Some English words refuse to stay on a vocabulary card because their sound offers a Mandarin-speaking learner no familiar handle. A keyword mnemonic supplies one temporary handle: connect the new sound to a familiar word or phrase, build a vivid image linking that cue to the meaning, then practise retrieving the English word. For a Taiwanese eighth-grader facing a stubborn set of 21 low-frequency words, the sound hook can make limited review time work harder.
Build a two-link cue
The keyword method uses two connections. The first links the unfamiliar English form to a familiar sound. The second links that sound to the meaning through an image or compact scene. Take the English word banish. A learner might hear the Mandarin phrase「搬椅」as a rough sound cue, then picture someone moving the last chair out of a room and sending it away. The phonetic match is imperfect. It only needs to prompt the target during early retrieval. The scene gives the learner somewhere to begin when the English form will not arrive unaided.
Temporary means temporary.
That example is invented to show the procedure. It has not been tested, and another learner may find it awkward. Personal associations differ. A cue that produces the word quickly earns its place; a cue that takes longer to decode than the English word should be discarded.
Taiwan evidence over four weeks
Chia-Hui Chiu and Chad Hawkins tested a combined keyword-and-retrieval procedure with 42 eighth-grade EFL learners in Taiwan. The study used 21 previously unknown English words divided into three sets. Three conditions rotated across groups: the combined method, retrieval practice, and a control condition. Productive vocabulary was tested immediately, after one week, and after four weeks.
The 2023 study found similar performance for the combined method and retrieval practice on the immediate and one-week tests. On the four-week delayed test, the combined method outperformed retrieval practice for productive recall. That delayed result is the interesting one for a learner who wants to produce English from a meaning rather than recognise it on sight.
The boundaries are narrow. The sample had 42 adolescents from Taiwan, the material contained 21 words, and the treatment lasted four weeks. A mnemonic that helps recall an isolated form has not yet taught its collocations, register, pronunciation in connected speech, or grammatical behaviour.
A second experiment with limited tests
A 2024 experiment recruited 110 Chinese university students learning English and assigned them to four conditions: restudy, retrieval practice, supplied keyword plus retrieval, or learner-generated keyword plus retrieval. Each item received only two retrieval attempts. The open-access report found that retrieval practice alone did not beat restudy under that limited schedule. Both keyword-plus-retrieval groups did, and the supplied and self-generated keyword conditions did not differ significantly. Out of 20 tested pairs, mean final recall was 6.46 in the restudy group and 4.64 in the test-only group. The supplied-keyword group reached 9.29, while the learner-generated-keyword group reached 11.39. Those are group averages from one controlled experiment, with 26 to 28 participants in each condition. The spread is large, though the small groups keep the estimates uncertain.
Direction matters.
Final testing asked learners to recall Chinese meanings from English cues. This direction measures receptive knowledge, unlike the productive test in the Taiwanese study. The two studies cannot be collapsed into one large claim. Together they show that a sound-and-image cue can support retrieval under specific, constrained conditions, in both directions of learning. There is a useful surprise in the 2024 result. Self-generation did not win. Teachers can offer a good cue when learners are stuck, then let each learner keep, change, or reject it. Choice remains with the person who must retrieve the word later.
Use the hook only where needed
Building a mnemonic for every word wastes time. Transparent compounds, familiar word families, and words that recur naturally in a text already carry useful support. Save the method for the item that has failed twice, lacks an obvious English relative, and matters enough to retrieve again.
A six-step card might look like this:
- Put the meaning or situation on the front:「把某人趕走，不准回來」.
- Try to retrieve banish before revealing anything.
- If retrieval fails, reveal the sound cue「搬椅」and the chair-removal image.
- Say banish aloud, then check its pronunciation from a reliable dictionary recording.
- Produce one short sentence that fits its grammar: “The king banished him from the city.”
- Test the card again after a delay, starting from meaning and returning to English.
Sound cues are scaffolding. The card still needs the target's pronunciation, a usable example, and delayed retrieval. Choc's article on vocabulary cards explains why a learner who wants to speak must practise the meaning-to-English direction.
Images need interaction
Placing a keyword and a meaning side by side may leave two separate facts. The image should make them act on each other. A chair merely placed beside the word banish is weak. Someone carrying every chair out while a guard closes the gate connects the sound cue with removal. Odd images can be memorable, though oddness alone has no educational virtue. The scene should be instantly interpretable. Five characters, three locations, and a joke that needs explaining produce a puzzle rather than a cue. One clear action is easier to replay during a timed recall attempt.
Keep the cue temporary.
Pronunciation deserves a check because Mandarin-based cues can distort English sounds.「搬椅」does not reproduce the vowels, stress, or final consonants of banish. Say the cue once to reach the word, then listen to the English recording and repeat the English form. The cue should fade as the target becomes familiar.
Know what the mnemonic has not taught
Recalling banish for「驅逐」does not show that the learner can choose between banish, expel, exile, and deport. It does not teach the pattern banish someone from a place. It says nothing about frequency. The method solves an early form-meaning problem. After successful recall on two separated days, move the word into a sentence-completion item, a short reading, or a speaking prompt. If the learner meets banish in a fantasy novel, the text can begin replacing the artificial cue with richer associations. Repeated context can now carry work that the invented image handled at first. The chair may quietly leave.
Keywords also fit some words badly. Abstract items with several senses, very short function words, and phrases whose meaning depends on context often resist a single image. Use a short dialogue, contrast, or repeated encounter instead.
Choose five cards that have failed at least twice. Build sound hooks for only two of them, with one clear image each. Test from Mandarin meaning to spoken English after 47 minutes, again tomorrow, and once more four days later. Retire any hook that slows the answer down.