A learner can answer eight comprehension questions by hunting for repeated words and still have a weak account of the passage. Writing one defensible question demands a different kind of reading. The learner must decide what matters, locate evidence, phrase the prompt, and predict an answer that the text can support. For a student preparing for the學測, that work exposes shallow reading before another score does.
The classroom claim is specific: after reading, learners should write and defend a small number of questions before they see the teacher's full question set. The activity needs modelling and criteria. “Write any question” produces spelling checks, trivia, and prompts whose answers sit in the same sentence.
A question leaves evidence
When a teacher asks the questions, the learner's thinking stays partly hidden. A correct option could come from understanding, elimination, or a lucky match between wording in the passage and wording in the answer. A learner-written question leaves a trail. Its focus shows what the learner treated as important. The answer key shows the evidence they selected. The distractors reveal which distinctions they noticed.
Khansir and Dashti studied question-generation instruction with 120 Iranian secondary EFL learners. Two experimental groups received ten weeks of instruction, and their reading-comprehension development exceeded the comparison groups. The context, materials, and assessment were specific, so the result supports taught question generation in that setting rather than a universal recipe.
Another experiment examined 98 adult EFL learners across several reading conditions. The group trained to use Questioning the Author, a method that probes what an author is trying to say and how well the text says it, achieved the strongest scores. The study of learner-generated questions and peer discussion also found advantages for the other supported reading groups over the individual control condition. Question writing came with training and talk. That combination matters.
Taiwan-based work offers a closer classroom reference. Yeh and Lai used an online system with organisation, composition, and peer-assessment modules among non-English-major university students. Their process study of online question generation linked stronger reading progress with particular patterns of organising, writing, and reviewing questions. It involved detailed traces from 19 participants inside a wider cohort, which is useful for examining process and too small for sweeping claims.
Teach two question moves
Start with two types. More labels create classification homework.
An evidence question asks for a claim that can be answered from two separated parts of the text. A relation question asks how a cause, contrast, example, or pronoun connects ideas. Both force the writer to look beyond a copied phrase.
Take a 640-word article about Taipei's YouBike system. A weak question asks, “When was YouBike introduced?” A better evidence question asks, “Which two changes made short trips easier for occasional riders?” A relation question might ask, “Why does the author mention rain immediately after discussing station density?” The answer needs a sentence reference and a short explanation.
Use one teacher model first. Show the paragraph, draft a poor question aloud, and revise it against four criteria:
- The text contains enough evidence for one best answer.
- The wording differs from the exact sentence that carries the answer.
- Answering requires at least two pieces of information or one named relation.
- A classmate can explain why an alternative answer fails.
Four criteria are plenty. Students then write one question in pairs and exchange it with another pair. The receiving pair answers, underlines the evidence, and marks any ambiguity. The writers revise the question once. Only then does the teacher display two or three selected examples for whole-class discussion.
The revision is where much of the reading happens, because an ambiguous answer forces the writers back into the passage to decide whether the problem lies in their wording, their evidence, or their interpretation. Keep both versions on the page; a teacher can learn more from the changed noun, added scope word, or removed distractor than from a polished final question alone.
Distractors deserve attention
Multiple-choice distractors often turn the task into accidental vocabulary testing. One option uses an unknown word, one contradicts common sense, and one copies the passage. The answer becomes visible without careful reading.
Ask writers to base each wrong option on a traceable reading error. One distractor may combine a cause from paragraph two with an outcome from paragraph four. Another may assign a view to the wrong person. A third may reverse a comparison. This step makes the writer model how a rushed reader could go wrong.
For learners around A2, open-response questions may be easier to compose than four balanced options. Let them write the question, a two-sentence answer, and the line numbers used. At B2, requiring two credible distractors can sharpen attention to scope words such as “mainly”, “some”, and “only”. Match the format to language level.
Choc's article on making comprehension checks produce an observable answer explains why “Do you understand?” gives a teacher little information. Learner-generated questions take the same principle further: the question itself becomes evidence of what the reader understood.
Keep scoring light
If every question receives a grammar mark, cautious learners will write the shortest possible sentence. Score the reading work separately from surface accuracy. A two-point check is enough: one point for a defensible answer, one for evidence that crosses a sentence boundary or identifies a relationship. The teacher can reformulate awkward English after the reasoning is visible.
There are clear limits. Question writing consumes time, and a poorly prepared class can circulate faulty interpretations. Difficult passages may overload a learner who is still decoding many words. In that case, give the answer and ask pairs to write the question, or provide the opening stem: “What explains…”, “How does paragraph 3 change…”, or “Which detail supports…”. Remove the stem in a later lesson.
I would use the method once a week, not after every text. Keep the final product small: two questions per pair, one revision, one chosen question added to the class bank. On Thursday, return a 517-word passage without the old worksheet. Ask each pair to produce one evidence question that another pair cannot answer by copying a single sentence.