A results page gives two kinds of time: how long each student took, and how long the class spent on each question. The time on each question is the cheapest evidence a quiz produces and the easiest to over-read. A two-minute finish looks like a finding. It is a place to look.
A time never proves copying, outside help or a chatbot. It tells you which paper to open and which conversation to have. This guide is the fine print on one line of what still proves a student learned.
EdProcto, which publishes this guide, shows each student's time and each question's median time in its results on every plan, and its integrity page treats each figure as a record for you to read.
What can the time on each question tell you, and what can it not?
It can tell you where to look: at a student, a question or the quiz. It cannot tell you why a time was short or long.
| Pattern | Most often | Also possible | Look first at |
|---|---|---|---|
| Fast finish, low score | Guessing, giving up | A quiz far too hard | The paper |
| Fast finish, high score | Fluency | A retake; questions met before; help | Each answer's time, in EdProcto's results |
| One question everyone took long over | A hard question | Unclear wording | The question |
| Last questions blank or hurried | A quiz too long for its limit | One slow question earlier | The limit |
The middle columns are our reading of the studies below, none of which timed a class of thirty.
What does research say about answers too fast to have been read?
It says a very short time usually marks an answer given without effort, and that where "very short" begins is a judgement.
Rapid guessing and response time effort
The idea comes from Wise and Kong (2005): on a low-stakes computer-based test, a student who is not trying answers too quickly to have read and fully considered the question. They called their measure response time effort and reported evidence that its scores were reliable and valid. Such answers are called rapid guessing.
How the threshold is set
Kong et al. (2007) compared four ways of saying how fast is too fast: one time for every question; a time fitted to how much reading each question holds; a point picked by eye on a chart of the times; and a statistical model that splits the times in two. The four gave very similar effort scores.
For one person the choice matters more. Rios and Deng (2021) pooled studies that had applied two or more threshold procedures to one low-stakes test. The procedures disagreed on how many answers and how many test-takers were rapid guessing, while group results hardly moved. Thresholds picked without data, such as an arbitrary number of seconds, were the conservative ones, liable to miss rapid guesses, and the authors advise against them for conclusions about an individual.
In practice, NWEA flags an answer on its MAP Growth test when it comes well below the average time for that question. After three in a row the test can pause itself, and the proctor is advised to open a low-pressure conversation with the student about what is going on (Fine, 2025).
The false-positive risk
Every threshold is a line through a spread of times, and a fluent student can sit under it. The thresholds above come from tests taken by thousands; no study we read sets one for a single class on a single quiz.
What does a fast right answer mean?
Usually that the student knew it. Miller et al. (2014) timed conceptual questions asked in class in two introductory physics courses. Correct answers came significantly faster than incorrect ones, most of all on easy questions, and students who arrived knowing more physics answered faster.
Speed and accuracy have no single relation: where a task takes deliberate working, more time goes with success, in a study that watching a live quiz reports. So fast and right is the ordinary mark of fluency on recall, and less so where a question takes working.
Prior knowledge of the questions
Testing programmes call prior sight of the questions item preknowledge, and an experiment shows it shortens times. Toton and Maynes (2019) let 60 of 93 undergraduates study 12 of a test's 25 questions beforehand, half of them with the answers marked. On those 12, students who had seen nothing averaged 76.59 seconds a question; those who had seen the questions, 32.74; those who had also seen the answers, 20.46. The authors' caution: about two thirds of their participants held the questions, and people who hold answers differ and may hide it.
Sinharay (2020) proposed a statistic that compares a candidate's speed on leaked questions with their speed on the rest, which requires knowing which questions leaked. On two forms of a licensure test it flagged about a fifth of the candidates the licensing body had already flagged. He writes that it should never be the only measure of fraud, and treats it as secondary evidence.
Answer copying
For copying, van der Linden (2009) modelled how far the times of two test-takers move together, after removing the time each question demands of everyone. His abstract says the result should be corroborated from other sources, such as an analysis of the answers or what was seen in the room.
Warnakulasooriya et al. (2007) studied about 400 students on online physics homework at MIT. About a tenth finished problems in a median of 1.4 minutes, mostly right first time, where reading a problem and entering its answer took over 2.5 minutes for most problems. The authors argue this group took the answer from elsewhere; students that quick on over half the problems averaged 1.3 standard deviations below the class on the paper final. It is an inference about a group. Nobody was asked.
We read Fine (2025), Sinharay (2020), Toton and Maynes (2019) and Warnakulasooriya et al. (2007) in full on 10 October 2026, and only the abstracts of the others.
