Skip to content

Authentic Assessment

Pattern · All patterns
Evidence · 22 claims (12 for, 9 mixed, 1 against) · 29 studies (9 quant-synthesis, 8 causal, 6 review, 3 qualitative, 2 design, 1 theoretical), q1–q4 · 9 of 29 report an effect size · 15 claims rest on one study

Description and scope

A reusable unit- or course-level policy for judging a capability through a task that asks for the performance itself, the kind of work the discipline, profession or community does (diagnosing a patient, designing for a client, arguing a case, writing for a reader who will use the text), scored against public criteria, with feedback and revision before the final judgement. The policy is: name the performance → build a task that demands it under realistic conditions → share criteria and exemplars before work starts → practise the components with support → draft or rehearse the whole performance → feedback → revise → final performance, to a real or realistic audience where the goal allows → judge against the criteria, with assessors checked for agreement. The intended change is in performing the capability (skill-performance on a complex-skill), and in carrying it to a new case (near-transfer, far-transfer).

It sits inside three converted pages and does not repeat their models. Constructive Alignment owns the general relationship, whether outcome, practice and assessment call for the same capability at the same level; authentic assessment is one way to make the assessed task match an outcome stated as a real performance. Competency-Based Assessment owns the quality of the judgement (does the task represent the capability, is the threshold defensible, do assessors agree). Assessment for Learning owns how evidence gathered during the work is read and used (shared criteria, stakes that let partial understanding show, a revision step). Authentic Audiences and Purposes owns the audience relationship and names this page as the nearest pattern for it, so the audience step below is where that principle's policy lives. This page owns the narrower configuration: an extended, realistic performance task with a draft-feedback-revise cycle, used as the unit's assessment.

No claim in this wiki tests authentic assessment itself: none compares learners assessed by a realistic performance task with learners assessed by a conventional test, on later performance or transfer, and none measures whether making a task more realistic changes what learners learn. The claims below test parts of the sequence (formative use of assessment, feedback, a real audience, rubric-referenced self-assessment, peer assessment, project work culminating in a product) and several limit it (guidance for novices, untrained self-assessment, scorer disagreement). Each is carried to this pattern by extrapolation, and the page says how far. The sequence below is the earlier page's, kept as the default; its doses and the response table are untested design proposals.

Inputs: establish the design brief

Field Record or elicit
Learner and starting observation Before the task is set, give a short version of the target performance (one patient vignette, one paragraph for a named reader, one small design request) and record what each learner produces and how they go about it. Record prior knowledge of the domain, experience of open-ended tasks, language of instruction, and any access needs. Preserve unknowns rather than labelling learners novice.
The performance What a practitioner actually does that the outcome names: the product or performance, its real conditions (time pressure, incomplete information, a client, a team), and which parts are the target and which are incidental (spelling in a clinical note, slide design in an engineering pitch). Whether the task is well defined or ill-structured.
Context and activities Setting (classroom, online-instructor-led, online-self-paced, workplace-clinical), unit length (single-session, days-weeks, term-plus), class size, assessor time available per product, whether peers, external partners or practitioners can take part, and how the final work will be shown.
Learner-valued goal Ask what the learner wants from the task (to be ready for placement, to build a portfolio piece, to pass, to help the partner it is for). Record disagreement with the designer objective.
Designer objective Which capability the final judgement certifies, at what threshold, and whether the result is formative only, contributes to a grade, or decides progression or licensure.
Outcome and instrument The rubric or rating scale and who applies it; whether assessors are calibrated and their agreement checked; whether a second, different task will check transfer; the horizon (end of unit, placement, later course).

A request to "add an authentic assessment" does not specify these. Ask what performance, what conditions, who judges it and what the result decides before choosing the task, the number of drafts or the audience.

Sequence and conditional policy

The default sequence is the earlier page's seven steps. Proposed doses are for a unit of about four to six weeks with two or three contact sessions a week; scale them to the brief. Steps 1, 2 and 7 have no claim that tests them; steps 3 to 6 have a neighbouring claim, carried by extrapolation.

