Skip to content

Summative Assessment

Principle · All principles
Evidence · 68 claims (15 for, 40 mixed, 13 against) · 57 studies (17 associational, 15 theoretical, 7 review, 6 causal, 6 qualitative, 4 design, 2 quant-synthesis), q1–q4 · 4 of 57 report an effect size · 66 claims rest on one study

Conditional relationship

This page owns the end-point judgment itself: when it is made, what it samples, under what stakes, and what comes before it. For learners at the end of a unit or course who have had instruction and practice on a stated capability (any knowledge_type, from a procedure or concept to a complex-skill), an end-of-sequence judgment is expected to report that capability fairly, and to steer study towards it beforehand, when four conditions hold:

  • It samples the capability at the level the outcome names, with several tasks, not one, and in the proportions the outcome implies (recall, application, analysis, whole performance).
  • The criteria and the cut are set before the work is judged, against the capability (criterion-referenced), not against the rank of other learners.
  • Learners have met the kind of thinking it asks for in low-stakes practice first: retrieval of the content during the course, and items or tasks at the same cognitive level, with feedback.
  • The stakes and conditions do not add a demand the capability does not include, such as speed, test anxiety or a single unrepeatable sitting, unless that demand is part of the capability.

The outcomes are general-achievement and skill-performance on the end-point measure, and how far a learner's result reflects their capability rather than the format (assessment-accuracy). The horizon is the end of the unit or course; nothing here forecasts retention beyond it.

No claim in the wiki tests summative assessment as a design against another way of reaching an end-point judgment, or compares summative designs built this way with designs that are not. The claims below test neighbouring parts: that low-stakes classroom quizzing during a course raised later classroom exam scores; that higher-order practice questions helped on lower-order exam questions; that collected university exam questions sat mostly at the lower levels of Bloom's taxonomy; that learners score lower on higher-order exam questions; that a lower-stress team-based format coincided with higher individual end-of-module scores in a cohort comparison; that test anxiety is associated with lower mathematics scores; and that one class's scorers disagreed using a rubric. The rest are arguments about grading, with no data.

How the converted siblings sit beside this: Validity, Reliability and Bias in Classroom Assessment owns whether any score supports the decision drawn from it, and this page does not repeat its measurement model; Competency-Based Assessment owns the "demonstrated / not yet" threshold decision and assessor agreement; Assessment for Learning and the Formative Assessment principle and pattern own evidence used during instruction, which this page treats as what comes before the end-point; the Constructive Alignment pattern owns whether outcome, activities and assessment ask for the same capability; the Authentic Assessment pattern is one design for a summative performance task; Mastery Learning holds checks with correction before moving on. Multiple Methods of Assessment and Standardized Test Fairness and Bias hold triangulation across methods and large-scale bias.

Default design, while the relationship is untested

The earlier pages' guidance (the principle and the folded pattern, at unit or course grain, using a summative assessment, a final exam or a performance-based assessment), kept as a concrete default. Each step is a design proposal; its evidence status is stated beside it.

