Designing an Integrated Performance Assessment Under Constraints

Task design (whole class, teacher device, two plays, no streaming needed):

1. Advance organizer, minimal pre-teaching. Students see a three-column organizer labeled ¿Quién habla / ¿A quién? / ¿Para qué? and one prompt: “Escucha dos veces. Toma notas en español sobre lo que se pide y por qué.”
2. First play: global listening, notes only, no questions visible.
3. Second play: students complete the organizer and answer three items that require inference, not transcription:
   • “¿Por qué la hablante menciona el horario del camión?” (reasoning about relevance)
   • “Según el anuncio, ¿qué pasa si la gente deja los materiales en la calle?” (inference from consequence)
   • “¿Qué tono usa la hablante: regaño, invitación o advertencia? ¿Qué palabra o frase te lo indica?” (tone + evidence)
4. Post-task transfer: one short written sentence in which students explain the main purpose to a friend, in Spanish.

Why this is interpretive, not a quiz: the items cannot be answered by matching a word heard to an option. They require constructing a mental model of the text-identifying speaker, audience, and purpose; inferring consequences; detecting tone from prosody and lexical choice. A comprehension quiz asks what did it say; an interpretive task asks what does it mean and how do you know.

On punto de acopio: tolerate the ambiguity first. The co-text (llévelos, no los deje en la calle, el camión pasa los martes) lets a listener infer “a drop-off place.” After the first play I will gloss it once on the board, so no item depends on a single unknown word-and so students practice the strategy rather than panicking. Teaching tolerance of ambiguity is part of the interpretive standard, not a workaround.

The colleague is right about one thing and wrong about the inference drawn from it.

Where they are right: a multiple-choice test is highly standardized, easy to score, and yields stable numbers. Reliability-consistency of scores-is genuinely a virtue, and performance assessment can be unreliable if it is badly built.

Where the argument fails: validity. The question is not “is it consistent?” but “does it measure what the course claims to teach?” Our adopted rubric has four rows: Interpretive, Interpersonal, Presentational, Language Control. A grammar MCQ cannot produce evidence for three of them. It measures metalinguistic knowledge about grammar, which is a different construct from the ability to understand, negotiate, and produce meaning. A student who can conjugate flawlessly and cannot follow a radio announcement would score at the top-plainly a validity failure.

And it is unfair to our population. Nine heritage speakers have never studied grammar terminology; their Spanish is real and their test score would understate it. Meanwhile strong test-takers with weak oral skills would be over-credited. That is construct-irrelevant variance, and it is exactly the sort of thing a parent or an administrator can legitimately challenge.

Parents. “What parents expect” is usually transparency and fairness, not a specific item format. A clear rubric, published in advance, with anchor examples and a plain-language explanation of what each score means, satisfies that expectation better than a percentage that predicts nothing about the student's actual Spanish.

The real work is building reliability into the performance assessment: common administration script, analytic rubric with descriptors, anchor samples calibrated department-wide, and double-scoring a sample. Reliability is not a property of multiple choice; it is a property of disciplined scoring.

Feasibility plan for the 52-minute period:

• Interpretive (whole class, ~15 min): played from the teacher device. No devices required. This solves the largest constraint.
• Presentational writing (~20 min): pen and paper, prompt projected. Collect all at the bell.
• Interpersonal (~12 min): station rotation. Six pairs at a time perform a two-minute task with me rating live on a grid; the other 16 students complete the written prompt simultaneously in their seats with a clear “do not interrupt” norm. Rotate three times, or split the interpersonal over two days with a second task type, since 14 pairs cannot all be heard in 12 minutes.
• Devices: reserve the 12 loaners for recording make-ups only.

Defensible scoring with this volume:

1. Score one dimension at a time across all students, not all dimensions per student. Unfocused scoring drifts.
2. Four-point analytic rubric matching the district rows, with concrete descriptors and a “3” anchor clip and a “3” anchor paragraph to calibrate against.
3. Live grid on a seating chart during the interpersonal task; no memory-based scoring after the fact.
4. Second-rate 10% with a colleague; discuss any gap of more than one point and adjust descriptors, not just scores.
5. Make-ups: because no teacher is present, use a fixed, recorded prompt with a written protocol administered by counseling staff; the interpersonal portion moves to a scheduled in-class slot. A make-up that changes the construct produces scores that cannot be compared.

What I will not do: dilute the task into a 40-item grammar test to make logistics easy. I will instead sample the construct carefully and document the administration so the scores are explicable.