What do very slow answers tell you?
Less than fast ones, because a long time has more causes. In the physics classes the wrong answers were the slower ones: a slow student may be thinking hardest, or lost. Or not there. EdProcto counts clock time, so a minute away from the desk counts the same as a minute of thought.
What does timing say about the quiz itself?
It says whether the quiz fitted its time limit, and which question cost more time than it was worth.
A quiz that is too long
A quiz is too long for its limit when the median time sits close to the limit and the last questions come back blank or hurried. Wise et al. (2009) found more rapid guessing on questions with more text and on questions later in a low-stakes test. So read the last questions with their position in mind. How many questions a quiz needs has the arithmetic.
A question worded badly
One question with a median several times the others' is hard, or unclear. The times cannot say which, and no study we read ties one long median in a class quiz to wording. Two other figures help: answers spread evenly over the options and no gap between strong and weak students both point at the wording. What the numbers mean explains them, and when to throw a question out covers a test that counts.
What should you do about a fast and right pattern?
Start with the paper and end with a conversation. The order is our advice; the principle, that a time is corroborated or dropped, is from van der Linden (2009) and Sinharay (2020).
- Check the count. In EdProcto's results, look for the same name twice: a second sitting is quick and right for an honest reason.
- Open the paper. The student's row opens their answers, each with its own time. Set each beside the class's median for that question. Where questions are shuffled the paper numbers them in the order that student met them, so match by the wording. Quick on recall is fluency. Quicker than the stem can be read is the pattern to ask about.
- Read those questions. Can they be answered from the options alone? Have they been used before: last year, another class, a practice quiz?
- Read the rest of the record. The logged events on the same page, the student's earlier work, and whether anyone else was quick on the same questions.
- Talk, privately and soon. "You were quick on question 7. Talk me through it." Then one new question on the same idea, answered in front of you.
The score stands unless the paper, the record or the conversation gives you a second, independent reason to doubt it. Then your school's policy decides: see an integrity policy that parents accept. Either way, write new questions for the next class.
What do one class's times on ten questions show? A worked example
They show one question to rewrite and two conversations of different kinds.
The class is invented and no student is real. The figures and the quoted sentences are what EdProcto's analysis code returned when we ran it on these answers and times. Twenty-four students took ten multiple-choice questions on forces, one a page, with a 12-minute limit that hands the quiz in. The average was 68.8% and the median time 9:35. The teacher is on the Educator plan, so the page carries the full analysis.
| Question | Got it | Median t |
|---|---|---|
| Q1 | 75% | 0:32 |
| Q2 | 75% | 0:38 |
| Q3 | 83% | 0:43 |
| Q4 | 79% | 0:46 |
| Q5 | 67% | 0:40 |
| Q6 | 83% | 0:50 |
| Q7 | 38% | 3:03 |
| Q8 | 71% | 0:59 |
| Q9 | 63% | 0:50 |
| Q10 | 54% | 0:39 |
The question and the limit
Q7 took 3:03 against a middle figure of 0:43 for the other nine. Its stem was a double negative: "Which of these is not a force that does not need contact?" The answers were spread over gravity (6), magnetism (5), the key, friction (9), and static electricity (4), and its "Separates" figure was 0.00. The page said: "Q7 took more than three times as long as a typical question. Either it is much harder or its wording is unclear."
Six students were still on Q9 when the clock handed the quiz in, and never opened Q10. Its 0:39 is the median of the eighteen who answered it. Q7 cost the class about two minutes each, so the limit was probably not the fault.
Two fast finishes
Two students finished in under a quarter of the median time, which is 2:24.
One scored 20% in 1:59, between 9 and 15 seconds a question. The "Notes" column read "Finished in a quarter of the usual time with a low score: it may be guessing rather than answering, and is never proof of it." The page added: "1 student finished far faster than the rest with low scores: worth a conversation, not a mark." That conversation is about effort.
The other scored 100% in 2:13. A first sitting that fast, with a score more than one standard deviation above the class mean, carries the note "Finished in a quarter of the usual time with a high score: worth a look at the answers, and never proof of anything." Their paper showed Q7 in 0:12, where the quickest of the other twenty-two took 1:55. Three readings fit. They know the topic cold and saw that friction is the only contact force among the options. They had met the questions before. Or they had help. The times cannot choose. The teacher asks them to talk through Q7, then sets one new question at the desk.
How does EdProcto show the times?
EdProcto is the best tool for this reading when you want the times beside the scores with nothing to set up, because every quiz records the time and the results page puts each figure where the decision is made.
