Introduction

Let’s start with a bit of honesty, because otherwise this becomes another polite CPD piece that changes nothing… do-now tasks are everywhere, they look good, they tick boxes but in many MFL classrooms they don’t actually move learning forward in any meaningful way… If that sounds a bit blunt it’s because, in my experience observing hundreds of lessons a year across WL, MFL and EFL learning contexts, what we often see is activity masquerading as learning

What are do-now tasks and where do they come from?

A do-now task is simply a short activity students complete as soon as they walk into the room, usually independently and without explanation, typically lasting three to five minutes; the idea was popularised by Doug Lemov in Teach Like a Champion, and it was originally about behaviour and efficiency, not learning in the deep cognitive sense. This is important because a lot of what we now attribute to do-now tasks was never part of the original design.

Only later did people start saying, “Ah, this is retrieval practice”, linking it to the work of Henry Roediger and Jeffrey Karpicke, or to activation of prior knowledge (Ausubel, 1968), and yes, in theory, it can do all of that, but – and this is th key point—only if it is designed with those principles in mind, which, in many cases, it simply isn’t! and this gap between theory and practice is where most of the damage happens…

What do-now tasks can do when they actually work

When they are done properly—and I stress properly, because this is where things usually go wrong—do-now tasks can be extremely powerful, precisely because they hit that moment at the start of the lesson when attention is relatively high and cognitive load is relatively low (Sweller, 2010), which makes them ideal for reactivating prior learning.

They can support retrieval practice, which we know strengthens memory (Roediger & Karpicke, 2006; Karpicke & Blunt, 2011), but only when retrieval is successful and followed by feedback (Butler & Roediger, 2008; Agarwal & Bain, 2019), they can enable spacing, which is critical for retention (Cepeda et al., 2006; Kang, 2016), and they can help students see learning as cumulative rather than a series of disconnected units, which is absolutely crucial in language learning where forgetting is the default state unless we intervene (Nation, 2013).

But there are a few other advantages that are often overlooked, and in my experience these are the ones that make a disproportionate difference when you get them right.

They create desirable difficulty at a low cost, in the sense that a well-designed do-now forces students to think just enough to engage memory without overwhelming working memory, which aligns with Bjork’s notion of “desirable difficulties” (Bjork & Bjork, 2011), and this matters because effortful recall, when it is successful, strengthens learning more than passive review.

They also provide a form of diagnostic assessment in real time, because within two or three minutes you can see very clearly what students can and cannot do, and this can inform the rest of your lesson in a way that end-of-unit tests simply cannot; in other words, they give you actionable data early, which is pedagogically far more useful than summative data late (Black & Wiliam, 1998).

Another benefit, which is rarely discussed but quite powerful, is that they help build fluency under low-stakes conditions, because students are asked to retrieve and produce language regularly without the pressure of formal assessment, and over time this repeated, low-stakes retrieval contributes to automatisation (DeKeyser, 2007), even if each individual task feels quite small.

They can also strengthen metacognitive awareness, because when students repeatedly encounter language they thought they knew but cannot fully retrieve, they begin—slowly, sometimes reluctantly—to recalibrate their own sense of what it means to “know” something, which is a key aspect of self-regulated learning (Dunlosky & Metcalfe, 2009).

And finally, they create lesson continuity, not just in the sense of revisiting content, but in signalling to students that what they learned yesterday, last week, or last term still matters today, and that learning is something that accumulates rather than resets, which, although it sounds obvious, is something many learners do not naturally perceive unless it is made explicit.

Let me give you a simple example of a do-now task that actually works.

Bad version (what I often see):
“Translate: I play football / I watch TV / I go to the park”
→ all from yesterday’s lesson
→ easy, quick, forgotten in 10 minutes

Better version:
“Translate: When I was ten I used to play football / Yesterday I went to the park / Tomorrow I will watch TV”

Now what’s happening?

• multiple tenses
• retrieval from different points in the curriculum
• actual thinking

Even better:
“Write one sentence about the past, one about the present, one about the future using these verbs: giocare, andare, guardare”

Now you’re getting:

• retrieval
• manipulation
• transfer

Same time, completely different impact.

