Do Now: A Teacher’s Guide to Lesson Starters

Walk into a secondary classroom at the start of period one and count the minutes before the first learner does any thinking. Four is common, and five is not unusual. Ofsted's own description of low-level disruption includes "being slow to start work or follow instructions", listed alongside chatting and calling out (Ofsted, 2014).

A Do Now is the standard answer to that problem. It is a short written task waiting on the desk or the board before the teacher says anything, which learners begin the moment they walk in, unprompted and without help. Doug Lemov named the technique in Teach Like a Champion, where it appears as technique 20 of 63, in the chapter on lesson structures. British schools frequently call the same slot a starter, or a Do Now task.

Where most guidance thins out is on the evidence. Ofsted's much-quoted figure came from a YouGov survey of 1,048 teachers and 1,024 parents, and it concerns lost learning time rather than starters. No study has ever tested the Do Now as a named practice, so everything defensible that anyone says about it is inference from three separate literatures: retrieval practice, distributed practice and cognitive load.

That inference is sound, and it remains inference. The practical consequence is that the design of the questions does all the work. In a Year 8 history class, three questions revisiting last week's material on the English Civil War are doing something the evidence supports; the same three minutes spent on a word search are not.

In a hurry? Go straight to the 24 named Do Now strategies, or to the worked examples by subject and year group. The research behind them is below that.

‍

What a Do Now Is, and What It Is Not

A Do Now is a short written task that learners begin unprompted as they arrive, work at independently and in silence for three to five minutes, and complete with nothing but a pen. What separates it from an ordinary starter is not the content but the independence. A task you have to launch cannot begin until the room is already settled, which is precisely the window a Do Now exists to occupy.

‍

Lemov's Four Criteria

Teach Like a Champion specifies four criteria, not five, and they are better read as a set than as a list of desirable features. Each removes a particular way in which the opening minutes leak away.

1. It sits in the same place every lesson, so finding it and starting becomes a habit rather than a decision.
2. It can be completed with no direction from the teacher, no discussion with a neighbour, and in most cases no materials beyond what you have already provided.
3. It takes three to five minutes and produces something written.
4. It either previews the lesson about to start or reviews recently covered content.

Criterion two is the one schools quietly drop, criterion three is the one they overrun, and criterion four is the one this guide deliberately narrows.

‍

The Folk Versions, and What the Technique Actually Says

Five versions circulate in staffrooms that the technique does not support. Each quietly removes the feature that made the routine effective.

What teachers pass round	What the technique specifies
"Any starter counts as a Do Now"	A task requiring explanation fails the independence criterion, and that is the entire distinction between them.
"It should be fun or engaging"	Neither word appears among the criteria. A written product does, because that is what makes the task checkable.
"It can be a paired discussion"	The criteria explicitly rule out discussion with classmates. Paired talk is valuable, and it belongs later in the lesson.
"Five to ten minutes is fine"	Three to five. Beyond that it becomes the lesson rather than the entry to it, and the review has nowhere to go.
"Going through the answers is the starter"	The review should be shorter than the task. A lengthy post-mortem costs more time than the disruption it replaced.

‍

The Five Design Criteria, and Why Each One Is There

Four of these five criteria come directly from Lemov. The fifth is ours, and we treat it as non-negotiable, because a task nobody checks is a task nobody learns from. Three to five minutes is brief enough for the slowest writer to produce something and to leave room for the review; the silence is what makes the written page diagnostic, since a learner who has conferred is no longer telling you what they personally know.

Criterion	What it looks like in the room	Where it comes from
1. Three to five minutes	Four short questions, timed, visible before the bell	Lemov's third criterion
2. Silent and independent	No teacher explanation, no conferring with a neighbour	Lemov's second criterion
3. Pen and paper, nothing else	No logins, no textbook hunt, no equipment to fetch	Lemov's second criterion, which excludes materials you have not already handed over
4. Reviews prior learning	Questions drawn from last lesson, last week, last term	A narrowing of Lemov's fourth criterion, which also permits previewing today
5. Self-checkable	Answers displayed and marked in under two minutes	Ours, not Lemov's. Rowland (2014) found testing with feedback outperformed testing without it

Lemov's first criterion, that the task lives in the same place every lesson, concerns routine rather than design, so it is handled further down.

‍

Prior Learning Rather Than New Material

A Do Now introducing unfamiliar content asks learners to think hard about something new at the one moment when no teacher is available. Sweller (1988) demonstrated that a demanding task can consume most of working memory while producing remarkably little learning. Reviewing partly secure material keeps the load manageable and the retrieval effortful, which is the combination that pays. Our guide to cognitive load theory covers the architecture, and our working memory guide sets out the practical limits.

