Why checking grammatical understanding is not the same as assessing grammatical acquisition

Introduction

One of the most worrying trends in MFL education, in my view, is the growing tendency to confuse successful performance in classroom assessment tasks with evidence of language acquisition, as though students correctly answering a diagnostic multiple-choice question, completing a translation, identifying a grammatical error or explaining a rule necessarily meant that they had ‘got it’.But have they? And, more importantly, what exactly have they got?

This is precisely the problem I see in Silvia Bastow’s recent blog post, “Do They Really Get It? Rethinking Formative Assessment and How to C4U (Part 1)”, which advocates mini-whiteboards, diagnostic questions, retrieval practice, interleaving and corrective feedback as ways of checking students’ understanding, yet pays insufficient attention to a fundamental distinction that decades of research in instructed second-language acquisition (ISLA) have highlighted: knowing a grammatical rule, recognising its correct application and being able to process and use it accurately, efficiently and automatically are very different things.

And surely this is a rather serious omission in an article specifically concerned with formative assessment? After all, if the whole purpose of formative assessment is to generate evidence that informs our next instructional decisions, what happens when we misunderstand what that evidence actually demonstrates? What happens when we take a correct answer as proof of learning, when all it may demonstrate is that a student can perform successfully under the very specific conditions created by the task?

Consider a student who correctly identifies elle est allée in a multiple-choice question but continues to produce elle a allé in spontaneous speech, or another who can confidently explain German verb-final word order after weil yet repeatedly gets it wrong in conversation. Or take a student who scores highly on a French tense quiz but cannot distinguish the same tenses when listening two connected speech… Has that student really ‘got it’? Or have we simply assessed the aspects of grammatical knowledge that are easiest to test, while leaving the more demanding aspects of linguistic processing and production largely untouched?

These are not minor discrepancies. They expose the limitations of such assessment tasks as indicators of developing linguistic competence, limitations which become particularly problematic when teachers interpret successful performance as evidence that a construction has been acquired and that they can safely move on.

And this brings me to a wider concern, one that I have raised repeatedly over the years: the increasingly uncritical transfer of techniques and practices from general education into MFL, often accompanied by the fashionable label ‘evidence-informed’, as though findings from cognitive psychology and general educational research could simply be transplanted into language teaching without first considering what decades of ISLA research tell us about input processing, form–meaning connections, developmental sequences and automaticity.

The danger, of course, is that we create an illusion of learning, whereby students appear successful because they perform well on the tasks we set, while crucial aspects of their developing linguistic competence remain unassessed. And if that happens, what are the consequences? Teachers may move on prematurely, misdiagnose persistent grammatical errors, devote precious classroom time to explaining rules students already know, or neglect the listening, reading and spontaneous production skills that students will eventually need when confronted with unfamiliar language in their GCSE examinations.

Because let’s not forget that, when a student sits a GCSE listening paper, there is no teacher pausing after every sentence to ask a diagnostic question, no mini-whiteboard on which to select the correct verb ending, and no helpful prompt reminding them which tense to retrieve. They have to process the language as it unfolds, identify grammatical distinctions under time pressure and construct meaning from what they hear… and if our formative assessment has consistently failed to reveal weaknesses in those processes, how well have we actually prepared them?

Badly interpreted formative assessment doesn’t merely fail to identify learning gaps; it can actively contribute to poor instructional decisions, precisely because what teachers believe students have mastered inevitably influences what they choose to teach, practise, recycle or abandon.

And that is why I find Bastow’s treatment of the issue problematic. The question is not whether mini-whiteboards, retrieval quizzes or diagnostic questions are useful, but whether the evidence they generate actually justifies the conclusions teachers draw from it. Because if we cannot distinguish between remembering a rule, recognising a form, processing it for meaning and deploying it spontaneously, then what exactly are we checking for understanding… and, more importantly, whose understanding of language acquisition really needs checking?

1. Knowing a rule is NOT the same as acquiring it

Take German weil clauses, for example, which students may have practised repeatedly through rule explanations, sentence corrections and mini-whiteboard exercises, and imagine asking a Year 10 class where the verb goes after weil. “At the end, Sir!” they confidently reply, and when you subsequently give them five sentences containing word-order errors, most correct them without difficulty. So far, so good… but what happens when, a few minutes later, you ask those same students to explain spontaneously why they dislike school?

Ich mag die Schule nicht, weil es ist langweilig.

Sound familiar? And if it does, should we automatically conclude that they haven’t understood the rule, that we need to explain it again, or that another round of corrective questions will somehow resolve the problem?

As I have often notices throughout my career, is that the issue is that students may know the rule perfectly well, and may even be able to explain and apply it successfully when their attention is explicitly directed towards it, while still lacking the processing efficiency required to retrieve and deploy the construction when simultaneously thinking about what they want to say, selecting vocabulary and organising their utterance.

Knowing ABOUT a construction and being able to USE it automatically are two very different things, as research on skill acquisition and automaticity reminds us (DeKeyser & Criado, 2013), and while explicit explanation may contribute to learning, it cannot be assumed that another explanation followed by another correct mini-whiteboard answer will produce the kind of fluent, independent performance we ultimately want.

So, when students repeatedly make the same grammatical mistake despite apparently knowing the rule, shouldn’t we at least consider whether the problem lies not in their understanding but in their ability to process and retrieve the construction efficiently?

2. What about processing grammar for meaning?

Another limitation of formative assessment that concentrates predominantly on translation, correcting errors, choosing endings and explaining rules is that it may tell us relatively little about whether students can actually process grammatical constructions for meaning when listening and reading, which, given the demands of GCSE receptive examinations, strikes me as a rather significant omission.

Consider the following French sentences:

Il a mangé.

Il mangeait.

Il mange.

Students may correctly identify the tense of each verb, explain how it is formed and translate the sentences accurately when presented in writing, but… can they distinguish these forms when they hear them in connected speech? Can they interpret the differences in meaning within a narrative, particularly when the context doesn’t make the answer obvious, or are they simply relying on familiar vocabulary and contextual clues while paying little attention to the grammatical morphology itself?

Imagine, for example, that students read a short account of a holiday and answer the comprehension questions correctly because they recognise words such as hier, plage and restaurant. Does that mean they have accurately processed the tense markers, or have they managed to reconstruct the message without really attending to them?

VanPatten’s research on input processing (1996, 2004) reminds us that learners often prioritise lexical information over grammatical morphology, particularly when the latter contributes relatively little additional meaning, while VanPatten and Oikkenon (1996) demonstrated the value of structured input activities in developing grammatical processing. And this is why I find it problematic when formative assessment of grammar focuses so heavily on whether students can identify, explain or produce forms, while paying insufficient attention to whether they can actually interpret them in meaningful input.

Before asking whether students can produce a grammatical construction, shouldn’t we establish whether they can process it accurately for meaning? And if we don’t assess that ability, how can we know whether students are developing the kind of receptive competence that GCSE listening and reading examinations demand?

3. Correct answers can be very misleading

Diagnostic multiple-choice questions are frequently presented as powerful tools for identifying misconceptions, and rightly so in certain circumstances. However, their diagnostic value depends entirely on what we are trying to establish and, more crucially, what conclusions we draw from students’ responses!

Consider a student who is asked to select the correct sentence from the following alternatives:

• Elle a allée.
• Elle est allée.
• Elle est allé.

The student selects the second option, and we congratulate ourselves because the question has apparently demonstrated that they understand the formation of the perfect tense with aller. But what exactly have we established? That they know the rule? That they remember an example from yesterday’s lesson? That they have successfully eliminated two implausible alternatives? Or simply that they have guessed correctly?

Now ask that same student to describe what their sister did last weekend without prompting, and will elle est allée emerge as readily and accurately? Possibly… but possibly not, which is precisely why a correct answer in a controlled task is evidence of successful performance in THAT task, not proof that the construction has been acquired.

The same problem arises when we interpret errors, because a wrong answer doesn’t automatically mean that the student has misunderstood the rule or that the most appropriate response is to explain it again. Perhaps the construction isn’t sufficiently automatised, perhaps competing forms are interfering with retrieval, perhaps the student is experiencing processing overload, or perhaps they simply need more meaningful exposure and appropriately designed practice.

And herein lies the limitation of diagnostic questions: although they can reveal where performance breaks down, they don’t necessarily tell us why it breaks down, particularly when the same observable error may arise from very different underlying causes.

If formative assessment is supposed to inform our instructional decisions, shouldn’t we be rather more cautious about diagnosing the cause of a problem from a single correct or incorrect answer?

4. Retrieval practice is useful. But WHAT are we retrieving?

Take a familiar translation exercise in which students areasked to translate “Yesterday we went to the beach” into Spanish and correctly produce Ayer fuimos a la playa. A successful retrieval, certainly, but let’s consider what the task actually demands, because the English prompt has already supplied the meaning, the temporal reference and much of the sentence structure, allowing students to concentrate on selecting and assembling the appropriate Spanish forms.

Now ask the same students to recount an unexpected holiday experience, without preparation, while responding to follow-up questions, and will fuimos emerge as readily? Will they select the appropriate tense while simultaneously deciding what to say, retrieving vocabulary and organising their ideas?

Not necessarily, and this is why the distinction between retrieving grammatical forms under controlled conditions and deploying them during communication matters so much.

Research by Nakata and Suzuki (2019), for example, provide evidence that interleaving grammatical exercises can improve delayed retention, while DeKeyser’s work on skill acquisition highlights the importance of appropriately designed practice in developing efficient language use. But neither finding justifies treating successful retrieval in a highly supported task as equivalent to fluent, independent production.

The type of retrieval practice matters, as does the type of performance we use to assess its effects, because while supported translation, sentence completion and grammaticality judgements can all contribute valuable evidence, they cannot, on their own, establish whether students can retrieve and manipulate the same structures when the prompts disappear.

And surely this should influence how we interpret the results of retrieval quizzes? If students consistently perform well on them but struggle to produce three spontaneous sentences, is the problem really that they need more retrieval quizzes, or that we have been assessing a rather narrow form of retrieval while neglecting the more demanding conditions under which language must eventually be used?

5. Grammar acquisition is not simply error elimination

Another concern of mine is the assumption that when students repeatedly make grammatical errors, the logical response is to explain the rule, correct the mistake and retest, as though grammar acquisition were essentially a process of identifying gaps in explicit knowledge and eliminating them through successive rounds of explanation and correction. Which, of course, is often an utter waste of time!

Don’t get me wrong; sometimes that may be precisely what is needed, but grammar development is considerably more complex, because learners do not simply accumulate rules until their internal grammar is complete; rather, they develop evolving linguistic systems in which certain constructions may remain unstable despite repeated explanations, apparently successful practice and considerable conscious knowledge…

This is precisely what Pienemann’s Processability Theory (1998) helps explain, since learners do not necessarily acquire grammatical structures in the order in which we teach them, and certain constructions require processing procedures that develop in a predictable sequence. Consequently, a student may know a rule, explain it accurately and even apply it successfully in a controlled exercise without necessarily being developmentally ready to deploy the construction spontaneously.

So, when a learner repeatedly produces weil es ist instead of weil es … ist, should we automatically assume that the original explanation was inadequate, that retrieval practice has been insufficient or that another corrective exercise will solve the problem? Or might the error reflect the current state of the learner’s developing grammatical processing system?

Of course, not every persistent error is evidence of developmental unreadiness, and we should be equally cautious about making that diagnosis without appropriate evidence, but Pienemann’s Teachability Hypothesis (1989) reminds us that instruction cannot necessarily override developmental constraints simply because we have explained a structure clearly and tested it repeatedly.

You cannot simply explain or quiz a learner into the next developmental stage, however appealing that possibility might be to teachers under pressure to demonstrate rapid progress.

Corrective feedback can certainly contribute to language development, as Lyster and Saito’s (2010) meta-analysis demonstrates, but its effectiveness does not mean that every error is best addressed through another explanation or another correction task, particularly when the learner already possesses the relevant explicit knowledge.

And if students repeatedly produce weil es ist despite being able to explain verb-final word order perfectly well, might they benefit more from meaningful processing, structured practice and opportunities to develop fluent retrieval than from yet another reminder of a rule they already know?

Explanation + correct answer ≠ acquisition, and until our formative assessment recognises that distinction, we risk responding to developmental and processing problems as though they were simply failures of memory or understanding.

6. The GCSE problem: Are we preparing students for the examination or for our classroom tasks?

And this is where the consequences become particularly serious, because while the limitations of controlled formative assessment may appear relatively harmless during an ordinary Year 10 lesson, they can become painfully obvious when students encounter the demands of GCSE examinations.

Consider a Year 11 student who consistently performs well in grammar retrieval quizzes, sentence translations and error-correction exercises, whose mini-whiteboard answers are usually correct, who can explain the difference between the perfect and imperfect tenses and who knows exactly where the verb goes after weil. Everything looks reassuring, doesn’t it? The teacher has evidence of progress, the student feels confident, and the class appears to be moving steadily towards examination readiness.

But what happens when that same student sits the GCSE?