  1. Name the performance and build the task (planning, before the unit). Write what a practitioner does, then a task that requires it (application), with realistic conditions for the parts that are the target and plain conditions for the parts that are not. Check it is not a recall test with a realistic wrapper: if a learner could pass by reproducing notes, rewrite it. Evidence status: untested; from the earlier page. A qualitative study of trainee nurses reported that training closer to real clinical scenarios was felt to transfer more smoothly [+W]; it is trainees' reports, not a comparison.
  2. Share criteria and exemplars before work starts (first session, about 30 to 45 minutes). Give the rubric in plain language with two or three exemplars at different levels; have learners score one exemplar and compare their scores with the teacher's. Evidence status: untested as a step on its own; the rubric studies below include it.
  3. Practise the components with support (the first third of the unit, about two to four sessions). Teach and practise the parts the whole task needs with worked examples and scaffolding, on cases simpler than the final one; fade the support across sessions, so early tasks give the data and the steps and later tasks leave learners to define part of the problem. Move a learner to the whole task when a short component check meets the threshold on the core criteria. Evidence status: guided rather than unassisted discovery [~S]: explicit instruction beat unassisted discovery, and discovery with feedback, worked examples or scaffolding beat other instruction; none of its studies used an authentic assessment task.
  4. First draft or rehearsal of the whole performance, with feedback (around the middle of the unit; feedback returned within a week). Learners produce work in progress; feedback names what meets the criteria, what does not and the next step, and comes from the teacher, from peers using the rubric, or both. Have learners self-assess the draft against the rubric before they receive feedback. Evidence status: assessment for learning [+S] (K–12 meta-analyses, small pooled effects of about .20 and .29, larger where students were supported in self-assessment); feedback [+S] (medium average effects, but more than a third of feedback interventions made performance worse, especially feedback aimed at the self); peer assessment [+M] (small-to-medium improvement over no assessment and over teacher assessment); rubric-referenced self-assessment [+M] (higher-rated essays and concept maps in two non-randomised comparisons, with more stress and performance-avoidance in one).
  5. Revise (one cycle by default, two for learners new to the domain or the task form; about one to two weeks). Revision is expected, not penalised; ask learners to say which criterion each change addresses (reflection). Evidence status: carried from the assessment-for-learning claim, whose effects belong to the use of evidence before the summative judgement; untested for this pattern.
  6. Final performance to a real or realistic audience (the last week). Present or deliver the product to the person or group it is for: a client, a community partner, a practitioner panel, a publication, or peers outside the class. Allow a private rehearsal first. Evidence status: authentic audiences [+M]: seventh-graders' compositions to distant peers, and second-graders' writing for an external audience, were rated higher than writing for the teacher; small within-subject comparisons in writing, product quality only.
  7. Judge against the criteria, with assessors checked (after the final performance). Two assessors score a sample of products independently; where scores differ by more than one level on a criterion, they discuss against the exemplars and rescore. Add a short reflection on what the learner would do next time, and where the decision matters, a second, different task to check transfer. Evidence status: untested as a remedy. In one inquiry lab, six scorers' rubric scores spread by about 1.2 of 9 points [~W], which the authors read as an unreliable rubric.
Observation Discriminating probe Proposed action, with its uncertainty
Draft is weak across most criteria Give one component on its own (a single calculation, one paragraph, one assessment step) under plain conditions. If the component also fails, the learner lacks the component: return to supported component practice (step 3). If the component succeeds, the difficulty is coordinating the parts under realistic conditions: give a partly completed version of the whole task. Neither branch is tested.
Strong on the product's form, weak on the reasoning the outcome names Ask the learner to explain one decision in the product and what they would change if one fact were different. If they cannot, the rubric or the feedback may be rewarding form; reweight the criteria and give feedback on the reasoning. If they can, the product may under-show what they know; allow an oral account beside it. Proposed, untested.
Self-assessment far above the teacher's or peers' scores Have the learner score an exemplar with the rubric and compare it with the teacher's score. If they also overrate the exemplar, they do not yet read the criteria as intended: model scoring with worked examples and recheck. If they rate the exemplar accurately, the gap is about their own work: ask them to find evidence in their draft for each level they gave. Supported in direction by the self-assessment training claim [~S], on genetics problems, not authentic tasks.
Assessors disagree on the same product Two assessors score three shared products and say which wording of the rubric led to each score. If the disagreement follows a criterion's wording, rewrite the descriptor and recalibrate with exemplars. If it follows the assessor, recalibrate that assessor. Untested.
Learner does not revise after feedback, or revises only surface features Ask which feedback point they acted on and what they did not understand. If the feedback was unclear or too much, cut it to two or three points tied to criteria, each with a next step. If they understood it but did not value the task, return to what the learner wants from it (inputs). Proposed.
Scores track language proficiency or access to resources rather than the target Compare one learner's written product with an oral or practical account of the same decision. If the account is markedly stronger, the task is measuring something other than the target: allow another form for the non-target parts, or score them apart. Untested.

