Optical mark recognition (OMR) is the technique that detects, on a scanned sheet, the boxes or bubbles a candidate has filled in and turns them into machine-readable answers. Historically it relied on dedicated readers measuring ink density at fixed positions on a form printed to the reader's exact layout. Modern software applies the same principle to an image from an office scanner or a smartphone: it locates the registration marks, straightens the image, then analyses each answer area. The output is one answer grid per paper, which the software then compares against the answer key to compute the grade.
See also:Bubble sheet, Scan resolution (dpi), Multifunction copier, flatbed scanner or smartphone
A bubble sheet is a printed form on which the candidate fills in, for each question, the bubble or box matching their choice instead of writing an answer. Designed for optical mark recognition, it aligns the answer areas on a regular grid framed by registration marks, which let the software find every box even when the sheet comes out of the scanner slightly rotated or shifted. It can be separate from the question paper, with one sheet for all the questions, or built into it, with the boxes printed under each question. The term is American in origin, popularised by the optical readers that made the format standard in US schools.
See also:OMR (optical mark recognition), Answer sheet, Scantron
The answer sheet is the document on which a candidate records their answers, whether or not it is separate from the question paper. For a multiple-choice exam it is usually a grid of boxes to tick; for a mixed exam it may also reserve areas for free-text answers. Keeping the answer sheet separate lets the question paper be reused, means only one page per candidate is scanned, and imposes an identical layout on every paper. Building it into the question paper removes transcription errors between the two and saves a distribution step. Either way, the answer sheet, and only the answer sheet, is what the grading software reads.
See also:Bubble sheet, Grading grid
The answer key is the list of expected answers to every question of an exam, the reference each paper is compared against to produce a grade. For a multiple-choice exam it states, for each question, the correct choice or choices and, where relevant, the scoring rule attached: points per correct answer, penalty per error, treatment of a blank. When papers are shuffled, the key must be attached to each question independently of its position, so the same question is graded the same way in every version. A printable answer key also serves as a proofreading document before the exam and as feedback material for candidates afterwards.
See also:Discordance-based scoring, Negative marking, Shuffled papers, A/B versions
Discordance-based scoring grades a question with several correct answers by counting the disagreements between the boxes the candidate ticked and the boxes expected, each box wrongly ticked or wrongly left blank counting as one discordance. A question with no discordance earns full marks; the score then falls with the number of discordances, often reaching zero at the second or third depending on the scale chosen. Used notably in French medical schools, this rule allows partial credit: a candidate who identified most of the correct answers is not graded as if they knew nothing. It only makes sense for questions with several correct answers; on a single-answer question it collapses to right or wrong.
See also:Negative marking, Single-answer versus multiple-answer question, Answer key
Negative marking is a scoring rule that deducts points for a wrong answer instead of simply counting it as zero. Its purpose is to discourage guessing: if a correct answer is worth one point and an error costs a fraction of a point, ticking at random no longer pays off on average. The penalty is often calibrated on the number of options offered, so that the expected gain from a random answer is nil. A blank is usually scored zero, leaving the candidate free to abstain. The rule must be announced before the exam, and the final grade is most often floored so it cannot fall below zero. In Papeez, point calculation with rewards and penalties is fully configurable.
See also:Discordance-based scoring, Answer key
Shuffled papers are versions of the same exam in which the order of the questions, of the answer options, or both, differs from one paper to the next, so that two neighbours cannot copy each other box by box. The simplest form is a pair of A and B versions handed out alternately; the most thorough generates a unique paper per candidate. Shuffling imposes two constraints: every paper must carry a version identifier the software can read, and the answer key must follow each question to its new position. It can also apply to exercises and sections, as Papeez does with its four-level shuffling, and combines with a barcode on every paper.
See also:Anonymity barcode, Answer key
An anonymity barcode is an identifier printed on each paper that links it to a candidate, or to a version of the exam, without showing any name in clear. The software reads it at scan time, which lets it match the paper to the right person and the right answer key automatically and detect duplicates. The candidate's name stays off the graded image, which keeps the marking anonymous. One barcode per paper also makes it possible to print a unique paper per candidate. Depending on the software, the identifier is printed in advance from the enrolment list or copied by the candidate into a dedicated area. Papeez prints a unique barcode on every paper.