In listening, they may struggle to distinguish j’ai joué from je jouais because they haven’t developed sufficiently rapid and accurate processing of tense morphology in connected speech; in reading, they may recognise familiar vocabulary yet misinterpret who did what, when or to whom because they overlook grammatical markers; in speaking, they may produce beautifully rehearsed answers but struggle when the examiner asks an unexpected question requiring independent retrieval and manipulation of language; and in writing, they may reproduce memorised structures accurately while finding it difficult to adapt them to an unfamiliar task.

The problem isn’t necessarily that they haven’t learned the grammar, but that their ability to retrieve grammatical knowledge in controlled conditions has developed further than their ability to process and deploy it independently, a discrepancy which research on input processing (VanPatten, 1996, 2004), skill acquisition (DeKeyser, 2013) and developmental constraints (Pienemann, 1998) gives us good reasons to anticipate.

And this is why formative assessment matters so much, because if our assessment repeatedly tells us that students have ‘got it’ simply because they can answer a diagnostic question, complete a translation or correct a grammatical error, we may seriously overestimate their examination readiness, move on prematurely and devote insufficient time to precisely those abilities that will determine how successfully they cope with unfamiliar language under pressure.

Worse still, students themselves may develop a false sense of confidence, believing that their success in classroom grammar tasks means they are well prepared, only to discover in the examination that the language they could manipulate so successfully with prompts and familiar examples is considerably harder to understand and use when those supports disappear.

And surely this is the real danger of misinterpreting formative assessment evidence? Not that students occasionally get a question wrong, but that teachers may be systematically reassured by the wrong kinds of success, thereby overlooking weaknesses in listening discrimination, grammatical processing, independent retrieval and spontaneous production until those weaknesses become visible in the GCSE examination itself.

Good formative assessment should expose these gaps BEFORE the GCSE examination does, which means assessing not simply whether students remember grammatical forms, but whether they can process and use them under conditions that increasingly resemble the demands of the examination.

Otherwise, are we really preparing students for GCSE, or merely making them increasingly proficient at the particular classroom tasks through which we have chosen to measure their progress?

Table 1. What classroom assessment tasks reveal, what they may conceal, and why the distinction matters for GCSE success.

Activity	What it assesses	What it may conceal	GCSE consequence
Vocabulary translation	Lexical recall	Difficulty recognising words in connected speech	Listening weaknesses
Sentence translation	Supported grammatical retrieval	Limited independent production	Speaking and writing weaknesses
Multiple-choice grammar	Recognition of correct forms	Inability to process or produce them independently	False confidence
Spot-the-mistake	Explicit error recognition	Lack of automatic grammatical control	Persistent errors
Rule explanation	Declarative knowledge	Inability to apply rules under pressure	Poor spontaneous performance
Retrieval quizzes	Recall under controlled conditions	Weak transfer to communication	Overestimated readiness
Interleaving	Discrimination and retrieval across structures	Difficulty processing structures in real time	Receptive weaknesses
Feedback and redrafting	Ability to improve supported output	Inability to reproduce improvements independently	Writing weaknesses
Short oral GCSE prompts	Prepared or cued production	Limited spontaneity and flexibility	Speaking weaknesses

7. So what should we actually be assessing?

If the problem lies not necessarily in the techniques themselves but in what we infer from their results, then surely the answer is to broaden the range of evidence we collect, ensuring that our formative assessment addresses the different dimensions of grammatical competence rather than treating them as interchangeable.

Research in instructed second-language acquisition provides good reasons for doing so, with VanPatten and Oikkenon (1996) highlighting the importance of structured input for developing grammatical processing, DeKeyser (2013) emphasising the role of practice in developing efficient language skills, Nakata and Suzuki (2019) providing evidence for the benefits of interleaving on delayed retention, and Lyster and Saito (2010) demonstrating that corrective feedback can contribute to language development.

So what might acquisition-informed formative assessment actually look like in a busy MFL classroom, particularly when teachers have limited preparation time and need techniques that can be implemented quickly without producing yet another elaborate worksheet?

I would suggest assessing at least five dimensions, using activities that require virtually no preparation and, importantly, making intelligent use of mini-whiteboards wherever they can provide useful evidence.

1. Form–meaning processing: Can students actually understand the grammar?

Before asking students to produce a construction, shouldn’t we establish whether they can interpret it accurately when listening and reading, particularly given what VanPatten’s input-processing research tells us about learners’ tendency to prioritise lexical information over grammatical morphology?

Minimal-prep activity: Listen and choose

Say two French sentences:

A. Quand j’étais petit, je jouais au foot.

B. Hier, j’ai joué au foot.

Then ask students which sentence describes a habitual activity in the past, requiring them to write A or B on their mini-whiteboards and reveal their answers simultaneously, before repeating the task with different sentences in which obvious temporal clues have been removed so that students must attend more closely to the grammatical forms.

What are we assessing here? Not whether students remember the rule for forming the imperfect, but whether they can interpret the meaning conveyed by the construction when they hear it, which is a rather different achievement.

The mini-whiteboard, in this case, becomes a tool for assessing form–meaning processing rather than merely checking explicit grammatical knowledge, although we should still be cautious about assuming that success with isolated sentences guarantees equally accurate processing in longer stretches of connected speech.

2. Controlled production: Can students retrieve the construction with limited support?

Once students have encountered and processed a construction, we can assess whether they can retrieve it in a relatively controlled context, while recognising that the amount of support provided will influence the demands of the task and therefore the conclusions we can draw from their performance.

Minimal-prep activity: Complete and transform

Write the following sentence on the board:

Ich mag Deutsch, weil es …

Ask students to complete it on their mini-whiteboards, then change the reason using interessant, before removing the written sentence starter altogether and asking them to produce the entire sentence independently.

By gradually reducing support, we obtain different kinds of evidence about students’ ability to retrieve and manipulate the construction, which is useful given DeKeyser’s (2013) emphasis on appropriately designed practice and the development of increasingly efficient access to linguistic knowledge.

But let’s not exaggerate what this demonstrates, because successful controlled production, however encouraging, is not proof that students can retrieve and use the same construction spontaneously, particularly when the task has explicitly directed their attention towards it.

3. Independent production: Can students use the construction without being told to?

This is where things become particularly interesting, because students who produce accurate grammar when explicitly prompted to use a particular construction may perform rather differently when the prompt disappears and they have to decide for themselves which linguistic resources to deploy.

Minimal-prep activity: Say something about yourself

Instead of asking students to “write a sentence using weil“, ask them why they enjoy or dislike learning German and invite them to give two reasons, allowing a brief thinking period before they write their responses on mini-whiteboards without a displayed model or sentence starter.

Now examine whether they choose to use weil, whether they use it accurately and whether they can formulate their answers without relying on the scaffolding that previously supported their performance; for a more demanding check, ask students to respond orally in pairs and listen to selected exchanges.

This distinction matters because Lyster and Saito’s (2010) meta-analysis found that corrective feedback can produce durable effects on language development, with particularly notable effects on measures requiring learners to construct their own responses.

The question, therefore, is not simply whether students can produce the target construction, but whether they can select and deploy it when the task doesn’t explicitly tell them to, which brings us considerably closer to the kind of independent language use required in GCSE speaking and writing.

4. Retrieval efficiency: Can students retrieve grammar accurately under time pressure?

Accuracy alone tells us relatively little about how efficiently students can retrieve grammatical knowledge, because a learner who takes thirty seconds to construct a correct sentence is demonstrating something rather different from one who can retrieve and assemble the same construction accurately within a few seconds. But how can we assess this efficiently across an entire class, without relying on oral responses that the teacher cannot possibly monitor simultaneously?

Minimal-prep activity: Timed Sentence Challenge

Display a familiar question on the board, such as Warum magst du Deutsch?, and give students five seconds to write a complete answer on their mini-whiteboards, without notes, sentence starters or other prompts. On your signal, everyone raises their boards simultaneously, allowing you to check who has produced an accurate response, who has made grammatical errors and who has failed to complete the sentence within the allotted time. Repeat with different questions and constructions, gradually adjusting the time available to the students’ proficiency level, while ensuring that the time limit remains realistic and does not simply penalise slower writers.

What does this tell us? Certainly more than an untimed grammar quiz about students’ ability to retrieve and assemble language efficiently, particularly when we compare their performance across successive lessons and observe whether accuracy can be maintained as the processing demands increase. However, speeded written retrieval is not the same as oral automaticity, since handwriting speed, spelling and the demands of written formulation also influence performance, which is why we should resist the temptation to interpret rapid, accurate mini-whiteboard responses as proof that students can deploy the same constructions spontaneously in conversation.

And isn’t that precisely the point? Good formative assessment doesn’t simply generate correct answers; it generates evidence whose limitations we understand, so that we can make better-informed instructional decisions rather than mistaking success in one type of task for mastery of the underlying linguistic competence.

5. Durability: Can students still use the grammar weeks later?

Perhaps one of the most neglected dimensions of classroom assessment is durability, because students frequently perform brilliantly immediately after a lesson, particularly when the target construction is fresh in their minds and the task makes it obvious which grammatical form they are expected to retrieve, but what happens three weeks later?

Minimal-prep activity: The unexpected comeback

Three weeks after teaching weil, include a question in an ordinary retrieval starter asking students to explain why they like or dislike one school subject, without mentioning weil, displaying the rule or announcing that subordinate clauses are being revisited.

Students write their answers on mini-whiteboards, allowing you to compare their performance with what you observed previously, before revisiting the same construction a week later through a different topic or communicative context.

Nakata and Suzuki (2019) found that interleaved grammar practice improved performance on a delayed grammaticality judgement test compared with blocked practice, providing support for the benefits of revisiting grammatical knowledge across time, although their findings should not be interpreted as evidence that delayed recognition necessarily entails spontaneous mastery.

If students can only use a construction correctly immediately after practising it, how secure is their learning? And if our formative assessment rarely examines what remains accessible after the immediate instructional context has disappeared, aren’t we at risk of confusing short-term performance with durable learning?

The point

Notice that mini-whiteboards feature in almost every example, which should make it abundantly clear that my criticism is not directed at the technique itself but at the narrow range of evidence that teachers may sometimes collect through it, and at the conclusions they subsequently draw.

Mini-whiteboards can be excellent tools for obtaining rapid, whole-class evidence of grammatical interpretation, controlled retrieval, increasingly independent formulation and delayed retention, provided that the questions are designed to assess those particular dimensions, although they are obviously less suitable for evaluating spontaneous oral fluency, where direct observation is essential.

And none of these checks, taken individually, proves that a construction has been fully acquired, which brings us back to the central issue of assessment validity: does the task actually provide evidence of the aspect of grammatical competence we claim to be assessing?

Processability Theory (Pienemann, 1998) adds another important consideration, since a learner may successfully apply a grammatical rule in a controlled task without having developed the processing procedures required for spontaneous production, reminding us that persistent errors should be interpreted cautiously rather than automatically treated as evidence that another explanation or quiz is needed.

So, by all means, use mini-whiteboards, diagnostic questions, retrieval quizzes and corrective feedback, but let’s be considerably clearer about what they tell us, because recognising grammar, retrieving grammar, using grammar spontaneously and using it automatically are not interchangeable achievements, and treating them as though they were risks making us increasingly proficient at measuring what is easiest to test while neglecting what students actually need to learn.

Concluding remarks

This is not the first time I have noticed what I consider to be an insufficiently critical transfer of findings and practices from general education into MFL in several MFL bloggers and commentators, and I suspect that this tendency reflects, at least partly, a superficial engagement with research in instructed second-language acquisition, combined with an eagerness to ride the latest wave of educational trends.

Retrieval practice, cognitive load, interleaving, mini-whiteboards, Rosenshine’s principles… all potentially valuable, certainly, but since when did the popularity of a technique in general education automatically make it appropriate for language learning, and why should we accept recommendations labelled ‘evidence-informed’ when the evidence being invoked has not been adequately considered alongside what we know about the distinctive processes involved in acquiring another language?

And before anyone interprets this as an ad hominem attack, let me be clear that my criticism is directed at a widespread misconception in MFL discourse, not at an individual, because the problem extends well beyond one blog post or one commentator and deserves serious scrutiny precisely because of its implications for classroom practice.

Those implications are far from trivial, particularly when we consider what happens to students at GCSE, because the danger is not that retrieval quizzes or mini-whiteboards are inherently harmful, but that overreliance on the evidence they generate can create a misleading picture of linguistic competence, encouraging teachers to move on prematurely, neglect essential processing skills and discover weaknesses only when students are confronted with the demands of an examination.

A class may appear successful because students consistently produce correct answers in controlled tasks, yet remain poorly equipped to process unfamiliar language in real time, interpret grammatical distinctions in connected speech or retrieve and manipulate structures independently when the scaffolding disappears… and if teachers have interpreted those correct answers as evidence of acquisition, what incentive have they had to address the weaknesses that their assessment never revealed?

This is why I would argue that poorly interpreted formative assessment is not simply an assessment problem but potentially a curriculum and instructional problem, because what teachers believe students have mastered inevitably influences what they choose to teach, practise, recycle or abandon, with consequences that may only become apparent when students struggle in their GCSE listening, reading, speaking and writing examinations.