Fitting the design to a situation

The three facts that most change the decision:

  • How much of the domain learners already have. An open-ended realistic task asks novices to coordinate parts they have not practised; unguided discovery learned less than explicit instruction. If unstated, ask: given a simplified version of the task, can they do each part on its own?
  • What the result decides. A formative task can tolerate inconsistent scoring and invite risk; a grade, progression or licensure decision needs calibrated assessors, a defensible threshold and more than one task. If unstated, ask: what happens to the learner if the score is wrong by one level?
  • Whether the outcome is a performance or a body of knowledge. Authentic tasks fit capabilities that are performances (complex-skill, procedure in context); for well-defined facts or procedures a conventional test may be more efficient and reliable. If unstated, ask: what does a practitioner do with this, and can it be shown without doing it?
If the brief says… Then change… Basis
Learners are novice in the domain (first course, first placement) Spend the first half of the unit on component practice with worked examples; use a simpler, well-structured version of the task for the first draft; plan two revision cycles; give feedback on each component before the whole task. Guided rather than unassisted discovery [~S]; its studies did not use authentic assessment tasks, and the meta-analysis does not break results down by prior knowledge
Learners are advanced or practitioners (final-year students, continuing professional education) Drop most component practice; give an ill-structured task in which learners define part of the problem; one draft with practitioner feedback; judge by a practitioner panel. Untested; from the earlier page
Learners are child (primary school) Keep the task short and concrete with one product, one audience the children know of (another class, families), criteria as two or three "I can" statements, and teacher-led feedback rather than peer scoring. Authentic audiences [+M]: second-graders wrote better for an external audience (one study, abstract only); the rest is untested
The goal is a well-defined procedure or factual knowledge Assess it with a conventional test or a short structured task; keep the authentic task for the capability that needs it, or use it as a capstone after the procedures are secure. Untested; from the earlier page's direct-instruction qualification
workplace-clinical or a placement-based course Use structured clinical or practical stations with checklists and global ratings, or a case simulation, and check with learners whether the simulated case behaves like real clients. Trainees' reports on clinical scenarios [+W] (one qualitative study of neurosurgical nurses); simulated cases judged unrealistic [~W] (10 graduate students, focus groups)
online-self-paced, no teacher present Give the rubric and scored exemplars up front; have learners score an exemplar and see the expert's score before self-assessing; use structured peer review in a timed window; reserve teacher or practitioner judgement for the final product. Peer assessment [+M] (robust to online or offline delivery in the meta-analysis); self-assessment training [~S]; carried to authentic tasks untested
Large class (more than about 60) or little assessor time Use peer assessment with the rubric for the draft; teacher feedback on a sample plus whole-class feedback on common problems; double-score only a sample of finals and products near the threshold. Peer assessment [+M]; structured criteria not shown to beat freeform peer review [~M], so do not expect the rubric alone to lift the peer effect
Short time (one or two weeks, or a single session) Shrink the task to one realistic decision or product, give one feedback point before submission, and drop the public exhibition. Untested proposal
High stakes (grade, progression or licensure) Use at least two tasks or stations, calibrate assessors on exemplars and check agreement before scores are released, set the threshold in advance, and keep a reassessment route. Rubric scores spread across scorers [~W] (one class, six teams); the remedy is untested
Learners anxious or new to public performance Rehearse privately with feedback first; make the first audience small and supportive; grade growth against criteria, not comparison with peers. Untested; from the earlier page. The rubric study with pre-service teachers [~M] found more stress and performance-avoidance with rubric use
Mixed language backgrounds or unequal access to resources Let learners choose the context or product form where the target allows; give the rubric in plain language; score the target reasoning apart from language mechanics; allow oral or multimodal products. Untested; from the earlier page
Learners use AI tools freely outside class Include a live or observed part (an oral defence, a supervised station, a recorded process) and ask learners to explain decisions in the product. Untested proposal
The real product goes to a client or audience who should not see the working (an advice letter, a patient leaflet, a pitch) Split the submission: the product as the audience would receive it, scored on audience criteria (relevance, clarity, usability), plus a short decision memo for the assessor (sources checked, priorities weighed, what is uncertain and how the product handles it), scored on the reasoning criteria; give feedback and the revision on both Untested proposal

