A Speech Rubric With 40% Delivery, Two Flagged Students, and Popularity Contaminated Peer Scores
Ms. Pruitt teaches a required high school communication course. The summative is a five-minute persuasive speech worth 20% of the semester grade, scored live by her on a 100-point rubric: content and evidence 30, organization 15, language 15, delivery 40. Peer evaluators - the audience of 27 - also submit scores that count for 10% of each speaker's grade. Two students are flagged. Devon has a documented fluency disorder and stutters more when reading from a script. Ana's IEP includes test anxiety and she has asked to present to a smaller group. Last term, peer scores correlated with how well liked the speaker was, and three students complained afterward that their delivery scores reflected their accents. Each speech day consumes a full class period, and the school has a cart of tablets that can record.
1. What is wrong with a 40% delivery weight, and how would you re-weight and re-describe the rubric?
Read the worked answer
Forty percent for delivery means a student can earn nearly half the grade for how the message sounds rather than whether it works - and in a communication course the construct is constructing and adapting a message for an audience. Delivery matters, but as a support for comprehension, not as a proxy for confidence or polish.
The re-weighting should reflect that hierarchy: content and evidence 35, organization 20, language 15, delivery 20, audience adaptation and Q&A 10. That keeps delivery consequential while making it impossible to pass on presence alone.
More important than the numbers is what the descriptors say. Rewrite delivery around observable, teachable, comprehension-relevant behaviors: volume audible at the back of the room; pacing that lets listeners process - some pauses, not a metronome; eye contact distributed across sections; notes used as support rather than read verbatim; clear articulation of key terms. Delete descriptors like smooth, natural, confident, fluid, and no fillers. Those words import accent bias and disability bias directly into the score, which is exactly what the three complaints last term described.
Finally, check alignment: every rubric line should name something that was taught and practiced. If a criterion was never taught, it does not belong on a summative. A quick audit of the delivery lines against the unit's lessons usually removes two or three items on its own.
2. Design the assessment conditions and accommodations for Devon and Ana, given one class period per speech day and the tablet cart.
Read the worked answer
The construct is speaking to an audience, so the accommodation must not remove the audience - a solo recording would be a different task. Ana's smaller-group request is compatible: she presents to five or six peers plus the teacher, with the same time limit, the same rubric, and the same Q&A. The audience shrinks; the demand does not disappear.
For Devon, target the conditions that worsen stuttering rather than the disorder. Script reading is the main trigger, so make a keyword outline the expected format for all speakers and let him speak from it rather than a verbatim script. Allow extra time without penalty. Remove fluency from the descriptors for everyone - a step already justified by the re-weighting - so his score cannot be depressed by disfluency that does not impede a listener. Do not require disclosure to the class and do not announce an accommodation.
Use the tablets for both: record every speech, then score from the recording rather than live. This buys three things - a second rater can score a sample, the teacher's standard stays stable across five speech days instead of drifting, and a student can review their own performance. Follow district consent and storage policy for recordings, and delete at the end of the term.
Keep the criteria identical across students. Changing conditions is an accommodation; changing criteria is only legitimate when the criterion itself was invalid, which is the case for fluency descriptors applied to everyone.
3. Keep or drop peer evaluation? Justify with the evidence in the scenario and specify the changes.
Read the worked answer
Keep it, but change what peers are asked and how much it counts. Dropping it would remove something genuinely part of the construct - whether a real audience understood and was moved - and would leave the teacher's single judgment as the only evidence. But in its current form the peer score is measuring likability, and 10% of a grade built on popularity is not defensible.
Changes:
- Change the question. Peers do not rate overall quality. They answer three observable audience questions: could I state the speaker's claim in one sentence; did I know what the speaker wanted me to believe or do and why; could I follow the organization. Comprehension by the audience is direct evidence about communication.
- Train with anchors. Two minutes of a previous speech, scored together, then discuss why responses differed. Untrained peers are the source of the popularity effect.
- Use robust aggregation. Drop the top and bottom 20% of peer responses or take the median, and make responses anonymous.
- Cap the weight at 5-10% and report it separately from the teacher score so it cannot mask or inflate the teacher's judgment.
- Check the correlation. If peer ratings still track social standing after training, convert them to formative feedback the speaker reads and drop them from the grade. That is a data-dependent decision, not a permanent one.
Social validity evidence like this is worth collecting every term; it is cheap and it tells you whether the audience measure is measuring the audience.