Why they go wrong in MFL (and they really do)

Now here’s where it gets slightly uncomfortable, because the issue is not the idea of do-now tasks, it’s how we use them… and, frankly, how casually we sometimes design them

1. They are too narrow

Most do-now tasks focus on the current unit, which makes sense superficially but is disastrous for retention; and this is why you get the classic situation where students can perform reasonably well in a unit test and then, three months later, that language has vanished as if it had never been taught, which is not a failure of teaching but a failure of retrieval over time (Cepeda et al., 2006; Kang, 2016).

Example of poor practice:
“Translate: I eat pizza / I drink water / I like ice cream”

Better:
“Translate: I ate pizza yesterday / I used to drink water every day / I will eat ice cream tomorrow”

Now you’re forcing:

• interleaving
• retrieval across time

2. Retrieval is introduced too early

This is a big one, and it links to recent discussions around the limits of retrieval practice (Carpenter, 2022; Nakatsukasa, 2023; Yang et al., 2021).

Teachers often ask students to retrieve language that has not been sufficiently processed, which leads to:

• guessing
• copying
• fossilising errors

I’ve seen this countless times. Students are introduced to a structure, then imediately asked to produce it from memory, and the teacher thinks, “Great, they’re retrieving”, but what is actually happening is approximation under pressure, which is not the same thing (Sweller, 2010; DeKeyser, 2007).

If they can’t do it with support, they won’t do it without support…

3. They are disconnected from the lesson

This one drives me slightly mad.

The do-now happens, students finish it, it is checked quickly, and then we move on as if it never happened; and at that point you’ve wasted one of the most cognitively valuable moments in the lesson (Rosenshine, 2012).

If you retrieve something, you must use it again.

Otherwise:

• it goes back into long-term memory
• and stays there… unused

4. They are too easy (or too mindless)

A lot of do-now tasks are designed to be “quick wins”, but in doing so they become mechanical, and here is the uncomfortable truth:

Effort matters (Bjork & Bjork, 2011)

If students are not thinking, they are not learning.

How to fix them (without reinventing the wheel)

You don’t need to scrap do-now tasks. You need to upgrade them… and the good news is that the fixes are not complicated, they just require intention.

1.Interleave previously learned (secure) items

The aim is to mix language that has already been stabilised in long-term memory, not to force complexity for the sake of it; interleaving strengthens retrieval when items are secure (Cepeda et al., 2006), but if you mix unstable material you simply overload working memory and trigger guessing (Sweller, 2010)… in other words, don’t interleave fragility.

Think:
“What is ready to be retrieved and recombined?”, not “How many tenses can I cram in?”

Year 7 (early KS3 – no interleaving yet, just stabilisation):

• I play football
• I watch TV
• I go to the park

Same structure, high repetition → building form–meaning mapping

Year 8 (emerging interleaving – safe and controlled):

• I play football every day
• Yesterday I played football
• I don’t play football

Mix:

• present + past
• affirmation + negation
  BUT only with high-frequency, well-rehearsed verbs

Year 9–11 (full interleaving — now it pays off):

• When I was ten I used to play football
• Yesterday I played football with my friends
• Now I play basketball instead
• Next week I will play in a match

Now you get:

• multiple time frames
• contrast
• transfer
  because the language is secure enough to handle it

2. Control difficulty

Aim for:

• ~70% success (Rowland, 2014)
• not guessing, not trivial

3. Recycle in the lesson

If the do-now includes:
Quando avevo dieci anni giocavo a calcio

Then later:
Ora invece gioco a tennis

That’s transfer.

4. Use partial support

Don’t go from full support → nothing.

Use:

• sentence builders
• prompts
• partial translations

Where do they fit in MARS EARS?

Do-now tasks sit after modelling and processing, not before.

What they are:
retrieval + structured production

What they are NOT:
– initial exposure
– blind production

In other words:

• language must be seen, heard, processed first
• THEN retrieved (VanPatten, 2015)

If you reverse that sequence, you are not doing retrieval practice, you are doing guessing practice…which will likely lead to fragile learning

Conclusion

Do-now tasks are not the problem. Poorly designed do-now tasks are. And if students forget what you taught them in Term 1 by Term 3, the issue is not that they are lazy or unmotivated, it is that the system has not forced that knowledge back into working memory often enough, and no amount of engagement or “fun” activities will fix that (Nation, 2013; Cepeda et al., 2006).

So next time you design a do-now, ask yourself: Am I just starting the lesson… or am I making memory work?

Because only one of those leads to learning.