‍

The Fifth Criterion: It Has to Be Self-Checkable

Design the task so answers can go up on the board and be marked in under two minutes. Rowland (2014) found in a meta-analytic review that effortful recall outperformed recognition, and that testing accompanied by feedback outperformed testing without it. A Do Now whose answers are never revealed is therefore the weakest available version, and also the most common, because it is the version that survives a rushed morning.

‍

‍

‍

The Research a Do Now Actually Rests On

No study has tested the Do Now as a named practice. There is no trial of it, no meta-analysis and no effect size, and anyone claiming otherwise is repeating something they have not checked. What exists instead is three well-evidenced literatures that a properly designed Do Now sits squarely inside, and reading them is what tells you which four questions to write.

‍

‍

‍

‍

Retrieval Practice

Roediger and Karpicke (2006) gave learners identical material, then had some re-read it while others were tested on it. After five minutes the re-readers looked better; after a week the tested group was substantially ahead. That reversal explains why the routine feels less productive than it is. Karpicke and Roediger (2008) extended the point: once something can be recalled, further restudying adds almost nothing while further retrieval continues to pay.

Both were laboratory studies with university participants, so the school-facing citation is Agarwal, Nunes and Blunt (2021), a systematic review of retrieval practice in real classrooms across a range of ages and subjects. The benefit holds up in schools, though the size of it varies considerably. Our guide to retrieval practice works through the classroom detail.

‍

Distributed Practice, and Why the Questions Should Not All Come From Last Lesson

Cepeda and colleagues (2006) synthesised a substantial body of work and established two findings that matter here. Spacing outperforms massing, and the optimal gap widens as the retention interval lengthens. If learners must hold something until a summer examination, revisiting it a week later is insufficient. See spaced practice for planning the gaps across a scheme of work, and for why they have to widen as material ages.

‍

Rosenshine's Daily Review

The first of Rosenshine's ten principles is to begin a lesson with a short review of previous learning, because daily review strengthens earlier learning and leads to fluent recall (Rosenshine, 2012). His illustration is an experiment in elementary school mathematics where teachers spent eight minutes a day on review, checking homework, working through errors and rehearsing skills that needed to become automatic. The study he points to is Good and Grouws (1979).

Two qualifications matter. That eight minutes included marking homework, so it was not a Do Now. And the illustration rests on a single experiment, in one American state, in fourth-grade mathematics, published in 1979. Our guide to Rosenshine's principles of instruction covers the complete set.

‍

A Bank of Do Nows That Actually Work

The quickest way to get a Do Now wrong is to write four questions about yesterday and stop there. The quickest way to get it right is to decide where each question comes from before deciding what it asks. Below is the structure we would recommend, followed by worked examples spanning primary, secondary and sixth form that you can lift and adapt directly.

‍

Where the Four Questions Come From

Allocate each question a job before writing it. This is the entire design decision, and it takes about a minute once the habit is established.

Question	Drawn from	What it is doing
1	Last lesson	Almost everyone succeeds, so nobody begins the day failing.
2	Last week	Distant enough to require effort, recent enough to remain fair.
3	Last half term	The spacing question, and the one that keeps older material alive.
4	Last year, or the prerequisite for today	Reaches furthest back, and connects to the lesson about to begin.

Worked through for that Year 8 history lesson: a cause of the English Civil War from last lesson, the year Charles I was executed from last week, what the Domesday Book recorded from the autumn term, and a power Parliament held beforehand. A retrieval grid does the same job in a format learners mark themselves.

‍

Examples by Subject and Phase

Every set below follows the identical four-slot structure. None requires a device, a worksheet or an explanation, and each can be marked from the board in under two minutes.

Class	The four questions
Year 2, mixed	Write the number 34 in words. What is ten more than 46? Circle the verb in "The dog ran quickly." Write the plural of "box".
Year 4, maths	What is 7 times 6? Write three quarters as a decimal. What is the perimeter of a rectangle 5cm by 3cm? Round 4,382 to the nearest hundred.
Year 6, English	Punctuate: "After lunch we went outside." What is a fronted adverbial? Write the past tense of "to bring". Name one feature of a formal letter.
Year 8, history	Name one cause of the English Civil War. In which year was Charles I executed? What did the Domesday Book record? Name one power Parliament held before the war.
Year 9, science	Write the word equation for photosynthesis. What does a catalyst do? What is the unit of force? What charge does an electron carry?
Year 10, English literature	Who says "Out, damned spot"? Define dramatic irony in one sentence. What is iambic pentameter? Give one quotation showing Macbeth's ambition.
Year 11, French	Give the past participle of "aller". Translate "I have lived here for three years." What gender is "problème"? Write a sentence using "depuis".
Year 12, psychology	Define standard deviation. What is an independent measures design? Name one ethical issue raised by Zimbardo's prison study. What are demand characteristics?

Notice what is absent. No question asks for an opinion, because opinions cannot be marked from the board, and none requires a text in front of the learner, because that counts as equipment.

