How much proctoring does an online quiz need? As much as its score decides. A practice quiz that settles nothing needs no monitoring. A final exam, or a certificate that lets someone do a job, needs a good deal. Most quizzes sit in between, and most of what protects them is not a camera.
This guide is for a teacher, lecturer or trainer torn between switching nothing on and switching everything on. Most of the research comes from single university courses, so each study is given with its sample.
EdProcto, which publishes this guide, treats academic integrity the same way: its integrity settings are separate switches, each off until the stakes call for it.
How much proctoring does an online quiz or exam need?
An online quiz or exam needs proctoring in proportion to what its result decides. The ladder below is our recommendation, not a research finding: nobody has tested these five rungs against one another.
| What the result decides | Switch on | Leave off |
|---|---|---|
| Nothing: practice, homework | Nothing. Show the answers at once | Every measure |
| A few per cent of a grade | A fitted time limit, shuffling, a short window, the tab logged | Full screen, camera, microphone |
| A unit test that counts | The rung above, plus full screen, copy and paste off, one question at a time | The camera |
| A final exam | The rung above, plus no going back, and a supervised room or a webcam with an alternative offered | Automatic penalties |
| A certification | The rung above, plus an identity check and a person reviewing every record | Decisions by software alone |
The reasoning behind each rung
- Practice. Looking an answer up only cuts the student's own practice short.
- A few per cent. The gain is small, and a log shows any pattern over the weeks. Our guide to weekly quizzes in a lecture course covers this case.
- A unit test. A grade now moves, so closing the second window is fair. A camera's costs still fall on everyone to deter a few.
- A final exam. Unwatched scores run higher in most comparisons, and here that changes outcomes. Flags err in both directions, so a person does the review.
- A certification. Others rely on the result, so who sat the test matters as much as how. Where identity must be certain, a test centre is the honest alternative.
Do students score higher on unproctored online tests?
In most comparisons they do, though a higher unproctored score is not in itself evidence of cheating.
Alessio et al. (2017) compared four tests across nine sections of an online medical terminology course at one American university (147 students). Tests recorded by webcam averaged 74.3% against 89.4% unrecorded (about 17 points in the authors' model) and used 20.4% of the time allowed against 41.2%. Instructors, not chance, chose which tests were recorded. In one section a lockdown browser with no video left scores and times much like the unmonitored sections.
In online psychology classes at one university (1,694 students), Daffin and Jones (2018) had one exam in each class watched live by webcam, and later rotated which one. Students averaged 64.4% watched and 77.1% unwatched, and took about 28 minutes against 48.
Not every study finds the gap: in a one-course experiment, Hylton et al. (2016) found no significant difference in scores, though the webcam group finished significantly sooner. Pooling 109 effect sizes from 49 studies of ability tests, Steger et al. (2020) found unproctored scores higher by 0.20 of a standard deviation, with significantly smaller gaps for tasks that are hard to look up online.
What a lower proctored score does and does not prove
It shows that the two conditions gave different numbers, not why. Alessio et al. (2017) write that their study did not assess cheating behaviour. Daffin and Jones (2018) offer two explanations, forbidden help and the anxiety of being watched, and say the first cannot be confirmed. Extra time may mean searching, or an interruption at home.
Unwatched scores may still rank students sensibly. When 18 courses (2,010 students) at one university moved from invigilated exams to unproctored online ones mid-semester in 2020, Chan and Ahn (2023) found the two sets of scores correlated at r = 0.59, against 0.74 between the halves of fully in-person semesters.
What does webcam proctoring cost the students being watched?
It raises test anxiety for some students, and in one study the most anxious scored lower under it.
Conijn et al. (2022) surveyed students in 105 courses at one faculty of a Dutch university during the pandemic (1,760 responses). On a seven-point scale, test anxiety averaged 4.42 for proctored online exams, 4.09 for unproctored ones and 3.53 for exams on campus and assignments. Part of that anxiety came from sitting an exam online at all; among online exams alone, proctored ones were still about 0.3 of a point higher. Proctoring made no difference to grades or to how tempted students said they were to cheat.
Woldeab and Brothen (2019) compared 44 introductory psychology students who sat their final with a live ProctorU proctor on webcam against 587 who sat it in a testing centre. The averages were one point apart out of 200. Among the 44, those at or above the median for trait test anxiety scored 144.6 against 162.3 for the rest. The groups were not randomised, and 44 is few.
Students' views are mixed. Among 30 computer science students, Bergmans et al. (2021) found a clear majority believed online proctoring would prevent cheating, while worrying about who would hold the recording and about the room scan.
