Listening comprehension questions put the information in a recording: a voicemail, an interview, a short talk, a patient describing a pain. Written well, the questions measure listening; written badly, they measure general knowledge, memory or spelling.
EdProcto, the publisher of this guide, carries the shape as listen and answer: a clip the student may replay, then a typed answer marked against the model answer you wrote. Whether a clip should play a single time is covered in our guide to the clip that plays once.
A caution before the evidence: almost all of it comes from people listening in a second language, and we read each study in its abstract only.
What do listening comprehension questions measure?
They measure four things: the gist, a detail, an inference, and the speaker's attitude or purpose. Cambridge English (n.d.), for one, lists what the first part of its B2 First listening paper has candidates listen for: feeling, attitude, opinion, purpose, function, agreement, gist and detail.
| Target | The question asks | A stem |
|---|---|---|
| Gist | what the recording is for | "Why is the museum calling?" |
| Detail | one thing that was said | "What time will the workshop start?" |
| Inference | what follows from what was said | "How many adults should come?" |
| Attitude | what the speaker thinks or feels | "Were four adults enough? What tells you?" |
The targets are not equally hard: Shohamy and Inbar (1991) tested 150 learners of English in their final school year and found better scores on local questions, which rest on one stretch of the recording, than on global ones about the whole.
Nor are they clearly separate skills. O'Connell (2019), in a doctoral study of 255 learners, found that questions on what a speaker implies were harder but did not form a separable skill: ordinary listening questions predicted them. Use the four targets to spread your questions across the clip, and read the total, not a sub-score for each.
What does research say about replays, speech rate, question preview and note-taking?
It says that each helps less, or less simply, than teachers tend to expect.
| Decision | What the studies suggest | What to do |
|---|---|---|
| Replays | Free replay did not raise scores, and it eased nerves | Let the clip replay, as EdProcto's listen and answer does |
| Speech rate | Fast speech hurts; slower than ordinary adds little | Record at your usual pace |
| Question preview | The question helps; options shown alone hurt | Put the question beside the player |
| Note-taking | Allowing notes helps sometimes; urging them hurts | Allow notes without pressing them |
Replays the student controls
In the one direct test we found, letting students pause and replay did not raise scores. Eberharter et al. (2023) gave 139 young Austrian learners of English 15 questions heard once and 15 self-paced, where they could pause or replay as needed. Self-pacing had no statistically significant effect on scores, and did not lift students with weaker literacy in their first language; the authors think young test takers may need training to use it. Kormos et al. (2024) add that students reported less listening anxiety in the self-paced half.
A fixed repeat is different. Chang and Read (2006) tried four kinds of support on 160 college students in Taiwan: information about the topic helped most, then hearing the recording again, and teaching the vocabulary first helped least.
Speech rate
Fast speech hurts, and slowing below an ordinary pace buys little. Griffiths (1990) played graded passages of 350 to 400 words to lower-intermediate adult learners at 100, 150 and 200 words a minute. Comprehension fell significantly at the fastest rate, and the slowest did no better than the middle one; he calls the findings tentative.
Question preview
Show the question first; options shown without it can hurt. Yanagawa and Green (2008) compared three versions of a multiple-choice listening test: question and options seen before listening, the options alone, and the question alone. The options-only version produced significantly fewer correct answers, and the other two did not differ. Their analysis points to test takers matching words in the options to words they heard. A typed-answer question, read before the clip, has no options to match.
Note-taking
Allow notes, and do not require them. Hale and Courtney (1994) found that allowing notes during the short talks of the TOEFL had little effect, and that urging students to take them significantly impaired performance. Carrell et al. (2004), with 234 students, found that notes helped on talks of two and a half minutes and on arts and humanities topics, and made no difference on five-minute talks or on physical science.
How do you choose or record the clip?
Choose speech like the speech your students need to understand, in a voice they can follow, and give everyone the same recording.
- A tidy script is easier than real talk. Wagner and Toth (2014) gave 85 university learners of Spanish unscripted recordings and 86 the same texts rewritten as textbook texts. The scripted group scored significantly higher.
- Written prose read aloud is harder than talk. Shohamy and Inbar (1991) presented the same content as a news broadcast, a short lecture and a dialogue. The more a version resembled talk, the easier it was, and the broadcast was the most written-like.
- Accent counts, even a light one. Ockey and French (2016) randomly assigned 21,726 TOEFL test takers to hear one lecture from one of nine speakers. Strength of accent, and familiarity with it, affected comprehension even when the accent was quite light.
To record your own, talk from notes in a quiet room at your usual pace. With replay allowed, half a minute to a minute of speech carries three or four questions. Both are working rules, not findings. Use only a recording you made or have the right to use.
How do you write questions that need the clip?
