Educational Assessment

Educational assessment is the practice of gathering evidence about what learners know and can do, and then using that evidence to make decisions.

14 frameworks, from 605 to today · Education · Social Sciences

still a working view ended · each leads to its entry below

Overview

Educational assessment is the practice of gathering evidence about what learners know and can do, and then using that evidence to make decisions1. Those decisions range from whether a student passes a course to whether a school is failing, from how to adjust tomorrow's lesson to which applicants get a university place. What makes the field more than a set of techniques is that every assessment embeds a claim about what counts as learning and a judgment about who gets to decide. The history of assessment is therefore a history of tools and statistics, but also of institutions, social selection, and arguments about what evidence should be trusted.

Examinations as instruments of selection

Formal testing is old. The Chinese imperial examination system, consolidated under the Sui and Tang dynasties from the seventh century and elaborated over the following millennium, selected officials through written examinations on classical texts and statecraft. It was a selection device of enormous reach: it promised advancement by merit of study rather than birth, and it shaped schooling across East Asia. Its limits were equally instructive—the curriculum narrowed to what the exam tested, and success depended heavily on access to tutors and texts.

In Europe, universities from the medieval period onward examined students orally for degrees, and by the eighteenth and nineteenth centuries competitive written examinations became the standard route into civil services and professions. The British civil service adopted competitive examination in the 1850s, and the model spread through colonial administrations and school systems. The underlying logic was consistent: a standardized task, administered under controlled conditions, scored against a common standard, allowed institutions to compare candidates at scale. What it could not do was say much about the meaning of a score—whether a difference of a few marks reflected real differences in ability or the accidents of the particular questions asked.

Mental measurement and the birth of the test

The late nineteenth and early twentieth centuries turned assessment into a science. Francis Galton, studying heredity in the 1880s, built anthropometric laboratories and tried to measure sensory discrimination as a proxy for intelligence. James McKeen Cattell brought these methods to the United States and popularized the term "mental test" in 1890. The decisive shift came with Alfred Binet and Théodore Simon, whose 1905 scale, commissioned by the French Ministry of Education to identify children needing special instruction, tested judgment, comprehension, and reasoning rather than simple sensory acuity. Binet's scale introduced the idea of mental age—the level of performance typical of a given chronological age—and, crucially, of tasks arranged in ascending difficulty and standardized on a comparison sample.

The idea traveled and mutated. Lewis Terman's Stanford revision of the Binet scale (1916) produced the intelligence quotient, originally a ratio of mental to chronological age multiplied by 100. During the First World War, the U.S. Army's Alpha and Beta tests assessed over a million recruits, demonstrating that group testing was feasible at enormous scale and embedding testing in the machinery of the state. The interwar decades saw aptitude testing spread into schools, industry, and college admissions; the Scholastic Aptitude Test, first administered in 1926, was built on this foundation. The same period produced a cautionary record: early intelligence testing was entangled with eugenic assumptions and with the misreading of culturally loaded scores as fixed hereditary capacity, a history the field has had to reckon with ever since.

Standardized achievement testing and the classical model

Alongside aptitude testing grew the standardized achievement test—measuring what had been taught rather than general capacity2. The development of machine-scorable answer sheets in the 1930s made mass testing cheap, and statewide and national testing programs expanded through the mid-twentieth century.

The statistical theory underpinning these instruments is now called classical test theory (CTT)3. Its core idea is simple: an observed score is the sum of a true score and error3. From this, practitioners derived reliability—the consistency of scores across occasions or item sets—and item statistics such as difficulty (the proportion answering correctly) and discrimination (how well an item separates high and low scorers). CTT made it possible to build tests with known properties and to estimate how much confidence a score deserved. Its limitation is that all these properties are sample-dependent: an item's difficulty depends on who took the test, and a person's score depends on which items were included. Comparing scores across different test forms required elaborate equating procedures.

Objectives, taxonomies, and the evaluation of programs

A parallel strand concerned what should be assessed. In 1949 Ralph Tyler's Basic Principles of Curriculum and Instruction framed education as the pursuit of stated objectives and evaluation as the check on whether they were met. Benjamin Bloom and colleagues' Taxonomy of Educational Objectives (1956, the cognitive domain) classified learning outcomes from knowledge through comprehension, application, analysis, synthesis, and evaluation, giving test designers and curriculum planners a shared vocabulary for specifying what a question was asking4. The taxonomy was widely adopted and widely criticized—for implying a hierarchy that research did not always confirm—but its influence on how objectives are written and assessed was lasting.

