Most learners who abandon a language blame willpower. The real problem is usually spacing. Words get met once, then dropped, then met again months later as strangers.
Games fix that by accident, and the evidence on how well they fix it is stranger than most people expect.
The card table got there first
Hermann Ebbinghaus spent the 1880s memorising nonsense syllables on himself, and he found that gaps between review sessions beat cramming everything into one night. Card players had been running the practical version for centuries without writing it down. Deal, call, fold, repeat.
One card game runs on maybe twenty words. Numbers, suits, the verbs for raising and passing, the muttering between hands.
Setting changes very little: a kitchen table, a student club, a casino in Kathmandu where the same handful of terms circles the room all evening. Repetition without a worksheet.
Beginners who play two hours of cards in a second language often outrun the ones who reread a vocabulary list twice. Same words, far more retrievals.
Quiz apps fixed the timing, mostly
Kahoot reports over 12 billion participants since launch, although that figure counts every repeat appearance, so read it as proof of ubiquity rather than a headcount. Gimkit and Quizizz run the same loop with different scoring gimmicks.
Teachers reach for them because thirty students can answer at once. Memory benefits for a completely different reason.
That reason has real evidence behind it. Cepeda and colleagues pooled 254 studies and more than 14,000 participants in 2006, and distributed practice beat massed practice across ages and materials, with the ideal gap scaling to how long you need the word to survive. Want it next month? Gaps of days work better than gaps of minutes.
Strip the leaderboard away and three things are left doing the work.
- Retrieval instead of recognition. The learner has to produce the word before the options disappear.
- Feedback lands in about a second, so a wrong guess never gets time to settle.
- Missed items return later in the same session. Crude spacing, wearing a party hat.
Chat servers teach the words nobody assigned
Discord exchange communities are hard to size. One listing puts the main Language Exchange server near 9,000 members, another around 74,000, depending which tracker you trust and when it last refreshed.
The exact number matters less than what happens inside: unscripted text, typos, slang, and a correction from a native speaker within a minute.
Interface language does the same job more quietly. MelBet runs its interface in dozens of language versions, and a user who switches one over meets the same nouns and buttons every session without opening a book.
Boring, repetitive, effective. Most people only notice how much stuck when they try to explain it in their first language and cannot find the word.
What to actually do with this?
Here is the part that surprised me. A 2022 meta-analysis of digital game-based language learning found that games built for entertainment outperformed games built to teach a language, though the confidence intervals overlap enough that nobody should brag about it.
Vocabulary gains sat in the medium range, around 0.5 for between-group designs and closer to 0.95 within groups.
Pick the game you would play anyway, then add a few constraints.
- Say the words out loud, even alone. Silent reading skips production entirely.
- Short and frequent beats long and rare.
- Five words that tripped you up go in a note. Not fifty.
- Human opponents are better than bots, because a person expects a reply.
Vocabulary size shows the scale of the job. Reading a novel or a newspaper without constant guessing takes roughly 8,000 to 9,000 word families on Paul Nation’s estimate, and no card game delivers that. Nobody claimed it would.
Games handle the first two or three thousand words, the ones that repeat constantly, and they handle them while the learner is busy trying to win. Hard to argue with free labour.