Seven rows rest on a claim (novices, children, clinical and simulated settings, self-paced online, large classes, high stakes, anxious learners), and each carries its claim beyond the learners, tasks or settings it studied, as its Basis cell says; the other five (advanced learners, well-defined goals, short time, mixed language and access, AI tools) are untested proposals or the earlier page's guidance.

Choosing configurations from evidence

  • Assessment information used before the final judgement. Assessment for learning improves achievement [+S]: a meta-analysis of US K–12 formative assessment interventions with control conditions (33 studies) found a small positive effect on learning (d = .29), larger where students were supported in self-assessment (d = .61); another K–12 meta-analysis (13 studies) found about .20; a narrative review began the claim. These are mostly classroom formative-assessment programmes on achievement tests, not authentic tasks; carry them to "a draft-feedback-revise cycle before the scored performance is worth building in", no further.
  • Feedback during the draft. Feedback improves learning [+S]: across 435 studies the average effect was medium (d = 0.48), and high-information feedback was larger (d = 0.99) than corrective feedback or praise and punishment; across 607 effect sizes the average was d = .41, but more than a third of the interventions reduced performance, more so when feedback drew attention to the self. This is the limit inside the support: feedback on the work against criteria, with a next step, not grades or praise alone.
  • A real audience for the final product. Authentic Audiences Improve Student Work [+M]: two small within-subject comparisons in writing (44 seventh-graders writing to distant peers against writing to the teacher for a grade, counterbalanced; second-graders writing for an external audience, abstract only) found higher-rated products, with no effect sizes; a review supplies the proposed mechanism. Product quality only, not content learning or a later horizon.
  • Rubrics, self- and peer assessment. Self-assessing against a rubric improved the quality of student work in two non-randomised studies, one of which also found more stress and performance-avoidance [+M]: rubric-referenced self-assessment produced higher essay scores among 162 middle-school writers, and higher expert-rated concept maps and more accurate self-assessment among 218 pre-service teachers, who also reported more stress and a performance-avoidance orientation; both non-randomised, no effect sizes. Peer assessment improves academic performance by a small average amount relative to no assessment or to teacher assessment [+M]: a meta-analysis of 54 control-group studies found peer assessment improved performance against no assessment (g = 0.31) and against teacher assessment (g = 0.28). Structured criteria in peer assessment [~M]: in the same meta-analysis, studies giving reviewers a rubric or script (g = 0.29) did not differ significantly from freeform review (g = 0.42). Self-assessment accuracy is low without training [~S]: secondary students who studied modelling examples of self-assessment were more accurate than those who did not (a randomised experiment, 80 students, genetics problems); in higher education, self-assessed and teacher marks agreed better in advanced than introductory courses (51 studies). So the rubric is a support for judgement that learners must be taught to use, not a guarantee.
  • Extended project work ending in a product. Project-based learning improves achievement [+M]: a meta-analysis of 30 articles (46 effect sizes, 12,585 students) found project-based learning, sustained work culminating in a product or performance, raised achievement against teacher-led instruction (d+ = 0.71), moderated by subject, location, instructional hours and IT support; read from the abstract. It tests the instruction around a product, not the assessment of it.
  • Limits. Guided rather than unassisted discovery [~S]: explicit instruction beat unassisted discovery (580 comparisons, d = 0.38), and discovery with guidance beat other instruction (360 comparisons, d = 0.30); a review argues guidance matters less only once learners have enough prior knowledge. An open authentic task set early to novices is the unguided case. Rubric scoring spread across six scorers [~W]: one class, six teams, a mean standard deviation of 1.2 points on a 9-point rubric. Simulated cases judged unrealistic [~W] and trainees' reports that realistic clinical training transferred more smoothly [+W] are single qualitative studies of learners' perceptions of realism; neither measures learning.