In 1967 Michael Scriven introduced the distinction between formative and summative evaluation: formative evaluation occurs during development to improve a program, summative evaluation at the end to judge it5. Benjamin Bloom, J. Thomas Hastings, and George Madaus applied the terms to student learning in their 1971 handbook, and the formative/summative pair became fundamental to the field5.

Item response theory and the measurement revolution

The problem CTT could not solve—sample-dependent item statistics and test-dependent scores—was addressed by item response theory (IRT), developed from the 1950s onward. IRT models the probability that a person of given ability answers an item correctly as a mathematical function, typically a logistic curve with parameters for item difficulty, discrimination, and sometimes guessing. The Danish mathematician Georg Rasch proposed a one-parameter model in 1960; Allan Birnbaum, Frederic Lord, and others developed two- and three-parameter models through the 1960s and 1970s. Because IRT places items and persons on a common scale, it allows test forms to be built to specification, scores to be compared across forms, and items to be characterized independently of any particular sample. It also enabled computerized adaptive testing, in which each item is chosen based on the examinee's running estimate—a method that became practical for large-scale admissions and licensure testing from the 1980s.

International large-scale assessment

After the Second World War, comparative assessment became an instrument of international policy. The International Association for the Evaluation of Educational Achievement, founded in 1959, conducted its first mathematics study in the early 1960s. The Programme for International Student Assessment (PISA), launched by the OECD in 2000, tested fifteen-year-olds' reading, mathematics, and science literacy every three years, and the Trends in International Mathematics and Science Study (TIMSS) continued a series dating from the 1960s. These surveys made national achievement comparable and put test results at the center of education policy debate. They also drew sustained methodological criticism: questions of cultural fairness in translated items, of whether literacy tests measure curriculum or general capability, and of the policy distortions that follow when rankings become political currency.

Accountability, standards, and the alternative-assessment response

The 1980s and 1990s saw assessment become a lever of system governance. Standards-based reform—most visibly in the United States, where the No Child Left Behind Act of 2001 required annual testing and attached consequences to results—tied school funding, staffing, and reputation to test scores. The theory was that measurable targets would drive improvement; the documented side effects included curriculum narrowing, teaching to the test, and, in some systems, outright cheating scandals. Similar accountability regimes took different forms in England's national curriculum assessments and in other countries' inspection and examination systems.

Against this, a reform movement argued that multiple-choice tests captured too little of what mattered. Alternative and performance assessment—portfolios, projects, extended writing, oral examinations, and authentic tasks—sought to measure complex performance directly. The movement gained institutional weight in the 1990s; in England, the Task Group on Assessment and Testing report of 1988 recommended a balance of formative, summative, and performance evidence, though its proposals were only partly implemented. The persistent difficulty was cost and reliability: performance tasks are expensive to score and hard to score consistently, which is why large-scale systems kept returning to standardized formats.

Assessment for learning

The most influential reframing of the era came from research on classroom practice. Paul Black and Dylan Wiliam's 1998 review, "Assessment and Classroom Learning," synthesized evidence that formative feedback—commenting on the quality of work and how to improve it—produced substantial learning gains, while grades alone often did not. Their work, followed by the booklet Inside the Black Box (1998) and the Assessment Reform Group's coining of "assessment for learning," shifted attention from assessment of learning to assessment for learning. The practical repertoire—clear learning intentions, success criteria, questioning, peer and self-assessment, feedback that moves learners forward—became standard in teacher education worldwide. The approach's critics noted that its evidence base was uneven and that "formative assessment" was sometimes reduced to a label for any classroom check.

Learning analytics and the data-rich present

The most recent turn is computational. Learning analytics—the measurement, collection, and analysis of data about learners and their contexts—emerged as a named field around 2010, drawing on educational data mining, learning management system logs, and, later, machine learning. Its promise is continuous, unobtrusive assessment: a system that tracks problem-solving attempts, time on task, and discussion participation can flag a struggling student or adapt the next task in real time. Its problems are familiar in new form—validity of the indicators, privacy, algorithmic bias, and the risk that what is easily logged is mistaken for what matters.

