: every new card creates between five and ten future reviews scheduled across the following weeks and months before it stabilizes at long intervals. Twenty new cards added today schedules roughly 100 to 200 future reviews over the next year.
For casual learners, the sustainable range I see working is five to ten new cards per day, which costs five to ten minutes. For a serious student treating Anki as a primary memorization tool, fifteen to twenty-five new cards per day produces fifteen to twenty-five minutes of daily work. Medical and law students who need extreme volume can sometimes push to forty new cards a day, but I have rarely seen anyone sustain more than that without burning out within a couple of months. The single hardest rule I enforce on myself and recommend to others is this: never let the daily review queue exceed what you can finish in twenty-five minutes. If the queue is creeping past that ceiling, the response is to lower the new-card rate, not to push harder for a few days hoping it normalizes. It will not normalize. It will compound.
Card design is everything
I cannot overstate this: bad cards cause everything else to fail. The first principle I teach is one fact per card. If a card asks “What is photosynthesis?” and the answer is a paragraph covering inputs, outputs, location, and stages, you will get it wrong on whichever component you forget and never reliably learn any of them. The same content split into six or eight atomic cards — one for inputs, one for outputs, one for the organelle, one for each stage name — gives the algorithm something it can actually schedule meaningfully, because each unit succeeds or fails on its own.
The second principle, which Wozniak called the minimum information principle, is that the smaller and more atomic the unit, the more reliably it stabilizes in long-term memory. I see students resist this constantly because writing many small cards feels less efficient than writing one big card. It is the opposite. One big card is roughly equivalent to no card.
The third tool I rely on heavily is cloze deletion. Anki accepts a syntax like “The {{c1::hippocampus}} is the brain region most associated with forming new {{c2::declarative}} memories,” which generates two cards from one note — one hiding the structure, one hiding the memory type. For sentence-shaped material, definitions, and lists, cloze cards are dramatically faster to author than traditional front/back pairs, and they preserve enough context to keep the recall feeling natural rather than decontextualized. About sixty percent of mi mazo of neuroscience vocabulary is in cloze format for exactly this reason.
For anything spatial — anatomy, geography, molecular diagrams, even software UI screenshots — image occlusion is the killer feature. I install the Image Occlusion Enhanced addon on every Anki instance I set up. You upload an image, drag rectangles over the labels you want to hide, and Anki generates one card per occlusion, testing you on each label independently. My brain anatomy deck would have taken me triple the time to author without it, and the retention is markedly stronger than for the equivalent text-only cards.
One mistake I want to name explicitly: do not make cards on material you do not yet understand. I see this constantly. A student finishes a confusing lecture, decides to “Anki their way through it,” and ends up with thirty cards whose answers are noises rather than meaning. Cards are not for learning. They are for not forgetting what you have already learned. Read first, understand first, take messy notes first. Then make cards on the specific facts you want to anchor. If a topic is unclear, no flashcard will rescue it, and you will spend weeks frustrated by reviews that feel arbitrary.
The daily rhythm
I run my reviews in the morning, before opening anything else on the computer. Willpower is highest then, and the mild cognitive load of fifty or sixty reviews acts as a warm-up for harder analytic work later in the day. This pairs naturally with the morning study routine, which I treat as a structural part of my workday. The alternative — reviewing within one or two hours of sleep — has its own empirical justification, since overnight consolidation strengthens the day’s encoding. I sometimes use it for material I really need to lock in, but I am careful not to schedule cognitively demanding reviews so late that they disrupt sleep itself.
What I would beg you not to do is batch reviews on weekends only, or attempt to clear a week’s accumulation in a single two-hour Sunday session. Both of these break the daily rhythm that makes the system function. The spacing algorithm assumes you are showing up each day; if you skip three days the queue grows in a way that is psychologically painful to face, and most people quit at exactly that point rather than work through it. I have lived this myself during fieldwork trips and the recovery is rough.
When reviews do pile up — and they will, eventually, for everyone — the protocol I follow is to stop adding new cards immediately until the queue is back below fifty per day, work through the backlog in twenty-five-minute Pomodoros rather than marathon sessions, and use the Hard rating liberally on cards that are barely recallable rather than the Again rating, which buries you in short-interval re-reviews. For backlogs above a thousand cards, Anki has a Reset and Reschedule option that redistributes due dates across the upcoming days rather than dropping them all on you in one avalanche. I have used it twice and it saved both decks.