‍

Formats That Work, and the One That Does Not

Four short-answer questions remains difficult to improve on. A retrieval grid weighted by how far back the content reaches works well once the routine is established, a recall column from a knowledge organiser works where the organiser is embedded, and labelling a blank diagram suits science and geography. The format to avoid is multiple choice whose answers are never revealed, because it surrenders both of the ingredients Rowland (2014) identified while still consuming five minutes. Twenty-four named formats, with their sources and their real running times, are catalogued below.

‍

‍

24 Do Now Strategies, Named and Sourced

There is no shortage of Do Now formats. There is a shortage of formats with a name, an author, and an honest account of how long they really take. Below are 24, grouped by what they ask a learner to do. Each row says where it came from, how many minutes it needs, whether it reviews old material or previews new, and whether learners can mark it themselves.

Eighteen of the 24 review old material only. Four work either way. Two preview new content, and they are flagged, because the placement evidence runs against them. McDaniel et al. (2011) ran quizzes before and after instruction in a real middle school science classroom across a full year: the quizzes placed after instruction produced large gains on the end-of-term exam, and the pre-lesson quizzes did not.

Where a format has no traceable originator, this guide says so rather than inventing one. That is the honest answer for seven of the 24, and it is worth knowing before you credit a colleague with something nobody can source.

‍

Jump to the group you need

• Recall from a blank page, 4 strategies
• Recall spread across time, 6 strategies
• Quizzes that mark themselves, 5 strategies
• Every learner answers, and you can see it, 3 strategies
• Starting from an error, 3 strategies
• Comparison and justification, 1 strategy
• The two that preview, and when that is defensible, 2 strategies
• Four that take longer than they look

‍

Recall from a blank page

These four give a learner nothing to recognise. No stem, no options, no word bank. That is the point: retrieving from nothing is harder and it sticks better. Karpicke and Blunt (2011) found free recall beat elaborative study with concept mapping on a delayed test, which is the strongest single result behind the brain dump.

Strategy	Where it comes from	Minutes	Reviews or previews	Learners can mark it
Brain dump, also free recall or blurting	Named by Pooja Agarwal and Patrice Bain in Powerful Teaching (2019). "Blurting" is UK revision slang with no traceable originator	Up to 5	Reviews	Nothing to mark, which is its weakness
Two Things	Pooja Agarwal, retrievalpractice.org	2 to 3	Either	Ungraded by design
Cops and Robbers	Described by Kate Jones in Retrieval Practice (2019). She may be popularising it rather than inventing it, so credit her with the description	5 to 8	Reviews	No, learners steal from each other instead
Knowledge organiser self-quiz, look, cover, write, check	Joe Kirby at Michaela Community School (2015). The look, cover, write, check routine itself is decades-old UK primary practice with no known inventor	5 to 10	Reviews	Yes

A brain dump needs a specific prompt or learners will not know how much to write, and it needs a check afterwards. On its own it cannot tell a learner what they failed to recall. Cops and Robbers solves that socially: learners write alone against a timer in a Cop column, then circulate and steal anything new into a Robber column.

‍

Recall spread across time

Six formats that deliberately reach further back than last lesson. This is the spacing effect doing the work, and Cepeda et al. (2006) is the synthesis behind it: the longer material has to last, the wider the gap between practices should be. A starter that only ever asks about yesterday keeps nothing alive.

Strategy	Where it comes from	Minutes	Reviews or previews	Learners can mark it
Retrieval grid	Kate Jones, UK history teacher, Retrieval Practice (John Catt, 2019)	5 to 10	Reviews	Partly, then teacher feedback
Last lesson, last week, last term	No traceable originator. It is the retrieval grid logic without the grid. Rosenshine’s daily review is the research ancestor	4 to 6	Reviews	Yes
The Review, 15 questions in three fives	Ben Newmark, "Nothing new; it’s a review", 13 November 2017	About 5	Reviews	Yes, read aloud then self-marked
Flashback 4	White Rose Education, premium tier. Revisits previous steps, blocks, terms and years	About 5	Reviews	Not stated on the source
Retrieval Roulette	Adam Boxer, "A Chemical Orthodoxy", 4 May 2017. A spreadsheet draws five questions from the whole course and five from the current topic	10 to 30	Reviews	Peer-marked on a folded slip
Flashcards and the Leitner box	Sebastian Leitner, So lernt man lernen (1972). Compartments of 1, 2, 5, 8 and 14cm; a correct card moves to a wider compartment, a wrong one goes back to the front	5 to 10	Reviews	Yes

Two honest notes. Kate Jones’s grids are recency-banded and colour-coded, but the specific colour scheme varies and some versions award escalating points for older material while others award none, so describe the mechanic rather than copying one scheme. And Boxer says the ten-question Retrieval Roulette "can take 25/30 minutes which can eat away at your lesson time". He later added six and eight question versions. The canonical form is not a five-minute starter.