  1. Write the end-point first, from the outcomes. Before teaching, list each outcome with the performance that would show it (recall a fact, apply a procedure to a new case, analyse, produce a whole piece of work) and draft the end-point tasks from that list. From the earlier page ("clear target outcomes", "defined outcomes"); untested as a sequence here; alignment itself is held by Constructive Alignment.
  2. Blueprint the sample before writing items. Make a two-way table of outcomes by level and decide what share of marks each cell carries. Check the draft against it: collected university exams were found to sit mostly at comprehension and knowledge level (University examination questions concentrate at the lower cognitive levels of Bloom's Taxonomy, dominated by comprehension and knowledge [~M]; Collected exam questions required no cognitive processes above apply and no originality in analysis or high-level interdependence [~M]); a RAND evaluation reported in a review found that only 2 percent of mathematics items and about 20 percent of English language arts items on NCLB-era state tests asked for higher-order thinking (State tests under NCLB measured little higher-order thinking, with only 2 percent of math items qualifying [~M], second-hand, descriptive). For a criterion-referenced end-point, choose items that represent the skills essential to the outcome and order them as the teaching hierarchy does, so a result can be read skill by skill (Criterion-referenced test items should represent skills essential to learning to read and be arranged in the hierarchy used for teaching [+W], a recommendation about reading tests, carried here to other outcomes). A proposed default: no outcome rests on one item; at least a quarter of marks at the level above "apply" where the outcomes ask for it; untested proposal.
  3. For a complex goal, include at least one whole task. Where the outcome is a complex-skill, include an integrated performance (a case, a project, a practical) alongside shorter items that check components, so a failure can be traced. From the earlier page; the whole-task argument is a design argument, not a tested comparison (A training-design argument, not tested by any study recorded here, holds that practising whole complex tasks improves transfer of complex skills [~W]).
  4. Set criteria and the cut in advance, against the capability. Write the criteria (a rubric for performances, a mark scheme for items) and the standard for each grade before marking, and share them with learners at the start of the unit. Criterion-referenced reporting says whether a capability was reached rather than where a learner ranks (Criterion-referenced tests indicate whether a particular skill or objective has been achieved rather than comparing performance to other test takers [+W], a definitional account). Treat scores just either side of a cut as not reliably different, and review them before deciding (Grade cutoffs are arbitrary because adjacent scores receiving different grades are not significantly different, per the author's statistical argument [~W], an argument). Where the same end-point is used with later groups, keep the cut and its meaning fixed, or say plainly that results are not comparable: in one district's state test, the number of correct answers needed to meet a grade-five mathematics standard fell from 36 to 32 over four years, and students matched on other tests scored higher on the state test in later years (ISAT scores are not comparable over time: students with identical earlier and later test scores earned different seventh-grade ISAT scores depending on the year tested [~W], a district cohort analysis); a promotion cut raised each year pushed eighth-grade retention from under 2 to 11.1 percent (Eighth-grade retention rates surged from under 2 percent prepolicy to 8.3 percent in the first policy year, peaking at 11.1 percent [~W], descriptive counts). Where a later outcome exists, check a cut against it: students who met one state test's college-ready mathematics standard were less likely to need remedial courses than those who met another test's proficient standard, while the two English standards could not be told apart (In mathematics, students meeting the PARCC college-ready standard are less likely to need remediation than students meeting the MCAS proficient standard; in English language arts the standards are not statistically distinguishable [~W], a policy brief). A cut's label ("proficient", "pass") says nothing until it is checked against what it is meant to predict.
  5. Run low-stakes retrieval in every session before the end-point. Five to ten minutes of quiz questions (clickers, paper or software) on earlier and current material each session, with answers given, counting for little or nothing. Classroom quizzing raised later classroom exam scores in middle-school and college courses, as reported in a review (Classroom quizzing delivered by clickers, computer software, or paper improves student performance on classroom exams in middle school and college courses [+M]); the dose of five to ten minutes is an untested proposal. In a survey reported in the same review, most middle and high school students who had had such programmes said frequent retrieval practice helped them feel less nervous about exams (Most middle and high school students surveyed after classroom retrieval practice programs viewed them positively and said frequent retrieval practice helped them feel less nervous about exams [+W], student beliefs, not measured anxiety).