See also:Anonymised paper, Shuffled papers, A/B versions
AMC-TXT is the plain-text file format of Auto Multiple Choice (AMC), offered to users who do not want to write their exams in LaTeX. The file describes the questionnaire with a lightweight syntax: header options, then the questions, whose correct and incorrect choices are told apart by a sign at the start of the line. AMC then converts it to LaTeX to produce the question papers and answer sheets. Its simplicity makes it a common exchange format for moving a stock of AMC questions to another tool; Papeez, for instance, accepts AMC-TXT files on import.
Source:AMC documentation, "AMC-TXT source file"
See also:Auto Multiple Choice (AMC)
Auto Multiple Choice (AMC) is free software for creating and automatically grading paper multiple-choice exams, distributed under the GPLv2+ licence, which lays out the papers with LaTeX and reads the scanned sheets by optical mark recognition. It installs locally, mainly on Linux, and chains the layout of the paper, printing, scanning, box recognition, scoring-rule entry and grade export. Exams are written in LaTeX or in the AMC-TXT format. Being free and accurate has made it a reference in French-speaking higher education, at the cost of a technical installation and learning curve.
Source:Auto Multiple Choice official site
See also:AMC-TXT, OMR (optical mark recognition)
Scantron is originally an American brand of optical readers and answer forms whose name has become a generic term: in the United States, "a Scantron" commonly means the bubble sheet itself. The system pairs a proprietary pre-printed form with a dedicated machine that reads it. Used generically, the word covers any device that grades multiple-choice exams by optical mark recognition; used as a brand, it names the company and its products. Outside the United States, "Scantron machine" is often shorthand for "MCQ grading machine", including for solutions that work with an ordinary scanner.
Source:Scantron company website
See also:OMR (optical mark recognition), Bubble sheet
Moodle Offline Quiz is an activity plugin for the Moodle platform that lets a Moodle quiz be taken on paper: question papers and answer sheets are generated from the course's question bank and printed, then the completed sheets are scanned and uploaded into the course for automatic grading. Grades go straight into the Moodle gradebook. The plugin is maintained by the Academic Moodle Cooperation. It suits institutions whose questions and students already live in Moodle, and requires the plugin to be installed by the platform administrator.
Source:Plugin page on moodle.org
See also:OMR (optical mark recognition), Answer sheet
QTI (Question and Test Interoperability) is a 1EdTech technical standard for exchanging questions, tests and results between assessment applications, each item being described in an XML format that any conformant tool can import. For a paper exam, QTI matters upstream, to retrieve questions written in an LMS or another platform.
Source:1EdTech, QTI specification
Go further:Evalmee's QTI converter
For online exams, see the Evalmee glossary entry
Item analysis is the statistical examination, question by question, of the answers given by all candidates to a test, in order to judge the quality of each question after the exam. It rests on a few indicators: the difficulty index (the proportion of correct answers), the discrimination index (does the question separate strong candidates from weak ones?), the point-biserial correlation and the spread of choices across each distractor. On a paper exam graded by optical mark recognition, these indicators are available as soon as the sheets are read, with no extra data entry. They serve to spot an ambiguous question, a wrong answer key or a distractor nobody picks, and to improve the question bank from one session to the next.
See also:Difficulty index, Discrimination index, Point-biserial correlation
For online exams, see the Evalmee glossary entry
The difficulty index of a question is the proportion of candidates who answered it correctly, expressed between 0 and 1 or as a percentage. Despite its name, the higher the index, the easier the question: 0.90 means nine candidates in ten found the right answer. A question close to 1 tells candidates apart from nobody; a question close to 0 often signals an ambiguous wording or an error in the answer key. In practice test authors aim for intermediate values, with a few easy questions to open. On a paper exam the index is computed as soon as the sheets are read and appears directly in the per-question statistics.
See also:Item analysis, Discrimination index
For online exams, see the Evalmee glossary entry
The discrimination index measures how well a question separates the candidates who master the subject from those who do not. The classic calculation compares the question's success rate in the group with the highest total scores (the top third or quarter, say) with its success rate in the lowest-scoring group; the difference, between -1 and 1, is the index. A high positive value indicates a good question; a value near zero, a question that adds nothing to the ranking; a negative value, a question the strong candidates miss more than the others, which almost always points to a problem in the wording or the answer key.