Where the field stands

Contemporary assessment is plural. Large-scale systems still rely on IRT-calibrated standardized tests and international surveys, now often delivered digitally. Classroom practice draws on assessment for learning and on performance tasks where reliability can be managed. Accountability regimes persist but have been softened in several countries after two decades of evidence about their side effects. Learning analytics and adaptive systems are expanding the range of evidence available, while raising unresolved questions about validity and equity. The oldest tension remains the live one: assessment is asked simultaneously to select, to certify, to diagnose, and to improve, and no single instrument serves all four well. How to combine multiple kinds of evidence—statistical, observational, and computational—without letting any one of them silently define what learning is, is the question the field has not yet answered.

References

  1. Educational assessment, Wikipedia
  2. Achievement test, Wikipedia
  3. Classical test theory, Wikipedia
  4. Bloom's taxonomy, Wikipedia
  5. Formative assessment (Origin of the term), Wikipedia

The frameworks of Educational Assessment

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 605-Present

    Examination-based selection

    Who should be selected for office and advancement?

    Fly to its tower in Nousfarer

  2. 1880-1960

    Mental measurement

    How can mental traits be measured by standardized tasks?

    Fly to its tower in Nousfarer

  3. 1905-Present

    Intelligence and aptitude testing

    How can general ability and aptitude be compared across people?

    Fly to its tower in Nousfarer

  4. 1930-Present

    Achievement test

    How much of what was taught has been learned?

    Fly to its tower in Nousfarer

  5. 1930-Present

    Classical test theory

    How much confidence does an observed score deserve?

    About Classical test theory
    Classical Test Theory (CTT) addressed a fundamental problem in early educational assessment: how to quantify a student's knowledge when any single test score is vulnerable to chance errors like guessing, fatigue, or lucky guesses. Before CTT, educators judged student performance mainly through oral exams, written essays, or informal observations—methods that were hard to compare across students or schools. CTT provided the first systematic, statistical framework for understanding test scores. Its core insight—that an observed score is a combination of a student's true standing plus random error—gave educators a precise way to talk about measurement accuracy. This framework emerged in the late nineteenth and early twentieth centuries as mass schooling and university entrance exams created a pressing need for standardized, comparable assessments.
    At its heart, CTT treats a test as a measurement instrument and a student's score as an estimate of an unobservable trait (the "true score"). The framework's concepts form a logical progression. First, the Total Score is simply the number of items a student answers correctly. This observed score (XX) is understood as the sum of the True Score (TT) and Random Error (EE): X=T+EX = T + E. The true score is defined as the average score a student would obtain over an infinite number of independent administrations of the same test—a theoretical construct, not directly observable. Random error is assumed to have a mean of zero across many test takers and to be uncorrelated with the true score. From this foundation, CTT derives its central concepts. Reliability is the proportion of observed-score variance that is due to true-score variance. A perfectly reliable test would give the same score to the same student every time (barring learning). Standard Error of Measurement expresses reliability in meaningful units: it estimates the band around an observed score within which the true score likely falls (roughly, SEM=sx1−rxx\text{SEM} = s_x \sqrt{1 - r_{xx}}, where sxs_x is the standard deviation of observed scores and rxxr_{xx} is reliability). Item Difficulty is simply the proportion of test takers who answer an item correctly. Item Discrimination measures how well an item separates high-performing from low-performing students, often using the point-biserial correlation between item score and total score. Correction for Guessing estimates how much a multiple-choice test score would be inflated by random guessing. Parallel Forms are two versions of a test designed to have equal true scores and equal error variances. Split-Half Reliability estimates reliability by correlating scores on two halves of the same test. Test–Retest Reliability assesses stability by correlating scores from the same test administered twice to the same group. Internal Consistency (estimated by methods like coefficient alpha) measures whether items on a test are measuring the same construct. Norm-Referenced Interpretation means that a student's score is interpreted by comparing it to the scores of a reference group. These concepts gave early test developers a practical toolkit for building, evaluating, and comparing tests without requiring complex mathematics beyond basic statistics.
    The intellectual history of CTT begins in the late 1800s with the work of Francis Galton (1822–1911), who pioneered the use of statistical methods to study human abilities. Galton did not develop CTT itself, but his work on correlation and regression provided essential precursors. The true originators of the framework were Charles Spearman (1863–1945) and his contemporaries. In a series of papers around 1904, Spearman laid out the core equation X=T+EX = T + E and introduced the concept of reliability. His work on the attenuation of correlation coefficients (correcting for measurement error) was a direct application of CTT logic. Edward L. Thorndike (1874–1949), a student of Galton's ideas, became a dominant figure in American educational measurement, writing textbooks that codified CTT practices for school testing. The framework reached its mature form in the 1940s–50s through the work of Lee J. Cronbach (1916–2001), who developed coefficient alpha (in 1951), giving CTT a practical method for estimating internal consistency reliability. Cronbach's 1951 paper, "Coefficient Alpha and the Internal Structure of Tests," is perhaps the single most influential methodological article in the CTT tradition. Other key contributions include L. L. Thurstone's (1887–1955) work on scaling and Frederic M. Lord's (1912–2000) 1952 book A Theory of Test Scores, which mathematically formalized the framework. The seminal works of this period include Spearman's 1904 paper "The Proof and Measurement of Association between Two Things" (which laid the foundation for reliability theory), Thorndike's 1918 The Measurement of Intelligence, and Cronbach's 1951 coefficient alpha paper. CTT remained the dominant psychometric framework for educational testing until the 1960s, when Criterion-Referenced Assessment introduced the idea of interpreting scores against fixed standards rather than norm groups, and Item Response Theory began to address CTT's fundamental limitation: its sample-dependency (test statistics like item difficulty and reliability change depending on the group of students tested). Today, CTT is not replaced but coexists with these later frameworks. It is still widely used in classroom testing, standardized test development (for initial item analysis), and situations where sample sizes are too small for IRT models. Its legacy is the fundamental vocabulary of measurement that every educator and assessment professional still uses: reliability, standard error, item difficulty, and the essential idea that observed scores always contain some degree of error.
    The 13 concepts its tower teaches
    • Correction for Guessing
    • Internal Consistency
    • Item Difficulty
    • Item Discrimination
    • Norm-Referenced Interpretation
    • Parallel Forms
    • Random Error
    • Reliability
    • Split-Half Reliability
    • Standard Error of Measurement
    • Test–Retest Reliability
    • Total Score
    • True Score

