Weekly quizzes in a university lecture course do two jobs that a midterm and a final cannot. They make students bring last week's material back to mind before it fades, and they tell the lecturer what three hundred people did not follow while there is still time to act. The old objection was the marking, and questions a rule can mark have removed it.
This guide is for a lecturer with 80 to 500 students. It covers what studies of frequent quizzing in real university courses found, how much a quiz should count, how to make it cumulative, and how to run, protect and read it. EdProcto, which publishes this guide, does the running and the reading: its page for universities shows how.
Do weekly quizzes work in a university lecture course?
Yes, on the evidence from real courses: classes that are quizzed often tend to do better on their exams, and more of their students pass. The caution is that most single studies compare one class with another, in psychology, at one institution.
The widest view is a meta-analysis: Sotola and Credé (2021) pooled 52 samples from real classes, 7,864 students, and found a moderate association between frequent low-stakes quizzes and academic performance (d = 0.42). In the five samples that reported it, quizzing went with much better odds of passing (odds ratio 2.566). The authors' word is association, and it should be read as one.
Three single courses:
- At the end of every lecture. Lyle and Crawford (2011) closed each lecture of a statistics course with a few questions on that day's material, answered from memory. Scores on the four exams were significantly and substantially higher than in a section taught without them.
- On the reading, every class. Batsell et al. (2017) compared two introductory psychology classes, one of which took a daily quiz on the assigned chapters. The quizzed class scored higher on exam questions about the textbook, whether they repeated a quiz question or were new but related.
- Online and unsupervised. In a web-based college course, McDaniel et al. (2012) had each student quizzed on some concepts, given others to read, and left alone on the rest. Quizzes with feedback beat rereading on identical exam questions, and beat unquizzed content on related ones.
Agarwal et al. (2021), reviewing 50 classroom experiments on retrieval practice, found smaller effect sizes in the 20 undergraduate experiments than in the middle-school ones, which our guide to low-stakes quizzing in the classroom covers.
The largest test: a quiz in every class for 901 students
Pennebaker et al. (2013) replaced the exams in two large introductory psychology classes at one university with a quiz at the start of every class: eight multiple-choice questions in the first ten minutes, on the students' own devices, with feedback at once. Seven covered the previous lecture and its readings. The eighth was one that student had answered wrongly before.
The 901 students were compared with 935 taught by the same instructors three years earlier. On 17 questions also set in earlier years they averaged 77.1% correct against 71.2%, about half a letter grade, though just short of the usual threshold for significance (p = .06). The grade gap between students whose parents had the most and the least education was 0.34 of a letter grade, against 0.71 before.
Those quizzes were 86% of the course grade, so the paper tests what its authors call "frequent consequential quizzing", not low-stakes quizzing. The comparison is with an earlier year, the textbook was swapped for online readings at the same time, and the authors call the project a case study.
Do cumulative exams and quizzes help students remember?
Cumulative testing helps students keep the material after the course has ended, and in one study it helped lower-scoring students most. Both studies are about exams, so carrying the idea into a weekly quiz is our inference.
Lawrence (2013) compared introductory psychology students who sat cumulative exams all semester with students who sat unit exams and then a cumulative final. Two months after the course, the kind of exam made no difference to strong students, but low-scoring students remembered more when every exam had been cumulative.
Khanna et al. (2013) looked at the final alone. Sections with a cumulative final scored higher on a content exam given straight after it, and students tested up to 18 months later had kept more if their final had been cumulative. Introductory students gained more than those in upper-division courses.
The likely mechanism is spacing. Cepeda et al. (2006), reviewing 317 experiments on distributed practice, found that the longer material has to be held, the wider the gap between sessions that served memory best. Those were verbal recall tasks, not lecture courses, so the review gives the direction and not the size.
How much should a weekly quiz count towards the grade?
Make each quiz worth one or two per cent, ten to fifteen per cent of the course in total, and count only each student's best scores.
The case for counting it at all comes from the meta-analysis: Sotola and Credé (2021) found a stronger association where quiz performance contributed to the class grade (d = 0.51). On how much, the research is nearly silent. Agarwal et al. (2021) note that most classroom studies did not report how the practice quizzes were graded. One or two per cent is our judgement, not a finding.
Best ten of twelve, and why to drop the lowest
Best-n-of-m means m quizzes are set and each student's best n count. Pennebaker et al. (2013) dropped the three lowest of 26. Take twelve weeks and 12% of the grade. If all twelve count at 1% each, a missed week costs 1% of the course and an email to you. If the best ten count at 1.2% each, it costs nothing, twice.