The mistakes I see most often
The single most common pattern of failure I see in students is cards that are simply too long. If the back of a card runs three or more sentences, the card is functioning as a paragraph to reread rather than a fact to retrieve, and it will not consolidate cleanly. The rule I enforce is that if it cannot fit comfortably on one phone screen without scrolling, it splits.
The second pattern is reliance on pre-made decks downloaded from AnkiWeb. The act of constructing the card is part of the learning — selecting what to extract, phrasing the prompt, deciding what context to preserve. Cards someone else wrote almost never stick, because they encode someone else’s mental model of the material, not yours. I use pre-made decks only as raw material from which I selectively rewrite cards into my own deck, and I tell every alumna who asks me about Anki the same thing.
The third is adding too many cards in bursts. Ten cards a day for thirty days produces 300 mastered cards. Fifty cards in a single weekend produces 300 cards added, most of which will be functionally forgotten by the end of the month because the review schedule cannot consolidate that many simultaneous introductions. The geometry of the algorithm punishes burstiness severely.
The fourth is cards without context. A card that asks “What is X?” with no subject area or chapter cue is harder to remember than the same card tagged with its course or topic. I prefix every card with a short context cue — “Neurobio:” or “French verbs:” or “EU law:” — and tag it appropriately. Recall is faster and reviews feel less arbitrary.
The fifth, which I noticed only after years of self-observation, is the environment in which reviews happen. Anki on your phone in a noisy café produces dramatically weaker encoding than Anki at a quiet desk. The same card studied in different conditions has measurably different retention curves. I do my reviews at the same desk every morning, with lofi music playing at a constant volume, and I credit a non-trivial portion of my consistency to the ritual itself.
The compound effect over years
What makes Anki feel almost unreasonable is that it compounds. After a year of consistent fifteen-minute daily reviews, I had something like 5,000 facts solidly in long-term memory across three decks. After two years, around 14,000. By the time I started my doctorate, my main neuroscience deck had grown past 22,000 cards and the daily review load was still around twenty minutes because the intervals on stable cards had stretched to multiple months. With a spaced repetition system maintained over years, this becomes mundane.
Anki is not a complete study system on its own. It pairs well with Pomodoro work blocks — I use the reviews as the warm-up at the start of a deep work session — and with Cornell-style note-taking, where cue column items convert naturally into card prompts. Active recall in essay form, problem solving for derivations, and adequate sleep for overnight consolidation are all complementary pieces.
If you are starting today
Install Anki Desktop from ankiweb.net — it is free on every platform. The mobile companion is a one-time $25 purchase on iOS, called AnkiMobile, or the free AnkiDroid app on Android. Sync your decks across devices through a free AnkiWeb account.
Set your new-cards-per-day limit to ten and your review queue limit to one hundred. Do your reviews every morning before anything else for seven consecutive days. After a week you will know whether the system fits your study style and your life. For most students the first two or three weeks feel deceptively easy — the cards are recent, the reviews are short, retention feels effortless. That is the trap of underestimating the system. By month three you will be reviewing material from week one and remembering it with no extra effort, which is the compound finally becoming visible. That compound is the entire point of the exercise.
Combine this with a calm study environment and study music, and you have, in my experience as both user and supervisor, the highest-leverage memory system available to a student today.
References
- Karpicke, J. D., & Roediger, H. L. (2008). The Critical Importance of Retrieval for Learning. Science, 319(5865), 966-968. The foundational paper showing testing dramatically beats re-reading.
- Cepeda, N. J., Vul, E., Rohrer, D., Wixted, J. T., & Pashler, H. (2008). Spacing effects in learning: A temporal ridgeline of optimal retention. Psychological Science, 19(11), 1095-1102.
- Bjork, R. A., & Bjork, E. L. (2011). Making things hard on yourself, but in a good way. Psychology and the real world, 56-64. The “desirable difficulties” framework that explains why retrieval practice works.
