Performance Rubrics: Criteria, Scoring, and Validity Checks

A performance rubric is a scoring guide for observable features of work on a defined task. It should represent the intended knowledge or skill, distinguish meaningful levels, and support the decision the score will inform. A polished grid does not make a task authentic, a criterion valid, or two scorers consistent.

Choose analytic or holistic scoring deliberately

Start with the performance task

Write the task before writing level labels. The task must elicit the evidence named in the objective. Authentic here means the work uses knowledge or skill in a purposeful performance; resemblance to adult work is not enough. Indiana University’s authentic-assessment guidance emphasizes applying learning through meaningful tasks and defined criteria.

Constructed task data: These values were created for rubric practice and are not actual school findings.

A four-by-four observable rubric

Score each criterion separately. The descriptions identify evidence in the response, not traits such as creativity, effort, or sophistication.

Replace weak criteria and level language

Score two constructed samples independently

The short samples below are written for rubric practice; they are not student records or reported classroom results.

Sample A

I recommend Plan A. It costs $18,000 instead of $24,000 and needs 6 maintenance hours each month instead of 10. Plan B shades 420 square metres, 120 more than Plan A, but the student-use table shows only 12 more users at midday. If maximum shade is the committee’s first priority, Plan B is stronger. Under a limited budget, Plan A adds substantial shade at lower cost and maintenance. The counts cover only three clear days, so they may not represent other weather or seasons.

Sample B

I choose Plan B because it has more shade and students will like it. The table says 420 square metres. It is the better plan even though it costs more. The school should choose quality.

Have scorers assign all criterion scores before seeing another score. The table shows a defensible independent application, including one disagreement to calibrate.

Calibrate before consequential scoring

1. Select two to four responses that show different patterns, remove identifying information when authorized, and have every scorer score them independently.
2. Compare criterion scores, not only totals. Each scorer points to the response feature and rubric phrase used.
3. Resolve whether the rubric, task, scorer interpretation, or sample caused each disagreement. Revise ambiguous descriptions before operational scoring.
4. Record annotated anchor responses. Recheck agreement after a scoring break and when a new response pattern appears.
5. Use a defined adjudication route for unresolved differences. Do not simply average scores when the difference reveals a criterion interpretation problem.

Agreement is evidence about consistency under those scoring conditions, not proof that the task measures the intended construct. Review alignment and access separately.

Check access and language demands

Identify language, reading, speaking, handwriting, motor, sensory, technology, and time demands in the task and rubric. Keep a demand when it is part of the objective; reduce or provide an approved route around it when it is incidental. Apply required accommodations and accessibility procedures. Student-friendly wording can improve usability, but do not replace necessary disciplinary terms; define them and provide examples.

Set weights after defining importance

Weight a criterion only when its importance to the objective and decision justifies the influence. In the example, data and reasoning might each receive 30%, counterargument and limitation 25%, and communication 15%. Publish the calculation before students begin. Check the effect with hypothetical profiles: a high communication score should not compensate for absent evidence if evidence is central.

Do not add points for the same evidence in several criteria. Report the criterion profile alongside any total when a minimum performance in one criterion matters. If a criterion is a threshold, such as a safety procedure, state the threshold rather than hiding it inside an average.

Copyable blank analytic rubric

Performance task:
Learning objective or construct:
Decision the scores will inform:
Allowed response routes and required accommodations:

Criterion 1:
4:
3:
2:
1:
Weight or threshold:

Criterion 2:
4:
3:
2:
1:
Weight or threshold:

Criterion 3:
4:
3:
2:
1:
Weight or threshold:

Criterion 4:
4:
3:
2:
1:
Weight or threshold:

Sample responses used for calibration:
Features marking adjacent levels:
Scorer disagreement and resolution rule:
Access or language barrier found:
Revision before use:

Copy blank performance rubric

Review the score before using it

After scoring, ask whether the task elicited every criterion, the descriptions fit actual response patterns, scorers applied them consistently, and access conditions allowed students to show the intended performance. Record revisions. A rubric score describes the sampled work under stated conditions; it should not be stretched into a judgment about a learner’s general ability or potential.

What changed: This revision adds a complete performance task and rubric, independent sample scoring, calibration and weighting guidance, access checks, and a copyable blank rubric.