  6. Practise at the level of the end-point. Include in those quizzes and in class tasks questions at the higher levels the end-point will ask for; one course found higher-order practice questions helped even on lower-order exam questions (HOCS clicker questions during learning improve performance on LOCS exam questions more than LOCS clicker questions do [+M]). Practice need not copy the end-point's item format exactly (The benefits of retrieval practice do not depend on an exact match between initial retrieval practice conditions and the final test format [+M], lab and classroom studies reported in a review), but give one full practice run in the end-point's format about a week before it, so the format itself is not new; untested proposal.
  7. Keep the end-point from being the only assessment, and lower the stakes of any single sitting. Spread the summative weight over two or more occasions or tasks (for example a mid-unit task and a final, or a portfolio plus an exam), and offer a reassessment route for a learner who fails one. From the earlier page ("one-shot stakes can distort performance"; "summative designs should not be the only assessment structure"); a lower-stress team-based format coincided with higher individual end-of-module scores in one cohort comparison (Dental students in a revised low-stress TBL format scored higher on the individual summative assessment (iSAT) than medical students in the original high-stakes format, in a non-randomised comparison of two cohorts [+W], two different cohorts, so the format is confounded with the programme; the same cohorts' surveys favoured the lower-stress format on stress, frustration and the perceived fairness of questions: Dental students in a revised low-stress TBL format reported lower stress and frustration than medical students in the original format, in a non-randomised comparison of two cohorts [~W]; Medical students in the high-stakes TBL format perceive questions as more difficult and less fair than dental students in the revised format [~W]). Make at least one of the occasions a different kind of demand (a written paper and an observed performance, or a product and an oral check of it), not two records of the same performance: a conceptual framework proposes that distinct demands support a capability claim better (Proposition 4: evidence gathered through distinct demands supports capability claims better than multiple records of the same performance [+W], a proposition, untested), and an essay argues that some check of whether each learner reached the objectives is needed, though tests are not the only way to make it (Abandoning assessment entirely is not an acceptable alternative to competitive grading, per the author's argument [+W], an argument).
  8. Remove demands the capability does not include. Allow enough time unless speed is part of the outcome; keep reading load and language below the level being assessed; give a calm, predictable setting. Test anxiety is associated with lower mathematics scores, most in timed and evaluative conditions (Math anxiety has a small-to-moderate negative correlation with math achievement (r = −.28) in a meta-analysis of correlations from 223 studies [~M], a meta-analysis of correlations). Untested as a design move here.
  9. Check the scorers before releasing results. For any task marked by judgment, have two scorers mark a sample of the same scripts against the rubric, compare, and settle differences before marking the rest; one class's six scorers using one rubric differed by more than a point in nine (In one community-college lab class, six scorers rating six teams with an engineering rubric differed by a mean of 1.2 of 9 points, which the authors read as a sign the rubric may not be reliable [-W]). Sample size: about one script in ten, at least five, is an untested proposal. How agreement is measured belongs to Validity, Reliability and Bias in Classroom Assessment.
  10. Report by outcome, not only as one grade. Return a result per outcome (reached, nearly, not yet) beside any overall grade, so the learner and the next teacher can see what the grade stands for. Untested proposal; the earlier page's "standards or criteria for judging performance". A faculty memo argues that transcript grades are a "blurred" record of performance, through inflation and changing grading conventions (Transcript grades are a blurred currency of performance information, partly due to grade inflation driven by student pressure [~W], an argument with no data), which is the reason for a per-outcome report beside the grade. Where learners start far below the cut, report their growth on the same measure beside whether they reached it: in one district's English-learner records, learners who had not reached the language-proficiency cut by grade eight had started grade one much lower, gained every year, and still never met a cut that rose with each grade (ELs who did not reach proficiency by eighth grade showed continued growth but started first grade with much lower ACCESS scores, and rising grade-level thresholds meant they never met the proficiency cut [~W], descriptive), and districts assessing broad "future-ready" skills reported more success when they focused on growth than on proficiency (Districts report more success with future-ready skill assessment focused on student growth than on proficiency [+W], district interviews, no learner outcome). Do not use learners' end-point scores to judge their teacher: a policy brief reports that test-based measures of individual teacher effectiveness proved too unreliable for that use (Standardized test data proved too statistically unreliable for determining individual teacher effectiveness, and support for its use is waning [-W], a review of evaluation research).