And surely this is what should concern us most? Not whether our lessons feature the latest fashionable assessment technique, whether students can confidently explain a grammatical rule or whether every mini-whiteboard displays the correct answer, but whether the evidence we collect actually tells us something meaningful about the linguistic competence our students are developing.

None of this requires rejecting cognitive psychology or dismissing the techniques Bastow advocates, but being genuinely evidence-informed requires considerably more than citing fashionable educational research and applying its recommendations to MFL, because it also demands an understanding of how those findings interact with decades of ISLA research on input processing, developmental readiness, form–meaning connections, skill acquisition and automaticity.

And if we fail to make those distinctions, who ultimately pays the price? Not the commentators promoting the techniques, nor the teachers who have been reassured by apparently impressive classroom performance, but the students who discover, when faced with unfamiliar language under examination conditions, that knowing the rules and passing the quizzes was never quite the same as being able to understand and use the language.

Students should not pay the price for our failure to distinguish between grammatical knowledge and grammatical acquisition, particularly when that failure may leave them less prepared for the very examinations towards which so much of their learning has been directed.

So perhaps the most important question isn’t whether students can demonstrate that they remember what we taught them, but whether they can actually understand and use the language when the prompts disappear, the scaffolding is removed and the demands of real-time processing take over…

If our formative assessment cannot distinguish between the two, we are not just misleading ourselves about how much students have learned; we risk failing the very students we claim to be helping.

Selected references

• DeKeyser, R. (2013). Automatization, Skill Acquisition, and Practice in Second Language Acquisition.
• DeKeyser, R. & Criado, R. (2013). Automatization, Skill Acquisition, and Practice in Second Language Acquisition.
• Lyster, R. & Saito, K. (2010). Oral Feedback in Classroom SLA: A Meta-Analysis. Studies in Second Language Acquisition, 32, 265–302.
• Nakata, T. & Suzuki, Y. (2019). Mixing Grammar Exercises Facilitates Long-Term Retention. The Modern Language Journal, 103, 629–647.
• Pienemann, M. (1989). Is Language Teachable? Psycholinguistic Experiments and Hypotheses. Applied Linguistics, 10, 52–79.
• Pienemann, M. (1998). Language Processing and Second Language Development: Processability Theory. John Benjamins.
• VanPatten, B. (1996). Input Processing and Grammar Instruction.
• VanPatten, B. (2004). Processing Instruction: Theory, Research, and Commentary.
• VanPatten, B. & Oikkenon, S. (1996). Explanation versus Structured Input in Processing Instruction. Studies in Second Language Acquisition, 18, 495–510.

Extensive Processing Instruction (EPI) is strongly aligned with Rosenshine’s Principles of Instruction, although EPI goes considerably further by providing a language-specific, psycholinguistically informed instructional sequence which operationalises many of those principles through the MARS EARS framework.

The important distinction is that, whereas Rosenshine identifies broad principles of effective instruction, EPI translates many of them into a coherent progression from modelling and input processing to structured production, automatisation and spontaneous communication.

How EPI aligns with Rosenshine’s 10 Principles of Instruction

Rosenshine’s principle	How EPI operationalises it through MARS EARS
1. Begin a lesson with a short review	Systematic retrieval of previously taught chunks, vocabulary and structures through starters, cumulative listening and reading texts, recycled sentence builders and interleaving. Retrieval is distributed throughout the lesson, not confined to its beginning.
2. Present new material in small steps	New language is introduced through carefully selected high-frequency chunks and sentence patterns. Cognitive load is managed by controlling novelty and gradually increasing processing demands, moving from supported recognition to independent production.
3. Ask many questions and check all responses	Frequent comprehension checks, meaning-focused questioning, aural discrimination, mini-whiteboards, true/false tasks and error detection ensure active participation and provide evidence of learning across the class.
4. Provide models	Sentence builders, teacher modelling, aural input, worked examples and repeated encounters with target patterns establish accurate phonological, lexical and syntactic representations before independent production.
5. Guide student practice	Structured Production progresses through paired chunking and reading aloud activities, vocabulary building and micro-writing, paired oral retrieval, mini-roleplays, automaticity activities and communicative tasks. Scaffolding is progressively withdrawn.
6. Check for student understanding	Diagnostic checks during Awareness-raising and Receptive Processing, alongside Autonomy checkpoints, establish whether learners can recognise, understand, manipulate and independently retrieve the target language. Frequent use of mini whiteboards in both the receptive and productive activities occurs throughout the MARSEARS sequence providing tons of formative data.
7. Obtain a high success rate	Comprehensible input, carefully controlled task demands, extensive scaffolding and gradual increases in difficulty maximise successful processing and retrieval without eliminating appropriate challenge.
8. Provide scaffolds for difficult tasks	Sentence builders, narrow input, partially completed texts, visual support, substitution tables and guided production reduce unnecessary cognitive demands. Scaffolds are systematically faded as competence increases.
9. Require and monitor independent practice	Autonomy, Routinization and Spontaneity progressively require learners to retrieve, assemble and use language without sentence builders. Teachers monitor accuracy, fluency, retrieval speed and independent communicative performance.
10. Engage students in weekly and monthly review	Spaced retrieval, cumulative recycling, interleaving and recurring high-frequency constructions ensure previously taught language is revisited across lessons, units and terms in varied communicative contexts.

Where EPI goes beyond Rosenshine

Although the correspondence is substantial, EPI should not be presented merely as Rosenshine applied to languages. It offers several additional dimensions which Rosenshine’s principles do not explicitly specify.

Additional dimension	What EPI contributes
Psycholinguistic sequencing	A principled progression from phonological and meaning-focused processing to syntactic encoding, retrieval, automatisation and spontaneous production.
Listening-first instruction	Systematic development of phonological representations, speech segmentation and aural processing before learners are expected to produce unfamiliar language independently.
Input processing before output	Extensive sentence-level and connected-text receptive work designed to strengthen form–meaning mappings before learners undertake increasingly demanding production.
Lexicogrammatical instruction	Vocabulary and grammar are taught together through high-frequency constructions and chunks, with explicit grammatical awareness developed in context.
Automaticity and fluency	Dedicated Routinization activities aim to increase retrieval speed, processing efficiency and the automatic deployment of familiar language.
Spontaneous communication	The instructional sequence culminates in less predictable communicative performance, requiring learners to retrieve, assemble and adapt language with decreasing external support.
Diagnostic autonomy checkpoints	Low-stakes assessments at the end of MARSEA sub-units check independent listening, reading, writing and speaking performance, generating evidence for responsive teaching.

An important qualification

The alignment is both conceptual and pedagogical, but, of course, not evidence that Rosenshine experimentally validated EPI. Whilst Rosenshine’s principles synthesise broad findings about effective instruction, EPI goes much further by integrating those principles with solid theories and evidence concerning second-language acquisition, cognitive processing, skill acquisition and memory.

In other words Rosenshine tells teachers what effective instruction should generally involve; EPI provides language teachers with a more specific account of how to sequence instruction so that learners move from understanding language to retrieving, assembling and using it fluently and spontaneously.

In that sense, Rosenshine provides many of the instructional principles, while MARS EARS provides a language-specific instructional architecture through which those principles can be enacted.

Engagement, attention, motivation and the challenges of teaching languages in the age of AI

Introduction

A few months ago, I delivered a keynote in Sydney entitled Language Teaching at a Crossroads: Engagement, Evidence, and the AI Disruption. In that talk, I explored some of the challenges confronting language education today, from declining uptake and growing disengagement to the impact of digital technology, fragmented attention and artificial intelligence on how young people learn. This article revisits the keynote’s central arguments, focusing particularly on something that I believe deserves much more attention: the growing mismatch between the cognitive and affective environment in which today’s students are growing up and the conditions required for successful language acquisition.

My central contention is that, while the fundamental mechanisms underpinning human learning have not changed, the habits, expectations and experiences that students bring into our classrooms may be changing in ways that make language learning increasingly challenging. We are dealing with learners who have grown up in an environment characterised by instant access to information, rapid stimulation, increasingly sophisticated digital assistance and opportunities for immediate gratification, yet we are asking them to engage in a learning process that demands sustained concentration, extensive repetition, tolerance of ambiguity, delayed rewards and a willingness to make mistakes… often publicly!

This is what I call the Great Mismatch.

Before proceeding, however, an important qualification is necessary. I am not suggesting that an entire generation has somehow become cognitively inferior, that all young people are less resilient than their predecessors or that digital technology is inherently detrimental to learning. Such claims would be both simplistic and difficult to substantiate.

Rather, my argument is that some of the environments in which today’s adolescents spend much of their time may be fostering habits and expectations that sit increasingly uncomfortably alongside the demands of learning a foreign language. The distinction matters, because what concerns me is not necessarily a change in students’ underlying cognitive capabilities, but a change in how those capabilities are habitually exercised, supported and challenged.

1. The Great Mismatch: when the demands of language learning clash with current teens’ expectations

Second-language acquisition is, by its very nature, a gradual and cumulative process. Learners need repeated encounters with language, opportunities to process it meaningfully, retrieval practice and extensive rehearsal before knowledge becomes sufficiently accessible for fluent, spontaneous communication. There are no shortcuts to automaticity. Knowing a word or grammatical structure is not the same as being able to retrieve and use it accurately, rapidly and effortlessly under communicative pressure, since the latter requires time, repeated successful processing and the gradual strengthening of the connections that support retrieval.

However, several characteristics of contemporary digital life may make these demands harder for some learners to accommodate. The constant availability of immediate answers, short-form content, rapid changes of focus and instant feedback can encourage expectations of quick results and continuous stimulation, whereas language learning frequently involves delayed rewards, slow progress, uncertainty and effort whose benefits are not immediately visible.

As we observe in our lessons day in, day out, learners may understand something today and struggle to retrieve it tomorrow; they may recognise a structure in a reading text but be unable to produce it spontaneously; they may perform successfully during a lesson only to discover a week later that much of what they appeared to know has faded. And these experiences are perfectly normal! They are part of learning. Yet students often interpret them as evidence of inability rather than as a sign that more processing and practice are needed.

One of the less obvious consequences of growing up in an environment of instant information is that students may have fewer opportunities, or feel less inclined, to tolerate not knowing. When virtually any question can be answered in seconds, unfamiliar words translated instantly and complex tasks completed with the assistance of AI, uncertainty becomes something that can often be eliminated almost immediately. Yet language learning is fundamentally an exercise in managing uncertainty! Learners must become comfortable with understanding only part of what they hear, inferring meanings from incomplete information, communicating with an imperfect linguistic repertoire and accepting that mastery will remain elusive for some considerable time. In other words: ambiguity tolerance.

Indeed, one of the hallmarks of a developing language learner is the ability to continue processing language despite not understanding everything. This creates a particularly interesting paradox: we may be teaching one of the most ambiguity-heavy subjects in the curriculum to a generation increasingly accustomed to having ambiguity resolved for them. A student who can instantly translate an unfamiliar French sentence may become less willing to persevere with interpreting it independently, while one who expects immediate clarity may find the normal experience of partial comprehension disproportionately frustrating.

The educational challenge, therefore, is not simply to make input more comprehensible, although this remains essential, but to develop learners’ capacity to operate confidently within gradually increasing levels of uncertainty. We need to scaffold ambiguity, not eliminate it, because the ability to tolerate partial understanding is itself an important component of communicative competence.

Another tension concerns the relationship between effort and reward. Much of contemporary digital life operates through exceptionally short feedback loops: an action produces an almost immediate response, a new stimulus is always available and an unrewarding experience can be abandoned with a swipe. Language acquisition, by contrast, operates according to a very different timetable. Students may spend weeks rehearsing a relatively small repertoire of structures before they can retrieve them reliably, months consolidating grammatical patterns before they can manipulate them with reasonable automaticity, and years developing the fluency required for genuinely spontaneous communication.

What concerns me is not simply that some students may be becoming less patient, but that the perceived relationship between effort and reward may be changing. When learners are accustomed to obtaining results with minimal delay, repeated practice without an immediately visible payoff can appear pointless, while the temporary deterioration in performance that sometimes accompanies more demanding retrieval may be interpreted as evidence that learning is not taking place.

This has profound implications for how we structure progression. We need to create more frequent opportunities for learners to recognise genuine improvement without sacrificing the extended practice on which durable learning depends. In other words, we must shorten the feedback loop without pretending that the learning process itself can be shortened to the same extent.

2. The illusion of engagement: are we entertaining students at the expense of learning?

One of the developments that concerns me most is the increasing pressure on teachers to make lessons constantly entertaining. Of course, enjoyment matters, and students who find lessons interesting and rewarding are generally more likely to participate and persist. However, we need to distinguish between engagement with an activity and engagement in the cognitive processes that produce learning.

A classroom can be noisy, energetic and apparently enthusiastic without much durable learning taking place! Students may enjoy a game, participate enthusiastically in a competition or complete an entertaining digital activity while engaging only superficially with the language itself. I have seen this in many lessons I have observed over the years. Conversely, a quieter activity involving careful listening, repeated reading, sentence reconstruction or retrieval practice may look less exciting to an observer while producing much more substantial learning.