Newmark’s Review is the clearest practitioner statement of the position McDaniel et al. support, and he arrived at it from the classroom rather than the literature. Five questions on last lesson, five on the current topic, five on anything covered since starting the subject. He reads them aloud rather than projecting them, which forces the pace and stops learners reading ahead. He calls it "a doddle to plan".

The Leitner box is the only format here where the spacing lives in the object rather than in your planning. That is why it survives a busy term when a teacher-scheduled rota does not.

‍

Quizzes that mark themselves

Self-marking is the difference between a starter that runs every day and one that dies in week three. These five all hand the marking to the learner. The cost is that you see the class picture rather than each individual, which is the right trade for a five-minute task.

Strategy	Where it comes from	Minutes	Reviews or previews	Learners can mark it
Carousel Learning quiz	UK company co-founded by Adam Boxer with Josh Perry and Jose Diaz, built 2020. The commercial descendant of Retrieval Roulette, and what it added was self-marking	5 to 15	Reviews	Yes, against a model answer
Low-Stakes Quiz	Craig Barton, Tips for Teachers. Ten mixed-topic questions, easiest to hardest, aiming at roughly an 80 per cent class average	15, plus 12 for review cards	Reviews	Yes, with a confidence rating per question
5-a-day	Corbettmaths. Five mixed questions from a dated page, no login. Primary, GCSE 9-1, Further Maths and legacy tiers	10 to 15	Reviews	Yes, though the source page does not state this explicitly
Numeracy Ninjas	William Emeny, launched 2015. Foundations for Years 1 to 6, Essentials for Year 7, Plus for Years 8 to 10	5	Reviews	Yes, against a projected answer deck
Cloze or gap-fill, no word bank	No traceable originator. The evidence is the specific part: Slamecka and Graf (1978) on the generation effect	3 to 5	Reviews	Yes

The cloze design rule follows straight from the generation effect. Generating a word yourself retains better than reading it, so a word bank turns generation back into recognition and gives most of the benefit away. If you want retrieval, remove the bank.

Two caveats worth printing. Barton’s full Low-Stakes Quiz routine runs closer to 30 minutes once the review cards are made; the quiz portion alone is what works as a starter. And Numeracy Ninjas was free at launch and is now a free trial then paid membership, which teachers who remember it from 2015 will not expect.

‍

Every learner answers, and you can see it

The problem with four questions in a book is that you find out who understood after the lesson. These three give you the class picture in the first five minutes, while you can still change what you teach. Do not conflate them: they are genuinely different techniques with different inventors.

Strategy	Where it comes from	Minutes	Reviews or previews	Learners can mark it
Mini whiteboard "show me"	Tom Sherrington, teacherhead, 28 August 2012. The ancestor is Dylan Wiliam’s all-student response systems in Embedded Formative Assessment	2 to 5	Either	No, you scan every board at once
Multiple-choice diagnostic question	Craig Barton, tipsforteachers.co.uk and the diagnosticquestions.com bank	2 to 5	Either	Yes, on the reveal
Show Call	Doug Lemov, Teach Like a Champion. "A type of Cold Call that involves taking students’ written work and displaying it to the class"	3 to 5	Reviews	No, the class works on it together

Barton’s design rules for a diagnostic question are the useful part. Every distractor must be plausible and each one should reveal a specific misconception, there must be no ambiguity in the stem or the options, and the question must not need several steps, because then you cannot tell which step went wrong.

Three things that look alike and are not. Sherrington’s mini whiteboards show you everyone’s board at once. Lemov’s Show Call displays one learner’s exercise book to everyone. Wiliam’s hinge question is a mid-lesson checkpoint by design, commonly repurposed as a starter. Boxer adds a fourth variant, the tick trick, where learners keep their board visible during feedback and tick each correct part against it.

‍

Starting from an error

Three formats that put a wrong answer at the centre rather than a right one. The evidence here is specific rather than general. Booth et al. (2013) found that studying incorrect worked examples alongside correct ones improved algebra learning, and Große and Renkl (2007) found the same effect depends on learners being told where to look.

Strategy	Where it comes from	Minutes	Reviews or previews	Learners can mark it
My Favorite No	Leah Alcala, via a Teaching Channel video. Keep the US spelling, it is a proper name	5 to 8	Reviews	Class-corrected, not self-marked
Find the error, or erroneous examples	No traceable originator as a named format, but the evidence base is real: Große and Renkl (2007), Booth et al. (2013)	5	Reviews	Partly
Exit ticket into the next Do Now	Exit tickets are Doug Lemov. The loop into the next lesson’s starter is not established to anyone, and it is often wrongly credited to Dylan Wiliam	5	Reviews	No, you read the slips

Alcala’s routine is worth the detail. Learners answer one warm-up question on an index card and hand it in. You sort the cards fast into a yes pile and a no pile, pick the most instructive wrong answer, and rewrite it on the board anonymously. The class first says what the answer got right, then finds and corrects the error. Learners correct the mistake, not you.