Across a unit or course (dose and progression). A proposed arc, untested: in the first session, show learners the outcomes, the criteria and one example end-point task; in every session after it, five to ten minutes of low-stakes retrieval, rising from mostly recall in early sessions to mostly application and analysis in late ones; at mid-unit, one graded task worth no more than a third of the total, returned with feedback within a week; about a week before the end-point, one full practice run in its format with worked answers; then the end-point, with a reassessment opportunity two to three weeks later for any outcome not reached. A learner moves on to the next unit when every outcome is reached or has a recorded plan for reassessment; for a course-level final, include earlier units in proportion to their weight in the outcomes rather than only the last unit.

When a condition is not met: if there is no time for low-stakes practice, at least make the end-point's format and criteria known at the start; if only one sitting is possible, widen its sample (more, shorter items and one whole task) rather than raise its difficulty; if no second scorer exists, re-mark a sample of your own scripts a few days later, blind to your first marks, and compare.

Fitting the design to a situation

The three facts that most change the decision:

  • What decision rides on the result, and for whom. A unit grade a teacher can revisit, a course grade on a transcript, and a licence or progression decision call for different breadth, scorer checks and reassessment routes. If unstated, ask: what happens to a learner who passes, and to one who fails, and can it be undone?
  • What the outcome asks the learner to do. Recalling facts, applying a procedure to a new case, analysing, and carrying out a whole performance need different tasks, and a test of one does not report another. If unstated, ask: what would a learner who has reached the outcome be able to do that one who has not cannot, and with what aids?
  • What the learners have met before the end-point. Learners who have practised retrieval and higher-level questions during the course, and seen the format, are judged on the capability; learners meeting the format or the level for the first time are partly judged on that. If unstated, ask: what low-stakes practice of this kind and this format did learners have, and how recently?
If the brief says… Then change… Basis
Learners are novice on the format (first exam of this kind, first-year students, a new kind of task) Add a full practice run in the end-point's format a week before, with worked answers; keep the first summative of the course lower in weight Classroom quizzing delivered by clickers, computer software, or paper improves student performance on classroom exams in middle school and college courses [+M] for quizzing during the course; the practice run itself is untested
Learners are child (primary age) Judge over several short occasions and observed work rather than one long test; read questions aloud where reading is not the outcome untested proposal
Learners are adult professionals, or in adult-workplace training Make the end-point a task from their work, observed or produced on the job, judged against a stated competency threshold; ask what they already do and test only what is new untested proposal; threshold decisions are held by Competency-Based Assessment
Goal is a complex-skill (a design, a case, a clinical or practical performance) Include a whole-task performance, scored with a rubric, plus short component items so a failure can be traced to a part A training-design argument, not tested by any study recorded here, holds that practising whole complex tasks improves transfer of complex skills [~W] (an argument about training, carried here to assessment)
Outcomes ask for analysis, evaluation or creation, not only recall Blueprint the paper so those levels carry marks; expect lower scores on them and practise them in class beforehand University examination questions concentrate at the lower cognitive levels of Bloom's Taxonomy, dominated by comprehension and knowledge [~M]; Students perform better overall on LOCS exam questions than on HOCS exam questions [~M]; HOCS clicker questions during learning improve performance on LOCS exam questions more than LOCS clicker questions do [+M] (one nonmajors biology course)
Stakes are high (certification, progression, licensing) Several tasks or occasions; criteria and cut set in advance; two calibrated scorers on every judged task; review all scripts within a few marks of the cut; a reassessment route In one community-college lab class, six scorers rating six teams with an engineering rubric differed by a mean of 1.2 of 9 points, which the authors read as a sign the rubric may not be reliable [-W]; Grade cutoffs are arbitrary because adjacent scores receiving different grades are not significantly different, per the author's statistical argument [~W] (an argument)
Learners have test anxiety, low confidence, or a history of failing exams Low-stakes quizzing through the course; untimed or generously timed papers where speed is not the outcome; weight spread over several occasions. A brief writing exercise about worries just before the exam is optional: evidence for it is mixed Math anxiety has a small-to-moderate negative correlation with math achievement (r = −.28) in a meta-analysis of correlations from 223 studies [~M] (correlational); Brief expressive writing before an exam can raise test-anxious students' scores, but across the wider literature its effect on test performance is negligible [~M] (helped highly anxious ninth-graders in one study; a meta-analysis found a negligible effect); Retrieval practice benefits learners regardless of trait anxiety level, but higher trait or induced anxiety is associated with smaller testing effects [~M]
