The word "break" has many meanings. You can break a glass, break a promise, or take a break. Learning all of them one by one is hard.

But they share one core image. Something whole is cut or stopped. A 2003 study found that learners guessed new meanings better when they knew the core meaning. But a study with Japanese high school students found it worked about as well as learning translations. So use both.

Words like "break" or "run" are polysemous, which means they have many related meanings. Instead of memorizing each meaning separately, learners can start from the core meaning and see how the others developed from it. In 2003, Verspoor and Lowie asked Dutch learners of English to guess figurative meanings of such words from context. Learners who were given the core meaning as a clue guessed best and also remembered the meanings better later.

However, results in Japanese classrooms were more modest. In a 2007 study, 58 high school students were taught "break" and "over" for 20 minutes using either image diagrams or translations. The image-based teaching worked about as well as translations and was clearly better on only one test. Core images help, but they are not magic.

Polysemous words such as "break," whose dictionary entries list senses ranging from "smash" to "pause" to "announce," are a notorious burden for learners. Cognitive linguists argue that these senses are not random but radiate from a core meaning, roughly "a whole being split or interrupted by force." Verspoor and Lowie tested whether this view has practical value.

Dutch learners of English guessed figurative senses of polysemous words from context, supported by either the core sense, another figurative sense, or no cue. Those given the core sense produced the most accurate guesses and retained the meanings better on a delayed test, which the authors attributed to more precise elaboration. Classroom evidence from Japan, however, is less dramatic. Morimoto and Loewen taught "break" and "over" to 58 high school students through image-schema diagrams, translation, or no instruction.

The image-schema group performed about as well as the translation group, outperforming it on only one measure. Because the instruction lasted only 20 minutes, the result may underestimate the approach, but it also warns against assuming it is superior. A balanced strategy is to use the core image to organize related senses while learning distant senses through example sentences.