Oral exam questions reach something no written paper does: whether the student can say it. An explanation given aloud, without notes, shows whether the reasoning belongs to the person giving it. The obstacle is the timetable: a viva for thirty students means thirty appointments.
A recorded spoken answer keeps most of the oral exam and drops the appointments. Every student answers the same question into a microphone, and you listen when it suits you. EdProcto carries the shape as its spoken answer template: the recording is made in the browser, kept with the attempt, and marked by you against your rubric.
Most studies below are of live oral exams or formal language tests, several from a single course. Short recordings in a class quiz have been studied far less.
What does a spoken answer show that a written one cannot?
A spoken answer shows three things: that a student can speak a language, that they can explain an idea without notes, and that the reasoning comes out in their own voice and words. Writing is a stand-in for the first and can be borrowed for the other two.
Joughin (1998), surveying the literature on oral assessment in higher education, identified six dimensions along which oral tasks differ: the kind of content, the interaction between examiner and student, authenticity, structure, the examiners, and orality, meaning how far the task rests on speech alone. A recording sits at the structured end with no interaction: each student meets an identical prompt, and nobody asks a follow-up. So the prompt has to ask for the reason itself.
The shape measures badly where speech is beside the point: a list of dates is quicker to write and to mark. Keep the microphone for questions where saying it is the skill, or where you need to hear the thinking.
What did research on oral exams find about scores, nerves and fairness?
In one direct comparison, students scored higher when they spoke, felt more nervous, and no group was shown to lose out. Huxham et al. (2012) split 99 biology students at random into an oral group and a written group and set both the same questions; a second cohort of 29 answered in both forms. In each cohort the mean mark was significantly higher in the oral assessment, and the authors found no evidence that particular groups were disadvantaged. Students tended to be more nervous before an oral, and many still judged it the more useful. It is a single study.
Iannone and Simpson (2015) report the views of first-year mathematics students in one module: despite worries about anxiety and fairness, they saw an oral assessment as pushing them towards understanding.
The warnings come from medicine. Memon et al. (2010), reviewing oral assessment in postgraduate medical examinations, reported enough concern about validity, reliability and fairness that some countries had moved away from it, and closed with 15 conditions under which it can be defended. We read the abstract only, which does not list them. The lesson: an unstructured oral marked on impression is the weak form.
Does a recording lose anything a live oral has?
It loses the conversation, and some students mind. Qian (2009) analysed the reactions of university students in Hong Kong who had just sat a face-to-face speaking test and a person-to-machine one. Many had no preference, but those who strongly favoured the face-to-face test far outnumbered those who strongly favoured the machine, mainly because nobody was there to interact with. Describe the recording in advance, and give one practice question that does not count.
How do you write a prompt for a recorded answer?
Write one task, state the length, and settle the rules on thinking time and retakes before anyone presses record.
- Set one task. "Explain why" joined to "give an example" is two answers needing two rubrics.
- State the length. Sixty to ninety seconds suits a single explanation. That is our working rule, not a research finding: less gives a fragment, and three minutes for every student is an evening of listening.
- Say whether there is preparation time. Wigglesworth and Elder (2010), studying the IELTS oral, gave 90 candidates speaking tasks with no planning time, one minute or two. Neither the scores nor an analysis of the speech showed a significant difference, yet the authors argued that one minute should still be considered, for fairness. A long pause invites a script.
- Say whether a second take is allowed. Allow one where the content is the skill. Refuse it where fluency is the skill, because a rehearsed third attempt is a different test.
- Say what is marked and who will listen. Name the criteria in a line.
- Build it and answer it yourself. In EdProcto, choose spoken answer, put the time in the question and the criteria in the rubric, then record your own reply on a phone. If you overrun, so will they.
A worked example
You have 30 seconds to think, then 90 seconds to speak. One recording, no second take. A frost destroys part of this year's coffee harvest. Using the supply and demand diagram from week 3, explain what happens to the price of coffee and to the quantity sold, and say which curve moves and why. Marked for accuracy, reasoning and clarity. Your tutor is the only listener.
| Criterion | 2 marks | 1 mark | 0 marks |
|---|---|---|---|
| Accuracy | Supply shifts left; price rises, quantity falls | One outcome right, or the curve not named | Demand moved, or both outcomes reversed |
| Reasoning | Ties the frost to less coffee offered at every price | Gives the result with no cause | No reasoning offered |
| Clarity | A listener follows it in one hearing | Can be followed with effort | Cannot be followed |
Hesitation, accent and a restarted sentence are not in the table, so they cost nothing.