    Fly to its tower in Nousfarer

  6. 1949-1980

    Objectives-based evaluation

    Were the stated educational objectives actually met?

    Fly to its tower in Nousfarer

  7. 1956-Present

    Bloom's taxonomy

    What kinds of learning outcomes can be specified and ordered?

    Fly to its tower in Nousfarer

  8. 1959-Present

    Large-scale learning assessments

    How do national education systems compare in achievement?

    Fly to its tower in Nousfarer

  9. 1960-Present

    Item response theory

    How can items and persons be placed on one scale?

    Fly to its tower in Nousfarer

  10. 1967-Present

    Formative and summative evaluation

    Is evaluation for improving or for judging?

    Fly to its tower in Nousfarer

  11. 1980-Present

    Standards-based accountability

    Should schools be held to measurable targets?

    Fly to its tower in Nousfarer

  12. 1988-Present

    Alternative assessment

    Can assessment capture what tests leave out?

    Fly to its tower in Nousfarer

  13. 1998-Present

    Formative assessment

    How can evidence during learning improve teaching?

    Fly to its tower in Nousfarer

  14. 2010-Present

    Learning analytics

    What can digital activity data reveal about learners?

    Fly to its tower in Nousfarer

How the ideas connect

Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.

Examination-based selectionMental measurementIntelligence and aptitude testingAchievement testClassical test theoryObjectives-based evaluationBloom's taxonomyLarge-scale learning assessmentsItem response theoryFormative and summative evaluationStandards-based accountabilityAlternative assessmentFormative assessmentLearning analytics

Beyond the field

Where Educational Assessment's ideas came from in other fields, and what they led to there: the strongest 12 of 88 links.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Educational Assessment is a planet in Nousfarer

Its city has a tower for each of these 14 frameworks. Land, walk in, and climb one concept at a time.

Land on Educational Assessment

Related fields

Fields its ideas are linked to

More in Education

See where Educational Assessment sits on the Map of Human Knowledge