The distinction between immediate performance and long-term learning, discussed by Bjork and Bjork (2011), is particularly relevant here, since what students can do successfully during an activity does not necessarily tell us what they will remember or be able to retrieve independently days or weeks later. Hence, the question we should be asking is not simply whether students enjoyed the lesson or appeared engaged, but what cognitive processing the activities required and whether that processing is likely to result in durable learning.

There is, however, another dimension to this problem. If students increasingly spend their leisure time in environments designed to maximise novelty, stimulation and immediate reward, schools may find themselves under growing pressure to compete for their attention on similar terms. Lessons are expected to be faster, more varied, more interactive and more entertaining, with the implicit assumption that any reduction in visible enthusiasm indicates a failure of teaching. Yet I wonder whether, by continually adapting instruction to students’ appetite for stimulation, we risk reinforcing the very attentional habits that make sustained learning more difficult.

After all, if every activity must provide immediate excitement, when do students learn to derive satisfaction from sustained effort, gradual improvement and eventual mastery? If we constantly change tasks to prevent boredom, when do learners develop the patience required to process the same linguistic patterns repeatedly until they become automatic? And if enjoyment becomes our principal measure of instructional quality, are we not in danger of confusing the experience of learning with the experience of being entertained?

Mind you, I am not arguing against fun. I am arguing against confusing fun with effectiveness! Nor am I suggesting that repetitive or cognitively demanding activities must necessarily be dull. On the contrary, carefully designed practice can be highly engaging, particularly when students experience growing competence and recognise that something they previously found difficult has become easier.

The point is that enjoyment should increasingly emerge from the experience of successful learning, rather than learning being subordinated to the pursuit of immediate enjoyment.

3. When failure breeds disengagement: the changing emotional landscape

We frequently talk about a crisis of motivation in language learning, particularly when discussing declining uptake at GCSE and equivalent qualifications. Yet I believe we sometimes reverse the direction of causality. We assume that students fail because they are disengaged, when in many cases they may become disengaged because they repeatedly experience failure.

Consider, for instance, the student in your class who cannot understand spoken French, struggles to remember vocabulary, finds grammatical explanations confusing and freezes when asked to speak. After enough unsuccessful experiences, that student may stop interpreting these difficulties as temporary obstacles and begin seeing them as evidence of a more permanent inability. Their internal narrative becomes not “I need more practice” but the much more tragic “I am rubbish at languages”.

This is where self-efficacy, in Bandura’s (1997) sense, becomes crucial, since learners’ beliefs about their ability to succeed influence the effort they invest, their willingness to persevere and their responses to setbacks. In the keynote, I proposed a model of sustained motivation built around three positive drivers: self-efficacy, relevance and enjoyment, with perceived cost acting as a potential counterweight. Students need to believe that they can succeed, see value in what they are learning and experience sufficient enjoyment to want to continue, while the perceived costs of learning — excessive effort, anxiety, embarrassment or repeated failure — must not become overwhelming.

The changing emotional environment in which adolescents are growing up introduces another challenge that deserves considerably more attention in discussions of language teaching. Research and educational commentary have highlighted concerns about adolescent anxiety, social comparison, perfectionism and sensitivity to failure, although the extent and causes of these developments remain contested. For language teachers, these concerns are particularly relevant because learning to communicate in another language involves a degree of vulnerability that many other school subjects do not expose quite so publicly.

Students must pronounce unfamiliar sounds, attempt sentences they cannot yet formulate fluently, risk making grammatical errors in front of their peers and sometimes communicate ideas in ways that make them sound considerably less articulate than they are in their first language. In an environment where social comparison is increasingly pervasive and public mistakes can carry considerable emotional significance, the perceived cost of participating may become disproportionately high.

A student who is perfectly capable of producing a simple French sentence in a paired activity may nevertheless refuse to do so in front of the class, not because the language is beyond their competence, but because the possibility of embarrassment outweighs the anticipated benefit of participation. This is where risk aversion becomes a pedagogical issue rather than merely a personality characteristic. If students increasingly associate mistakes with threats to their self-image, they may withdraw from precisely those activities through which communicative confidence and automaticity are developed.

The answer, however, cannot be to remove every possibility of failure or avoid challenging students to speak spontaneously, since doing so would merely reinforce the idea that uncertainty and imperfect performance are experiences to be avoided. Instead, we need to design a progression in which linguistic and emotional demands increase together, beginning with highly supported rehearsal, moving through paired retrieval and controlled interaction, and gradually introducing less predictable communicative tasks.

The aim is not to protect students indefinitely from the discomfort of speaking imperfectly, but to give them enough successful experiences to make that discomfort manageable. Psychological safety should be the starting point for developing communicative courage, not a reason for avoiding communicative risk!

Perhaps the most challenging implication of this changing affective environment concerns students’ relationship with cognitive and emotional discomfort. Learning a language necessarily involves experiences that are not immediately pleasurable: struggling to retrieve a word, listening to a passage that cannot yet be fully understood, repeating a structure that has not become automatic or discovering that yesterday’s apparent mastery has faded. These experiences are not necessarily indicators of poor teaching. Within appropriate limits, they are normal features of acquiring complex knowledge and skills.

Yet if learners increasingly interpret frustration, uncertainty or temporary failure as reasons to disengage, the very experiences that should lead to further practice may instead trigger avoidance. We therefore need to distinguish carefully between unproductive struggle, which overwhelms learners and undermines confidence, and productive struggle, which remains within their capabilities and contributes to learning.

Our responsibility is not to make language learning effortless, but to ensure that effort is sufficiently structured, purposeful and successful for students to recognise its value. As I have said many times before in my workshops and blog posts, success is not merely the consequence of motivation. It can also be one of its most powerful sources!

4. Fragmented attention, fragmented learning

Another major concern I explored in the keynote was the impact of increasingly fragmented attention, particularly in light of research on media multitasking, smartphone interruptions and attentional switching, which suggests that frequent distractions can interfere with sustained cognitive processing. Repeatedly switching between tasks incurs cognitive costs, especially when learners are processing unfamiliar information and must devote considerable mental resources to understanding, interpreting and retaining it.

There is, however, an important distinction between distraction and attentional fragmentation that is often overlooked. Traditional distraction involves attention being diverted from a task by an external stimulus, whereas fragmentation describes a more pervasive pattern in which attention is repeatedly redirected between competing stimuli, activities and sources of information. The concern is not simply that students occasionally stop paying attention, but that environments characterised by rapid switching and constant novelty may encourage habits of engagement that are poorly suited to sustained cognitive processing.

This distinction matters enormously in language acquisition because processing language requires continuity of attention. When listening to connected speech, learners must identify sounds, segment the speech stream, recognise words, access their meanings, process grammatical relationships and construct a coherent interpretation, often under considerable time pressure. Crucially, they must also retain earlier information while interpreting what follows, establishing relationships between successive pieces of language and integrating them into a meaningful representation.

If attention is repeatedly interrupted, the problem is not merely that some information is missed; the learner may also lose the contextual and linguistic information needed to interpret subsequent input. Something similar occurs during reading, where constructing meaning across sentences depends on maintaining and updating a mental representation of the text. Speaking and writing place similarly substantial demands on working memory, particularly during the early stages of acquisition, when many of the component processes have yet to become automatic.

Consequently, attentional fragmentation may interfere not only with the amount of language processed, but also with the depth, continuity and coherence of that processing. Students may encounter more information, complete more activities and appear continuously occupied while processing relatively little of that information deeply enough for durable learning to occur. This is one of the paradoxes of our information-rich age… more exposure does not necessarily mean more learning!

We should be careful, however, not to suggest that an entire generation has permanently lost the capacity to concentrate, since the evidence is considerably more nuanced. What research on attentional switching and interruption does establish is that environments characterised by frequent distractions can undermine the conditions required for sustained cognitive processing. Modern adolescents may not be inherently less capable of attention, but many are growing up in environments that repeatedly encourage attention to shift towards novelty and competing stimuli. This is something which is very close to home, having battled with severe ADHD throughout my life and ADHD being characterised by an extreme form of fragmented attention.

A particularly worrying possibility is that schools may inadvertently reinforce the very attentional habits they should be helping students overcome. If we respond to declining concentration by continually shortening activities, accelerating transitions, introducing more novelty and replacing sustained engagement with a succession of entertaining micro-tasks, we may make lessons more immediately palatable while providing fewer opportunities for students to practise concentrating for extended periods.

This does not mean that short activities are inherently ineffective. Carefully sequenced processing cycles, particularly with beginners, can be highly beneficial because they manage cognitive load and allow students to concentrate on specific learning demands. The problem arises when brevity and novelty become ends in themselves rather than serving a clear instructional purpose.

Our pedagogical response should therefore be to protect attention while gradually developing attentional stamina, reducing unnecessary distractions, avoiding excessive multitasking, controlling cognitive load and designing activities that require sustained but manageable engagement with language. In other words, we should adapt instruction to learners’ current attentional needs without allowing those needs to become the permanent ceiling of our pedagogical expectations.

5. Cognitive overload is often mistaken for lack of effort

One of the most important implications of cognitive science for language teaching is that behaviours we commonly interpret as disengagement, inattention or lack of effort may sometimes be the consequence of excessive cognitive demands rather than insufficient motivation.

Consider, for instance, a beginner who is expected to learn unfamiliar vocabulary, manipulate a new grammatical structure, remember word order, monitor agreement and produce a complete sentence simultaneously. Although each of these operations may be manageable in isolation, their combined demands can exceed the learner’s available working-memory resources, resulting in hesitation, avoidance, errors or apparent inattention. In such circumstances, the underlying problem may lie less with the student’s willingness to engage than with the way the learning has been structured.

This becomes particularly significant when we consider the interaction between cognitive overload and the changing emotional environment described earlier. A learner who repeatedly experiences tasks as overwhelming may not merely perform poorly; they may begin to anticipate failure, experience anxiety before attempting similar tasks and eventually avoid participation altogether. What initially presents as a cognitive problem can therefore develop into an affective one, creating a cycle in which excessive task demands undermine performance, unsuccessful performance weakens self-efficacy, and diminished self-efficacy reduces willingness to engage.

This is precisely why instructional sequencing matters so much, and why we need to distinguish between productive difficulty, which encourages meaningful cognitive processing, and unnecessary difficulty arising from poorly structured tasks, insufficient modelling or premature demands for independent production — one of the most common causes of student failure in secondary school settings, in my opinion.

The solution, however, is not to avoid teaching vocabulary and grammar together. On the contrary, presenting grammatical structures through meaningful, carefully selected lexical patterns can support learning by allowing students to process form and meaning in an integrated way. Problems arise when learners are required to handle too many unfamiliar elements simultaneously without sufficient support, particularly when they are expected to produce language independently before the underlying patterns have become sufficiently familiar.

In practical terms, this calls for greater control over the amount of new language introduced, highly comprehensible input, systematic familiarisation with target patterns and extensive receptive processing before learners are expected to produce language independently. Rather than reducing challenge indiscriminately, the objective should be to manage its complexity so that students can devote their limited cognitive resources to the processes most conducive to learning.

Indeed, if there is one thing we should resist, it is the tendency to interpret every instance of hesitation, reluctance or apparent disengagement as evidence that students need more entertainment, more stimulation or greater motivational incentives. Sometimes what they need is simply better instructional design!

6. AI: a shortcut to performance, but not necessarily to learning

Artificial intelligence adds another dimension to this problem, not least because students can now generate sophisticated texts, obtain instant translations, produce apparently fluent answers and complete many language-learning tasks without necessarily engaging in the cognitive processes those tasks were originally designed to elicit.

This creates a potentially serious disconnect between observable performance and underlying competence, whereby the quality of the finished product may tell us remarkably little about what the learner can actually do independently. A student may submit an accurate paragraph in French without having retrieved a single grammatical structure independently, just as another may produce a polished translation without having processed the original text sufficiently to learn from it. Although the resulting work may be impressive, the crucial question is what, if anything, the learner has actually acquired through producing it.

This is very concerning because retrieval, formulation, error detection and repeated language processing are not merely incidental obstacles that stand between students and a successful outcome; they are often integral components of the learning process itself. When AI consistently performs these operations on students’ behalf, it risks depriving them of precisely those opportunities for cognitive engagement through which independent competence develops.

There is also a less obvious affective dimension to this development. When students become accustomed to producing polished work with technological assistance, the gap between what they can achieve with AI and what they can accomplish independently may become increasingly uncomfortable. A learner who can generate an impressive French essay in seconds may find the experience of struggling to formulate a relatively simple paragraph unaided particularly frustrating, especially if they begin to regard the AI-assisted product as the standard against which their own competence should be judged.

In this way, AI may not only reduce opportunities for independent cognitive processing; it may also alter students’ expectations of what successful performance should look like and how much effort achieving it ought to require. The danger is that learners may become increasingly dependent on external assistance precisely because their independently produced language feels inadequate by comparison.

