Educational assessment: align a judgement with its evidence
An assessment combines a purpose, an evidence task and a decision rule. A correct percentage calculation does not establish that the task measures the intended learning, and feedback is formative only when it can inform a next step. This route treats design, marking and decisions as connected questions.
Course topic: Assessment and Evaluation
Study route and exercise: University Reading List editorial project. How we curate · Sources checked:
On this page
Scope and responsibility
For education students and teachers who can identify a learning outcome and calculate percentages. McConlogue is situated in higher education; Wiliam emphasizes classroom evidence and response; Brookhart/Nitko supplies a broader measurement and task-construction route. Scores and policies below are entirely fictional, with no actual learner records or claim about a real institution’s rules.
Choose your starting point
I need a free design-and-marking spine
Use McConlogue for course-level alignment, academic standards, calibration and feedback. Translate the setting carefully if your work is outside higher education.
I need evidence that changes the next lesson
Use Wiliam to design a prompt, interpret responses and choose an instructional response. Recording a grade alone does not specify that response.
I need measurement and score interpretation
Use Brookhart/Nitko for validity, reliability, item/task construction and grading. Keep the intended interpretation of a score visible at each calculation.
Compare the three books
Assessment and Feedback in Higher Education: A Guide for Teachers
Teresa McConlogue · 2020
Free higher-education design spine
- Choose it when
- You want to connect learning outcomes, assessment tasks, shared standards and marking.
- Use another route when
- You need a universal school-testing manual or expect a rubric to remove all judgement.
- Selected reading
- 2020 free PDF: chapters 3 course-unit design,6 marker reliability,7 peer assessment and 8 feedback.
- Limits and trade-offs
- The selected ISBN 9781787353640 is the PDF, not the paperback. It is a complete free reading route with specific reuse conditions; higher-education examples require adaptation of the educational context elsewhere.
Try before choosing
Write a two-criterion rubric, then describe one disagreement that needs calibration rather than averaging.
Check your work: Each criterion names observable evidence and a quality distinction. A numeric total alone does not explain the disagreement.
Embedded Formative Assessment
Dylan Wiliam · Edition: 2 · 2017
Formative evidence-and-response companion
- Choose it when
- You need practical ways to elicit understanding and use it during teaching.
- Use another route when
- You want a techniques list to prove an effect without checking its use or purpose.
- Selected reading
- Second edition,2017, ISBN 9781945349225: select the strategies for eliciting evidence, feedback and learner participation by title.
- Limits and trade-offs
- Its practical classroom focus differs from a psychometrics text. The separately published handbook is another work; no unverified chapter/page assignment or free-full-book claim is made.
Try before choosing
Design a brief response prompt and two different next teaching actions for two plausible response patterns.
Check your work: The prompt discriminates the intended misconception; the planned response uses that evidence while learning can still change.
Educational Assessment of Students
Susan M. Brookhart; Anthony J. Nitko · Edition: 8 · 2018
Measurement and task-construction spine
- Choose it when
- You need to distinguish validity, reliability, feedback, performance tasks and grading.
- Use another route when
- You expect one reliability number or a published cut score to justify every use.
- Selected reading
- Selected eighth-edition 2018 print ISBN 9780134807072: chapters 3–4 validity/reliability,7–8 formative assessment/feedback,15 grading.
- Limits and trade-offs
- The same product page also advertises later digital offerings; match the selected print date and format. Access codes and courseware are separate entitlements.
Try before choosing
Calculate the original weighted score, then write what interpretation its task design would and would not support.
Check your work: Separate the arithmetic rule, achievement evidence and decision threshold; none substitutes for the other.
Check your preparation
What is being judged?
Name the learning outcome and the evidence that could demonstrate it. A recall quiz and an application project can have different roles even when both yield percentages.
Validity or reliability?
Validity concerns evidence for a proposed interpretation/use; reliability concerns consistency of the result. Consistent marking can still consistently reward the wrong thing.
What is the uncertainty interval below?
It is a stipulated set of plausible marks, not a confidence interval, estimated standard error or probability of passing. A robust decision must be the same for every mark in the set.
A six-step study plan
Steps completed: 0 / 6
These are your own study marks, not an assessment. They are not sent to this site.
-
Map purpose to evidence
Selected reading: McConlogue design or Brookhart/Nitko goals/validity.
What to do: Write an outcome→task→criterion→decision map.
Check your work: Every claimed outcome has actual evidence; an attractive task is not enough.
-
Make the rule explicit
Selected reading: Grading discussion in the chosen spine.
What to do: Record maxima, achieved marks, weights, thresholds and any essential-component rule.
Check your work: Weights sum to 1, and raw maxima do not silently become intended weights.
