Revising a Performance Rubric After the Data Comes In
You teach a high school mixed chorus of 32 students and just finished a solo-performance assessment using a 4-point rubric with five criteria: pitches, rhythm, tone, diction, and expression. You scored every student live during class. Mean scores came out at 3.4 for pitches, 3.3 for rhythm, 3.0 for tone, 2.9 for diction, and 2.2 for expression. Two things bother you. First, the tone criterion as written reads 'tone, posture, breath support, and vowel shape' in a single descriptor, and students who sang beautifully but stood badly scored low. Second, expression was the lowest score, but you are not certain you ever taught it. Two students have IEP accommodations allowing them to submit a recording instead of singing live, and you are the only rater.
1. Diagnose the rubric's problems and rewrite the criteria so the scores mean something.
Read the worked answer
Three defects. The tone criterion is double-barreled - it bundles a sound outcome with three physical behaviors, so a student who stands poorly drags down a score that is supposed to describe sound. The criteria have no observable descriptors, so a '3' means whatever you were feeling that period. And 'expression' as written is not measurable.
Rebuild it as five single-construct criteria with three or four levels each, described in behavior a student could self-assess:
- Pitch accuracy: 'accurate throughout, including unaccompanied entrances' (4) versus 'errors cluster at the largest leaps' (2).
- Rhythm and pulse: steadiness across the phrase.
- Vocal tone: resonance and freedom of production, judged by ear alone.
- Diction and text clarity: intelligible consonants, unified vowels.
- Musical interpretation: a stated phrasing plan, dynamic contrast, and evidence the student made a choice.
Move posture, breathing, and alignment into the rehearsal-participation grade, where they belong as behaviors. Publish the descriptors before the assessment so students can self-assess against the same language.
2. The mean for expression is 2.2. What does that number actually tell you, and what do you do with it in the next unit?
Read the worked answer
A low mean on a criterion you never taught is a curriculum finding, not a student finding. It says the assessment measured something outside the instruction, which makes the score invalid as evidence of student learning.
The next unit has to teach interpretation explicitly before it is assessed. Sequence: have students mark a phrasing plan on their own scores - where the phrase peaks, where the breath goes, which word carries the stress - and defend one choice aloud. Model the same passage two ways and have them describe the difference in concrete terms (dynamic, tempo, articulation, vowel color) rather than in adjectives. Use short recorded examples and ask what the performer decided. Build a low-stakes formative check on phrasing before the graded performance.
Then re-examine the 2.2. If expression still comes out low after explicit instruction and practice, the rubric descriptor is probably too vague and needs a rewrite; if it rises, the original number measured a gap in teaching, and you now have the evidence to say so.
3. You are the only rater and two students submit recordings. What do you change to make the scores trustworthy?
Read the worked answer
Single-rater, live, one-shot scoring has two problems: drift within a class period and no way to check yourself.
Fix drift with anchor recordings. Before scoring anyone, select three performances from last cycle - a clear 2, a clear 3, a clear 4 - and score them at the start of each period; if your score on an anchor has moved, recalibrate before continuing. Score no more than a set number of students per period, and rescore a random handful afterward without looking at the first score.
Fix the reliability gap with a second rater on roughly one in five performances, with a stated agreement standard - for example, scores within one point on at least four of five criteria. Disagreements are the useful data; they locate the descriptors that are still ambiguous.
For the two students with recording accommodations, score the recordings the same day, in randomized order with names hidden, against the same anchors, so the accommodation changes the mode of performance and nothing about the standard. Document the accommodation and the scoring procedure in your records.