A listening question keeps its information in the sound. The student cannot skim a recording or search it, so each decision about the audio changes the test: how often it plays, when the question appears, how long it runs. The first decision matters most.
EdProcto has a shape built on it. The listen once template plays the clip a single time, asks for one option and marks the answer at once. This guide covers when one play is right, when a second is fairer, how to write the question, and dictation.
One caution. Nearly every study below is of people listening in a second language. A music teacher testing an interval can borrow the reasoning, but the figures were not measured there.
Should a listening test play the recording once or twice?
Play it once when a single hearing is the real task, the clip is short and the question asks for one thing. Play it twice when the recording is long, the class is new to the language, or nerves would otherwise decide the score. A second play changes more than the difficulty.
It does make the test easier, modestly. Holzknecht and Harding (2024) gave 306 Austrian secondary students four tasks from past school-leaving papers, each on a recording of about three to four minutes. Scores were higher on all four when the recording played twice, and every effect was small (Cohen's d from 0.11 to 0.36). Jones (n.d.), reviewing earlier studies for Pearson, reports one unpublished dissertation, with 62 test takers, that put once-heard tasks 10% to 14% harder.
What a second play changes besides the score
It changes how students listen from the first second. In the Austrian study, students who knew a second play was coming left 31.6% of the multiple-choice answers blank after the first, against 0.7% for students who knew there was only one. Told they have one chance, people commit.
It changes how they feel: with two plays the same students reported less anxiety (d = 0.35) and slightly less reliance on test tactics.
And it costs time. Jones (n.d.) notes that repeating every recording nearly doubles the length of a test, so a once-heard test fits in more questions, and more questions make a more reliable score.
Where the studies disagree
They disagree about which version measures better. Ruhm et al. (2016) tested 1,266 eighth-grade learners of English and concluded that the results favoured listening once; we could read only the abstract, which does not say in what way. Holzknecht (2019), in the thesis behind the Austrian study, found questions more reliable and better at separating candidates with two plays. The dissertation in Jones's review found the reverse on the second point.
So a second play raises scores a little and calms the room, and whether it sharpens the measurement is unsettled. Holzknecht and Harding (2024) advise choosing by purpose: once where time is short or where catching it first time is the skill, twice for ordinary classroom assessment, or some of each. They note one limit: their students were used to two plays at school.
| Play it once when | Play it twice when |
|---|---|
| Real life gives one hearing: an announcement, a phone call | Real life allows a replay: a recorded lecture, a podcast |
| The quiz counts and nothing should be left to copy | The class is anxious, or the quiz is practice |
Should students see the question before the clip plays?
Yes, tell them what to listen for, and accept that it makes the question easier. Koyama et al. (2016) gave 206 Japanese university students a video-based listening test, scored out of 41, in three versions. Students who previewed the questions and options averaged 31.27, those who previewed the questions alone 30.43, and those who saw nothing in advance 28.13.
Preview also changes what is measured. The same authors suggest no preview when the aim is general comprehension, since a student who has read the options can listen for matching words. Sherman (1997) tried four arrangements with 78 learners; the one with the most correct answers put the questions between two hearings. She concluded that previewing helps less than students believe.
For a clip that plays once, the fair middle is one line of context and the task, read first ("A station announcement. Listen for the platform."), and short options.
How long should a clip be?
Long enough to carry one idea and no longer; the research gives no number. Bloomfield et al. (2011) reviewed the studies of passage length and found only weak evidence that length by itself makes listening harder. What the length is made of matters more: faster speech hurts second-language listeners, although slowing it down does not reliably help, and ideas repeated inside the passage help.
What length does is load memory. He and Jiang (2020), reviewing 87 studies of listening assessment, describe one in which questions on specific details drew more on working memory than questions on the main idea. Add a single play, and a long recording with a detail question at the end measures how much a student can hold.
Our working rule, which is practice and not a finding: one clip, one question, one idea, and for a single play usually under half a minute. Cut a three-minute recording into parts, or allow a second play.
What should a listening question ask, and in which shape?
