Why pictures

Picture the Words is built on a small number of findings about how vocabulary actually sticks. This page sets out what those are, and what the product does with them — so you can judge the method rather than take it on faith.

Two memory traces beat one

Allan Paivio's dual coding theory holds that the mind stores verbal and visual information in separate but connected systems. A word learned with an image is encoded twice, and either trace can retrieve the other. This is why a picture attached to a word is not decoration: it is a second route back to the same memory.

Every concept here is built around a purpose-drawn illustration rather than a stock photo, because the picture has to show the concept unambiguously — not merely accompany it.

Translation pairs teach the wrong thing

Learning ringo = apple teaches a mapping between two words. Learning りんご alongside a picture of an apple teaches a mapping between a word and a thing. The first has to be translated in your head every time you meet it; the second doesn't.

This is also why English is not designed to be the permanent hub of this product. The next pairs connect directly to one another without passing through English, because routing every language through English rebuilds exactly the intermediary step the method is trying to remove. This release is Japanese–English; the architecture behind it is not English-centred.

Recall beats recognition, and timing beats repetition

Retrieving a word from memory strengthens it far more than seeing it again does — the testing effect. And a review scheduled just before you would have forgotten does more work than ten reviews crammed together, which is the spacing effect.

The review system implements SM-2, the algorithm behind SuperMemo and Anki. Each card carries its own ease factor and interval; answering well pushes the next review further out, answering poorly pulls it back in. Cards you find hard come round more often, cards you know get out of your way.

Words met in the wild are worth more

Vocabulary encountered while reading something you actually wanted to read carries context, motivation and a memory of where you met it. Vocabulary from a list carries none of those.

That is the whole reason the browser extension exists. Looking a word up where you found it and saving it in one action means your review deck is built from your own reading rather than from someone else's idea of what you should know.

Sound is part of the word

A word you have only ever seen is a word you cannot use. Every concept carries audio for both the word and its meaning, in all three languages, so pronunciation is learned at the same time as the form rather than repaired later.

How the content is made

The concept list is ordered by frequency, so early cards are words you will actually meet. Illustrations are generated against a fixed style reference and reviewed by hand. Japanese is checked by the creator, who holds JLPT N1 and has worked professionally in Japanese–English translation. Each further language is reviewed by a native speaker before it ships — which is why Hindi, though already built, is not in this release. Dictionary data draws in part on JMdict/EDICT, credited in the footer of every page.

Where this is still incomplete. Example sentences are not in yet, which is the largest gap between the method described here and the product as it stands. Frequency ranking is coarse. Pitch accent is not marked. These are named on the What's included page rather than glossed over.