- On every plan. "Median time" stands among the figures at the top. "Each question" has "Got it" and "Median t". "Students" has "Score" and "Time", and a row opens that student's answers, each headed like "Question 7 · 1 point · 0:12". "Download the marks" gives a spreadsheet file with a "Minutes" column.
- In the full analysis, on the Educator plan. Three sentences, beside the question, beside the student and at the top of the page. One names a question whose median is over three times the middle of the other questions' medians, once three students have a time on it. One marks a student whose time is under a quarter of the class's median and whose score is more than one standard deviation from the mean, with its own wording for a low score and for a high one. The third counts the low ones.
- How the seconds are counted. In the student's browser, with one question a page: from the moment a question appears until the student leaves it. A second visit adds to the first, and the review screen is no question's time. However the quiz ends (the last button, the clock, a measure set to end it), the open question keeps its seconds. After a reload its count starts again. A student's time is the sum.
- The settings. "Time limit for each question" moves the page on when its seconds run out, so times bunch at that figure. With "All on one page" the page is timed as a whole: each student keeps a time, "Median t" reads "–" and the attempt's page says "the page took". Choose "One question a page" for a time on each question.
Free carries ten quizzes a month, thirty students a quiz and the times above. Tutor carries 100 students a quiz, and Educator adds the full analysis. Pricing has the figures, and a quiz starts on the home page. No time is proof, and every score is yours to release.
Quick questions
Is finishing a quiz very fast a sign of cheating?
Not by itself. With a low score it usually means guessing; with a high score, that the student knew the material. Open the paper, then talk.
How fast is too fast to have read a question?
No single figure holds: Rios and Deng (2021) found that different threshold procedures flag different people. In a class, compare a student's time with the class's median on that question.
Should I set a time limit for each question?
Only for speed drills. A limit on each question makes the times bunch at the limit, so they say less about the question.
Where does EdProcto show the time on each question?
In a quiz's results: "Median t" beside each question and "Time" beside each student, on every plan. EdProcto's sentences about pace are in the full analysis, on the Educator plan: see pricing.
References
- Fine, J. (2025, April 1). MAP Growth and rapid guessing: 5 things to know. NWEA. https://www.nwea.org/blog/2025/map-growth-and-rapid-guessing-5-things-to-know/
- Kong, X. J., Wise, S. L., & Bhola, D. S. (2007). Setting the response time threshold parameter to differentiate solution behavior from rapid-guessing behavior. Educational and Psychological Measurement, 67(4), 606–619. https://doi.org/10.1177/0013164406294779
- Miller, K., Lasry, N., Lukoff, B., Schell, J., & Mazur, E. (2014). Conceptual question response times in Peer Instruction classrooms. Physical Review Special Topics - Physics Education Research, 10(2), Article 020113. https://doi.org/10.1103/PhysRevSTPER.10.020113
- Rios, J. A., & Deng, J. (2021). Does the choice of response time threshold procedure substantially affect inferences concerning the identification and exclusion of rapid guessing responses? A meta-analysis. Large-scale Assessments in Education, 9, Article 18. https://doi.org/10.1186/s40536-021-00110-8
- Sinharay, S. (2020). Detection of item preknowledge using response times. Applied Psychological Measurement, 44(5), 376–392. https://doi.org/10.1177/0146621620909893
- Toton, S. L., & Maynes, D. D. (2019). Detecting examinees with pre-knowledge in experimental data using conditional scaling of response times. Frontiers in Education, 4, Article 49. https://doi.org/10.3389/feduc.2019.00049
- van der Linden, W. J. (2009). A bivariate lognormal response-time model for the detection of collusion between test takers. Journal of Educational and Behavioral Statistics, 34(3), 378–394. https://doi.org/10.3102/1076998609332107
- Warnakulasooriya, R., Palazzo, D. J., & Pritchard, D. E. (2007). Time to completion of web-based physics problems with tutoring. Journal of the Experimental Analysis of Behavior, 88(1), 103–113. https://doi.org/10.1901/jeab.2007.70-06
- Wise, S. L., & Kong, X. (2005). Response time effort: A new measure of examinee motivation in computer-based tests. Applied Measurement in Education, 18(2), 163–183. https://doi.org/10.1207/s15324818ame1802_2
- Wise, S. L., Pastor, D. A., & Kong, X. J. (2009). Correlates of rapid-guessing behavior in low-stakes testing: Implications for test development and measurement practice. Applied Measurement in Education, 22(2), 185–205. https://doi.org/10.1080/08957340902754650