How the method works
A foreign word on its own is a sound with nothing attached to it. The method gives it something to hold: an English sound-alike, then an image of that sound-alike doing something to the meaning. Here is why that works, why the images are loud, rude and ridiculous, and where the research says the trick runs out.
Two steps: a sound, then an image
It's called the keyword method, and it has been studied since the 1970s. The first careful test, at Stanford in 1975, happened to use Spanish. It builds a bridge in two steps:
- The sound link. Find English that sounds like the Spanish, or like most of it. Perro sounds like "pear oar". Not the meaning: the sound.
- The image link. Build one image in which that sound-alike and the real meaning act on each other. A dog, rowing a boat carved out of a giant pear, with a pear-shaped oar.
Recall runs back along the chain. You hear perro, it sounds like "pear oar", the pear boat comes back, and there's a dog in it. Or you see a dog in the street, the absurd boat turns up, and "pear oar" hands you perro.
An image, not a picture
The app shows you a drawing, but the drawing isn't what you remember. What does the work is the image in your head: the one you can close your eyes and replay, with the oars dripping pear juice and the dog looking pleased with itself. The drawing is a starting point. The first time you meet a word, give its image five seconds: make it move, make it louder, make it yours. That's the step that turns a card you've seen into a word you know.
The rule that matters most: make them interact
Across fifty years of research on the keyword method, the strongest single factor is interaction: the sound-alike and the meaning have to be doing something together in the same scene, or be the same object.
- Weak: a casserole sitting next to a house.
- Strong: a house that is a casserole, bubbling, with a spoon sticking out of the chimney. Casa.
Two things side by side are two memories. One thing that is both is a single memory, and pulling on either end brings back the other.
The image has to be something you can see
A syllable isn't a hook. For libro (book), "lee bro" is just raw sound; there's nothing to look at. Push "lee" one step to flea and you have a flea reading a book as big as a house. That jump, from a sound to something you can draw, is the creative step, and it's the one that decides whether a sound-alike works.
Why the images are loud, rude and ridiculous
Memory isn't neutral. The brain keeps what is vivid, emotional, surprising and physical, and throws away what is bland. "Casa means house" is bland, and it's gone by tomorrow. Each of these gives an image more to hold on to:
Action beats a still
A thing doing something is remembered far better than a thing sitting there. Splashing, burning, crushing, rowing. A dog rowing beats a dog standing near a pear.
The impossible beats the sensible
The brain flags anything that breaks the rules of reality. A house that is also a casserole can't be mistaken for any other house you've seen.
Funny sticks
A joke is a surprise with a small emotional hit attached, and both mark a memory as worth keeping.
Rude sticks hardest
The more outrageous or slightly inappropriate an image is, the more firmly it tends to lodge. Laughter and mild embarrassment engage the brain's emotional centres, including the amygdala, which helps regulate how memories are consolidated.
So when a card is rude, dark or undignified, that is the method working. Abuelo (grandfather) sounds like "a boil, oh!", and the image is a grandfather sitting in a pot of boiling water, saying exactly that. You won't forget it, which is the whole point. Doctors have known this for generations: anatomy students still learn the nerves and bones from mnemonics that can't be printed on a hospital wall.
Where it stops working
This part matters, and most memory advice leaves it out. In 1986 McDaniel and Einstein found that bizarre images improved recall only when they were mixed with ordinary ones. In lists that were all bizarre, or all ordinary, the advantage disappeared. Later work added a second limit: the benefit shows up mostly when you recall freely, and less in other kinds of test.
The reason is simple. Bizarre works because an item stands out from its neighbours. If every image is outrageous, none of them stands out. So:
- Interaction is never optional. Every image has the sound-alike and the meaning acting on each other. That's what the research backs most strongly.
- Tone varies across the deck. Some images are gentle and some are absurd.
- The rudest and most shocking images are saved for the hardest words, the ones that don't stick any other way.
And the app's test never shows the image. When it checks whether you know a word, you get the Spanish and give the meaning, or the other way round. If the drawing were on screen you'd be recognising a drawing, not recalling a word. How learning works.
Where the line is
Three separate tests, and they must not be merged:
| If the image is… | Then | Because |
|---|---|---|
| Dark, rude, painful, undignified | Use it | That's the mechanism. Taste is never a reason to reject an image. |
| Dated | Reject it | Only because nobody will get it. A mnemonic that needs a footnote has already failed. |
| A stereotype | Reject it | Is the joke doing the work, or the group? Swap in a different group: if the gag disappears, the group was the gag. |
Separately from all of that, every image on the community site passes an automated safety scan before anyone sees it. The scan looks for nudity, sexual content and violence. It lets cartoon slapstick through on purpose, and cartoon weapons too: ganso (goose) sounds like "guns", so the goose is armed. What a machine can't reliably judge (hate, stereotypes, anything involving a child) is caught by moderators and by members' reports, and none of it stays up. No vote can change either. Reports, penalties and how they lift.
Make your own
You can do this for any word, in any language, without the app:
- Say the word out loud. Slowly. Listen for English hiding in it.
- Catch a sound-alike for the start of the word, or most of it. It doesn't have to be exact; it has to be close enough to lead you back.
- Make it something you can see. If it's only a sound, push it to a nearby thing: lee → flea.
- Make it interact with the meaning. Same object, or one doing something to the other.
- Make it move, and make it too much. Bigger, louder, messier, ruder, especially for a word that keeps slipping.
- Run it once, eyes closed, for five seconds.
- Test yourself later from the word, not the image. If the image comes back and brings the meaning with it, it works. If not, the image needs to do more.
The same two moves (make it an image, make it do something) work on shopping lists, speeches, numbers and names. Remember anything has a tutorial for each.
Read the research
- Raugh, M. R. & Atkinson, R. C. (1975). A mnemonic method for learning a second-language vocabulary. Journal of Educational Psychology, 67(1), 1–16. The foundational paper on the keyword method, tested on Spanish.
- McDaniel, M. A. & Einstein, G. O. (1986). Bizarre imagery as an effective memory aid: the importance of distinctiveness. Journal of Experimental Psychology: Learning, Memory, and Cognition, 12(1), 54–65. Why bizarre works, and why it stops working when everything is bizarre.
- Why anatomy's naughtiest mnemonics work so well. The Conversation, 2026. Rudeness, distinctiveness and the emotional side of memory, in plain language.
For books, not papers, see books worth reading. Start with Moonwalking with Einstein.