We therefore need to ask whether the widespread availability of AI risks creating a new form of learned dependence, in which students become progressively less willing to tolerate the effort, uncertainty and imperfection associated with genuine skill development. This is not an inevitable outcome, but it is a possibility that educators should take seriously!

None of this means that AI has no legitimate place in language education. Used judiciously, it can generate varied input, create additional retrieval opportunities, support comparison activities and provide material for learners to analyse, critique, correct and improve. Students can evaluate alternative formulations, identify errors, adapt texts for different communicative purposes and use AI-generated language as a springboard for deeper processing.

What matters, therefore, is not simply whether AI is being used, but whether its use supports the cognitive processes underpinning acquisition or substitutes for them. In my view, the guiding principle should be straightforward: AI should enrich the learning process, not remove the thinking that makes learning possible!

7. Designing against the grain: what needs to change?

If the modern cognitive and affective environment increasingly works against some of the conditions required for language acquisition, we need to reconsider how language curricula are designed and delivered. I believe this calls for slimmer, more coherent, more cumulative and more memory-conscious curricula, particularly given the tendency, historically, to prioritise coverage over depth.

Students are frequently exposed to large quantities of vocabulary and numerous grammatical structures without sufficient time being allocated to consolidating them or developing the automaticity required for fluent use, after which we wonder why they struggle to remember what they have been taught or apply it independently.

A more effective approach would begin by prioritising depth over superficial coverage, recognising that a smaller repertoire of high-frequency, useful language which students can retrieve and manipulate confidently is often more valuable than a much larger repertoire that remains fragile and inaccessible.

Equally important is the careful sequencing of progression, whereby learners move gradually from highly supported comprehension towards increasingly independent production, with sufficient processing and rehearsal at each stage. Rather than treating recycling as an occasional revision activity, we should make it a central curriculum design principle, ensuring that previously taught language is revisited cumulatively and systematically through spaced retrieval, interleaving and varied contexts (see the retrieval schedule in the image below, for instance)

Another potentially significant consequence of the contemporary appetite for novelty is a reduced tolerance for repetition, particularly when learners perceive repeated activities as evidence that they are not progressing. Yet repetition is not an unfortunate necessity that teachers should disguise or minimise; it is one of the principal mechanisms through which initially effortful linguistic processing becomes increasingly rapid, reliable and automatic.

Students may understand a structure after a few encounters, but understanding is not the same as being able to retrieve it spontaneously under communicative pressure. The transition from recognition to fluent production requires substantially more practice than learners often anticipate, and it is precisely this extended period of consolidation that can appear particularly unrewarding when students expect visible progress at every step.

This creates a pedagogical dilemma. If teachers continually introduce new content to sustain interest, students may experience a satisfying sense of progression while accumulating increasingly fragile knowledge. Conversely, if repetition is insufficiently varied or poorly structured, learners may disengage before the benefits of rehearsal become apparent. The solution lies in distinguishing between repeating the same activity and repeatedly processing the same language.

Through changes in context, modality, retrieval demands and communicative purpose, teachers can provide extensive exposure to a relatively stable repertoire of high-value language without making learning unnecessarily monotonous. In fact, one could argue that the more novelty-driven students’ wider environment becomes, the more deliberately we need to design opportunities for meaningful repetition.

Alongside these priorities, we need to make success more visible so that students regularly experience evidence of their growing competence, rather than simply completing another topic or worksheet. Carefully designed retrieval activities, cumulative assessments and opportunities to compare current performance with earlier attempts can help learners recognise progress that might otherwise remain invisible.

We also need to cultivate students’ tolerance of uncertainty, gradually increasing the unpredictability of tasks as their linguistic competence develops, while ensuring that the demands placed on them remain manageable. Similarly, if students are increasingly reluctant to expose themselves to public error, speaking activities should be sequenced so that confidence develops through repeated successful rehearsal before learners are expected to communicate spontaneously.

This is where I believe the principles underpinning Extensive Processing Instruction (EPI) become particularly relevant. By systematically moving learners from modelling and awareness-raising through extensive receptive processing, structured production, expansion, autonomy, routinisation and spontaneity, the MARS EARS sequence attempts to reconcile the cognitive demands of acquisition with learners’ need for comprehensibility, confidence and progressive independence.

Its emphasis on repeated processing, carefully scaffolded input-output progression, cumulative recycling and gradual automatisation is not simply a response to the limitations of working memory; it also addresses several of the affective challenges discussed in this article. When students experience early success, develop familiarity with linguistic patterns before being required to manipulate them independently and gradually discover that previously difficult tasks have become manageable, they are more likely to develop the self-efficacy needed to persevere.

However, I would argue that even the best instructional sequence will fall short unless we also pay attention to the habits and expectations that learners bring to the classroom. We need to teach students not only the language itself, but also how to engage productively with the processes through which languages are acquired: how to tolerate partial understanding, how to persist when retrieval is effortful, how to recognise the value of repetition and how to interpret mistakes as information rather than personal failure.

Ultimately, designing against the grain of the modern environment does not mean rejecting technology, abandoning enjoyment or making learning unnecessarily difficult. It means recognising that effective teaching must sometimes counterbalance, rather than merely accommodate, the habits encouraged by students’ wider experiences.

Conclusion: the learner’s environment has changed… but the architecture of learning has not!

We are teaching languages in a world that has changed profoundly, with students whose experiences, expectations and opportunities differ in important ways from those of previous generations. Digital technologies have transformed how information is accessed, how attention is allocated and, increasingly, how tasks are completed, yet the fundamental cognitive mechanisms underpinning human learning have not changed at anything like the same pace.

Attention remains essential, working memory remains limited, repetition and retrieval continue to play a crucial role in acquisition, and fluency still requires extensive practice over time. No technological innovation, however sophisticated, has removed these fundamental requirements.

What makes the present situation particularly challenging is that the cognitive and affective pressures discussed in this article do not operate independently. A learner accustomed to rapid stimulation may find sustained rehearsal increasingly unrewarding; limited rehearsal may then produce fragile knowledge and unsuccessful retrieval; repeated failure may undermine self-efficacy, increase anxiety and encourage further avoidance of challenging tasks. In this way, fragmented attention, impatience with delayed rewards and fear of failure can potentially reinforce one another, creating a cycle in which students become progressively less willing to engage in precisely the kinds of processing that would help them succeed.

Against this backdrop, we should resist the temptation to respond to every emerging challenge by making lessons faster, louder, easier or more entertaining, just as we should avoid romanticising difficulty or assuming that resilience can be developed simply by exposing students to repeated failure.

What is needed instead is instructional design that combines meaningful challenge with appropriate support, protects and gradually strengthens attention, fosters self-efficacy, makes progress as tangible as possible and systematically develops learners’ tolerance of uncertainty, repetition and delayed gratification.

In other words, rather than responding to learners’ potentially decreasing tolerance of cognitive effort by removing that effort from learning, we should seek to make it more manageable, purposeful, successful and rewarding.

Perhaps the future of language teaching depends less on finding ever more entertaining ways of presenting content than on rediscovering how to help students experience the satisfaction of genuinely mastering something difficult. After all, if meaningful engagement is partly sustained by the experience of growing competence, we may need to reconsider whether our increasingly entertainment-oriented approaches to teaching are always serving our students as well as we imagine.

The real challenge is not to redesign language learning around every habit that the modern environment may be fostering in our students, but to design learning experiences that help them develop the habits, confidence and cognitive stamina that successful language acquisition demands.

Because, ultimately, the world may have changed, the learner’s expectations may have changed, and the technology certainly has… but the fundamental architecture of human learning has not!

Introduction

From what I see daily on social media and read in the specialized literature, much has been said about how artificial intelligence can help teachers teach more efficiently and students learn more efficiently. Yet very little is being said about its potential negative effects on teachers themselves: on their pedagogical competence, professional judgement, creativity and autonomy.

What I see as the greatest potential threat posed by AI to teacher expertise is professional deskilling: the gradual weakening of the knowledge, judgement and decision-making capacities that teachers develop through deep and sustained engagement with planning, teaching, assessment and reflection. AI may make teachers more productive without necessarily making them more competent. When teachers routinely delegate important pedagogical decisions to AI, they risk losing opportunities to develop and refine the expertise underpinning effective instruction.

Ten potential negative effects

1. Erosion of pedagogical content knowledge (PCK)

Teachers develop expertise by continually deciding how to transform subject knowledge into explanations, examples, tasks and representations that learners can understand.When AI routinely makes these decisions, teachers engage less deeply in this process. In MFL, this includes, for instance, choosing which grammatical structures to prioritise, anticipating linguistic difficulties, selecting appropriate examples, etc.

Research principle potentially undermined: Shulman’s (1986, 1987) concept of pedagogical content knowledge, emphasising the integration of subject knowledge and knowledge of how to teach it.

2. ‘Atrophy’ of instructional design skills

Designing effective lessons requires teachers to identify learning objectives, sequence content, anticipate misconceptions, manage cognitive demands and plan opportunities for practice and assessment with their students and learning setting in mind. Outsourcing lesson planning to AI obviously reduces opportunities to exercise these skills, particularly among novice teachers who have yet to develop robust instructional knowledge. This is very concerning, as the crisis we are currently experiencing in recruiting language learners at GCSE and A-level demands teachers with exceptional instructional design skills!

Research principle potentially undermined: Constructive alignment (Biggs, 1996), backward design (Wiggins & McTighe, 2005) and cognitive load theory (Sweller et al., 1998).

3. Weakening of diagnostic expertise

An important chunk of a teacher’s professional growth refers to the development ofone’s ability to interpret student errors, distinguish between different causes of misunderstanding and identify appropriate interventions.If AI increasingly diagnoses learning difficulties and recommends corrective actions, teachers will eventually become less practised at independently interpreting evidence of learning! For example, an MFL teacher might accept an AI-generated explanation of listening difficulties without investigating whether the problem concerns phonological decoding, lexical recognition, grammatical parsing or working-memory demands. Another MFL teacher delegating corrective feedback on a written piece to a bot, will miss out on an incredibly useful window into their students’ interlanguage development.

Research principle potentially undermined: Teacher noticing and professional vision (van Es & Sherin, 2002), and formative assessment as eliciting, interpreting and acting upon evidence (Black & Wiliam, 1998).

4. Reduced capacity for critical pedagogical judgement

AI-generated recommendations appear very convincing, particularly when accompanied by plausible explanations or research references. Teachers may consequently become less inclined to question instructional choices, compare alternatives or scrutinise the assumptions underlying a suggested approach. The risk is a shift from independently evaluating pedagogical options to selecting among solutions generated by an algorithm.

Research principle potentially undermined: Reflective professional judgement (Schön, 1983), educational judgement (Biesta, 2015) and automation bias (Parasuraman & Riley, 1997).

5. Reduced reflective practice

Professional expertise develops to a great extent through reflection on instructional decisions: why an activity succeeded, why students struggled and what should be changed. If AI routinely generates lesson evaluations, identifies weaknesses and recommends improvements, teachers may engage less actively in the reflective processes through which professional understanding develops.

Research principle potentially undermined: Schön’s (1983) reflection-in-action and reflection-on-action, and Kolb’s (1984) experiential learning cycle.

6. Weakening of curriculum design expertise

Curriculum planning requires complex decisions about progression, sequencing, knowledge selection, cumulative practice and assessment. AI-generated units may be superficially coherent while failing to reflect learners’ prior knowledge or the developmental demands of the subject. When teachers increasingly adopt rather than construct curriculum sequences, they may lose opportunities to understand how knowledge accumulates over time.

Research principle potentially undermined: Curriculum coherence, constructive alignment (Biggs, 1996), and spacing and distributed practice (Cepeda et al., 2006).

7. Reduced pedagogical creativity

Professional creativity involves generating, testing and refining instructional approaches in response to particular learners and contexts. AI can support creativity by suggesting alternatives, but habitual reliance on ready-made materials may reduce teachers’ opportunities for inventive problem-solving. A related danger is pedagogical homogenisation, as teachers draw on similar AI-generated lesson structures and activities.

Research principle potentially undermined: Teacher agency (Priestley, Biesta & Robinson, 2015) and adaptive expertise (Hatano & Inagaki, 1986).

8. Declining professional autonomy

As AI platforms increasingly recommend instructional strategies, assessment approaches and learning interventions, decisions traditionally made by teachers may shift towards technological systems. This becomes particularly problematic when schools institutionalise AI-generated planning or algorithmic recommendations, reducing teachers’ discretion to adapt instruction to their learners.

Research principle potentially undermined: Teacher agency and professional autonomy (Priestley, Biesta & Robinson, 2015), and educational judgement (Biesta, 2015).

9. Reduced development of novice teachers

Novice teachers acquire expertise partly by undertaking demanding professional tasks, receiving feedback, reflecting on outcomes and progressively refining their decisions. If AI routinely performs lesson planning, resource development, assessment design and diagnostic reasoning for them, beginning teachers may miss opportunities to build foundational professional skills. They may also lack the expertise required to recognise weaknesses in AI recommendations.