The exit ticket loop is the only format in this catalogue that closes the assessment loop, which makes it the strongest answer to the question teachers actually ask: how do I know what to put in tomorrow’s Do Now. Read the slips after the lesson, find the common gap, write tomorrow’s four questions to target it.

‍

Comparison and justification

One format, and it is the odd one out in this catalogue because it cannot be marked. Structured comparison has a real evidence base: Alfieri, Nokes-Malach and Schunn (2013) meta-analysed learning through case comparisons and found reliable benefits. The cost is that it needs talk, so it needs longer than four questions in a book.

Strategy	Where it comes from	Minutes	Reviews or previews	Learners can mark it
Odd One Out, and Which One Doesn’t Belong	Which One Doesn’t Belong is Christopher Danielson’s book; the classroom site was built by teacher Mary Bourassa. Odd One Out generally has no single originator	5 to 8	Either	No, it is discussion-marked

The design intent of Which One Doesn’t Belong is that a case can be made for every one of the four items, so there is no answer to reveal. That is a feature for oracy and a problem for a five-minute self-marking starter. Use it when you want justification, not when you want coverage. One practical note: the wodb.ca site now redirects to a domain sale page, so do not send learners there.

‍

The two that preview, and when that is defensible

Both of these introduce new material rather than strengthening old, which is the placement McDaniel et al. (2011) found did not produce a gain. They are still defensible, for reasons worth stating plainly rather than dodging, and Lemov’s own fourth criterion for a Do Now explicitly permits preview.

Strategy	Where it comes from	Minutes	Reviews or previews	Learners can mark it
Vocabulary and etymology starter	Alex Quigley, Closing the Vocabulary Gap, for the root, prefix and suffix routine. Vocabulary Ninja publishes the word-of-the-day format in two tiers	3 to 5	Previews	No
Thunks and Big Questions	Ian Gilbert, The Little Book of Thunks. Gilbert credits Matthew Lipman and Philosophy for Children as the lineage	5 to 10	Previews	No right answer to mark

The vocabulary starter earns its place because vocabulary is cumulative: the routine reviews known word families while introducing one new word, so it is doing more review than it looks. It is still not doing what a review quiz does, and this guide will not pretend otherwise.

Thunks are the weakest fit on this guide’s own argument, and the Association for Science Education is right about why: a Thunk is a settling activity to run before the actual starter. It gets a class through the door. It does not strengthen memory. That is a real job, but it is a different job.

The honest reconciliation on placement comes from Pan and Sana (2021), who compared pretesting against posttesting directly. Pretesting reliably helps the specific material the prequestion asked about, and does little or nothing for the material it did not ask about, while review quizzing strengthens everything it covers. So the usable rule is this: if you preview, you are buying attention on one point and paying for it elsewhere.

‍

Four that take longer than they look

These four appear on every list of starters and none of them is really a five-minute task. They are listed separately rather than dropped, because they are genuinely useful in the right slot, and because a teacher who tries to run one as a Do Now will lose the first quarter of the lesson.

• Silent reading, DEAR or ERIC. Typically 20 minutes, so it is a timetabled block rather than a starter. Alex Quigley’s critique is the useful material: learners fake-read or pick narrow, easy texts, the fluent get more fluent while weaker readers gain less, and teachers assume a quiet class is reading. His alternative is structured silent reading: audit what is being read, top and tail the session with a summary and a check, and use the time to work with the weakest readers.
• SSDD problems, same surface and different deep. Craig Barton, ssddproblems.com, created 2014. Sets of problems that look identical but need different processes, built explicitly on the interleaving research of Rohrer, Dedrick and Burgess (2014). More often an end-of-topic task than a starter, but the interleaving logic makes it strong for a mixed-topic class.
• Image and source analysis. What can you see, then what can you infer. Common in history, English and art, and not established to any originator. Usually previews, because the point is normally to open a new source.
• Prediction starters, predict, observe, explain. Previews by design. The attribution to the POE framework is plausible but this guide could not verify it, so no source is claimed here.

Two more formats are widely used and could not be traced to anyone: the retrieval clock and the retrieval mat. Both are real and both are sold by resource sites, but no originator could be established, so this guide names neither an inventor nor a canonical version.

‍

‍

The Review, Where Most Do Nows Fail

The task is the straightforward half. Most Do Nows fail after the pens go down, because the review is either skipped or allowed to swell until it has consumed the first quarter of the lesson. Set a rule and hold it: a three-minute task earns a two-minute review, which means marking all four questions while discussing only the one or two the room actually got wrong.