Ask for the gist or a detail, one to a clip, and choose the shape by how much help the page should give. Gist (who is speaking, what they want) survives a missed word. Detail (the time, the price, the name) is the purest test of hearing and the hardest on memory. Keep inference for recordings that force it: with one hearing there is no second chance to test an interpretation.
| Shape | The student | Marked | Good for |
|---|---|---|---|
| Choice after the clip | picks one option | by rule, at once | gist and detail: EdProcto's listen once |
| A typed answer | writes what they understood | against your rubric | explanation: listen and answer |
| Dictation | types what was said | by rule, against accepted versions | every word: dictation |
He and Jiang (2020) report that test takers did better on multiple-choice listening questions than on open ones, because options cue recall. Nor are options read equally. Holzknecht et al. (2021) tracked the eyes of 30 test takers on four-option listening questions: the second option got 30% less viewing time than the first, the third 56% less and the fourth 75% less. Across 200 live test questions, each answered by about 6,000 candidates, those with the key in fourth place were significantly harder than those with it in first. The authors drew two lessons: three options suit listening better than four, and shuffling the options for each candidate makes one question harder for some than for others.
How do you write a fair play-once question?
Write it so that one attentive hearing is enough for a student who knows the material.
- Decide the one thing the clip is for. Two facts mean two clips.
- Record it clean. One voice at a natural pace, nothing playing under it. Noise turns a listening mark into luck.
- Say that it plays once. A student expecting a replay holds back, as the blank answers above show.
- Say what to listen for before the sound starts.
- Write three short options from real mishearings. EdProcto can shuffle the options for each student; with that off, everyone meets the same order, which is fairer by the finding above and weaker against copying.
- Read the question without the audio. If good sense alone finds the answer, rewrite it.
- Build it and sit it yourself. In EdProcto, choose listen once, add the clip's address to the question, and take the quiz on a phone with ordinary earphones before the class does.
A worked example
You will hear this once. A station announcement: listen for the platform.
Clip, about eight seconds: "The 14:50 to Leeds will now leave from platform fifteen, not platform five."
Which platform does the train leave from? A. 5 B. 15 C. 50
Each wrong option is a real mishearing. A student who picks 5 caught the last number and missed "not". A student who picks 50 confused fifteen with fifty, helped by the fifty in the departure time.
Who needs a transcript, captions or a second play?
Students who are deaf or hard of hearing, students for whom one hearing tests a processing difficulty and not the subject, and anyone whose headphones fail on the day. What to give depends on what the question is for.
If hearing is the skill (a language test, an interval, a heart sound), a transcript for everyone turns it into a reading test. The web's accessibility standard allows for this: where content is a test that would be invalid as text, the text alternative need only identify it (World Wide Web Consortium, 2024). If the clip merely delivers content, such as a recorded lecture in a history quiz, give everyone the transcript or captions.
For one student with a recognised need there is no settled rule: He and Jiang (2020) note that test developers still dispute which adjustments keep a test fair without changing what it measures. Your school's policy and the law come first. Then adjust for that student only:
- A second hearing. In EdProcto the once-only rule belongs to the listen once template, so a quiz set for that student in listen and answer can ask the same thing with a typed answer.
- A text version, where hearing is not possible, sent to that student only and recorded for what it is.
- For everyone, afterwards. Paste the transcript into the question's explanation, which EdProcto shows after submitting by default.
EdProcto's "Accessibility" settings cannot stand in for the hearing a listening question tests, so the transcript and the alternative are yours to supply.
Why can a play-once clip not be pasted into a chatbot?
Because there is no text to paste. A clip that plays once leaves only the options on the page, and they mean little without the recording. EdProcto holds the rule on the student's screen with a player of one button, "Play the clip", and a line that fills: nothing on it pauses the clip or takes it back. The template has no going back to change an answer, and the key stays on the server, where the mark is made.
That is a first defence and not a promise: a second phone beside the speaker can still record a clip. For a quiz that counts, add the measures the stakes need from the integrity settings, with students told what is on before they begin.
How do you set a dictation, the strictest listening test?
Give one short sentence per clip and decide in advance every version you will accept. Dictation takes away all the help: no options, no question hinting at the answer, every word counted.
It is sometimes dismissed as a spelling exercise. Cai (2013) tested that with 367 university learners of English in China, setting partial dictation (a text with phrases missing, filled in from the recording) beside other listening tasks. The best-fitting model supported the view that partial dictation measures the same ability as tasks thought to need higher-order listening. It is one study, of the partial form.
In EdProcto's dictation shape the typed sentence is marked by rule against your accepted versions, right or wrong as a whole, so keep it short. List the variants: numbers as figures and as words, contractions both ways. Capitals, spacing and stray punctuation are ignored. A one-letter slip can be allowed; switch that off where a word's ending is the grammar being tested. The guide to gap questions covers accepted answers in detail.