Research principle potentially undermined: Development of teacher expertise (Berliner, 2004), pedagogical reasoning (Shulman, 1987) and deliberate practice (Ericsson, Krampe & Tesch-Römer, 1993).

10. The illusion of professional competence

This is what I keep seeing more and more when I visit schools. AI can make relatively inexperienced or less competent teachers appear highly competent by enabling them to produce sophisticated lesson plans, assessments, presentations and instructional materials. Yet the quality of these products does not necessarily reflect underlying understanding. A teacher may become increasingly efficient at generating materials while remaining unable to explain why particular activities are appropriate, diagnose their limitations or adapt them effectively.

Research principle potentially undermined: The distinction between performance and learning (Soderstrom & Bjork, 2015), applied cautiously to professional learning, and pedagogical reasoning (Shulman, 1987).

Summary: How AI may erode teacher expertise

Dimension of expertise	Potential negative impact
Pedagogical content knowledge	Less independent transformation of subject knowledge into teachable content
Instructional design	Reduced practice in sequencing, scaffolding and planning
Diagnostic competence	Greater reliance on automated interpretations of learner difficulties
Critical judgement	Increased acceptance of algorithmic recommendations
Reflective practice	Less independent analysis of instructional successes and failures
Curriculum expertise	Reduced engagement with progression and cumulative learning
Creativity	Dependence on ready-made solutions and homogenisation of teaching
Professional autonomy	Transfer of pedagogical decisions towards AI systems
Novice teacher development	Fewer opportunities to acquire foundational professional skills
Professional self-evaluation	Confusion between AI-generated outputs and genuine teaching expertise

What does the evidence actually establish?

There is a crucial distinction between well-established research on how professional expertise develops and direct evidence that generative AI causes teacher deskilling. Research by Shulman, Schön, Berliner and others provides strong theoretical reasons to be concerned about outsourcing professional thinking. Research on automation bias demonstrates that people can become excessively reliant on automated recommendations. However, longitudinal evidence showing that regular generative AI use causes measurable deterioration in teachers’ pedagogical content knowledge, diagnostic competence or professional judgement remains limited. These are credible risks, not established universal outcomes!

The distinction also matters between AI that replaces professional reasoning and AI that stimulates it. A teacher who critically evaluates, challenges and improves AI-generated lesson plans may strengthen rather than weaken their expertise.

The central paradox

AI can dramatically increase the speed at which teachers produce lessons, worksheets, assessments and feedback. But the intellectual work involved in creating those materials is not merely an administrative burden to be eliminated. It is often part of the process through which teachers develop pedagogical expertise.

The danger is that schools may celebrate gains in teacher productivity while overlooking a gradual reduction in opportunities for teachers to practise the complex professional reasoning that makes them effective.

AI may help teachers do more, faster, while giving them fewer opportunities to become better at what they do.

That is arguably the most important long-term concern about AI and teacher professional development.

Selected research references

Berliner, D. C. (2004). Describing the behavior and documenting the accomplishments of expert teachers. Bulletin of Science, Technology & Society, 24(3), 200–212.

Biesta, G. (2015). What is education for? On good education, teacher judgement, and educational professionalism. European Journal of Education, 50(1), 75–87.

Black, P., & Wiliam, D. (1998). Assessment and classroom learning. Assessment in Education, 5(1), 7–74.

Cepeda, N. J., et al. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354–380.

Ericsson, K. A., Krampe, R. T., & Tesch-Römer, C. (1993). The role of deliberate practice in the acquisition of expert performance. Psychological Review, 100(3), 363–406.

Hatano, G., & Inagaki, K. (1986). Two courses of expertise. In Child Development and Education in Japan.

Parasuraman, R., & Riley, V. (1997). Humans and automation: Use, misuse, disuse, abuse. Human Factors, 39(2), 230–253.

Priestley, M., Biesta, G., & Robinson, S. (2015). Teacher Agency: An Ecological Approach. Bloomsbury.

Schön, D. A. (1983). The Reflective Practitioner. Basic Books.

Shulman, L. S. (1986). Those who understand: Knowledge growth in teaching. Educational Researcher, 15(2), 4–14.

Shulman, L. S. (1987). Knowledge and teaching: Foundations of the new reform. Harvard Educational Review, 57(1), 1–22.

Soderstrom, N. C., & Bjork, R. A. (2015). Learning versus performance: An integrative review. Perspectives on Psychological Science, 10(2), 176–199.

van Es, E. A., & Sherin, M. G. (2002). Learning to notice: Scaffolding new teachers’ interpretations of classroom interactions. Journal of Technology and Teacher Education, 10(4), 571–596.

Why doing more GCSE listening papers isn’t necessarily going to make your students better listeners, and why understanding what goes wrong when they listen may be the key to improving their exam performance.

1. We keep testing listening. But are we actually teaching it?

Picture this: it’s a Year 10 French lesson and your students have just completed a GCSE-style listening comprehension task which, judging by the expressions on their faces and the rather disappointing scores you’ve just collected, hasn’t exactly gone according to plan. A few students have scored eight or nine out of ten, a sizeable group have managed four or five, and a handful have barely got anything right. We have all been there!

So what happens next?You play the tricky bits again, explain what the speaker said, perhaps put the transcript on the board, underline the relevant words and point out the clues they should have picked up. You might even tell them to listen more carefully next time, as though careful listening were something they could simply switch on at will…

Then, in the following lesson, another listening comprehension exercise, and guess what? The same students struggle again!

I’ve seen this scenario countless times in my teaching career and I suspect most MFL teachers will recognise it, but here’s the question I would like you to consider: when a student scores 4/10 on a listening comprehension task, what exactly have we learnt about their ability to listen?

That they didn’t understand six answers? Certainly, but why didn’t they understand them, and more importantly, what are we going to do about it?

Was it because they didn’t know the words, because they knew the words but couldn’t recognise them when spoken, because they couldn’t work out where one word ended and the next began, because they missed a negative or a tense marker, or because they understood the individual bits but couldn’t hold everything together long enough to answer the question?

Different problems, different causes and potentially very different interventions, yet we often respond to all of them in exactly the same way: replay, explain, correct, move on… and then we wonder why listening scores don’t improve as much as we’d like!

The problem, in my view, isn’t necessarily that we’re not doing enough listening, because in many classrooms students complete plenty of listening comprehension exercises, but that we often use those exercises to assess comprehension without doing enough to investigate the cognitive and linguistic processes responsible for comprehension breakdowns.

Are we teaching students how to listen, or are we simply finding out, over and over again, how much they fail to understand? This is where diagnostic listening comes in.

2. What exactly is diagnostic listening?

Put simply, diagnostic listening is what we do after a listening comprehension task to find out why students got particular answers wrong, rather than merely giving them another listening test, displaying the transcript and showing them where the answers were, or replaying the same recording until everyone eventually gets it right.

It involves a systematic investigation of comprehension breakdowns, undertaken with the specific purpose of deciding what needs to happen next instructionally, and rests on three questions which, although closely related, need to be distinguished:

• Assessment: What did my students understand?
• Diagnosis: Why did they fail to understand certain parts?
• Intervention: What can I do to address those difficulties?

Consider a student who hears Je suis allé au cinéma avec mes amis samedi dernier and gets the question about when the speaker went to the cinema wrong, even though, at first glance, the sentence appears relatively straightforward.

Perhaps they don’t know what samedi dernier means, in which case we’re dealing with a vocabulary knowledge problem; perhaps they know the expression perfectly well when they see it written down but fail to recognise it in the recording, suggesting a possible spoken-word recognition problem; or perhaps they recognise samedi but fail to pick up dernier, which might involve segmentation, lexical recognition or insufficient familiarity with the expression.

Alternatively, they may understand the entire sentence when you play it in isolation but lose the information when it’s embedded in a longer passage, because the effort required to process the incoming language leaves insufficient attentional capacity for retaining and integrating what they’ve heard.

The same wrong answer, four different possible explanations, and potentially four different instructional responses, which raises a rather obvious question: would you prescribe the same treatment to four patients presenting the same symptom but suffering from four different conditions?

Of course not, yet mutatis mutandis, that is precisely what we sometimes do when we give students more generic listening practice without investigating what is actually going wrong!

And before anyone accuses me of suggesting that teachers should become amateur cognitive psychologists, let me clarify that diagnostic listening doesn’t require sophisticated laboratory equipment or lengthy individual assessments, merely a sound understanding of listening processes and a repertoire of carefully chosen activities that allow us to investigate plausible causes of failure.

The sequence, in essence, is straightforward: comprehension task, identification of errors, diagnostic investigation, targeted intervention and independent reassessment, although putting it into practice consistently is rather less straightforward than describing it…

Because, as we say in Italian, tra il dire e il fare c’è di mezzo il mare. Between saying and doing lies the sea.

3. What actually goes wrong when students listen?

Before we can diagnose listening difficulties, we need at least a basic understanding of what the brain has to do when processing speech, because without some understanding of the processes involved, how can we possibly know what to investigate?

As I have often written on this blog and in my books, listening isn’t simply recognising words and translating them into English, but a highly demanding, time-pressured activity in which several cognitive and linguistic processes must operate simultaneously in a very limited timed window, often under conditions that are far less forgiving than those encountered during reading.

First, learners need to discriminate the sounds of the language and identify relevant syllabic and prosodic patterns, since stress, rhythm and intonation provide cues that help listeners organise the incoming speech stream and, in some languages, distinguish meanings that would otherwise be difficult to identify.

They then need to segment the continuous speech stream into recognisable words and chunks, a process which, especially in French, can prove extremely challenging because liaison, enchaînement and other connected-speech phenomena may make familiar expressions sound rather different from the forms students have encountered in isolation.

Next comes lexical recognition, whereby the learner identifies familiar spoken words and retrieves their meanings quickly enough to keep pace with the recording, followed by grammatical processing, which involves attending to the little words, endings and structural cues that tell us who did what, when, to whom and under what circumstances.

Finally, learners must integrate the information they have decoded, connect ideas across sentences, track references, interpret discourse markers and make inferences where necessary, all while the recording continues playing…

Now, if a student is using up most of their attentional resources simply trying to recognise individual words, how much capacity will they have left to construct and retain the overall meaning? Not much, I would suggest, which is one reason why vocabulary knowledge alone doesn’t guarantee successful listening, and why a student who can translate every word in a sentence when presented in writing may nevertheless fail to understand that same sentence when spoken at a natural pace.

Of course, these processes don’t operate in a neat, linear sequence, but interact continuously, with prosody supporting segmentation, lexical knowledge helping listeners identify word boundaries, and contextual information facilitating recognition, although the latter can sometimes encourage inaccurate guessing.

Distinguishing these processes is nevertheless pedagogically useful because it provides us with a framework for investigating why comprehension has failed.

4. Eight listening problems worth investigating

I am certainly not suggesting that teachers should diagnose eight different processing skills after every listening comprehension exercise, which would be neither practical nor necessary, but rather that when a pattern of errors emerges, we should ask which processing difficulty might plausibly explain it before selecting one or two targeted activities to investigate.

1. They couldn’t discriminate the sounds

Sometimes the problem starts at the phonological level, as English-speaking learners of French may struggle with the distinction between /y/ and /u/, as in tu and tout, while German learners may have difficulty with vowel length contrasts and Spanish learners may fail to perceive certain consonantal distinctions or stress patterns that contribute to meaning.

If a comprehension error appears to involve a particular sound contrast, I might use Faulty Echo, an activity I frequently use in my listening workshops, in which I say a familiar word or short chunk, then produce a second version which is either identical or slightly different, and students decide whether the echo is correct or faulty.

For example, I might produce Tu viens ? followed by Tout vient ?, asking students to identify whether the second utterance faithfully reproduces the first.

The important thing is to choose the contrasts carefully, because what we want to establish is whether students can actually perceive the relevant phonological distinction, rather than whether they can infer the answer from the surrounding context.

If they consistently fail to discriminate the sounds, what is the point of immediately giving them another long listening passage containing the same contrast, when what they need first is focused perceptual practice?

2. They struggled with syllables, stress or prosody

This is an area I think we have neglected far too much in MFL teaching, because prosody isn’t some decorative extra that we teach when we have time, but a set of features, including stress, rhythm, intonation and syllabic organisation, that contribute to how listeners process the speech stream and whose role in segmentation and lexical recognition deserves considerably more attention than it typically receives.

Take Spanish hablo and habló, where failure to perceive the stress distinction may result in a missed grammatical contrast, or German words such as AUFstehen and verSTEHen, where stress contributes to the identity of the spoken word.

One activity I use is Syllable Building Blocks, in which students are presented with a selection of eight to ten familiar syllables before listening to words or short chunks containing one of them, their task being to identify which syllable they heard, an apparently simple exercise which can nevertheless reveal whether learners are able to detect familiar syllabic units when embedded in larger stretches of speech. Another is Stress Detective, where students identify the stressed syllable or compare two spoken versions.

Simple activities? Certainly, but potentially revealing ones, particularly when the teacher has reason to suspect that the original comprehension breakdown involved failure to perceive relevant syllabic or prosodic information, and why should focused phonological work be boring anyway?

