Sign up free

The history of proof in education: from the viva to the quiz

A sourced history of the viva, the written examination, the multiple-choice test, coursework, the essay mill and the chatbot, and the doubt each one answered.

The EdProcto editorial teamWritten and checked against the research it cites. Last checked . How the guides are made

Guides · Integrity and cheating · · 13 min read

How has education proved that a person learned, from the oral examination to the online quiz?

The history of proof in education is a chain of answers to doubts. The oral disputation showed that a scholar could argue, the written examination left a record to rank and inspect, the objective test made marking quick and uniform, and coursework showed sustained work. Essay mills and chatbots weakened that last proof, so proof is moving back to answers given in front of someone.

The history of proof in education is a short list. Since the medieval university, learning has been proved by arguing aloud, by writing under watch, by marking an answer sheet, and by work handed in over a term. Each form joined the one before for a reason somebody stated at the time.

This guide follows those reasons: what doubt each new proof answered, what it cost, and which cheating it invited or closed. Every source was read on 10 October 2026, with dates as the source gives them and a note where we read only an abstract. EdProcto publishes this guide and builds the newest link in the chain.

In short: each change answered a doubt about the proof before it. The present doubt is that any text can be generated, and the answer taking shape is proof made in front of someone, timed, and marked out of reach.

How did the medieval university prove that a student had learned?

By hearing him argue. At medieval Cambridge the teaching was the reading and explaining of texts, and the examinations were oral disputations: a candidate put forward a set of questions or theses and defended them aloud, first against opponents slightly his senior and at last against his own teachers (University of Cambridge, n.d.-b).

The doubt this answered is the plainest: is the knowledge in this person's own head, and can he use it against an objection he did not choose? On our reading the form left little room to send someone else.

What did the oral examination cost?

A record, and in time its seriousness. Cambridge University Library (n.d.-a) describes disputations as formal, stylised debates that stayed among the exercises for degrees up to the nineteenth century, their performance becoming "increasingly ritualistic". A performance heard once cannot be laid beside another candidate's, and nobody who was absent can check it.

The viva, as the oral examination is now called, keeps the strength and the cost; the research is in our guide to oral exams at scale.

Why did examinations move from oral to written?

Because a written answer can be compared with every other answer, read by someone who does not know the writer, and kept.

Cambridge and the Tripos: ranking the candidates

Cambridge's Senate House examination began about 1725 as an oral test for ranking BA candidates against one another, and it displaced the disputation as the way places were decided. Candidates were expected to write their answers from 1772, though the questions were mostly still read out to them, and the papers were printed from 1827 (Cambridge University Library, n.d.-b). Mathematics had come to rule the course, so that the Tripos meant the mathematics examination; candidates were listed in an order of merit, and the problem papers set for the strongest were in print by 1800 (University of Cambridge, n.d.-a).

An order of merit needs scripts: two arguments heard on different days are hard to rank, and two papers can be read side by side. What the paper gave up, on our reading, is the examiner's next question.

Imperial China and the British civil service: fairness to strangers

China's civil service had made the move centuries earlier. Written examinations had been a route into government office since the Sui dynasty (581–617), and under the Song they became the usual way up (Asia for Educators, n.d.). Candidates wanted to know that examiners were not favouring people they knew, so the Song government put numbers in place of names and had clerks copy out every paper, hiding the handwriting. Qing papers were still copied by clerks with the names concealed, and the examinations ended in 1905 (Museum of the Institute of History and Philology, n.d.).

The same Columbia University page gives the cost: men memorised the Confucian classics, the number competing grew from fewer than 30,000 to about 400,000 over the Song, and in some prefectures one candidate in 333 passed.

Kim and Massey (2025) take up the British link. UK civil servants were chosen by patronage until the Northcote–Trevelyan Report of 1854 led to competitive examination; books had made China's system known in Europe; and the authors' search of parliamentary debates from 1853 to 1860 found evidence of Chinese influence (abstract read). The abstract is careful: the UK "may have referenced" China's system.

