Rebuilding kanji from the radical up

Learners who quit Japanese almost never quit because of grammar. They quit somewhere around the six hundredth kanji, when memory stops paying interest and every new character feels like the last one seen through fog.

The problem is the unit. If a character is a picture to be memorised whole, then two thousand characters are two thousand separate acts of memory. If a character is an arrangement of about two hundred components, then most of the work is recombination, and recombination is something the mind does cheaply.

Recognition is a system, not a feat of memory.

So we spend the first month on parts rather than characters: water, hand, thread, roof, speech. I ask students to guess meanings of characters they have never seen. They are often wrong, and being wrong is the point - the guess forces them to look at the structure instead of the silhouette.

Readings still need drilling; I will not pretend otherwise. But we drill them in families. Characters that share a phonetic component are learned together, so the sound is attached to a shape rather than to a list, and a new word gives up half its pronunciation before it is looked up.

After nine years of teaching, the difference I see is not speed. It is what happens on an unfamiliar page. A student who memorised characters stops. A student who learned components makes a reasonable guess, keeps reading, and checks later - which is, more or less, what reading in a second language is.