How does EdProcto run a listening quiz?
EdProcto is the best tool for a play-once question because the rule is part of the template and not a request to the student. That matters for the reason the Austrian study gives: students listen differently when they know how many hearings they have, so the number has to be certain and the same for everyone.
Start in the quiz maker on the home page: paste the transcript of your recording, choose the shape, and Procto AI drafts questions for you to edit. Listen once sits among 111 templates. The other ready listening shapes are listen and answer, read by you or, when you ask, by Procto AI against your rubric; dictation; recognise the sound; and match the sound, with pictures as options for children who cannot read yet. Pick the clip, whose options are clips, and compare two clips are marked by rule.
The Free plan has the core templates, marking by rule and results question by question. Tutor opens all 111 templates. Educator adds the full analysis, which shows how many students took each wrong option: an option nobody picks is not a mishearing anyone makes. The pricing page has the detail.
Quick questions
How many times should a listening test recording be played?
Once for a short clip with one question; twice for a long recording or an anxious class. Holzknecht and Harding (2024) found that a second play raised scores slightly and lowered anxiety, and advise choosing by the purpose of the test.
Should students read the questions before they listen?
Usually yes. Koyama et al. (2016) found that previewing made a listening test easier, and Sherman (1997) that it helped less than students thought. For a clip that plays once, say what to listen for.
Can a student replay a listen once question?
In EdProcto the listen once template gives the student one button, "Play the clip", and a line that fills as it runs: the page has no pause and no going back, and when the clip ends it reads "The clip has played its one time." A student with a recognised need can be set the question in another shape.
Is dictation a real test of listening comprehension?
For the partial form, one study says yes: Cai (2013) found it measured the same ability as tasks thought to need higher-order listening. Whole-sentence dictation tests spelling too, so say whether spelling counts.
References
- Bloomfield, A. N., Wayland, S. C., Blodgett, A., & Linck, J. (2011). Factors related to passage length: Implications for second language listening comprehension. Proceedings of the Annual Meeting of the Cognitive Science Society, 33, 2317–2322. https://escholarship.org/uc/item/2pr8m0xc
- Cai, H. (2013). Partial dictation as a measure of EFL listening proficiency: Evidence from confirmatory factor analysis. Language Testing, 30(2), 177–199. https://doi.org/10.1177/0265532212456833
- He, L., & Jiang, Z. (2020). Assessing second language listening over the past twenty years: A review within the socio-cognitive framework. Frontiers in Psychology, 11, Article 2123. https://doi.org/10.3389/fpsyg.2020.02123
- Holzknecht, F. (2019). Double play in listening assessment [Doctoral thesis, Lancaster University]. https://doi.org/10.17635/lancaster/thesis/812
- Holzknecht, F., & Harding, L. (2024). Repeating the listening text: Effects on listener performance, metacognitive strategy use, and anxiety. TESOL Quarterly, 58(1), 451–478. https://doi.org/10.1002/tesq.3249
- Holzknecht, F., McCray, G., Eberharter, K., Kremmel, B., Zehentner, M., Spiby, R., & Dunlea, J. (2021). The effect of response order on candidate viewing behaviour and item difficulty in a multiple-choice listening test. Language Testing, 38(1), 41–61. https://doi.org/10.1177/0265532220917316
- Jones, G. (n.d.). Once or twice? A critical review of current literature on the question how many times the audio recording should be played in listening comprehension testing items [Research summary]. Pearson. https://www.pearsonpte.com/content/dam/ELL/pte/pearsonpte/general/Listen-Once-Or-Twice--Glyn-Jones.pdf
- Koyama, D., Sun, A., & Ockey, G. J. (2016). The effects of item preview on video-based multiple-choice listening assessments. Language Learning & Technology, 20(1), 148–165. https://hdl.handle.net/10125/44450
- Ruhm, R., Leitner-Jones, C., Kulmhofer, A., Kiefer, T., Mlakar, H., & Itzlinger-Bruneforth, U. (2016). Playing the recording once or twice: Effects on listening test performances. International Journal of Listening, 30(1–2), 67–83. https://doi.org/10.1080/10904018.2015.1104252
- Sherman, J. (1997). The effect of question preview in listening comprehension tests. Language Testing, 14(2), 185–213. https://doi.org/10.1177/026553229701400204
- World Wide Web Consortium. (2024). Web Content Accessibility Guidelines (WCAG) 2.2. https://www.w3.org/TR/WCAG22/