A worked case, with each quiz out of 8. A student scores 6, 7, 0 (ill), 5, 8, 6, 7, 0 (a timetable clash), 6, 7, 8 and 5. The two zeros drop, and the other ten make 65 out of 80, or 9.75 of the 12 marks. Nobody filled in a form, and nobody judged whether the illness was real.
None of the studies above tested this rule against counting every quiz, so treat it as a sensible default. Say in the syllabus that the dropped quizzes are the whole allowance for illness and travel.
How do you make a weekly quiz cumulative?
Reserve a fixed share of every quiz for earlier weeks, and print that share in the syllabus. For eight questions: five on this week's lectures, two on the three weeks before, one on anything older.
- Ask the old idea in a new case. Batsell et al. (2017) and McDaniel et al. (2012) both found gains on new, related exam questions, so nothing is lost by rewording.
- Pick the older questions from the results. The concept half the room missed in week four belongs in weeks five and seven. Pennebaker et al. (2013) did this student by student.
EdProcto's spiral review shape mixes this week, last month and older material in set proportions, a fresh draw for each student; or build the mix by hand.
How do you run and mark a weekly quiz for hundreds of students?
Use question shapes a rule can mark, put the quiz on the students' own devices, and let no answer wait for a person. Pennebaker et al. (2013) ran 26 quizzes for 901 students with about one teaching assistant to every 200.
- Fix the slot. The first ten minutes, the last five, or an online window that closes before the next lecture.
- Write from the lecture. Give the week's notes to the quiz maker on EdProcto's home page. Procto AI drafts questions from them and, if you ask, keeps only those the notes themselves answer. Edit every one.
- Choose shapes a rule can mark. Multiple choice with wrong options taken from real errors (see writing multiple-choice questions that measure), a fill in the blank where the term should be produced, an estimate scored by closeness where the answer is a number.
- Set the rules once. EdProcto's Standard quiz preset is the starting point: a time limit, questions and options shuffled, one attempt, answers shown after submitting. It also asks for full screen, which you can switch off for a quiz this small.
- Open it with a code. Students join on a phone or a laptop, and each answer is saved as it is given, so a dropped connection loses nothing.
An online window costs no lecture time: in EdProcto you publish the quiz when the window opens and close it when it ends. Written answers can be marked by Procto AI against your rubric from the Tutor plan up, one answer at a time as you ask, but for hundreds every week that is a queue to clear.
How much academic integrity does a quiz worth one per cent need?
Very little: shuffled questions, a time limit and a noticed tab are proportionate to a quiz worth one per cent, and a webcam is not. The student who looks an answer up loses the practice the quiz was for, and in the web-based course above, unsupervised quizzes still improved exam performance.
Students are not suspects in this design. What fits the stakes:
- Shuffling, so each student sees the questions and options in a different order.
- A different set for each student, written by Procto AI on the same concepts, is a setting on the Educator plan.
- Light integrity: leaving the quiz's tab is noticed and logged, and every student sees what is switched on before starting.
Heavier measures belong to the final, and EdProcto's integrity page sets out every measure and the level it belongs to. A log entry is a record for you to read, not proof: a student may have left the tab to silence a notification, and no flag becomes a penalty you did not set. No setting makes copying impossible, and a weekly quiz does not need one.
What about missed quizzes and accessibility?
Let the dropped quizzes absorb ordinary absences, and settle adjustments once, before week one.
- Missed quizzes. Best ten of twelve needs no make-ups. For a longer absence with good reason, moving that weight to the final is one remedy, and your institution's rules decide it.
- Extra time. Speed is not what a weekly quiz measures. A limit generous enough for everyone removes most separate timings. Where a student's support plan says otherwise, the plan wins.
- Reading the screen. The quiz is a web page, so a student's own zoom and contrast settings apply to it, and EdProcto's "Larger text", "High contrast" and "Screen-reader mode" are buttons the student turns on. A question built on an image or a clip needs an alternative for a student who cannot see or hear it.
- Devices. Ask in week one who has no phone or laptop, and arrange a loan or a paper copy.
What should you do with the results before the next lecture?