Learners have low literacy, or are assessed in a second language Keep the language of items below the level being assessed; offer an oral or demonstrated response where the outcome is not reading or writing; gloss terms that are not the target untested proposal for the item language; for language learners measured on a rising grade-level cut, also report growth on the measure, since learners who start low can gain every year and never cross it (ELs who did not reach proficiency by eighth grade showed continued growth but started first grade with much lower ACCESS scores, and rising grade-level thresholds meant they never met the proficiency cut [~W], one district's records); the measurement case is in Validity, Reliability and Bias in Classroom Assessment
Setting is online-self-paced, phone-only, or some learners remote while others sit in the room Give remote and in-room learners the same tasks, time and permitted aids; prefer open-book items that ask for application to a new case over recall that can be looked up; follow a sample of submissions with a short oral check of the learner's own work; make tasks readable on a phone screen untested proposal
One teacher marks for a large group, or marking time is scarce (volunteers, a busy shift, one tutor) Machine-score the knowledge items; judge a smaller set of performance tasks with a short rubric; double-mark a sample of about one in ten, not all; return results by outcome with one comment each. A computerised adaptive test can shorten a mastery test while reaching nearly the same decision An adaptive mastery testing strategy reduces test length by 30% to 81% while reaching the same mastery decision as the conventional test for 96% of trainees [+W] (a military-training simulation); the marking plan is untested
Time is a single session only (a one-day workshop) Do not report a summative judgment of the capability; give an end-of-session check, report it as immediate performance, and say what it cannot show untested proposal
The end-point is a product or performance for a real audience (an exhibition, a client report, a recital) Judge against the published criteria by trained assessors; record audience response as separate information, not as the grade; keep a component check for each criterion untested proposal; the task design is in Authentic Assessment
Learners work in teams (team-based learning, group projects) Keep an individual component in the summative grade; avoid grading on a curve, which sets team members against one another; consider lowering the stakes of in-module team tests Dental students in a revised low-stress TBL format scored higher on the individual summative assessment (iSAT) than medical students in the original high-stakes format, in a non-randomised comparison of two cohorts [+W] (two cohorts, confounded); Competitive (norm-referenced) grading pits students against one another and discourages cooperation, according to the author's argument [-W] (an argument)
The end-point is an external test the teacher does not set (a state accountability test, an exam-board paper), with stakes for the school or programme Teach and run coursework against the full outcomes, not only the part the test samples; keep time for subjects and outcomes the test does not cover; limit format preparation to one practice run in the test's format, not weeks of test practice; give the course its own end-point on the untested outcomes and read the external score as one sample of what learners can do Approximately 40 percent of educators reported spending less time on social studies and science as a result of the promotion policy [-W]; Some CHSRI small schools struggle to align their themes with CPS graduation standards and PSAE content, and accountability pressure is reshaping instruction [~W] (teachers began modelling their own finals on the state test); Incorporating the ACT into high-stakes accountability has not, by itself, produced instructional practices teaching more high-level skills [~W]; High-stakes testing without investment did not improve achievement and narrowed instruction [-W]; Standardized tests measure only part of what teachers expect students to learn and are not taken in all subjects [~W]; Districts' most frequently stated testing weaknesses are lack of alignment with standards and curriculum and time taken from instruction, each cited by 24 percent [~W]. All are school- or district-level surveys, interviews and trend reports; none tests the moves in this row
The brief names its own end-point (its own exam, rubric, grading scheme or portfolio) Keep it. Change only what the evidence here speaks to: check its items against the outcomes and levels (one blueprint pass), add low-stakes retrieval before it, and calibrate scorers on a sample untested proposal; the three changes rest on the claims cited in the rows above
The capability is one integrated performance (a clinical scenario, a full repair, a live service) and the brief asks for an authentic assessment Keep a single whole-task scenario as the judgment; do not split it into paper items. Put recall and recognition checks into the low-stakes practice before it; inside the scenario, prompt the learner to say decisions aloud from the start and tell them the assessor will not confirm answers, so "is that what you wanted?" gets no reply and is not marked down the whole-task claim [~W] is a design argument; the rest is untested

