Writing rubric descriptors that markers agree on

When two tutors give the same essay a 58 and a 71, the rubric is usually part of the cause. Descriptors like "good critical analysis" feel precise to the person who wrote them and mean something different to everyone else. This guide shows how to rewrite descriptors so that markers can point to evidence in the script, and gives a set of before and after rewrites to work from.

Why vague descriptors cause disagreement

Most disagreement between markers is not about standards. Experienced tutors usually agree on which scripts are strong and which are weak. They disagree about the middle, and about which words in the rubric the script satisfies.

Vague descriptors cause this in three ways.

• Evaluative adjectives with no anchor. "Clear", "thorough", "sophisticated" and "appropriate" all depend on what the marker is comparing against. A new sessional tutor and a unit coordinator with ten years of scripts behind them will read "sophisticated" differently.
• Quantity words. "Some", "several", "most" and "limited" invite counting, and markers count differently. Is three references "several"?
• Levels separated only by intensity. If the difference between Credit and Distinction is "good" versus "very good", the marker has no way to decide except gut feeling.

What an observable descriptor looks like

An observable descriptor names something the marker can find in the work. It usually answers one of these:

• What does the student do? (compares, justifies, identifies, applies)
• Where does it appear? (in the introduction, in each paragraph, in the discussion)
• What is present or absent? (a counterargument, a limitation, a unit of measurement)
• How is it connected? (evidence linked to a specific claim, recommendation follows from findings)

You do not need to remove judgement. "Explains why the evidence supports the claim" still needs a marker to decide whether the explanation works. But it tells them what to look for, and that is where most agreement comes from.

Before and after rewrites

Notice what the rewrites do. Some replace adjectives with actions. Some specify location. Several describe what separates this level from the one below. None of them are longer than two lines.

Techniques that work

Write the difference between levels, not the level alone

Put two adjacent descriptors side by side and ask: what would I see in a Distinction script that I would not see in a Credit script? Write that difference into the higher descriptor. If you cannot name it, you may have one level too many.

Replace counts with functions

Instead of "uses at least five sources", say what the sources should do: "uses sources to support each main claim, including at least one that challenges the position". If a minimum number really matters, state it as a requirement outside the rubric so it does not distort the levels.

Describe the fail level by what is missing

"Does not meet the requirements" gives neither marker nor student anything to work with. "No clear position is stated, or the position changes between sections" does.

Keep one idea per cell where you can

A descriptor that bundles three features ("clear, well evidenced and critically engaged") forces a marker to decide what to do when a script has two out of three. If the features matter separately, they may belong in separate criteria. If they genuinely go together, say which one is essential.

Borrow the words students will see

If the assessment brief asks students to "evaluate two policy options", the rubric should talk about evaluating two policy options, not about "critical engagement with alternatives". Matching language helps students and markers read the rubric the same way.

Testing whether descriptors work

The only real test is to have more than one person use them.

1. Choose four or five scripts that span the range. Past submissions (with permission and identifiers removed) or samples you write yourself both work.
2. Ask two or three markers to mark them independently, recording the level per criterion and, briefly, the words in the descriptor they relied on.
3. Compare. Where markers agree, move on. Where they differ by a level, ask each to point to the evidence they used.
4. Rewrite the descriptor so that the evidence one marker used is named explicitly.

A useful sign that a descriptor is still too vague: when asked to justify their level, markers reach for different words from the ones in the cell. If three tutors all say "it didn't really answer the question", that phrase probably belongs in the rubric.

A short worked example

A second year history rubric had this Credit descriptor for "Use of primary sources": "Makes good use of a range of primary sources." In calibration, one tutor gave a script Credit because it quoted six primary sources; another gave it Pass because every quote was used as illustration, not analysed.

The team rewrote the ladder:

On the next calibration round, the same two tutors placed that script at Pass, and the conversation took two minutes instead of fifteen.

Keep the rubric readable

It is possible to overcorrect and write descriptors so long that nobody reads them. Aim for one or two lines per cell. If you need more, the extra detail can go in a marker guide that sits beside the rubric: worked examples, common borderline cases, and notes on what not to penalise. Students see the rubric; markers see both.