A learner chooses postpone from four options and gets a green tick. Ten minutes later, the same learner writes, “Can we move the meeting to Friday?” because postpone never comes to mind. The quiz recorded recognition. The message demanded recall. A useful vocabulary check should sample both routes for the same word. For a B1 office worker in Taipei who needs English for scheduling calls, one multiple-choice item can confirm a form-meaning link. A second item, with the answer absent, shows whether the word is available when a real sentence needs it.
One word has several test scores
Batia Laufer and Zahava Goldstein tested four degrees of form-meaning knowledge with 435 learners of English. Their 2004 study arranged the tasks from passive recognition through active recall. The predicted difficulty order appeared at every frequency level, and passive recall was the best predictor of classroom language performance among their four measures. A learner's score also grew at a different rate in each mode. That result gives teachers a warning about the word “known”. A correct recognition answer establishes something real, though narrow. It does not establish that the learner can retrieve the English form from a meaning, spell it in a message, fit it into a clause, or combine it with a natural neighbour such as postpone a meeting. Each demand needs evidence.
The test format changes the result.
Benjamin Kremmel and Norbert Schmitt examined four common form-meaning formats in two studies. Their 2016 paper found that scores from those formats gave poor grounds for claiming that learners could employ the words in reading or knew derivatives and collocations. The authors were careful about the boundary: these tests may measure a form-meaning link, yet the score interpretation often reaches further than the item supports.
Build a two-route check
Take six target words from one lesson. Test each word once through recognition and once through recall or constrained use. Separate the two encounters so the first item does not hand the answer to the second. A twelve-item check can fit into 8 minutes. For postpone, the recognition item might show a short email and ask which option means “move to a later time”. The recall item can appear four questions later: “The client cannot join on Tuesday. Complete the reply with one word: Could we ___ the call until Friday?” Accept postpone and any other answer that genuinely satisfies the task, then score the target word separately from the sentence.
Now the teacher has two marks:
| Route | What the learner sees | What a correct answer supports |
|---|---|---|
| Recognition | English word plus choices | The learner can distinguish a meaning in this context |
| Recall | Meaning or situation, with no target form | The learner can retrieve an English form for this context |
One correct cloze answer offers some evidence of retrieval and grammatical fit.
It still says little about pronunciation, rapid speech, another meaning of the word, or a collocation that the prompt never required.
Treat disagreement as useful data
The two results can disagree in four ways. Two correct answers suggest that the form-meaning link is available in both directions under these conditions. Recognition correct and recall wrong points to an item that still needs retrieval practice. Recognition wrong and recall correct may reveal a confusing distractor, an unfamiliar sense, or lucky production from the sentence frame. Two wrong answers send the item back to teaching. This is diagnostic evidence, not a label for the learner. I'd rather keep those two small marks than convert them into one percentage. A score of 9 out of 12 hides whether three words failed twice or six words failed on recall alone. The next lesson differs sharply. The first pattern calls for revisiting meanings. The second calls for opportunities to produce forms from meanings. Our article on giving a vocabulary card a way back to English shows how to practise the weaker direction. Put the Mandarin meaning or a picture on the prompt side, say or write the English form before turning the card, and keep recognition cards for items whose meaning is still unstable.
Add use only when the lesson taught use
A teacher may want a third item asking for a sentence. That item earns its place after the lesson has covered a grammatical pattern or collocation. Asking for free sentence writing immediately after presenting a word can load the task with content, grammar and spelling decisions. The resulting error becomes hard to interpret. Use a constrained prompt. After teaching postpone, postponed and postponement, give: “Write one sentence explaining why the launch moved from 14 May to 27 May. Use one form of POSTPONE.” Score word form and sentence meaning in separate columns. An accurate postponed with a missing article shows a different teaching need from postpone used to mean cancel.
Kremmel and Schmitt's study offers a useful restraint here. A form-meaning test cannot carry a claim about derivatives or collocations that it never sampled. The reverse holds too. One successful sentence cannot certify every sense and pattern of the word.
Keep the quiz teachable
Twelve items already create enough information for six words. Mixing 37 target words into one Friday test makes the marking tidy and the follow-up vague. Use the result to choose the next action while the words still belong to a live unit. For the hypothetical office worker, suppose recognition is correct for all six words, while recall fails for postpone, attach and confirm. Monday's review can start from three situations: move a call to 27 May, add a receipt to an expense email, and tell a supplier that an order has been received. Hide the target forms. Ask for the message, reveal the answer, and return to the same three prompts near the end of class in a changed order.
Friday's green tick remains valid. It simply keeps its proper size: evidence that one learner recognised one meaning of postpone in one item.