One Writing Prompt, Many Spanishes

The prompt is a timed handwritten production task, so scores combine writing proficiency with transcription fluency, test familiarity, and register control. For the heritage speakers, 'accuracy' is probably penalizing colloquial or regional forms while missing their real strength: discourse complexity and vocabulary range. If the rubric counts every nonstandard form as an error without separating register from grammar, it mislabels dialect features as incompetence. For the student from Honduras, 22 words in ten minutes is not a good estimate of Spanish proficiency; oral fluency plus third-grade literacy suggests strong basic interpersonal communication skills but underdeveloped academic language and limited practice with school writing. The dyslexia profile explains slow orthographic retrieval and avoidance, not a lack of ideas. The repeaters may have specific gaps, but one timed prompt cannot locate them. The fairness problem is construct-irrelevant variance: the assessment ranks students on speed and handwriting as much as on writing. Fairness does not require identical tasks; it requires comparable evidence of the same construct. A defensible response is to separate the construct into ideational content, organization, language control, and register, then collect multiple samples: planned, revised, timed, and untimed.

Week 1: diagnostic and community building. Students write a 15-minute ungraded baseline on a familiar topic, then analyze two mentor texts: a personal narrative and a short opinion. Small groups compare how the writers address audience. Flexible grouping: heterogeneous groups for oral rehearsal and idea generation; short, targeted groups for students who need orthographic or decoding work. The heritage speakers lead a 'language noticing' activity on regional versus school register; they are not asked to tutor the whole class. The student from Honduras gets a parallel text at an accessible reading level with the same ideas, plus oral rehearsal and sentence-combining. Students with dyslexia get keyboard access, planning templates, and word banks; the goal is still composing. Week 2: guided drafting. Use a planning frame (audience, purpose, three ideas, evidence), oral rehearsal recorded on a phone, then a draft. Mini-lessons: ten minutes on transitions, ten on verb endings, ten on register. Conference rotation: six students per day, three minutes each, focused on one goal. Week 3: revision and publication. Students revise one piece, write a short reflection, and prepare a one-minute oral presentation. Text types are narrative, opinion, and a short informational paragraph so students practice transfer. Justification: volume and feedback matter more than covering many grammar points; the heritage speakers need explicit academic register, the recently arrived student needs literacy scaffolds, and the dyslexic students need access without changing the composing construct.

Build the common assessment around a shared prompt type and shared rubric, not identical accommodations. Use a two-part task: 15 minutes of planning with a graphic organizer, then 25 minutes of writing, with a choice of keyboard or handwriting. Prompt: a short opinion about a school issue with an authentic audience. Rubric: four criteria-content and ideas, organization and cohesion, language control, and register and audience-each on a four-point scale, with descriptors that separate dialect-appropriate colloquial forms from errors in academic register. Do not count spelling twice. For the 504 plan, provide extra time, keyboard access, spell-check if the team allows it, and a separate setting; do not scribe, because scribing changes the writing construct. For the student from Honduras, allow oral planning and a bilingual glossary, but score the same writing criteria; the student may need a lower score now while showing growth. Keep common: prompt type, planning time, rubric criteria, and anchor papers. Vary: access supports and, if necessary, topics. Compare fall and spring samples for growth, and report the distribution by subgroup without labeling students.