‍

Self-Checking and Mini Whiteboards

The quickest review is answers on the board and a green pen. It takes ninety seconds, every learner marks their own work, and you circulate reading margins rather than collecting books. Where you need the whole class simultaneously, mini whiteboards give a show-me on the single question you most need answered, and cold calling keeps the sample honest when volunteers would distort it.

‍

When Most of the Class Got It Wrong

Rosenshine puts the optimal success rate during instruction at approximately 80 per cent. A serviceable rule is that if roughly four in five answered correctly, review the question and move on. If substantially fewer did, you have identified a reteaching problem rather than a review problem, and it belongs inside the lesson. This is the moment a Do Now becomes genuine formative assessment rather than a settling routine.

‍

Making It a Routine That Survives

A Do Now is a habit before it is a task, and habits are established in the opening weeks or not at all. Teach the routine as explicitly as you would teach content. In week one, narrate it: this is where it lives, this is when you start, this is what silence looks like. In week two, prompt rather than narrate. By week three, say nothing and let the room run it.

‍

Same Place, Every Lesson

This is Lemov's first criterion and the one teachers break most frequently, usually with good intentions. The slide deck changes, the task migrates to a handout, then the board, then a shared document. Every move reintroduces the question the routine exists to answer. Choose one location and defend it for a term, because a consistent lesson format is easier to sustain than a clever one. Our guide to classroom routines covers the broader pattern.

‍

The Greeting at the Door, and the Learner Who Cannot Start

Cook and colleagues (2018) evaluated greeting learners at the classroom door in American middle schools, using a randomised design, and reported increased academic engaged time alongside reduced off-task behaviour. It is a small single-country study, so read it as encouraging rather than settled. The greeting and the Do Now are the same routine viewed from two sides, since you can stand at the door precisely because the task does not need you at the front. Decide in advance that a late arrival joins the review, and keep a two-question version for the learner who cannot begin.

‍

Adapting It for SEND, EAL and Mixed Prior Attainment

One identical task for an entire class is a design decision carrying a cost, not a neutral default. Kalyuga and colleagues (2003) established the expertise reversal effect, whereby instructional support that helps a learner who knows little can actively hinder a learner who knows more. A Do Now pitched at the middle is therefore slightly wrong at both ends of the room, which is manageable provided the adjustments are cheap enough to make every lesson.

‍

Processing Speed and the Learner Who Never Finishes

The instinct is to allow more time. Reduce the number of questions instead. Two questions completed and marked are worth considerably more than four abandoned, and the timing of the routine stays intact for everyone else. Prepare the reduced set in advance so no negotiation happens at the door, because the aim is that this learner finishes something inside the same window as the class.

‍

Reading Demand, Separated From Retrieval

If the question concerns photosynthesis, the sentence should not be the difficult part. Strip the wording back until the only remaining difficulty is recalling the answer. For learners with English as an additional language this is the highest-value adjustment available, and it costs nothing: shorter stems, one clause, no embedded conditionals, and subject vocabulary retained, because that is precisely what you want retrieved.

‍

When Writing Itself Is the Barrier

The written product criterion exists so the task is checkable, not because handwriting is the objective. For a learner with significant writing difficulty, circling, annotating, matching or answering to a scribe all satisfy the underlying purpose. Agarwal and colleagues (2017) found benefits from retrieval practice were at least as large for participants with lower working memory capacity, though those participants were at university rather than at school, so read it as a reason to include rather than as evidence about SEND.

‍

The Do Now in an International School

Two things differ in an international setting and both press on the fourth criterion. Intakes arrive mid-year from entirely different curricula, so reviewing previously covered content can be meaningless for a third of the room. Staffrooms also mix teaching traditions, so a British-trained colleague says starter, an American-trained colleague says bell work, and each assumes the other means the same thing.

The remedy is to draw questions from the current course rather than an assumed prior curriculum, and to keep a knowledge organiser for the unit so a learner who joined in March has somewhere to recover content from. Where a class spans a wide range of English proficiency, the reading-demand adjustment above matters more than any other.

‍

Limitations and Critiques

The Do Now has no evidence base of its own, borrows one from laboratory studies conducted largely with university participants in the United States, and belongs to a school of practice that has attracted serious academic criticism. None of that makes it a poor routine. All of it should temper how confidently anyone recommends it, and four limits are worth knowing before building a department policy around it.

‍

Low Stakes, or It Does Not Work

Hinze and Rapp (2014) compared high-stakes and low-stakes quizzing directly. Performance on the quizzes themselves was equivalent, so on the day everything appeared healthy. Final test performance, however, was better after the low-stakes version, and only the low-stakes version outperformed re-reading. Grading a Do Now, or allowing it to feel graded, can remove the benefit entirely while leaving every visible sign of it intact.