Ask for what the recording says and an outsider could not guess. The check is quick: hand the questions to someone who has not heard the clip, and rewrite every one they answer.
Tsui and Fullilove (1998) sorted seven years of listening questions from Hong Kong's public examinations into two kinds. In one, the recording opened with an expectation that the rest overturned, so a candidate had to keep following the words; in the other a sensible guess from the opening was enough. The first kind told stronger from weaker candidates significantly better, whatever the type of question. So ask where the speaker corrects, changes or surprises.
The typed answer helps too. In'nami and Koizumi (2009) pooled 56 data sources and found multiple-choice formats easier than open-ended ones in second-language listening, by a medium to large margin. With nothing to recognise or guess from, expect lower marks than the same question would earn with options.
- Listen and list what a listener learns only from the clip.
- Give each question one target: the gist, a detail, an inference or the attitude.
- Ask where the speaker turns. After a correction, the first figure heard is the wrong answer.
- Use your own words. A question that repeats the speaker's phrase can be answered by matching sounds.
- Keep the order of the clip, so a student who replays knows where to listen.
- Write the model answer first, with what earns each point and whether spelling counts: see short answer questions and their rubric.
- Build it and sit it. In EdProcto, choose listen and answer, paste the address of the clip, and take the quiz on a phone before the class does.
A worked example: one clip, four questions and the key
Hello, this is the education office at the harbour museum, calling about your class visit on Thursday. I'm sorry to say the ten o'clock workshop is now full, so we have moved your group to half past one. The coach can still stop at the main gate, but please bring the children in by the side door on Mill Lane, because the front steps are being repaired. One more thing. You told us thirty-two children are coming. With that many, I would bring a fifth adult, to be honest. Four was a stretch last year.
The message is 95 words, about 38 seconds at 150 words a minute, the middle rate in Griffiths (1990).
| Question | Target | Model answer | Points |
|---|---|---|---|
| 1. Why is the museum calling? | Gist | To change the arrangements for Thursday's visit | 1 |
| 2. What time will the group's workshop start? | Detail | Half past one, in any form | 1 |
| 3. How many adults does the caller think should come? | Inference | Five | 1 |
| 4. Did the caller think four adults were enough last year? Give the words that tell you. | Attitude | No: four "was a stretch" | 2 |
Question 1 accepts "to move the workshop" and gives nothing for "to confirm the visit". Question 4 gives one point for the view and one for the caller's words.
Nobody who missed the message can say why a museum rang. Questions 2 and 3 sit where a guess goes wrong: "ten o'clock" and "four" are both said aloud and both are wrong answers, so a student who types either caught a number and missed the message.
How does EdProcto build, play and mark a listening question?
EdProcto is the best tool for listening comprehension questions when the recording is your own and the answer should be typed, not picked. The question sits on the page with the player, there are no options to match against the sound, and the mark is made against the model answer you wrote: the conditions the preview and format studies above point to.
- Building. In the quiz maker on the home page, paste the transcript, choose "Listen and answer", and Procto AI drafts questions, each with a model answer and a rubric, for you to check against the recording. Each drafted question arrives with "It is" set to "Audio": paste the link into "Audio clip address". The link goes to the student's browser as it stands, so it must lead straight to the sound file and open without a sign-in. Each question keeps its own address: for a set on one clip, paste the same link into each. Then fill "Model answer (what a full answer contains)" and "Rubric (what earns the points)".
- Playing. The student gets the browser's own audio player above the question, which is in view before the first play, and may pause, go back and replay at will. Leave "Plays once, no replay" off. To cap the replays, turn "Time limit for each question" to "Set": 150 in "Seconds a question" allows the 38-second message two hearings and over a minute to write. Switch off "Shuffle the questions for each student" so the set follows the recording; "Notes" at "A blank page" puts the permission on the start screen.
- Marking. At hand-in a typed answer is held for a reader and counts as zero until it is read. On the grading page you mark it beside your model answer and rubric, or, on Tutor and above, press "Ask Procto AI" and keep or change the mark it saves. The grade is made against what you wrote, so put into the model answer everything from the clip that a mark depends on.
- A detail marked by rule. For a one-phrase answer marked at hand-in, add the question in the dictation shape, which takes the same clip address and a list of accepted answers. We ran question 2 through EdProcto's marking function with "half past one", "1.30", "1:30", "13.30" and "13:30" accepted: "half pass one", "1.30 pm" and "at half past one" all came back wrong. One slip is forgiven only inside a word of five letters or more, and never where a figure is involved, so a short word such as "past" has to be exact. A time has many typed forms, which is the case for a reader.
- Afterwards and on paper. Paste the transcript into "Explanation shown to the student afterwards", which appears after submitting by default, and keep it out of "Material under the prompt (optional)", which students read with the question. On paper the question prints with answer lines, so play the clip to the room yourself.