How do you mark speech consistently?
Mark against an analytic rubric with three or four criteria, listen once at normal speed before scoring, and send borderline recordings to a second listener.
Jonsson and Svingby (2007) reviewed 75 studies of rubrics. They concluded that a rubric can make scoring more reliable, above all an analytic one written for the topic and backed by exemplars or by training for the raters, and that no rubric makes a judgement valid by itself. Our guide to short answer questions and a rubric to grade them covers building one from a model answer.
Three habits suit recordings; they are practice, not research findings.
- Hear it once, at normal speed, first. Replay only to settle a criterion, and never speed it up, since pace is part of what you are judging.
- Keep three anchor recordings, strong, middling and weak, and return to them when your standard drifts.
- Use what a recording has and a live oral lacks. A colleague can hear the same ninety seconds. Agree beforehand which marks count as borderline, and have those heard twice.
Mark the criteria, not the accent
Accent is not a criterion unless you can point to what a listener failed to understand. Munro and Derwing (1995) had 18 native speakers of English listen to unscripted speech from 10 Mandarin speakers, write down what they heard and rate it. The accent ratings ranged widely, a notable share at the heavily accented end, yet the speech was mostly transcribed correctly and rated easy to understand. A strong accent, they concluded, does not necessarily reduce intelligibility.
Markers bring their own ear. Winke et al. (2013) analysed 107 raters' scores on 432 speech samples from 72 test takers of English. Raters who had themselves learned Spanish were significantly more lenient with Spanish speakers, and raters who had learned Chinese with Chinese speakers. Word the criterion as "a listener can follow it".
Which shape for a language class: a free answer, pronunciation or reading aloud?
Choose by what the student must produce: their own sentences, a set phrase or a set text.
| Shape | The student | It shows | Mark for |
|---|---|---|---|
| Free answer: EdProcto's spoken answer | replies to a question in their own words | vocabulary, grammar and fluency at once | the task done, range, accuracy, being understood |
| Pronunciation | says a given word or phrase | particular sounds and stress | the features you named and nothing else |
| Read aloud | reads a given passage | decoding and delivery, with no composing | accuracy, pace and expression, each on its own line |
For a pronunciation test, write the target into the rubric. "Je voudrais deux billets pour Lyon, s'il vous plaît" might be set for two features: the vowel of deux and the nasal ending of Lyon. A student who produces both and sounds the r differently has full marks.
For reading aloud, do not fold the mark into one number. Valencia et al. (2010) compared the common one-minute measure, scored as words correct per minute, with separate measures of rate, accuracy, prosody and comprehension for pupils in grades 2, 4 and 6. Assessments built on several indicators predicted comprehension better, and the usual benchmarks applied to the single score flagged some readers wrongly and missed others. That is school reading, not a language class; take it as a reason to keep three lines in the rubric.
Where the student should hear and not speak, see our guide to listening questions. For longer pieces EdProcto has presentation and sing or play. Speak and compare, interview, roleplay and the oral defence are in the builder too.
Who is disadvantaged by speaking to a microphone, and what should students be told?
Students who stammer, students with a speech or hearing impairment, anxious students, and anyone without a quiet room or a working microphone. For them a recording can measure the obstacle in place of the subject.
The marker is a risk too. Werle and Byrd (2022) showed 158 college instructors one of four videos of a student presenting, with or without stuttering and with high or low communication competence, to score on a standard rubric. Where competence was low, the student who stuttered was rated more favourably than the one who did not, which the authors read as overcorrection. Instructors, they concluded, evaluate students who stutter differently from their peers.
What to offer depends on what the question is for:
- Speech is not the skill (the coffee question): offer the same task as a typed answer, or a short conversation in person.
- Speech is the skill (a language oral): keep the task and change the conditions, with more time, a second take or a private room.
Your school's policy and the law where you teach come first, and none of this is legal advice.
Before anyone records, tell students who will hear it, what it is for, how long it is kept and how to ask for it to be removed. In EdProcto the recording stays with the attempt on the server for marking. EdProcto's "Accessibility" settings do not replace speaking, so the alternative task is yours to set.
Why is a recorded answer hard to hand to a chatbot?
Because the answer has to come out of the student's own mouth, inside the time, on your material. A chatbot can write a script, but the student still has to deliver it, and a script read out often sounds like one: even pace, written sentences, no searching for a word.
That makes the shape a first defence and no guarantee. A recording does not prove who is speaking, and a student with a second screen can still read an answer aloud. Tie the prompt to something said in your own class, and keep the thinking time short.