How often is a proctoring flag a false flag?
Nobody has measured a general rate, and the two studies we could read show errors in both directions.
Yoder-Himes et al. (2022) examined the reports Respondus Monitor gave instructors for about 357 students in four STEM courses at one American university in autumn 2020. Students with darker skin tones averaged 6.07 flags an assessment, against 1.19 for lighter tones, most of them for a face "missing from frame". When the researchers watched 298 recordings they saw possible cheating behaviour in 12, and eight of those carried the software's lowest priority for review. It is one product, in that term's version.
Bergmans et al. (2021) tested the other direction with Proctorio: six of their 30 students were asked to cheat and five to act nervous while working honestly. The software flagged none of the six, and a human reviewer caught one. No honest student was marked suspicious by the software; the reviewer wrongly suspected one.
So false flags land on students who did nothing, and a clean record clears nobody. A flag begins a look, never a penalty.
What should you switch on before any camera?
Switch on the measures that cost students nothing: they shrink what outside help can gain, where a camera only watches for it.
- Questions that cannot be looked up. The meta-analysis found smaller gaps on such tasks. Ask for a principle applied to a new case (see multiple-choice questions that measure), or ask from your own handout: the quiz maker on EdProcto's home page writes from your material.
- A shape with nothing to paste. A clip that plays once, a click on an image or a spoken answer leaves a second device little to work with. See the guides to listening and hotspot questions, and every shape under templates.
- A time limit that fits. Unwatched students took far longer in both course studies. Set the limit by what a prepared student needs.
- Shuffled questions and options, so neighbours see different screens. EdProcto shuffles both for each student.
- One question at a time and, where the test counts, no going back. Our reasoning, not a tested result: it slows the forwarding of a whole paper.
- A short window and, where your tool allows it, a different set for each student (in EdProcto, "A different set of questions for each student"). In one course of 425 students with exams sat over several days, Chen et al. (2018) found that students judged to have been passed information gained 13 percentage points when all met the same problem, and 2 to 3 with a pool of two or four.
None of this is a guarantee. Alessio's course already used timed tests, a shared question pool and shuffled answers, and the gap remained.
What can each kind of monitoring see, and what can it not see?
Each kind of monitoring records one thing and is blind to the rest, which is why none amounts to proof. The rows are EdProcto's levels above none, each with what it switches on.
| Level, and what it switches on | Records | Cannot tell you |
|---|---|---|
| Light: leaving the tab | The tab being left, each time with its time | Why; what is on a phone beside the keyboard |
| Medium: full screen; copy, paste, right-click, printing and shortcuts off | Leaving full screen; an attempt to copy, paste, print or use a shortcut | Another person in the room |
| Strict: webcam, microphone, the screen shared, a photo every few minutes, other apps noticed, a phone watching the room | The face, the room's sound, the screen and the desk | What is outside the frame |
| Maximum: a photo and an ID card at the start; faces and gaze watched; a voice sample and the typing rhythm kept | Who began; no face, a second face, eyes off the screen | Whether a lost face is absence or poor detection |
What should you tell students, and what do you do with a flag?
Tell students every measure that is on before they start, and treat a flag as a record to examine once the student has been heard.
Before the quiz
Say what is on, what it records, who sees it, how long it is kept, and what a student who cannot meet a rule should do. EdProcto does this ahead of the first question: the screen lists each measure that is on with its consequence and asks for consent. A student who declines does not start, so the alternative must exist before the day.
When a flag appears
- Read the record: what happened, when, for how long.
- Set it beside the attempt: the pace, the questions involved, earlier work. Two brief exits from full screen are not eleven, each just before a right answer.
- Watch any recording yourself.
- Ask the student what happened.
- Decide under your institution's procedure, and note why.
In EdProcto each measure carries the consequence you choose (log it, warn the student, reduce the score, end the quiz, flag for review), and each defaults to logging. Given the errors above, we would keep anything a camera or an algorithm judges, in any tool, at log or flag.
Who is harmed by camera and full-screen rules, and what can you offer?
Those most exposed are students the software sees badly, those without a quiet room or reliable equipment, and those already anxious.
- Faces the software loses. Besides the gap by skin tone, Yoder-Himes et al. (2022) saw that a head resting on a hand or leaning towards the screen made a face hard to detect.
- Rooms and equipment. Conijn et al. (2022) found higher test anxiety in online exams among students without a dedicated study space or reliable technology, and among those in financial hardship.
- Assistive tools. Try the quiz with the set-up of anyone who uses a screen reader, magnification or dictation: rules about shortcuts and full screen were not written with those tools in mind.
