← Unit 3 review hub

Exam-question integration — design proposal (2026-09-07)

Key takeaways

The corpus is codified: 245 FRQ parts from six exams (2021–2026), every part classified, point-mapped, and CED-mapped; the scoring guidelines and four chief-reader reports distilled into a source-cited playbook of how points are actually earned and lost. Three design decisions are yours; everything else can proceed on the recipes below.

What the evidence says, in one paragraph: calculations carry 37% of FRQ points and justify/explain/predict-with-reasoning another 31%, so two skills dominate — showing work the way AP wants it, and the three-move justification (claim → named principle → applied to THIS system's particles/values). The corpus reduces to 24 recurring archetypes (calorimetry chain, titration curve + Ka, thermodynamic-favorability chain, rate-law determination, Lewis/VSEPR suite, experimental-error critique…), several appearing in ALL six years. Unit 3 alone owns ~19% of FRQ points; Units 6–9 together own ~47% — the FRQ ramp must be strong before the second half of the course. The chief readers repeat the same advice yearly: students lose points to vague justifications, missing work, dropped units/signs, and restating the prompt — all trainable habits, none of them chemistry.

Decision 1 — the question-types module ("How the AP exam asks questions")

Proposed: a slim 4-lesson module, taught early, in the existing line-by-line format:

  1. The two question types — exam anatomy (60 MCQ / 90 min; 7 FRQs / 105 min: 3 long at 10 points, 4 short at 4), what each section rewards, one real MCQ and one real short FRQ shown.
  2. Reading an FRQ — task verbs as contracts: what identify / calculate / predict / justify / explain / estimate each demand structurally (from the playbook's grammar).
  3. Earning the point — the three-move justification with a real scored sample (2023–25 packets include genuine student answers with commentary): why "because of IMFs" earns 0 and what the point-earning version adds; restating the prompt earns nothing.
  4. Calculation habits — work shown or no credit (even when correct), units and sig figs as point-bearing, the consistency doctrine (wrong answers carry forward if work is visible).

Your call: placement. "Toward the start" most naturally means after Unit 1's first gate (students have seen checks before meta-lessons about them). Options: (a) standalone module between Units 1 and 2 — needs nothing from other sessions if built as its own small artifact; (b) inside Unit 1 — that session owns the surface, so it's a handoff; (c) before Unit 3, in my territory, weaker fit with "toward the start". I recommend (a).

Decision 2 — the FRQ capability ramp across units

Proposed staging (skills accumulate; each unit's exam-practice thread uses REAL past parts mapped to it by the registry, delivered through the approved pen-and-paper reveal-gated pattern — student writes, reveals the model answer plus a rubric checklist built from the actual scoring guidelines, self-scores):

Stage Units New FRQ skill layered on
1 1–2 (retrofit-light) single calculation parts; identify-with-evidence parts
2 3–4 the justification grammar; particulate-representation parts; 2-3-part chains
3 5–6 full short FRQs (4 pts) under the real rubric; data/graph analysis parts
4 7–8 full long FRQs (10 pts); experimental-design and error-critique parts
5 9 + review integration archetypes crossing units (favorability chain, titration+thermo); timed sets

Mechanically each unit gains 2–4 "exam practice" lessons at natural points (typically after a gate), each anchored on one real FRQ part or archetype from the registry, teaching the relevant playbook scaffold before the attempt. Units 1–2 retrofits are one light lesson each (their sessions' call).

Decision 3 — end-of-topic tests (~20 items, typically 2 FRQs)

Proposed blueprint per unit test: - 18 MCQs, our authoring, AP-style (the gates' AP-mimicry pattern extended), sampled across the unit's sequences with fresh species — note: College Board publishes no MCQs, so these are authored unless you supply secure-area practice exams (drop PDFs in past_exams/secure/ and they join the codification). - 2 FRQs: one short (4 pts) + one long (10 pts) — a REAL past question where a unit-pure one exists (the registry lists candidates per unit; Unit 3 has several), otherwise authored to a named archetype under the item contracts. Scored by self-check against an SG-style rubric (one checklist line per rubric point, real point values shown). - Delivered as a distinct test page in the course app (same engine; a "test" surface like the gate pages, with the FRQ reveal-gated). Gates stay as they are — slim mid-unit checkpoints; the test is a new end-of-unit artifact. Say the word if you'd rather merge. - Unit 3's test gets retrofitted first (its material is the largest FRQ pool); Units 4–9 get tests built in-line with each build.

What I need from you

  1. Module placement — (a) standalone between Units 1–2 (recommended), (b) in Unit 1, (c) later.
  2. Ramp staging — sign off or adjust the table above.
  3. Test blueprint — 18 MCQ + 1 short + 1 long FRQ per unit: confirm counts; real-FRQs-where- available vs all-authored; gates stay separate?
  4. Do you have secure-exam MCQ PDFs to contribute? (Changes the MCQ sourcing answer.)
  5. Sequencing: settle 1–3 before Unit 4 starts so its test and exam-practice lessons are built in-line (Unit 3 retrofit happens either way).

The evidence base (for reference)