Akimov and Malin (2020), in a case study of an oral examination used for online assessment, concluded that it can help with some challenges of online courses, among which they list academic dishonesty, if it is planned for validity, reliability and fairness. For a quiz that counts, add what the stakes need from the integrity settings. One measure is ruled out for these templates: "Microphone on", because the microphone is the answer. Students are told what is on before they begin.
How does EdProcto run oral exam questions?
EdProcto is the best tool for oral exam questions when the appointments are what stop you setting them, because it supplies what the studies above keep asking for: one structured prompt for every student, and your rubric behind every mark.
EdProcto builds a quiz from your own material in any of 111 templates. Spoken answer, pronunciation and read aloud are in the builder now. These shapes are marked by a reader, by hand: you listen and score against the rubric you wrote. No score is released until you release it.
The Free plan has the core templates and light integrity. All 111 templates, the recorded ones among them, open on Tutor, with medium integrity. Educator adds the full analysis. The pricing page sets out each plan.
Quick questions
How long should a recorded spoken answer be?
Sixty to ninety seconds for one explanation. That is a working rule, not a research finding.
Should students get preparation time before an oral exam question?
A short pause, yes. Wigglesworth and Elder (2010) found no significant difference in scores between no planning time, one minute and two, and still argued that a minute should be considered, for fairness.
Should accent count in a speaking test?
Only where it stops a listener understanding. Munro and Derwing (1995) found heavily accented speech that was still highly intelligible.
Can AI mark a recorded spoken answer?
In EdProcto the spoken answer, pronunciation and read aloud shapes are marked by you, by hand, against your rubric, and the score is yours to release.
References
- Akimov, A., & Malin, M. (2020). When old becomes new: A case study of oral examination as an online assessment tool. Assessment & Evaluation in Higher Education, 45(8), 1205–1221. https://doi.org/10.1080/02602938.2020.1730301
- Huxham, M., Campbell, F., & Westwood, J. (2012). Oral versus written assessments: A test of student performance and attitudes. Assessment & Evaluation in Higher Education, 37(1), 125–136. https://doi.org/10.1080/02602938.2010.515012
- Iannone, P., & Simpson, A. (2015). Students' views of oral performance assessment in mathematics: Straddling the 'assessment of' and 'assessment for' learning divide. Assessment & Evaluation in Higher Education, 40(7), 971–987. https://doi.org/10.1080/02602938.2014.961124
- Jonsson, A., & Svingby, G. (2007). The use of scoring rubrics: Reliability, validity and educational consequences. Educational Research Review, 2(2), 130–144. https://doi.org/10.1016/j.edurev.2007.05.002
- Joughin, G. (1998). Dimensions of oral assessment. Assessment & Evaluation in Higher Education, 23(4), 367–378. https://doi.org/10.1080/0260293980230404
- Memon, M. A., Joughin, G. R., & Memon, B. (2010). Oral assessment and postgraduate medical examinations: Establishing conditions for validity, reliability and fairness. Advances in Health Sciences Education, 15(2), 277–289. https://doi.org/10.1007/s10459-008-9111-9
- Munro, M. J., & Derwing, T. M. (1995). Foreign accent, comprehensibility, and intelligibility in the speech of second language learners. Language Learning, 45(1), 73–97. https://doi.org/10.1111/j.1467-1770.1995.tb00963.x
- Qian, D. D. (2009). Comparing direct and semi-direct modes for speaking assessment: Affective effects on test takers. Language Assessment Quarterly, 6(2), 113–125. https://doi.org/10.1080/15434300902800059
- Valencia, S. W., Smith, A. T., Reece, A. M., Li, M., Wixson, K. K., & Newman, H. (2010). Oral reading fluency assessment: Issues of construct, criterion, and consequential validity. Reading Research Quarterly, 45(3), 270–291. https://doi.org/10.1598/RRQ.45.3.1
- Werle, D., & Byrd, C. T. (2022). Professors' perceptions and evaluations of students who do and do not stutter following oral presentations. Language, Speech, and Hearing Services in Schools, 53(1), 133–149. https://doi.org/10.1044/2021_LSHSS-21-00069
- Wigglesworth, G., & Elder, C. (2010). An investigation of the effectiveness and validity of planning time in speaking test tasks. Language Assessment Quarterly, 7(1), 1–24. https://doi.org/10.1080/15434300903031779
- Winke, P., Gass, S., & Myford, C. (2013). Raters' L2 background as a potential source of bias in rating oral performance. Language Testing, 30(2), 231–252. https://doi.org/10.1177/0265532212456968