Do not rank authentic assessment against tests from these claims: none compares them, none measures the assessor time an authentic task costs, and none follows learners to the practice the task imitates. Do not read a strong final product as the capability attained without a check on a second task, or a single assessor's score as reliable without an agreement check. A precise abstention names the missing comparison (an authentic performance task with revision against a conventional test of the same outcome, on a transfer task at a stated horizon) and the local observation that would stand in for it (scores on a second, different task; agreement between two assessors).

The evidence by question

Further evidence, not yet read against this model

Claims this page cited before the 2026-10-05 rewrite and not used above. Each keeps its old direction, capped by the claim's recorded evidence.

Illustrative design instance and observation record

Invented illustration, not a tested unit. A second-year undergraduate nursing module (postsecondary, adult, about 90 students, classroom with a skills lab) wants to assess handover communication. Learners have completed one short placement. The table changed the default: because the decision contributes to progression and the class is large, the designer uses two short observed stations (a handover from a written chart, and one from a simulated patient) rather than one long product, scored with a rubric and a global rating; two lecturers calibrate on three recorded exemplars before the stations and double-score every handover near the threshold. Weeks 1–2 practise the handover structure on worked examples with given patient data; week 3 is a recorded rehearsal with peer feedback against the rubric, after learners have scored an exemplar and compared their score with the lecturer's; week 4 is revision and a second rehearsal for those below the threshold; week 5 is the assessed stations, followed by a short reflection. A learner says she wants to stop freezing when a senior nurse asks questions; the designer's objective is a complete, safe handover. They agree to add a question-and-answer minute to the rehearsal, scored only formatively. None of the claims above tested handover, nursing students or this horizon; these are local choices to observe.

Record: performance named and task conditions → criteria and exemplars shared, and learners' exemplar scores against the teacher's → component checks passed and support given → draft or rehearsal, self-score, peer and teacher scores by criterion → feedback points given and which were acted on → revised work → final performance, audience and conditions → each assessor's scores and their agreement → candidate interpretations (component missing, coordination under realistic conditions, form rewarded over reasoning, non-target demands, task not valued) and their basis → second task or later reobservation at the stated horizon. Preserve the learner's stated purpose and any change in it. An agent may summarize this record but must not fabricate missing inputs or assign numerical probabilities of competence without a calibrated model.

Elements and limits

Application, performance-based assessment, rubrics, exemplars, assessment, scaffolding, worked examples, feedback, self-assessment, peer assessment, reflection, audience analysis, simulations, case studies and accommodations. Principles the pattern draws on: Assessment for Learning, Authentic Audiences and Purposes, Competency-Based Assessment, Validity, Reliability and Bias in Classroom Assessment, Scaffolding and Fading and Active Learning. Theories it draws on: Situated Learning, Constructivism and Self-Determination Theory, qualified by Cognitive Load Theory.

