Spaced repetition for French: fixing decay and the wrong storage format
Learners of French lose words in a particular way. You revise a list, you recognise every entry on paper, and then a native speaker says one of them and you hear nothing. The word was in your memory. It simply was not stored in a form you could retrieve at speed, from sound.
Spaced repetition fixes the first half of that problem, the forgetting. Card design fixes the second half.
Why do French words fade the way they do?
Failure one is decay. Ebbinghaus measured it in 1885 and the numbers have held: about 70% of what you learn is gone within 24 hours, more than 75% within a week. A word met once looks like noise to your brain, which discards it.
Failure two is the wrong storage format. French written and spoken forms diverge more than in most languages. Ils parlent and il parle are identical to the ear. Liaison welds words together, final consonants fall silent. A card learned in silence trains reading and leaves listening untouched.
When should you review a word?
Every time you retrieve a word just before losing it, the trace strengthens and decays more slowly. So the sequence is one review today, then 3 days, 7, 14, 30, 90. Correct answers push the next one further out; a miss brings it back sooner.
Placement decides the outcome here. Ten repetitions in one evening produce a memory measured in days. Six repetitions spread across three months produce one measured in years, for less total effort.
Fixing the format: three rules for French cards
- Audio on every card, from the first exposure. Not as a bonus afterwards. The sound is the form you will meet in conversation.
- The article travels with the noun. Store une table, never table. Gender learned later is gender fought forever.
- Test in both directions. Recognising rendez-vous is easy; producing it under pressure is the skill you actually need.
Recall beats rereading, and it is not close
Rereading a list feels efficient because familiarity builds fast. But familiarity is recognition, and conversation demands retrieval. Researchers call the gap the testing effect: reaching for the answer before seeing it produces markedly stronger memory than reviewing it. Flashcards work for that single reason.
Why this needs software
Six review dates per word, several hundred words, each on its own clock. The arithmetic defeats paper within a fortnight. French Progress keeps that ledger: it tracks every word, schedules its reviews at the right intervals, attaches the audio, and shows you only what is due today, which runs to about five minutes.
For the other half of the equation, how many words you are aiming at, see the numbers on vocabulary size.
Gender is the first thing to go
English has no grammatical gender, so there is nothing in your memory for it to attach to. Learn table and three weeks later you have the word and not the la. The word survives, the gender does not, and you produce le table without hearing anything wrong.
The only reliable fix is to make the article part of the word from the first exposure. Never store pain, store le pain. If a card already exists without it, rewrite the card rather than adding a second one for the gender, because two cards for one word means the two answers drift apart.
Colour or position helps some learners more than repetition does: keep all masculine nouns on one side of the sentence you write, or say the article aloud with a fixed rhythm. The point is to give the gender a hook, since French itself gives you almost none.
False friends drift back to English
You learn that attendre means to wait. You get it right for a fortnight. Then one day you read it and think "attend", and from that point the card keeps failing. English never stops pulling, so these words do not behave like the others: they relapse rather than fade.
Treat them as a separate class. Review them from English to French, so the prompt "to wait" has to produce attendre and not assister. And keep their intervals shorter for longer than the algorithm would choose on its own, because a relapse after a correct answer is a different failure from simple decay.
What does a week of reviews look like?
Ten new words on Monday brings Monday's reviews plus whatever the previous fortnight scheduled. In a settled deck that lands around forty to sixty cards a day, which is four or five minutes. The number stays flat because words graduate out of the short intervals at roughly the rate you add them.
Change the input and the arithmetic changes fast. Thirty new words a day triples the review load within two weeks, and the session crosses twenty minutes. Most people stop there and blame their discipline, when the cause was an intake they set themselves a fortnight earlier.
Sources
- Henry L. Roediger III & Jeffrey D. Karpicke, Test-Enhanced Learning: Taking Memory Tests Improves Long-Term Retention, Psychological Science 17(3), 2006, pp. 249-255.
- Hermann Ebbinghaus, Memory: A Contribution to Experimental Psychology (1885), English translation, Classics in the History of Psychology, York University.
French Progress keeps the review ledger. You bring the 5 minutes.