Ten of the sixteen rows cite a claim. Only the novice, higher-order and anxiety rows rest on studies with learners and outcomes; the rest carry an argument, a cohort comparison, one class's scorer data, a description of exam papers, a simulation, or school- and district-level reports to this setting, and say so. The child, adult, mode, single-session, audience and named-routine rows are untested proposals, as is the item-language part of the low-literacy row.

Observation, state and explanation

What can be observed is a response to a task under stated conditions: the item or task and its level, the time and aids allowed, the answer, the score each scorer gave, and how this compares with the same learner's coursework and low-stakes quizzes. "Has reached the outcome", "panicked" and "was taught to the test" are inferred states. A summative score shows what a learner did on these tasks on this occasion; it is a sample of the capability, not the capability.

Response Plausible explanations Observation that could change the interpretation
High scores on recall items, low on application or analysis items The capability at that level was not reached; higher-level questions were not practised; the higher-level items were badly worded or longer to read Compare with class tasks at the same level; give one higher-level item orally and ask for reasoning; check item wording with a colleague.
Much lower on the final than on coursework and quizzes Anxiety or time pressure on the day; disengaged answering (rapid guessing) on a test that carries little for the learner; coursework was easier or helped; the final sampled different outcomes; forgetting since the unit Compare item by item with coursework on the same outcomes; look at response times on a computer-delivered test for runs of implausibly fast answers; offer a reassessment under lower stakes; ask the learner what happened in the sitting.
Two scorers give the same script different grades Criteria are vague; scorers read a criterion differently; the script is genuinely between levels Have both scorers explain their mark against each criterion; score three anchor scripts together and re-mark.
Most of the group fails one item The item is flawed or ambiguous; that outcome was not taught or practised; it sits above the level taught Read the item with a colleague; check the teaching record for that outcome; look at wrong answers for a common pattern.
A learner just below the cut on overall score Below the standard; within measurement error of it Review the script against the criteria outcome by outcome, and look at other evidence on the same outcomes, before confirming.

These rows are proposals, untested as diagnostics. The disengagement explanation rests on an analysis of a state summative test, which found disengagement had a material impact on individual scores and possibly a minor one on aggregates (Disengagement has a material impact on individual state summative test scores, though its impact on score aggregations may be relatively minor [~M], no statistic printed in the source text read); a second analysis of the same kind found school means, proficiency rates and rankings only minimally affected (Disengagement had only minimal impact on school mean performance, proficiency rates, and relative ranking on the state accountability test [~W]), while another found that even a small amount of rapid guessing could change institutional rankings (Even a small amount of rapid-guessing behavior can impact institutional rankings [~W]). So check engagement before reading one learner's low score, and before ranking groups whose scores are close. An unaided end-point shows only what a learner can already do alone, not what they can do with help, as a Vygotskian argument puts it (Static individual assessment misses developing functions; ZPD-based assessment reveals potential competencies [~W], an argument); report it as that. A reassessment is itself further practice, so record that it took place and when; keep the explanations that remain open rather than choosing the one that fits the grade.

Evidence and qualifications

No claim compares summative designs built on the conditions above with designs that are not, or tests an end-point judgment against another way of reaching one. The claims below test neighbouring parts.

Learners, settings and outcomes differ across these (eighth-graders, university nonmajors, dental and medical students, exam papers rather than learners), so do not rank them by effect label. None measures what an end-point design does to learning beyond the end of the course.