This pattern is scoped to judging a performance capability through an extended, realistic task with revision. Knowledge that a test checks well, and decisions that need many independent observations, need other configurations or a mix (direct instruction for the components, formative assessment for the moment-to-moment loop, competency-based learning for progression). The policy supports observation and design reasoning; whether assessing through an authentic task, rather than a conventional test of the same outcome, changes what learners can do later, for which learners, at what cost in assessor time and with what reliability, remains to be tested.

Claims

  • Constructive Alignment — authentic assessment is the assessment half of alignment; outcomes, tasks, and criteria must point at the same competence
  • Cognitive Apprenticeship — authentic tasks provide the realistic practice context in which modeling, coaching, and fading unfold
  • Case-Based Learning — cases supply the realistic problem contexts that authentic assessments often draw on
  • Competency-Based Learning — shares the focus on demonstrated performance against explicit criteria rather than seat time or test scores
  • Formative Assessment — the moment-to-moment loop of reading a response and choosing the next action, which runs inside the draft-feedback-revise cycle
  • Direct Instruction — teaches the components a novice needs before the whole authentic task

Examples

Teacher education — edTPA: Pre-service teachers plan, teach, and video-record a real lesson sequence, then analyze their own teaching against standardized performance criteria scored by trained external evaluators. https://www.edtpa.com

Writing across the curriculum — National Writing Project: Students write for genuine publication and community audiences, revising through peer response cycles rather than submitting single drafts for grades. https://www.nwp.org

Engineering — capstone design courses: Teams deliver a working prototype to an actual client sponsor, assessed by faculty and practicing engineers against professional design criteria.

Medical education — OSCEs (Objective Structured Clinical Examinations): Students perform standardized patient encounters assessed with structured checklists and global ratings — a hybrid that adds scoring reliability to authentic performance. - Blueprint-aligned OSCE station assessment for consent-taking skills - Scenario-based standardized assessments with simulated peer and teacher interaction

Key Sources

  • Wiggins, G. (1989). A true test: Toward more authentic and equitable assessment. Phi Delta Kappan, 70(9), 703–713. doi:10.1177/003172171109200721
  • Wiggins, G. (1998). Educative assessment: Designing assessments to inform and improve student performance. Jossey-Bass.
  • Gulikers, J. T. M., Bastiaens, T. J., & Kirschner, P. A. (2004). A five-dimensional framework for authentic assessment. Educational Technology Research and Development, 52(3), 67–86. doi:10.1007/BF02504676
  • Black, P., & Wiliam, D. (1998). Assessment and classroom learning. Assessment in Education: Principles, Policy & Practice, 5(1), 7–74. doi:10.1080/0969595980050102
  • Sweller, J., van Merriënboer, J. J. G., & Paas, F. (2019). Cognitive architecture and instructional design: 20 years later. Educational Psychology Review, 31, 261–292. doi:10.1007/s10648-019-09465-5
  • Dean, A. C. (2014). The Interactional Dimension of LOA: Within and Beyond the Classroom. Teachers College, Columbia University Working Papers in TESOL & Applied Linguistics. https://eric.ed.gov/?id=EJ1176847
Page metadata
Field Value
type pattern
id authentic-assessment
title Authentic Assessment
description Assessing a capability through an extended task that mirrors real practice, with public criteria, feedback and revision before the final judgement, is expected to improve performance and transfer of that capability for learners who already have its components; no claim here tests authentic assessment against conventional tests, and guidance for novices, untrained self-assessment and scorer disagreement limit it.
status review
grain_size unit, course
author Wiggins (1989, 1998)
moved_from principle/interactional-authenticity-large-scale-assessment
canonical True
generated by claude/unspecified on 2026-10-05
trust unverified — no one has confirmed this page's content
sources 1 — dean-2014

Frontmatter is the page's machine-readable record — see the Schema & Guide.