A clip that can be replayed can also be recorded, so for a quiz that counts, use listen once or add what the stakes need from the integrity settings. Where the student should speak, see our guide to recorded spoken answers.
On Free you get ten quizzes a month, the core templates and marking by rule. Tutor opens all 111 templates and adds grading by Procto AI; Educator brings the full analysis. The pricing page sets the plans side by side.
Quick questions
How many times should students be allowed to replay a listening clip?
For practice and ordinary classwork, as often as the time allows: Eberharter et al. (2023) found that free replay did not raise scores, and Kormos et al. (2024) that it lowered anxiety. For a clip that must play a single time, EdProcto has the listen once shape.
Should students read the question before the clip plays?
Yes. Yanagawa and Green (2008) found that seeing the options alone produced fewer correct answers than seeing the question, with or without its options.
Should students take notes during a listening test?
Let them, and do not insist. Hale and Courtney (1994) found little effect from allowing notes and a significant loss from urging them, while Carrell et al. (2004) found a gain on short talks.
How does EdProcto mark a typed answer to a listening question?
In listen and answer the typed answer waits on the grading page, where you mark it against your model answer and rubric or press "Ask Procto AI" and keep or change the mark.
References
- Cambridge English. (n.d.). B2 First for Schools and B2 First exam format. Cambridge University Press & Assessment. Retrieved October 10, 2026, from https://www.cambridgeenglish.org/exams-and-tests/first/exam-format/
- Carrell, P. L., Dunkel, P. A., & Mollaun, P. (2004). The effects of notetaking, lecture length, and topic on a computer-based test of ESL listening comprehension. Applied Language Learning, 14, 83–105. https://eric.ed.gov/?id=EJ1069316
- Chang, A. C.-S., & Read, J. (2006). The effects of listening support on the listening performance of EFL learners. TESOL Quarterly, 40(2), 375–397. https://doi.org/10.2307/40264527
- Eberharter, K., Kormos, J., Guggenbichler, E., Ebner, V. S., Suzuki, S., Moser-Frötscher, D., Konrad, E., & Kremmel, B. (2023). Investigating the impact of self-pacing on the L2 listening performance of young learner candidates with differing L1 literacy skills. Language Testing, 40(4), 960–983. https://doi.org/10.1177/02655322221149642
- Griffiths, R. (1990). Speech rate and NNS comprehension: A preliminary study in time-benefit analysis. Language Learning, 40(3), 311–336. https://doi.org/10.1111/j.1467-1770.1990.tb00666.x
- Hale, G. A., & Courtney, R. (1994). The effects of note-taking on listening comprehension in the Test of English as a Foreign Language. Language Testing, 11(1), 29–47. https://doi.org/10.1177/026553229401100104
- In'nami, Y., & Koizumi, R. (2009). A meta-analysis of test format effects on reading and listening test performance: Focus on multiple-choice and open-ended formats. Language Testing, 26(2), 219–244. https://doi.org/10.1177/0265532208101006
- Kormos, J., Eberharter, K., Guggenbichler, E., Baumgartinger, S., Ebner, V., & Kremmel, B. (2024). First language literacy skills, listening strategy use, and anxiety in self-paced versus single-play L2 listening tests. Language Assessment Quarterly, 21(4–5), 303–323. https://doi.org/10.1080/15434303.2024.2435811
- Ockey, G. J., & French, R. (2016). From one to multiple accents on a test of L2 listening comprehension. Applied Linguistics, 37(5), 693–715. https://doi.org/10.1093/applin/amu060
- O'Connell, S. P. (2019). Comprehension of conversational implicature: Examining evidence of its separability as a listening subskill [Doctoral dissertation, University of Maryland]. Digital Repository at the University of Maryland. https://doi.org/10.13016/rain-xgj3
- Shohamy, E., & Inbar, O. (1991). Validation of listening comprehension tests: The effect of text and question type. Language Testing, 8(1), 23–40. https://doi.org/10.1177/026553229100800103
- Tsui, A. B. M., & Fullilove, J. (1998). Bottom-up or top-down processing as a discriminator of L2 listening performance. Applied Linguistics, 19(4), 432–451. https://doi.org/10.1093/applin/19.4.432
- Wagner, E., & Toth, P. D. (2014). Teaching and testing L2 Spanish listening using scripted vs. unscripted texts. Foreign Language Annals, 47(3), 404–422. https://doi.org/10.1111/flan.12091
- Yanagawa, K., & Green, A. (2008). To show or not to show: The effects of item stems and answer options on performance on a multiple-choice listening comprehension test. System, 36(1), 107–122. https://doi.org/10.1016/j.system.2007.12.003