3. They couldn’t segment the speech stream

This is a big one, because students may know every word in a sentence and still fail to recognise those words when they occur in connected speech.

Consider Il est allé au cinéma, a sentence which, on paper, is hardly challenging for a Year 10 student, but which, when spoken naturally, requires the learner to identify plausible boundaries before accessing the relevant lexical items.

One way of investigating this is through Break the Flow, where students listen to a short extract and identify the words or familiar chunks they can hear, or through Chunk Tracking, in which students listen for specific chunks embedded in longer utterances.

For example, students might hear Le weekend | je joue au tennis | avec mes amis | parce que c’est amusant, with the teacher investigating whether they can identify the familiar chunks within the continuous speech stream.

If students recognise the chunks individually but fail to locate them in the recording, segmentation is a plausible source of difficulty, although lexical recognition and familiarity with the chunks may also contribute, and the intervention should therefore involve repeated exposure to familiar chunks in connected speech, with support gradually faded, rather than another generic comprehension worksheet.

4. They knew the words but couldn’t recognise them

This is probably one of the most common problems I encounter, because you can ask students what souvent means and they’ll tell you immediately, or write je voudrais on the board and they’ll recognise it, but play a recording containing those same expressions and suddenly they seem to disappear into the speech stream.

Why? Because knowing a word’s written form and meaning is not the same as having a sufficiently robust spoken representation of it, or being able to retrieve that representation rapidly enough for real-time comprehension.

To investigate this, I might use Word Grab, where students listen to a short extract and identify familiar words or chunks from a selection of plausible alternatives, or Sentence Bingo, where they listen for familiar expressions embedded in different sentences.

If the problem is slow or unreliable recognition, activities such as Audio Boxing and the Language Gym Listening Trainer can provide repeated opportunities to process familiar language under increasingly demanding conditions, with the aim not simply of making students faster but of making recognition accurate, reliable and eventually more automatic.

And why does this matter? Because when word recognition becomes less effortful, more attentional resources become available for constructing meaning!

5. They missed the grammar

Here’s another familiar scenario in which the student hears Fußball, spielen and gestern, but fails to process the grammatical structure that establishes exactly what happened, or recognises the main verb in a French sentence but misses the negative, or hears hablo when the speaker actually said habló.

In these cases, the student may have understood the topic without understanding the message accurately, which is why I might use Grammar Detective, playing two sentences containing largely the same vocabulary but different grammatical meanings, such as Er spielt Fußball and Er hat Fußball gespielt, before asking students to identify the time reference.

Another possibility is Faulty Transcript, where students hear Nous avons mangé au restaurant but see Nous allons manger au restaurant written down and have to identify the discrepancy.

But here’s the crucial question: did they fail to hear the grammatical distinction, or did they hear it without understanding its significance? Two very different problems, and therefore two different starting points for intervention.

6. They couldn’t process the language quickly enough

Sometimes students understand an utterance perfectly well when it’s played in isolation, but fail when the same language occurs in a longer recording, because they are still processing one chunk when the next one arrives, with information being lost and comprehension compromised as a result.

To investigate this, I might compare recognition of familiar expressions in short and longer utterances, or use timed recognition tasks, although we need to be careful because a student who struggles with fast speech doesn’t necessarily have a general processing-speed problem, as their lexical representations may simply not be sufficiently established.

Similarly, slowing down the recording doesn’t always solve the problem, since artificially slowed speech can distort natural rhythm and prosody, which is why the instructional response should focus on repeated, varied processing of familiar language, gradually increasing demands as accuracy improves.

7. They couldn’t retain or integrate the information

A student may understand the first part of a sentence but lose it by the time the speaker reaches the end, as might happen with Je suis allé au cinéma avec mes amis samedi dernier, mais nous sommes rentrés tôt parce que le film était ennuyeux, where the learner recognises the important expressions but struggles to retain and integrate the information.

We often call this a working-memory problem, but I’d be cautious, because what looks like poor memory may actually be inefficient decoding or lexical recognition, with the effort required to identify each word leaving fewer resources available for maintaining the message.

A useful diagnostic approach is to compare performance on shorter and longer utterances containing similarly familiar language, while recognising that increasing length also increases linguistic and integration demands, so that if students manage the shorter versions but struggle as length increases, we have a reason to investigate further rather than immediately attributing the problem to memory.

8. They understood the sentences but not the whole text

Finally, some students understand individual sentences but struggle to connect them, as in Paul voulait aller à la piscine. Cependant, il faisait trop froid. Il a donc décidé de rester chez lui. Son frère, lui, est sorti avec ses amis, where learners may recognise the vocabulary but lose track of who stayed home, who went out and why Paul changed his plans.

Activities such as Pronoun Chain, Connector Snap and Track the Structure can help us investigate these difficulties, for example by replaying the relevant section and asking students to identify what il refers to, or whether cependant introduces a contrast.

If they understand the individual sentences but fail to integrate them, we need to work on discourse processing rather than simply revising vocabulary.

Eight possible sources of comprehension breakdown, one incorrect answer, and still convinced that simply replaying the recording and explaining the answer is enough?

5. The transcript: A diagnostic goldmine

I have nothing against using transcripts after listening comprehension, quite the opposite, but what I object to is using them simply to show students where the answers were, because when we display the transcript, underline the relevant expression and hear students say, “Oh, yes, I know that!”, we should be asking why, if they knew it, they failed to recognise it.

Here’s a routine I would recommend, starting with students completing the comprehension task independently, followed by a replay of the problematic extract during which they identify what they think they heard, before the transcript is displayed and they highlight words and chunks they already knew but failed to recognise in the recording, using a different colour for genuinely unfamiliar vocabulary.

This simple distinction is enormously useful, because a student who didn’t know cependant needs something different from a student who knew it perfectly well but failed to recognise it in connected speech.

This is also the stage at which metacognition can be explicitly scaffolded, because simply asking students to reflect on why they got an answer wrong is unlikely to produce particularly useful insights, especially when they lack the vocabulary or understanding of listening processes needed to explain their difficulties. Rather than expecting students to diagnose their own problems independently, we can provide simple, structured prompts that guide them through a sequence of questions: What did I fail to understand? What does the transcript reveal? Did I not know the language, or did I know it but fail to recognise it when spoken? Was the problem identifying individual sounds, separating words, noticing grammatical clues or keeping track of meaning? By asking students to select one or two plausible explanations, identify the specific word or chunk that caused difficulty and decide what they need to practise next, we gradually develop their ability to monitor and evaluate their own listening, while giving the teacher valuable information to investigate further. The important thing is that these scaffolds support metacognitive reflection without turning it into an exercise in guesswork or excessive form-filling! The two pupil-friendly diagnostic forms below, one for Years 7–8 and the other for Years 10–11, are designed precisely for this purpose and can be used during the post-transcript and audio-processing phase, ideally followed by a brief teacher-led diagnostic probe and targeted intervention.

HERE ARE TWO READY-TO-USE PUPIL LISTENING DIAGNOSTIC FORMS: ONE FOR YEARS 7–8 AND ONE FOR YEARS 10–11. USE THESE AFTER STUDENTS HAVE LISTENED AGAIN AND COMPARED THE AUDIO WITH THE TRANSCRIPT, TO HELP THEM IDENTIFY WHAT WENT WRONG AND WHAT TO PRACTISE NEXT.

(1) Form 1 — Years 7–8

LISTENING DETECTIVE (Years 7 to 8)

My Listening Check-Up

Name: ____________________ Class: __________ Date: __________

I’ve listened again and looked at the transcript. Now I want to find out what made listening difficult.

1. What went wrong?

Tick up to TWO boxes.

☐ I didn’t know what some words meant.

☐ I knew the words when I read them, but didn’t recognise them when I heard them.

☐ The words sounded joined together.

☐ I mixed up two sounds or words.

☐ I missed a little word or ending that changed the meaning.

☐ I understood some bits, but forgot them or got lost.

2. Find the tricky bit!

Copy ONE word or short phrase from the transcript that was difficult to hear or understand.

---

---

3. What should I practise next?

Tick ONE box.

☐ Learning the meaning of words

☐ Recognising words I already know

☐ Hearing sounds and syllables

☐ Finding words in a sentence

☐ Noticing little words and endings

☐ Following longer sentences

4. Did the practice help?

Listen to a NEW example without looking at the words.

☐ Yes! I understood it on my own.

☐ A little, but I still need practice.

☐ Not yet. I still need help.

Remember: A wrong answer is a clue! Finding out WHY you got it wrong helps you become a better listener.

---

(2) Form 2 — Years 9-11

WHY DID I MISS THAT?

GCSE Listening Diagnosis

Name: ____________________ Class: __________ Date: __________

Question(s) I got wrong: __________________________

You’ve completed the listening task, checked the transcript and listened again. Now identify WHY you lost marks and what you need to improve.

1. What was the likely problem?

Tick ONE or TWO boxes.

☐ Vocabulary knowledge: I didn’t know the meaning of a key word or expression.

☐ Spoken-word recognition: I knew the word or expression in writing but didn’t recognise it when spoken.

☐ Sound or syllable recognition: I confused similar sounds, syllables or stress patterns.

☐ Segmentation: I couldn’t separate the stream of speech into familiar words or chunks.

☐ Grammar processing: I missed a negative, tense marker, pronoun or another grammatical clue.

☐ Processing and retention: I recognised parts of the message but couldn’t keep up or remember enough information.

☐ Connecting ideas: I missed a contrast, reason, sequence of events or who was being referred to.

2. What is your evidence?

a. Copy the exact word, chunk or sentence from the transcript that caused difficulty.

---

---

b. Why do you think you missed it?

I knew / recognised ____________________________

but when listening, I ___________________________

---

3. What should I practise next?

With your teacher, choose ONE priority.

☐ Sound or syllable discrimination

☐ Recognising familiar words and chunks in connected speech

☐ Identifying word boundaries in continuous speech

☐ Noticing and interpreting grammatical clues

☐ Following and remembering longer utterances

☐ Tracking meaning and connections across sentences

☐ Learning unfamiliar vocabulary

4. Did the intervention work?

After practice, listen to a NEW extract containing a similar challenge, without the transcript.

☐ I understood it independently.

☐ I improved, but still missed some details.

☐ I still needed support.

My next listening target:

---

---

Remember: Knowing WHAT you got wrong is useful. Understanding WHY you got it wrong is what helps you improve!

Don’t put the transcript away just yet!

There is another important point I would make about the post-comprehension phase, because once students have compared the audio with the transcript and reflected on their errors, why should we simply put the transcript away and move on to another text? In my view, this is a missed opportunity, since the original recording and its transcript can provide the basis for a further round of purposeful listening activities, either targeting the processing difficulties identified through diagnosis or simply giving students additional opportunities to process familiar language more accurately and efficiently. If students struggled to recognise familiar chunks, for example, I might use Word Grab or Chunk Tracking; if they missed grammatical cues, Faulty Transcript or carefully designed contrastive listening activities; and if segmentation proved problematic, tasks requiring them to identify familiar words and chunks within the continuous speech stream. Even when comprehension was successful, revisiting the same recording through different processing activities can be valuable, because understanding a text once doesn’t mean that students can necessarily recognise its language rapidly and reliably under less supportive conditions.

But there is also a powerful motivational dimension to this practice which, in my view, we should never underestimate, especially for those students who initially performed poorly and may have come away from the comprehension task feeling frustrated, inadequate or simply convinced that listening isn’t their thing. Think about it: a student who has just scored 3/10 or 4/10 is given another opportunity to engage with the very same recording, this time through carefully scaffolded activities that make previously inaccessible language increasingly recognisable, until eventually they can follow stretches of speech that only minutes earlier seemed like an incomprehensible blur. What a difference that can make to their sense of competence! Instead of walking away with yet another disappointing score, they experience progress, and crucially, progress they can actually hear for themselves, which may help strengthen their self-efficacy and willingness to persevere with future listening tasks. Of course, success with a familiar recording isn’t proof that they can now understand unfamiliar speech, but as a confidence-building experience, particularly when the support is gradually withdrawn and students realise they can manage independently, it can be immensely valuable. And isn’t that precisely the sort of experience we want our weaker listeners to have, rather than repeatedly exposing them to failure and then moving on?

The important thing is that each replay should have a clear processing purpose, rather than becoming another passive listening exercise, and that we gradually move from audio supported by the transcript to audio alone, before checking whether students can process similar language in a new recording. After all, why waste a perfectly good listening text when we’ve only just begun to exploit its learning potential?

We can then go further, asking students to identify where the speech stream became difficult to segment, where they missed grammatical cues or where they lost the thread of the message, although we should remember that students’ explanations aren’t always reliable, since they may think they missed a word because it was spoken too quickly when the real issue was weak lexical knowledge.

That’s why we need to follow up with short diagnostic probes and, after the intervention, check whether students can process similar language in a new recording without transcript support, otherwise we may simply have taught them to understand the original recording…

A rather important distinction, wouldn’t you agree?