-
Calculate both summaries
Selected reading: The original scoring packet.
What to do: Compute the published weighted percentage and the raw pooled percentage.
Check your work: Explain which weighting each number actually implements.
-
Inspect mark uncertainty
Selected reading: McConlogue marker reliability and the follow-up.
What to do: Propagate the stated project-mark range through the decision rule.
Check your work: Do not label a stipulated interval a statistical confidence interval.
-
Plan calibration and feedback
Selected reading: McConlogue calibration/feedback or Wiliam’s evidence strategies.
What to do: Write one calibration comparison and an actionable feedback response tied to a criterion.
Check your work: The comparison targets a disagreement; the feedback gives a next step, not just another number.
-
Audit the decision
Selected reading: Brookhart/Nitko grading and validity.
What to do: State whether the decision is stable, the evidence needed to resolve it and the limits of the score.
Check your work: Do not average away disagreement or infer an unmeasured competence from a composite.
Worked exercise
Original scoring packet: the maxima are not the weights
An invented course awards a quiz 18/20 and project 48/80. Its published composite weights are 40% quiz and 60% project. A pass requires overall≥70% and project≥60%. Compute the composite and decision; compare with pooling raw points.
Assumptions
The two marks are exact in this main case. Percentages are not rounded before a decision and equality passes each threshold. The quiz is stipulated to elicit recall and the project application; no empirical validity or reliability estimate is supplied. The rule is invented, not an institutional policy.
- Quiz percentage=18/20=90%; project=48/80=60%. Weighted composite=0.40×90+0.60×60=72%.
- Overall 72≥70 and project 60≥60, so the packet’s decision is pass. Both conditions must be checked; the total alone is not the rule.
- Raw pooling gives (18+48)/(20+80)=66%. This is correct arithmetic for 20% quiz and 80% project weighting, which is a different rule from the published 40/60 weights.
- The score does not itself establish whether the tasks justify every proposed learning claim. Report recall/application evidence separately when that distinction matters.
Answer and interpretation
Published composite 72%, project 60%, therefore pass under both stipulated thresholds. Raw pooling 66% implements a different weighting.
Test the result
Verify that 0.40+0.60=1. If a quiz weight w were chosen instead, the same exact marks would produce 60+30 w percent; changing weights changes the decision basis, not the underlying work.
Change an assumption
Changed assumption: a plausible mark range crosses the rule
Now the project’s true mark is only known to lie between 44 and 52 inclusive out of 80; quiz 18/20 stays exact, and the same 40/60 weights and pass thresholds apply. Find the composite range and whether a pass is robust.
- Project percentages range 55% to 65%. The composite is 36+0.75 p, where p is the raw project mark; hence the composite range is 69% to 75%.
- The overall threshold needs 36+0.75 p≥70, or p≥45⅓. The essential project threshold needs p/80≥0.60, or p≥48, which is the binding requirement.
- At p = 44 the result fails; at p = 48 it passes with 72 overall; at p = 52 it passes with 75. The allowed interval crosses 48, so neither pass nor fail is robust across all allowed marks.
Answer and interpretation
The composite range is 69–75%, and the decision switches at raw project 48. The stated evidence does not support one stable classification; calibration/moderation is needed to resolve the mark.
Test the result
Check both endpoints and the binding threshold. The interval is not a probability distribution, so no probability of passing can be calculated without further information.
Common mistakes
Raw pooling masquerades as stated weights
Adding points uses the maxima as implicit weights; that can change the policy.
One total replaces a learning claim
A composite can conceal different evidence across outcomes and essential components.
Uncertainty averaged away
Averaging two disputed marks does not establish which judgement is justified; examine criteria and evidence.
Before moving on
- I can map a learning claim to a task and decision rule.
- I can reproduce the weighting and binding-threshold calculations.
- I can distinguish a stable decision, a conditional mark and the evidence needed for moderation.
Sources and edition checks
Sources checked:
- UCL Discovery: free complete assessment and feedback book
The free 2020 book supports course design, standards, calibration and feedback in higher education.
- UCL Press: edition, format, contents and licence
The PDF contents and format-specific ISBN anchor the named reading route.
- Solution Tree: Embedded Formative Assessment second edition
The selected second edition is a formative evidence-and-response companion, not a measurement manual.
- Solution Tree: primary second-edition catalogue receipt
The publisher distinguishes the selected book from its separate handbook; no classroom effects are invented here.
- Pearson: selected eighth-edition print and assessment contents
The selected paperback is explicitly the 2018 eighth-edition update; later digital dates on the page do not overwrite that identity.
These are proposed study activities, not official course assignments or a claim that every book has been read cover to cover.