Read four things in about ten minutes, and let each one change the next lecture or the next quiz.
| The results show | Before or at the next lecture |
|---|---|
| Most of the class missed one question | Open with it, and put the concept in next week's cumulative slot |
| One wrong option beat the right answer | Name the misconception and say why it is wrong |
| Stronger students missed a question more than weaker ones | Check the key and the wording; if it is at fault, give everyone the mark |
| A student has missed two in a row | A short message offering help |
The last row matters most. Sotola and Credé (2021) found that quiz scores correlated strongly with later academic performance (r = 0.57 across 19 samples), so by week four the quiz record is an early warning, with two thirds of the term left to act on it.
How does EdProcto run a weekly quiz for a lecture course?
EdProcto is the best tool for this job when the quiz has to come from this week's lectures, because it does the three things the designs above depend on: questions from the course's own material, marking at the moment of the answer, and a reading of what the answers mean. Pennebaker's team had to build its own platform to deliver that to 901 students.
Students open the quiz with its code, and you can require a name and an email address before it starts. The questions reach each device without the answer key, and the marks are made on the server, by rule, at once. The full analysis then gives each question's difficulty, how well it separated stronger from weaker students, who took each wrong option, and the class's mean on each topic. Every score is yours to release, and counting the best ten of twelve is arithmetic for your gradebook.
The Free plan takes thirty students a quiz, enough for a pilot in one tutorial group. Tutor takes 100, which covers a course of 80. Educator takes 500 and adds the full analysis. A department is the School plan and a whole institution the Institution licence, each arranged by a call: see the page for schools. The pricing page has the rest.
Quick questions
How often should a lecture course have a quiz?
Once a week at least. Lyle and Crawford (2011) quizzed at the end of every lecture and Pennebaker et al. (2013) at the start of each. None of the studies above compared weekly with every lecture.
Should weekly quizzes count towards the final grade?
Yes, a little: one or two per cent each, with the lowest two dropped. Sotola and Credé (2021) found a stronger association with performance where quizzes contributed to the grade.
Do weekly quizzes need proctoring?
No. Shuffling, a time limit and a logged tab are proportionate to one per cent. EdProcto's light level logs a student leaving the tab and no more, and the integrity page lists every measure.
Can one lecturer run a weekly quiz for 400 students?
Yes, if every question is marked by rule. On EdProcto's Educator plan a quiz takes up to 500 students, and the page for universities shows how it runs.
References
- Agarwal, P. K., Nunes, L. D., & Blunt, J. R. (2021). Retrieval practice consistently benefits student learning: A systematic review of applied research in schools and classrooms. Educational Psychology Review, 33(4), 1409–1453. https://doi.org/10.1007/s10648-021-09595-9
- Batsell, W. R., Jr., Perry, J. L., Hanley, E., & Hostetter, A. B. (2017). Ecological validity of the testing effect: The use of daily quizzes in introductory psychology. Teaching of Psychology, 44(1), 18–23. https://doi.org/10.1177/0098628316677492
- Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354–380. https://doi.org/10.1037/0033-2909.132.3.354
- Khanna, M. M., Badura Brack, A. S., & Finken, L. L. (2013). Short- and long-term effects of cumulative finals on student learning. Teaching of Psychology, 40(3), 175–182. https://doi.org/10.1177/0098628313487458
- Lawrence, N. K. (2013). Cumulative exams in the introductory psychology course. Teaching of Psychology, 40(1), 15–19. https://doi.org/10.1177/0098628312465858
- Lyle, K. B., & Crawford, N. A. (2011). Retrieving essential material at the end of lectures improves performance on statistics exams. Teaching of Psychology, 38(2), 94–97. https://doi.org/10.1177/0098628311401587
- McDaniel, M. A., Wildman, K. M., & Anderson, J. L. (2012). Using quizzes to enhance summative-assessment performance in a web-based class: An experimental study. Journal of Applied Research in Memory and Cognition, 1(1), 18–26. https://doi.org/10.1016/j.jarmac.2011.10.001
- Pennebaker, J. W., Gosling, S. D., & Ferrell, J. D. (2013). Daily online testing in large classes: Boosting college performance while reducing achievement gaps. PLoS ONE, 8(11), Article e79774. https://doi.org/10.1371/journal.pone.0079774
- Sotola, L. K., & Credé, M. (2021). Regarding class quizzes: A meta-analytic synthesis of studies on the relationship between frequent low-stakes testing and class performance. Educational Psychology Review, 33(2), 407–426. https://doi.org/10.1007/s10648-020-09563-9