A comprehension check needs an observable answer. A teacher who asks, "Do you understand?" usually receives a nod, a quiet "yes," or no visible response. None of those reactions shows whether a learner can follow the instruction, distinguish two forms, or use the new phrase. For a teacher planning a 90-minute English lesson with fourteen adults and a pair task waiting on the next slide, the better check asks for a small performance that reveals the exact knowledge the next activity requires. This article is for a teacher working with an A2 or B1 class in Taiwan, where learners may be polite, alert, and reluctant to announce confusion in front of classmates. The aim is modest: obtain usable evidence before the class moves on.
Why yes feels easy
Yes-or-no checks are quick. They also let a teacher close an explanation without changing the lesson plan. The social shape of the question causes trouble. "Do you understand?" makes agreement the easy response, especially when the teacher has already turned towards the next slide or begun handing out a worksheet. A learner may understand the words in the question and still lack the language needed for the task. The problem becomes sharper in an English-only lesson. A beginner may need clarification yet have no ready sentence for requesting it. Silence can mean comprehension, uncertainty, deference, or an attempt to work out what comes next. The teacher sees one behaviour with at least four possible causes.
One nod is weak evidence.
What classroom evidence shows
Hansun Zhang Waring analysed 28 hours of interaction in one adult ESL classroom. Her conversation analysis examined questions such as "Do you have any questions?" in the turns where they actually occurred. These questions sometimes checked understanding, and sometimes worked as signals that an activity was ending. Learners treated them accordingly. A fast "no" could help the class proceed without demonstrating what anyone had understood. The study describes interaction in one setting, so it cannot supply a universal effect size. It does show why the same sentence performs different jobs depending on timing, gaze, and the teacher's next move.
Taiwanese evidence points in the same practical direction. Shin-Mei Kao and Wan-Chi Weng studied 12 recorded lessons taught by four teachers in a Taiwanese cram school. They coded eight question types across high- and low-level classes. The learners' proficiency level was associated with the kinds of questions teachers used. The teacher's first language showed no such association. Classes dominated by instructional questions gave learners less room for active interaction. The study was observational and involved only four teachers, which limits any causal claim. Its local value lies in showing that an English-only policy, by itself, did not create communicative classroom talk.
Evidence from another teaching setting supplies a useful comparison. Michelene Chi and colleagues analysed 44 one-to-one tutoring sessions on research-methods topics. Students' answers to questions such as "Do you understand?" were misleading indicators of achievement, while the quality of their answers to content questions gave tutors better evidence. Those undergraduates were learning research methods rather than English. The common mechanism is still relevant: a claim of understanding contains less information than a sample of the understanding itself. The result concerns evidence quality across teaching, a narrower claim than language acquisition.
Make the answer visible
A useful comprehension check copies one mental step from the coming task. If learners will sort requests by politeness, give them two lines and ask which one fits an email to a supervisor. If they will follow route instructions, ask one learner to trace "past the bank, then the second left" on a simple map. If they will use the present perfect to connect past events with now, offer two short situations and ask which one still matters at 3:17 p.m. The prompt should expose one choice without rehearsing the full activity for them.
The response can stay small:
- Learners point to one of four pictures.
- Pairs rewrite a single instruction in their own words.
- One student demonstrates the first step while a partner checks it.
- Everyone types a five-word example in the chat at the same time.
These assessment samples belong inside the teaching sequence. Each answer tells the teacher what to do next. If eleven of fourteen learners choose the wrong picture, another explanation may help. If three learners hesitate over the same word, a brief lexical check may be enough. If one pair has misunderstood the procedure while the language is secure, the teacher can fix the instruction and start. A result should change the teacher's next move, even if the change amounts to releasing the class into practice. Record no more than the decision it produced.
Match the check to the lesson
Different lesson goals require different evidence because recognition tasks can use pointing or sorting, production goals need a spoken or written sample, and a pronunciation check needs sound that the teacher or partner can actually hear. Asking learners to identify the stressed syllable in photographer cannot show whether they can produce its four syllables under time pressure. The reverse also holds: one accurate repetition may come from imitation, with no evidence that the learner hears the contrast in a new voice. I would rather use one precise check than a chain of five broad questions. Precision protects class time. Before a speaking task, the teacher might check the purpose, the information gap, and the stopping rule in three different ways. A 23-word instruction can then replace two minutes of rescue once every pair has started the wrong task. This sits beside wait time. Pausing after a question gives learners room to assemble an answer, while the comprehension check determines whether that answer exposes the needed knowledge. More silence cannot repair a question whose only available answers are yes and no.
The evidence should fit the decision.
Where a yes-or-no check fits
Yes-or-no questions still have ordinary classroom uses. "Can everyone see slide 17?" checks access. "Have you submitted the file?" checks completion. "Would you like another example?" offers a choice. The answer directly matches the information requested, and the learner can answer without admitting an academic weakness. Public demonstration also has a boundary. Asking one learner to perform in front of the class may measure confidence along with understanding. Simultaneous responses, private mini-whiteboards, chat messages, or pair rehearsal reduce that extra pressure. A teacher also needs to avoid checking every sentence. Constant interruption can break the meaning of a story or discussion into crumbs.
Try one check tomorrow
Choose the point in tomorrow's lesson where a misunderstanding would waste the next ten minutes, then write one prompt that makes the needed knowledge visible before pencils, microphones, or breakout rooms begin moving. Decide in advance what response will trigger a second example, a shorter instruction, or immediate release into practice. Keep "Do you understand?" for a different job. At 3:17 p.m., the map, sentence, sound, or first completed step will tell you more than fourteen nods.