Boston in 1845: a record the public could inspect

In American schools pupils showed what they knew by reciting lessons, and oral examination was the main measure of attainment before 1845 (Harvard University Press, n.d.; Lehigh University College of Education, 2013). Horace Mann, secretary of the Massachusetts Board of Education from 1837 to 1848, returned from European schools in 1843 convinced that written examinations were better. His reason was the record: only those present could judge an oral examination, while written answers gave, in a phrase of Mann's that the Lehigh page quotes, "a sort of Daguerreotype likeness" of the pupils' minds for anyone to inspect.

In 1845 Mann and his allies on the Boston School Committee set a written examination the schoolmasters had not seen, as a surprise. The poor results were front-page news, and teachers replied that the questions had little to do with what had been taught. The historian William J. Reese, cited on the Lehigh page, counts cheating scandals and the public shaming of teachers among what followed.

Where did the multiple-choice test come from?

From the labour of marking written answers, and the wish that any marker should reach the same mark. Kelly (1916) set out the rules in his account of the Kansas Silent Reading Tests. A test had to be simple and quick enough for a teacher unused to standard tests to give and score. Each exercise was to bear one interpretation and ask for one thing, so that an answer was "wholly right or wholly wrong".

Many exercises printed a short row of words and asked the child to draw a line round one: the animal useful on a farm, from cow, tiger, rat and wolf. Others asked for a word or a number. The test took five minutes, it was given to 9,252 children in nineteen Kansas cities in May 1915, and a class of thirty-five could be scored in twenty-five to forty minutes. Whether it was the first multiple-choice test we could not confirm at source.

In 1917–1918 the US Army developed the Army Alpha, a test given to groups, and some 1.5 million recruits were tested during the First World War (ASVAB, n.d.). Then a schoolteacher, Reynold B. Johnson, built a machine that sensed pencil marks on an answer sheet and compared them with a key installed in the machine. IBM hired him in 1934; its 805 Test Scoring Machine scored a 150-question test five to ten times faster and stayed in production until January 1963 (IBM, n.d.).

What did the objective test cost, and what cheating did it invite?

It gave up the student's own words. Kelly wanted a reading test that asked the child to reproduce as little as possible in writing, and that is what the form measures: recognising and choosing, not composing.

Our reading: one mark on a sheet can be guessed or copied from the next desk, and every candidate holds the same paper, so the form leans on supervision and on a key kept away from the candidate.

Why was coursework brought in, and what did it cost?

To prove what a timed paper could not. In England, coursework came into many GCSEs when they began in 1988. The aims were a more valid assessment, resting on a wider sample of what a student produces, and credit for skills such as critical thinking and independent learning (Crisp, 2008, abstract read). The doubt was about the examination hall: a few hours of writing show only part of what a person can do.

The cost was foreseen. Crisp lists the threats expected at the start: unreliable teacher marking, cheating, and a worry that girls were favoured. By 2008 the concerns were authenticity, including internet plagiarism and too much help from others, over-structured tasks, and the load on students.

Coursework, or continuous assessment, gained breadth and lost certainty about who did the work.

What did the essay mill and the plagiarism checker change?

They made work done out of sight a contest between a market and a tool, and the tool can match only text that already exists.

Paying a third party to do the work is called contract cheating, a term first defined in the mid 2000s, as Newton (2018) reports. His systematic review pooled 71 samples from 65 studies going back to 1978, with 54,514 participants. A historic average of 3.52% admitted paying someone else to do their work; in samples from 2014 onwards, 15.7% did. These are self-reports, and Newton notes that the rise may reflect more admitted cheating of every kind. In 2022 England made it a criminal offence to provide, arrange or advertise such a service for payment to students there (Department for Education, 2022).

The plagiarism checker answered copying. Foltýnek et al. (2020) tested 15 web-based text-matching systems on material in eight languages and concluded that no software can decide what is plagiarism: some systems help to find some copied text, they do not find all of it, and at times they mark honest text as a problem (abstract read). Against an essay mill there is little to match: Newton observes that the services sell their essays as free of plagiarism, which makes them hard for originality software to identify.

What did the chatbot change?