Offer a supervised room, another time, or the unit-test rung plus a short spoken check of two or three answers. This guide is not legal advice: recording a face, a voice or a room is regulated differently in each country and institution, so ask whoever handles data protection where you teach before a camera goes on.
How does EdProcto match monitoring to the stakes?
EdProcto is the best tool for this job when the assessments in one course carry different stakes, because its monitoring is a stack of separate switches, not a session that is on or off. The research above points the same way: the measures before the camera do much of the work, and a flag needs a person.
Each measure is switched on by itself with its own consequence, and the levels are presets over them. The browser measures take a quiz to Light or Medium; the camera, microphone and identity measures take it to Strict and Maximum, and the integrity page lists each one. Practice, Standard quiz and Exam are three starting points a click away. As the page on why it is built this way explains, the shape of the question is the first defence, the answer key never reaches the student's device and the marks are made on the server.
Light integrity is on the Free plan and medium on Tutor; strict and maximum will come with Educator and the Institution licence, arranged by a call (for schools). Pricing has the rest. No setting makes cheating impossible, and no flag is a penalty you did not set.
Quick questions
Does an online quiz need a webcam?
Rarely. A quiz worth a few per cent needs a fitted time limit, shuffling and a logged tab. Keep the webcam for a final or a certification.
Is a lower proctored score proof that students cheated without proctoring?
No. Alessio et al. (2017) and Daffin and Jones (2018) both found lower proctored scores, but neither measured cheating, and anxiety is a rival explanation.
Is a proctoring flag proof of cheating?
No. Yoder-Himes et al. (2022) found flags falling unevenly by skin tone. EdProcto's integrity page treats every flag as a record for your review.
Does online proctoring make test anxiety worse?
For some students. Conijn et al. (2022) found higher test anxiety in proctored online exams, and Woldeab and Brothen (2019) found lower scores among anxious students watched live.
References
- Alessio, H. M., Malay, N., Maurer, K., Bailer, A. J., & Rubin, B. (2017). Examining the effect of proctoring on online test scores. Online Learning, 21(1), 146–161. https://doi.org/10.24059/olj.v21i1.885
- Bergmans, L., Bouali, N., Luttikhuis, M., & Rensink, A. (2021). On the efficacy of online proctoring using Proctorio. In Proceedings of the 13th International Conference on Computer Supported Education (pp. 279–290). SciTePress. https://doi.org/10.5220/0010399602790290
- Chan, J. C. K., & Ahn, D. (2023). Unproctored online exams provide meaningful assessment of student learning. Proceedings of the National Academy of Sciences, 120(31), Article e2302020120. https://doi.org/10.1073/pnas.2302020120
- Chen, B., West, M., & Zilles, C. (2018). How much randomization is needed to deter collaborative cheating on asynchronous exams? In Proceedings of the Fifth Annual ACM Conference on Learning at Scale (pp. 1–10). Association for Computing Machinery. https://doi.org/10.1145/3231644.3231664
- Conijn, R., Kleingeld, A., Matzat, U., & Snijders, C. (2022). The fear of big brother: The potential negative side-effects of proctored exams. Journal of Computer Assisted Learning, 38(6), 1521–1534. https://doi.org/10.1111/jcal.12651
- Daffin, L. W., Jr., & Jones, A. A. (2018). Comparing student performance on proctored and non-proctored exams in online psychology courses. Online Learning, 22(1), 131–145. https://doi.org/10.24059/olj.v22i1.1079
- Hylton, K., Levy, Y., & Dringus, L. P. (2016). Utilizing webcam-based proctoring to deter misconduct in online exams. Computers & Education, 92–93, 53–63. https://doi.org/10.1016/j.compedu.2015.10.002
- Steger, D., Schroeders, U., & Gnambs, T. (2020). A meta-analysis of test scores in proctored and unproctored ability assessments. European Journal of Psychological Assessment, 36(1), 174–184. https://doi.org/10.1027/1015-5759/a000494
- Woldeab, D., & Brothen, T. (2019). 21st century assessment: Online proctoring, test anxiety, and student performance. International Journal of E-Learning & Distance Education, 34(1). https://www.ijede.ca/index.php/jde/article/view/1106
- Yoder-Himes, D. R., Asif, A., Kinney, K., Brandt, T. J., Cecil, R. E., Himes, P. R., Cashon, C., Hopp, R. M. P., & Ross, E. (2022). Racial, skin tone, and sex disparities in automated proctoring software. Frontiers in Education, 7, Article 881449. https://doi.org/10.3389/feduc.2022.881449