How do you mark consistently against a QCAA ISMG?

What is an ISMG actually asking you to do?

An instrument-specific marking guide is not a checklist of things a response has to contain. Each QCAA senior syllabus supplies an ISMG for each summative internal assessment, and the guide describes the characteristics evident in student responses at each performance level, aligned to the assessment objectives that instrument is assessing. Your job is to match the evidence in front of you against those characteristics and decide which level the work best fits.

The structure is consistent across General subjects. The guide splits into criteria, each criterion splits into performance levels, and each level carries a band of marks rather than a single mark. ISMGs also identify the relative contribution of each internal assessment to a student’s overall subject result, so the marks you award are not a school record that gets translated later. They feed the reported result directly.

Two things follow from that. First, you are making an on-balance judgment about best fit, not accumulating ticks; QCAA describes teachers matching the evidence in student work to the descriptors and judging the best match. Second, because levels are bands, every judgment has two moves in it: which level, then where inside that level. Most disagreement between two competent markers lives in the second move, and almost none of it is visible until someone else looks at the pile.

Why do performance-level descriptors drift across a class set?

Because the descriptors are qualitative. Words like discerning, considered, effective and adequate do real work in an ISMG, but they carry no word count, no length and no worked example. You calibrate them against the responses you have already read. So the meaning you attach to a descriptor at script three is not the meaning you attach at script twenty-eight, unless you do something deliberate to hold it still.

The drift patterns are predictable enough to plan around. Marking in class-list order lets the first handful of scripts set an anchor that the rest of the pile gets measured against, and that anchor is random. Marking a whole response at a time carries an overall impression into every criterion, which drags the criteria toward each other instead of letting each one rest on its own evidence. Fatigue compresses the middle: two hours in, most responses start to look like the same band.

The cost lands in two places. Students receive feedback calibrated to where their name fell in the alphabet, which is not a standard anyone would defend out loud. And the pattern of marks you submit carries variation you cannot account for if a reviewer asks you to walk through it criterion by criterion.

How do you hold a performance level steady across the pile?

Fix your anchors before you mark anything for record. Read five or six responses without awarding marks, choose one that clearly sits in the top band for a given criterion and one that clearly sits mid-band, and write a single sentence for each naming what in that response gave you the characteristic. Those two scripts become reference points. Every later judgment on that criterion is then a comparison against a specific piece of student work rather than against your memory of an adjective.

Then mark criterion by criterion instead of response by response. Take criterion one across the whole class, then criterion two. It feels slower and it usually is not, because you hold one set of descriptors in your head at a time and every comparison is like for like. It also surfaces the halo effect instead of hiding it: when a response with strong analysis scores lower on communication, you see that as evidence rather than an inconsistency to quietly smooth over.

Finish with a boundary pass. Pull out every response you placed at the very top or very bottom mark of a band and re-read that criterion alone against your anchor script. This is where marks actually move, and moving them now costs you twenty minutes. Annotate while you go, noting which part of the response supplied the characteristic you matched, because reconstructing that from memory three weeks later is close to impossible.

What does QCAA check, and when does it check it?

Two separate quality-assurance points sit either side of your marking. Endorsement comes first: schools develop their internal assessment instruments and submit them to the QCAA, where assessors check the instrument against syllabus requirements before it is administered with students. Confirmation comes after you have marked. The QCAA analyses the provisional marks, identifies which samples it wants, and trained assessors review those student responses to check the accuracy and consistency of teachers’ judgments.

Confirmation is not a re-marking exercise, and it helps to say that plainly to a nervous faculty. The QCAA describes it as a quality-assurance process examining whether teachers have used the ISMG accurately and consistently, so results are comparable across Queensland. Schools submit the requested sample responses and annotated ISMGs through the Confirmation application in the QCAA Portal. Each school gets its own sampling design, built from the spread of achievement in the cohort, its size, and the school’s previous confirmation decisions.

This is why annotation matters more than a tidy mark sheet. A confirmed mark can differ from the provisional mark a school submitted, and what makes your judgment hold up is a visible line between a characteristic in the descriptor and the evidence in the response. If your annotation names the paragraph that earned the descriptor, an assessor can follow your reasoning. If it is a column of highlighted boxes, there is nothing to follow.

• Before the task: instrument developed by the school, submitted for endorsement, checked against syllabus requirements
• During marking: evidence matched to descriptors, on-balance judgment per criterion, annotations recorded as you go
• After marking: provisional marks entered, samples selected by the QCAA, responses and annotated ISMGs submitted via the QCAA Portal
• After confirmation: the school receives confirmed results, and marks may move from the provisional position

Does any of this change for Applied subjects or junior classes?

The mechanics change, the discipline does not. Applied subjects use instrument-specific standards rather than ISMGs. Those standards describe the characteristics of responses at five levels, A to E, they are supplied in the syllabus, and the QCAA states schools may not alter them. You are still matching evidence to characteristics and making an on-balance judgment. You are placing the work at a standard rather than inside a mark band.

Junior secondary is where drift runs hardest, because nothing external ever checks it. Year 7 to 10 tasks are marked against Australian Curriculum achievement standards, usually through a rubric the faculty wrote itself, and the anchor-and-criterion routine pays off there twice over: junior reporting gets more consistent, and by the time those teachers pick up a senior class they already mark the way an ISMG expects.

It is worth pushing the same language out to students. Publishing the criteria in plain terms across task sheets, exemplars and study guides narrows the gap between what you actually mark and what students believe you mark, which cuts down the post-results conversations that start with a student telling you they did everything on the sheet.

Where does software fit into marking an ISMG?

A tool cannot make the on-balance judgment for you, and under the QCAA process it should not try. You are the marker of record, and the annotated guide that goes to confirmation has to carry your reasoning. What software can genuinely take off you is the two mechanical causes of drift: the arbitrary order you read in, and the effort of restating the same descriptor forty times in slightly different words.

In practice that means entering the ISMG criteria and their performance-level descriptors as the rubric, then having a first-pass draft produced against those exact words for every response in the set, so the wording applied to script twenty-eight is the wording applied to script three. Jeddle works this way for extended written responses, including expert essay marking against a rubric you enter yourself, returning a suggested level with the evidence it matched. You edit, override and confirm every one before it counts.

The sharper use is as a consistency check rather than a shortcut. Mark a class set yourself, then compare your levels against the drafted ones: wherever JeddAI and you disagree, you have a ready-made shortlist of boundary scripts to re-read before the confirmation sample gets selected. That is the boundary pass described earlier, run with a second pair of eyes that never gets tired at script twenty-eight.