See also:Item analysis, Point-biserial correlation
For online exams, see the Evalmee glossary entry
The point-biserial correlation is a discrimination indicator that measures the correlation between success on one question (a binary variable: right or wrong) and the total score obtained on the test. It ranges from -1 to 1, like an ordinary correlation coefficient, and is read like the group-based discrimination index, of which it is the continuous counterpart: it uses every candidate rather than the extreme groups only. A clearly positive value means the candidates who get the question right are also those who do well on the test. To avoid bias, the total score is sometimes computed without the question under study, in which case it is called the corrected item-total correlation.
See also:Discrimination index, Item analysis
For online exams, see the Evalmee glossary entry
Scan resolution, expressed in dots per inch (dpi), is the number of pixels the scanner produces for each inch of paper; it sets the fineness of the image and therefore how reliably the boxes can be read. Too low a resolution makes the registration marks and small boxes hard to detect; too high a resolution bloats the files without improving the reading. For an answer sheet, 200 to 300 dpi in greyscale is the usual setting, and the default on most office copiers. It also matters to print the papers at 100% scale, without fit-to-page adjustment, so that the proportions the software expects are respected.
See also:OMR (optical mark recognition), Multifunction copier, flatbed scanner or smartphone
Three kinds of device can digitise multiple-choice papers: the multifunction copier, the flatbed scanner and the smartphone. The multifunction copier, found in most institutions, has an automatic document feeder that swallows a stack of papers and produces a single PDF, which makes it the fastest tool for a whole cohort. The flatbed scanner requires placing the sheets one at a time and suits small groups or damaged papers. The smartphone photographs each paper; the application must then correct perspective and lighting, and quality depends on how the picture is taken. None of the three is a dedicated optical reader: the software does the mark reading. Papeez accepts scans from all three.
See also:Scan resolution (dpi), OMR (optical mark recognition)
An MCQ (multiple-choice questionnaire) is an exam made of closed questions for which the candidate selects one or more answers from written options rather than writing an answer. The format is graded objectively: two markers, or a machine, produce the same grade. On paper, answers are recorded on a grid of boxes or a bubble sheet, then read by scanning and optical mark recognition. MCQs measure knowledge and understanding well; they measure the ability to produce or justify a line of reasoning poorly. Their quality depends on how the questions and distractors are written, and on a scoring rule suited to the number of correct answers expected.
See also:Single-answer versus multiple-answer question, Bubble sheet, OMR (optical mark recognition)
A single-answer question admits exactly one correct answer among the options, whereas a multiple-answer question may admit several: the candidate must then tick all the correct ones and none of the wrong ones. The distinction governs the scoring rule: a single-answer question is graded right or wrong, possibly with negative marking; a multiple-answer question calls for partial credit, for instance by discordance. It also governs the instructions: candidates must know whether they may tick more than one box. On an answer sheet the two formats use the same grid; the answer key and the scoring rule make the difference.
See also:MCQ (multiple-choice questionnaire), Discordance-based scoring, Negative marking
The grading grid is the table that, for one paper, matches each question with the answer read and the points awarded, and whose sum gives the grade. In MCQ grading software it is produced automatically from the optical reading and the answer key, then shown paper by paper for review: this is where a teacher checks a misread box, a crossed-out answer or a double tick, and adjusts by hand if needed. By extension the term also names the grid of boxes printed on the answer sheet, nominative or anonymous, that the candidate fills in.
See also:Answer key, Answer sheet
An anonymised paper is an exam paper on which the candidate's identity is not visible at marking time, so that the grade cannot be influenced by the name. On paper, anonymity is traditionally achieved with a folded and glued corner, or with an anonymity number written on the paper and on a list kept separately. With a multiple-choice exam graded by scan, the barcode printed on each paper plays that role: it identifies the paper for the software without displaying the name, and the match is only restored when grades are published. Anonymous marking is a common requirement in French higher education for final exams.
See also:Anonymity barcode, Grading grid