‍

It Strengthens What It Asks, and Little Else

Pan and Rickard (2018) reviewed whether the benefit of testing carries beyond the exact material tested. It does transfer, but less reliably than the direct benefit on the tested content. For a Do Now that means the four questions you ask are the four things you have strengthened, and the rest of the topic is not covered by proxy. It is an argument for planning questions across a term rather than selecting them the night before.

‍

Does the Effect Hold for Complex Material?

This is a live dispute rather than a settled question, and the two decisive papers were published back to back in the same journal issue. Van Gog and Sweller (2015) argued that the testing effect diminishes or even disappears as material becomes more complex. Karpicke and Aue (2015) replied that it is alive and well with complex materials. For a Do Now the reading is reassuring, since short factual questions are the territory where the effect is least contested.

‍

The Compliance Question

Cushing (2021) examined Teach Like a Champion techniques in an English secondary school and argued that they operate as a form of language and behaviour policing, falling hardest on learners whose speech is furthest from the school norm. Golann (2015) studied a high-performing no-excuses school and argued that its scripted regime produced deference rather than the independence it aimed at. These are serious arguments from serious researchers and they deserve stating at full strength.

Where we land is this. The objection targets a whole-school regime of scripted compliance, and a three-minute silent retrieval task is not that. It acquires teeth when the Do Now is one of forty compliance moments in a day, when it is the only silent activity in an otherwise talkative classroom, or when enforcement falls unevenly. The technique itself does not police speech; the culture assembled around it can. Balancing a silent start with a routine that gives talk back, such as turn and talk, is the straightforward answer.

‍

References and Further Reading

Every reference below was checked twice: once to confirm the record exists, and again against Crossref to confirm it is the paper being cited. A deliberately malformed control was run through the same check and correctly failed, so the check is known to fail when it should. Links go to the publisher or the DOI, never an aggregator.