The evidence by question

Further evidence, not yet read against this model

Read against the model on 2026-10-15 and again on 2026-10-19; these bear on it only at a distance and are left here.

Objective and learner-valued goal

The designer's objective at the end-point is usually a defensible judgment that a capability was reached, at a stated level, for a record others will read. Elicit separately what the learner wants from it: a pass, a grade for a scholarship or progression, a licence to practise, proof to themselves that they can do it, or to be able to do the thing in their own life or work. Agreement and divergence are both observations. A learner who values the grade may study the format and the likely questions; one who values the capability may resent a paper that tests recall of what they can already do.

For example, a nurse in a medication-safety course may want to calculate doses confidently on a busy ward, while the course's end-point is a timed written paper of calculation items. The paper serves the course's record; an observed calculation during a simulated ward round, judged against the same criteria, tests what the nurse came for. Record both, and do not read a pass on the paper as the capability on the ward.

What would revise this model?

The expectation should weaken if end-point judgments built with broad sampling, advance criteria, prior low-stakes practice and lowered stakes turn out to report capability no better, by later performance or by agreement with other evidence on the same outcomes, than a single conventional final. If randomised comparisons show that spreading summative weight over several occasions does not raise scores or lower anxiety-linked gaps, step 7 is a preference, not a recommendation. If blueprinting papers towards higher levels lowers scores without improving how well they predict later performance, the level shares should be revisited. If calibrated double-marking adds little agreement over a single trained scorer with a clear rubric, the marking plan can be lighter. Do not protect the model by calling every poor result "not enough practice" or "the wrong format" after the fact.

A summative score at the end of a course, retention months later, transfer to a new setting and use in the setting the learner values are separate claims. The present evidence sets no number of items or occasions, no share of higher-order marks and no scorer-agreement threshold.

Claims

Examples

Key Sources

  • Harlen, W. (2005). Teachers' summative practices and assessment for learning. The Curriculum Journal, 16(2), 207-223. https://doi.org/10.1080/09585170500136093
  • American Federation of Teachers. (2006). Smart testing: Let's get it right. AFT.
  • Popham, W. J. (2004). America's "failing" schools: How parents and teachers can cope with No Child Left Behind. RoutledgeFalmer.
  • Bracey, G. W. (2004). Setting the record straight: Responses to misconceptions about public education in the U.S. (2nd ed.). Heinemann.
  • Linn, R. L. (2005). Fixing the NCLB accountability system. CRESST Policy Brief, National Center for Research on Evaluation, Standards, and Student Testing.
  • Jerald, C. D. (2006). The hidden costs of curriculum narrowing. Center for Comprehensive School Reform and Improvement.
  • Recht, D. R., & Leslie, L. (1988). Effect of prior knowledge on good and poor readers' memory of text. Journal of Educational Psychology, 80(1), 16-20. doi:10.1037/0022-0663.80.1.16
  • Pedulla, J. J., Abrams, L. M., Madaus, G. F., Russell, M. K., Ramos, M. A., & Miao, J. (2003). Perceived effects of state-mandated testing programs on teaching and learning: Findings from a national survey of teachers. National Board on Educational Testing and Public Policy, Boston College.
  • Arduini-Van Hoose, N. (2020). High-stakes testing. In Educational psychology. Retrieved from https://edpsych.pressbooks.sunycreate.cloud. CC BY-NC-SA 4.0.
Page metadata
Field Value
type principle
id summative-assessment
aliases high-stakes-testing-accountability-effects
title Summative Assessment
description An end-of-unit or end-of-course judgment is expected to report a capability fairly, and steer study towards it, when it samples the capability at the outcome's level with several tasks, uses criteria and a cut set in advance, follows low-stakes practice of the same thinking, and adds no demand such as speed or anxiety that the capability lacks; no claim here tests summative assessment as a design.
status review
moved_from pattern/summative-assessment
canonical True
generated by claude/unspecified on 2026-10-19
trust unverified — no one has confirmed this page's content
sources 1 — harlen-2005

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