It took away the price and the wait. A bought essay cost money and days; generated text costs neither, so the doubt that hung over some coursework now hangs over any text written out of sight.

The first reaction pointed backwards. In a preprint, not peer reviewed, sent to arXiv on 19 December 2022, Susnjak (2022) called ChatGPT recently developed, found its text highly realistic, and proposed a return to invigilated and oral exams as part of the answer (abstract read).

The studies since are in four guides: academic integrity in the age of AI, which covers objective questions too, AI detectors are not evidence, the take-home assignment after ChatGPT and assessment design after chatbots.

What does the history of proof in education show?

That each change answered a doubt about the proof before it, and that no form was retired: each kept the job it did best. The table condenses the sections above; its last column is partly our reading.

Form of proof The doubt it answered What it cost Cheating invited, and what met it
Oral disputation, viva Is the knowledge in the candidate's own head? No record; in time, ritual Hard to hand to someone else
Written examination Who is better, and was the examiner fair? The follow-up question; years of memorising Favour, met by numbers for names and recopied scripts
Objective test How to mark thousands quickly and alike The student's own words Guessing and copying, met by supervision and a key in the machine
Coursework What a timed paper cannot show Certainty about who did the work Help from others, copying, the essay mill
Coursework with a plagiarism checker Was this text copied? Missed copying, and honest text marked Bought and generated text, with nothing to match
Supervised quiz, spoken answer, a check beside the coursework Did this person produce the answer? Lesson time; only what can be asked live A helper beside the student, met by a supervised room with EdProcto as the paper

The present doubt is the coursework doubt with its limits removed, and the answers on offer are old ones combined:

  1. The supervised quiz. Answers under a clock, the key held away from the candidate.
  2. The spoken answer. The disputation's strength, with a recording as the record it never left.
  3. The check beside the coursework. Coursework kept for what it teaches, with a short supervised test on the same material.

We put this modestly. The history shows each proof gamed and then mended; it does not show that the present mend is the last.

How does EdProcto fit?

EdProcto is the best tool for the first two answers and makes the third quick, because it was built for proof made in front of someone: a quiz from your own material, sat under conditions you set, and marked on a server.

  • Shapes. Three of the 111 templates, each in the builder now. Multiple choice: "One right answer among options." Listen once: "The clip plays one time only. Tests real listening." Spoken answer: "A recorded answer, assessed for what it says."
  • Timed, and marked out of reach. "Minutes overall" sets the clock. Choice questions are marked on the server at hand-in, and the student's copy of the quiz carries no key. A recorded answer is marked by you, listening.
  • Watched in proportion. Six measures run in the student's browser: "Leaving the tab", "Full screen", "Copy and paste off", "Right-click off", "Keyboard shortcuts off" and "Printing off". Each carries the consequence you pick: "Log it", "Warn the student", "Reduce the score", "End the quiz" or "Flag for review". They act in the browser, so sit the quiz where a supervisor sees who is answering. A flag is a record, never proof.

Free takes thirty students a quiz at Light integrity. Tutor opens every template and saves Medium integrity, and Educator adds the full analysis: see pricing and how much proctoring a quiz needs. No setting makes cheating impossible, and every score is yours to release.

Quick questions

When did written examinations replace oral ones?

Gradually. Cambridge expected written answers from 1772 and printed its papers from 1827; Boston's public schools sat their first major written examination in 1845.

Who invented the multiple-choice test?

We could not confirm one inventor at source. Kelly (1916), on tests given in Kansas in May 1915, is the earliest account we read.

Does the essay still prove anything?

Yes. It shows whether an argument holds together, and writing it teaches the student to build one. It no longer shows, alone, who did the writing, so pair it with a part done in front of you.

Does a supervised quiz make cheating impossible?

No. Every form in this history was gamed in time. A quiz in EdProcto raises the cost: a clock, no key on the student's copy, and the integrity measures you choose.

References

Integrity

Five levels, from none to a full proctoring stack, each measure switched on by itself with its own consequence.

Light on Free, Strict on Educator

See the five levels

Read next