1. Test-Enhanced Learning. Roediger, H. L., and Karpicke, J. D. (2006). Psychological Science, 17(3), 249-255. Re-reading looked better after five minutes and testing was well ahead after a week. If a Do Now feels less productive than reteaching, this study explains why.
2. The Critical Importance of Retrieval for Learning. Karpicke, J. D., and Roediger, H. L. (2008). Science, 319(5865), 966-968. Once something can be recalled, restudying adds very little while retrieval keeps paying. The argument for asking rather than re-explaining.
3. Retrieval Practice Consistently Benefits Student Learning: a Systematic Review of Applied Research in Schools and Classrooms. Agarwal, P. K., Nunes, L. D., and Blunt, J. R. (2021). Educational Psychology Review, 33(4), 1409-1453. The citation to reach for when someone asks whether this works outside a laboratory. School-based, across ages and subjects, and honest that the benefit varies widely in size.
4. Distributed practice in verbal recall tasks: A review and quantitative synthesis. Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., and Rohrer, D. (2006). Psychological Bulletin, 132(3), 354-380. Spacing beats massing, and the optimal gap grows with how long the material has to last. The reason a Do Now should reach back beyond last lesson.
5. Test-enhanced learning in a middle school science classroom: The effects of quiz frequency and placement. McDaniel, M. A., Agarwal, P. K., Huelser, B. J., McDermott, K. B., and Roediger, H. L. (2011). Journal of Educational Psychology, 103(2), 399-414. The paper that settles where a quiz goes. Quizzes after instruction produced large gains; the same quizzes before instruction did not. Run in a real science classroom across a full year, which is why it outranks a laboratory study here.
6. Pretesting versus posttesting: Comparing the pedagogical benefits of errorful generation and retrieval practice. Pan, S. C., and Sana, F. (2021). Journal of Experimental Psychology: Applied, 27(2), 237-257. The honest counterweight. Pretesting does help, but mostly on the exact material the prequestion asked about. Read alongside McDaniel before deciding that preview never works.
7. Finding and fixing errors in worked examples. Große, C. S., and Renkl, A. (2007). Learning and Instruction, 17(6), 612-634. Studying incorrect worked examples helps, but only when learners are told where to look. The design rule behind a find-the-error starter.
8. Using example problems to improve student learning in algebra: Differentiating between correct and incorrect examples. Booth, J. L., Lange, K. E., Koedinger, K. R., and Newton, K. J. (2013). Learning and Instruction, 25, 24-34. The school-based version of the same finding, in algebra. Correct and incorrect examples together beat correct examples alone.
9. The generation effect: Delineation of a phenomenon. Slamecka, N. J., and Graf, P. (1978). Journal of Experimental Psychology: Human Learning and Memory, 4(6), 592-604. Generating a word yourself retains better than reading it. The reason a cloze starter should not carry a word bank.
10. Learning Through Case Comparisons: A Meta-Analytic Review. Alfieri, L., Nokes-Malach, T. J., and Schunn, C. D. (2013). Educational Psychologist, 48(2), 87-113. The evidence under Odd One Out and any starter that asks learners to compare cases and name the difference.
11. The benefit of interleaved mathematics practice is not limited to superficially similar kinds of problems. Rohrer, D., Dedrick, R. F., and Burgess, K. (2014). Psychonomic Bulletin & Review, 21(5), 1323-1330. Why a mixed-topic starter beats a same-topic one, and the research SSDD problems are built on.
12. The Missouri Mathematics Effectiveness Project: An experimental study in fourth-grade classrooms. Good, T. L., and Grouws, D. A. (1979). Journal of Educational Psychology, 71(3), 355-362. The study underneath Rosenshine's daily review example. Read it before quoting the eight minutes, because that eight minutes included marking homework.
13. Cognitive Load During Problem Solving: Effects on Learning. Sweller, J. (1988). Cognitive Science, 12(2), 257-285. Establishes that hard thinking and learning are not the same thing. The case against a demanding, unscaffolded task at the door.
14. The Expertise Reversal Effect. Kalyuga, S., Ayres, P., Chandler, P., and Sweller, J. (2003). Educational Psychologist, 38(1), 23-31. Support that assists a novice can hinder a more expert learner. The evidence that one identical starter for a whole class carries a cost.
15. Positive Greetings at the Door: Evaluation of a Low-Cost, High-Yield Proactive Classroom Management Strategy. Cook, C. R., Fiat, A., Larson, M., Daikos, C., Slemrod, T., Holland, E. A., Thayer, A. J., and Renshaw, T. (2018). Journal of Positive Behavior Interventions, 20(3), 149-159. A randomised evaluation of greeting learners at the door in American middle schools. Small and single-country, so read it as encouraging rather than settled evidence.
16. Retrieval (Sometimes) Enhances Learning: Performance Pressure Reduces the Benefits of Retrieval Practice. Hinze, S. R., and Rapp, D. N. (2014). Applied Cognitive Psychology, 28(4), 597-606. Quiz scores were identical under high and low pressure, but later learning was better after the low-stakes version. The single best reason to keep a Do Now ungraded.
17. Transfer of test-enhanced learning: Meta-analytic review and synthesis. Pan, S. C., and Rickard, T. C. (2018). Psychological Bulletin, 144(7), 710-756. Asks whether the benefit of testing spreads beyond what was tested. It does, but less reliably, which is an argument for planning questions across a term.
18. The effect of testing versus restudy on retention: A meta-analytic review of the testing effect. Rowland, C. A. (2014). Psychological Bulletin, 140(6), 1432-1463. Recall beat recognition, and testing with feedback beat testing without it. Together those findings condemn the multiple-choice starter that never gets reviewed.
19. Not New, but Nearly Forgotten: the Testing Effect Decreases or even Disappears as the Complexity of Learning Materials Increases. van Gog, T., and Sweller, J. (2015). Educational Psychology Review, 27247-264. One half of a live dispute about whether retrieval practice holds up on complex material. Read it alongside the reply below, never on its own.
20. The Testing Effect Is Alive and Well with Complex Materials. Karpicke, J. D., and Aue, W. R. (2015). Educational Psychology Review, 27317-326. The reply, published in the same issue. Presenting either paper without the other converts an open question into a settled one.
21. Benefits from retrieval practice are greater for students with lower working memory capacity. Agarwal, P. K., Finley, J. R., Rose, N. S., and Roediger, H. L. (2017). Memory, 25(6), 764-771. Frequently quoted as an equity finding. Note before citing it that the participants were university students, so it is not evidence about learners with SEND.
22. Language, discipline and 'teaching like a champion'. Cushing, I. (2021). British Educational Research Journal, 47(1), 23-41. Open access, and the only peer-reviewed article naming Teach Like a Champion directly. Argues from fieldwork in an English secondary school, and worth reading before any department policy.
23. The Paradox of Success at a No-Excuses School. Golann, J. W. (2015). Sociology of Education, 88(2), 103-119. The strongest peer-reviewed version of the compliance critique. It concerns a whole-school regime rather than a five-minute task, which is the distinction that matters here.
24. Principles of Instruction: Research-Based Strategies That All Teachers Should Know. Rosenshine, B. (2012). American Educator, 36(1), 12-19. Principle one is daily review, and this is the source teachers actually reach for. It is a union magazine article with no DOI, so any DOI attached to it elsewhere is wrong.
25. Below the radar: low-level disruption in the country's classrooms. Ofsted (2014). Reference 140157. The origin of the widely repeated 38 days figure, which came from a YouGov survey rather than any timed observation. Its list of typical low-level behaviour is the more useful part.
26. Teach Like a Champion 3.0: 63 Techniques that Put Students on the Path to College. Lemov, D. (2021). Jossey-Bass. The Do Now is technique 20, in the chapter on lesson structures. Go to the book rather than the summaries, because the folk versions have drifted a long way from the criteria.