Draft a retrieval starter for tomorrow's lesson

Draft a retrieval starter for tomorrow's lesson

The retrieval starter is the single best-evidenced five minutes of a school lesson, and roughly nobody in a busy Australian staffroom runs it properly every day. Rosenshine put daily review as Principle 1 for a reason. The Bjorks have thirty years of data behind it. Kate Jones has a book on it. What stops us running it Monday to Friday is not that we don't believe in it. It is that writing nine specific questions, three from yesterday, three from last week, three from last term, takes fifteen minutes at 7:40am, and by 7:40am tomorrow you will not have fifteen minutes. So the class gets whatever's on the slide deck from last year, or the "any questions?" opener, which is not retrieval. The prompt below writes the nine questions in ninety seconds.

The prompt

You are helping me draft a five-minute retrieval starter for tomorrow's [YEAR LEVEL] [SUBJECT] lesson in an Australian school. It is a "do now" activity the class completes silently in the first five minutes, in their books, before I go over the answers on the board.

You are working in the tradition of Rosenshine's Principles of Instruction (Principle 1: daily review), Robert and Elizabeth Bjork on desirable difficulties and the spacing effect, Pooja Agarwal and Patrice Bain's Powerful Teaching, and Kate Jones on retrieval practice in the UK and Australian classroom. That means:
- Retrieval, not re-teach. The questions test what the class already knows, not new content.
- Spaced on purpose. The mix pulls from three time horizons.
- Each question is answerable in 30 to 60 seconds. If a question needs a paragraph, it belongs in a lesson, not a starter.
- Cognitively demanding as a set. "What year did Federation happen?" is fine as one question, but the set overall needs application, not just facts.
- Written in Australian English (realise, colour, analyse). Assume ACARA v9.0 where relevant, and the relevant state study design (VCE, NSW, QCAA, WACE, SACE, ACT, NT, TAS) where I have named one.

CONTEXT

- Year level: [YEAR LEVEL, e.g. Year 8, or VCE Unit 3]
- Subject and topic of tomorrow's lesson: [TOPIC, e.g. Year 9 Science, introducing chemical reactions]
- What we covered LAST LESSON (yesterday or the most recent lesson): [ONE PARAGRAPH, or bullet points, on the learning intention and the two or three key ideas.]
- What we covered LAST WEEK (three to five lessons ago): [ONE PARAGRAPH on the key ideas the class should still be holding.]
- What we covered LAST TERM (or earlier in this term) that this lesson quietly depends on: [ONE PARAGRAPH on the prior knowledge tomorrow's lesson assumes.]
- Anything the class specifically struggled with the first time: [OPTIONAL. If you tell me nothing, I will not fabricate a misconception. If you tell me one, I will design a question that probes it.]

TASK

Write exactly nine questions, in this order:

1. Three questions from LAST LESSON. These are fluent-recall questions. If a student was paying attention yesterday, they should get all three right.
2. Three questions from LAST WEEK. These are firmer. A student who has not thought about it since last Thursday will feel the effort.
3. Three questions from LAST TERM. These are the hardest of the set. This is where spacing is doing its work.

Each question is on its own line, numbered 1 to 9. Below each question, on the next line, put the expected answer in one short sentence, and, where relevant, the most likely wrong answer a student would give, so I can watch for it as I circulate.

Mix formats across the set: at least two multiple choice, at least four short-answer, at least one "explain in one sentence why", and at least one "given [x], calculate or identify or classify [y]". Do not exceed nine questions total.

Then, after the numbered list, add a two-line "watch for" note: the one misconception this starter is most likely to surface, and the one follow-up question I could ask in the whole-class debrief to press on it.

Do not include an activity introduction, learning intention, or any classroom management scaffolding. I already have those. Give me the nine questions, the answers, and the "watch for" note. Nothing else.

What it does well

It writes retrieval, not comprehension, if you give it a specific scope for each of the three time horizons. It will produce nine calibrated questions in the time it takes to open your slide deck, and it will flag likely wrong answers in a way that makes the whole-class debrief more useful than the questions themselves.

Where it falls over

It only knows what you tell it. If your "last term" paragraph is vague, the last-term questions will be vague too. It cannot see what the class actually got wrong last lesson, so if you skip the "struggled with" line you will get a generic set. On niche content (say a specific VCE English text or a state-specific study design point) it will sometimes hallucinate a plausible-sounding but wrong reference, so eyeball the answers before you put them on the board.

Try it like this

For VCE English Unit 3: last lesson's chapter, last week's essay-structure focus, last term's context and metalanguage. Ask it to bank a short quote in question 6 that students would have highlighted three weeks ago.

For Year 7 maths: last lesson's operation, last week's number properties, last term's fractions. Set the "watch for" question to the classic order-of-operations trap.

If you only have three minutes tomorrow: ask it for six questions instead of nine, and drop the "last term" row. Better a real three-minute retrieval than a skipped five-minute one.

Ninety seconds of setup on the way to class beats ninety minutes of re-teaching next Friday.