6. But what about a class of 30?

Whenever I discuss diagnostic listening with teachers, this question inevitably comes up, and it’s a perfectly reasonable one, because we don’t have the time to interview every student individually after every comprehension exercise, although fortunately we don’t need to!

Let’s say 18 students in your Year 10 class have misunderstood a question involving past time reference, in which case, instead of immediately explaining the answer, you could spend three or four minutes investigating the problem by playing two contrasting verb forms and asking whether students can hear the difference, before presenting the forms in short, familiar sentences to establish whether they can recognise them in context.

Next, ask students to identify the time reference to see whether they can interpret the grammatical meaning, and finally embed the forms in slightly longer utterances to check whether they can still process them accurately, collecting responses through mini-whiteboards, response cards or another whole-class response system.

Within minutes, you have a much clearer picture of what might be going wrong, because perhaps most students can hear the distinction but don’t interpret it correctly, perhaps they manage the forms in isolation but fail when they occur in connected speech, or perhaps only a small group continues to struggle, suggesting the need for more targeted support.

The secret is not to diagnose everything. It’s to investigate one suspected bottleneck at a time!

Over several lessons, patterns emerge, and you begin to see which students consistently struggle with phonological discrimination, which ones have weak spoken lexical recognition, and which ones have difficulty integrating information across longer utterances, with that information influencing what happens next in your teaching.

No elaborate spreadsheets required, no individual diagnostic interviews, just a clear hypothesis, a well-chosen activity and careful observation.

7. From diagnosis to intervention: Don’t just give them another listening test

This is where diagnostic listening becomes genuinely useful, because once we’ve identified a likely cause of comprehension breakdown, we need to do something about it, and the intervention should match the difficulty.

If students struggle with phonemic discrimination, use focused contrastive listening activities; if they can’t segment familiar language, provide repeated opportunities to identify words and chunks in connected speech; if lexical recognition is slow, use repeated, varied recognition activities, including Audio Boxing and other rapid-processing tasks; if they miss grammatical cues, use structured input-processing activities that require them to attend to those cues to interpret meaning; and if they struggle with discourse, use short connected texts requiring them to track references, contrasts and causal relationships.

The key is to avoid the all-too-common response of simply giving students another listening comprehension passage and hoping for the best, because the purpose of diagnosis is precisely to identify what needs strengthening.

And, as we say in Italian, “Non si può curare un male senza conoscerne la causa” (you cannot cure an ailment without knowing its cause), which, when you think about it, captures precisely what diagnostic listening is about: how can we possibly expect to remedy a comprehension difficulty if we haven’t first made a serious attempt to understand what caused it?

There are three principles I would insist on.

First, target the suspected cause rather than the symptom, because a wrong answer involving past time reference doesn’t automatically mean students need another lesson on the past tense.

Second, gradually remove support, since transcripts, repeated exposure and shorter extracts can be useful during intervention, but students ultimately need to understand unfamiliar recordings independently.

Third, check for transfer, because if students can recognise an expression after hearing it six times, that’s encouraging, but does it necessarily mean the underlying difficulty has been resolved, and can they recognise similar language in a new recording, with a different speaker, or a week later?

That’s the evidence we’re looking for, and a fortiori, if a student cannot recognise a familiar expression in a short, carefully controlled utterance, why would we expect them to process it successfully in a longer, faster and less predictable GCSE recording?

8. Where does diagnostic listening fit within MARS EARS?

For teachers familiar with my Extensive Processing Instruction framework, diagnostic listening fits naturally within MARS EARS, because the entire approach is built around the principle that students need extensive, carefully sequenced processing of language before we can reasonably expect accurate and increasingly spontaneous performance.

During Modelling and Awareness-raising, we anticipate potentially challenging phonological, prosodic and grammatical features and make them salient, while during Receptive Processing, we provide extensive, carefully sequenced opportunities to process familiar language through listening and reading, using brief diagnostic probes when difficulties emerge.

During Structured Production and Expansion, we continue reinforcing relevant sound–meaning and form–meaning relationships while gradually developing students’ productive control, whereas at the Autonomy checkpoint, we establish what students can understand independently, without sentence builders, transcripts or substantial scaffolding.

These low-stakes assessments are particularly useful starting points for diagnostic investigation, because they tell us what students can actually process without the support that may have helped them succeed during earlier stages.

During Routinization, we aim to make recognition increasingly accurate, rapid and automatic, while as learners move towards Spontaneity, they must process language under progressively less predictable conditions.

Within the EPI schemes of learning I’ve been developing, the Autonomy checkpoints at the end of successive MARSEA sub-units assess listening, reading, writing and speaking respectively, with the listening checkpoint providing an opportunity to identify comprehension difficulties and use the findings to inform subsequent instruction.

But diagnostic listening isn’t confined to that checkpoint, since it can follow any listening comprehension task that reveals a meaningful pattern of errors.

One important principle of EPI is that we don’t simply move on because we’ve taught something; we need evidence that students can process it independently and increasingly efficiently, otherwise what exactly are we moving on from?

9. A ten-minute diagnostic listening routine you can try tomorrow

Let’s make this concrete by imagining that your Year 10 French class has just completed a listening comprehension task containing the extract Samedi dernier, je suis allé au cinéma avec mes amis, mais nous sommes rentrés tôt parce que le film était ennuyeux, and that several students have misunderstood when the event occurred or why the group returned home early.

Here’s what I would do.

Minutes 1–2: Go back to the error. Replay the relevant extract and ask students to identify which part of the message they failed to understand, without giving them the answer yet.

Minutes 3–4: Check recognition. Establish whether students can recognise samedi dernier, sommes rentrés, mais and parce que when heard in short, familiar utterances, using mini-whiteboards to collect responses.

Minutes 5–6: Investigate the suspected difficulty. If students recognise the words but fail to interpret the relationships between events, use short contrastive sentences such as Nous sommes rentrés tôt parce que le film était ennuyeux and Nous sommes rentrés tôt, mais le film était intéressant, asking students to identify whether the second clause gives a reason or introduces a contrast.

If the problem is lexical recognition instead, use short Word Grab or Chunk Tracking activities targeting the expressions students missed.

Minutes 7–8: Use the transcript intelligently. Display the relevant section, ask students to identify precisely what they failed to hear or interpret, and provide a brief, focused intervention.

Minutes 9–10: Check again. Play a new sentence containing similar features, such as Dimanche dernier, nous sommes allés au parc, mais nous sommes partis tôt parce qu’il pleuvait, and establish whether students can now identify when the event occurred and why the group left.

This isn’t a definitive test of learning, of course, because immediate improvement doesn’t necessarily translate into durable acquisition, but it gives us an indication of whether our intervention is moving things in the right direction, and surely that is preferable to simply replaying the original recording, displaying the transcript and moving on?

10. Ten mistakes to avoid when diagnosing listening

Before wrapping up, here are ten pitfalls worth keeping in mind, particularly because even well-intentioned diagnostic work can become misleading if we draw conclusions too quickly or fail to control the demands of the tasks we use.

1. Assuming every wrong answer is a vocabulary problem, when students may know the words but cannot recognise them in speech.

2. Confusing a comprehension score with a diagnosis, since a score tells us how well students performed, not why.

3. Jumping to conclusions after one error, rather than looking for patterns and converging evidence.

4. Assuming that understanding the transcript means the listening problem has been solved, when reading and listening are not interchangeable.

5. Always slowing down the recording, even though slower speech may alter natural prosodic cues and conceal rather than resolve the underlying problem.

6. Blaming working memory for everything, when what looks like a memory problem may actually be inefficient lexical recognition.

7. Using diagnostic tasks that are too complicated, so that reading, memory, grammar and lexical demands become difficult to disentangle.

8. Failing to distinguish unfamiliar language from familiar language that students cannot decode, despite the fact that these require different responses.

9. Diagnosing without intervening, which rather defeats the purpose of finding out what went wrong.

10. Checking improvement only on the original recording, instead of establishing whether students can process similar language independently in new contexts.

11. So, how do we get from 4/10 to 8/10?

Let’s return to the student who scored 4/10 and suppose that, after investigating their errors, we discover that their main difficulties involve recognising familiar chunks in connected speech, processing spoken tense markers and interpreting relationships between clauses.Instead of giving them another ten GCSE listening papers, we spend the next few weeks systematically targeting those difficulties, using short, engaging micro-listening activities, recycling familiar language in varied contexts, gradually increasing processing demands and checking whether recognition is becoming more accurate and automatic.

We continue to use comprehension tasks, of course, but not just to measure performance, because we also want to establish whether our interventions are making a difference, whether students can process similar language independently and whether any improvement survives beyond the immediate lesson.

Perhaps the student eventually scores 6/10, perhaps 8/10, although there are no guarantees and I certainly wouldn’t claim that diagnostic listening automatically produces a particular increase in exam marks, because that would require evidence from controlled studies which, as far as I am aware, we do not yet have for this particular instructional approach.

But here’s what I would argue: if a student repeatedly fails at listening comprehension, giving them more listening comprehension exercises without investigating the causes of failure is unlikely to be the most efficient way of helping them improve.

After thirty years of teaching languages, one thing I’m increasingly convinced of is that we need to spend less time simply checking whether students understand and more time developing the processes that enable them to understand, because without that shift in emphasis we risk perpetuating a cycle in which students repeatedly experience failure without ever receiving the targeted support they actually need.

As we say in Italian, non basta sapere dove si cade, bisogna capire perché, meaning that it’s not enough to know where you fall; you need to understand why, which is precisely what diagnostic listening is about.

Not abandoning comprehension tasks, not turning every lesson into a laboratory experiment and certainly not drowning students in metacognitive questionnaires or lengthy self-assessment forms, but simply making better use of the information their errors provide.

The next time your students complete a GCSE listening exercise, don’t just look at the scores, but examine the wrong answers, ask yourself what might have caused them, replay the relevant extracts, investigate the suspected difficulties and select activities that address them, before checking whether students can process similar language independently…

Ultimately, the point of diagnostic listening isn’t to become better at explaining why students got questions wrong, but to use that knowledge to help them get more questions right next time and, more importantly, to become better listeners!

Research and further reading

• Field, J. (2008). Listening in the Language Classroom. Cambridge University Press.
• Vandergrift, L., & Goh, C. C. M. (2012). Teaching and Learning Second Language Listening: Metacognition in Action. Routledge.
• Cutler, A. (2012). Native Listening: Language Experience and the Recognition of Spoken Words. MIT Press.
• Rost, M. (2011). Teaching and Researching Listening (2nd ed.). Pearson.
• Conti, G., & Smith, S. (2019) Breaking the Sound Barrier: Teaching Learners How to Listen. Piefke Tradimg

Introduction

What research tells us about developing accurate, fluent and independent readers.

Reading aloud is a staple of the modern foreign languages classroom Yet it is often poorly sequenced, insufficiently scaffolded and used without a clear understanding of the cognitive processes involved. Effective reading aloud requires learners to coordinate orthographic recognition, phonological decoding, lexical retrieval, articulation and prosody. For novice learners, these processes can place considerable demands on working memory.Here are ten common mistakes, organised around the progression from initial phonological preparation to independent reading fluency.

1. Moving to reading aloud without sufficient receptive processing and explicit phonics instruction

Many teachers I have observed over the years including my former self, model the pronunciation of words and sentences before asking students to read them aloud. However, hearing a model once or twice is not the same as having processed the language sufficiently to reproduce it accurately and fluently. Learners may recognise words when they hear them but still lack sufficiently stable phonological representations, secure sound–spelling connections and rapid access to the relevant lexical sequences.

In my recent post on how to scaffold dictation fluency, I emphasised the importance of carefully sequencing activities along a continuum from receptive to productive processing, gradually withdrawing support as learners become more proficient. The same principle applies to reading aloud: productive performance should emerge from extensive, carefully structured receptive processing, not merely follow a brief demonstration.

An important component of this progression is explicit phonics instruction. However, before expecting learners to apply grapheme–phoneme correspondences independently, we should ensure that they have had sufficient opportunities to perceive, discriminate between and become familiar with the target sounds.

This is where carefully designed auditory discrimination and phonological awareness activities become particularly valuable. For example, Faulty Echo, in which learners detect deliberate discrepancies between a model and its repetition, can encourage close attention to phonological detail. Similarly, aural sorting, sound contrasts and syllable-identification tasks can help learners notice distinctions they might otherwise overlook.

These activities can prepare learners for explicit attention to sound–spelling relationships. In French, for instance, learners might first discriminate between the vowels represented by u and ou, before learning to recognise and apply the corresponding graphemes in familiar and unfamiliar words.

The progression can be conceptualised as follows:

Auditory modelling (teacher models target sounds, words and phrases through repeated listening)

↓

Phonological discrimination (e.g. Faulty Echo, identifying whether two words sound the same or different, aural sorting)

↓

Phonological awareness (e.g. identifying syllables, recognising individual sounds, locating sounds within words)

↓

Explicit sound–spelling mapping (e.g. matching sounds to graphemes, identifying recurring spelling patterns, contrasting French u and ou)

↓

Supported decoding (e.g. echo reading, choral reading, reading familiar words and phrases with teacher support)

↓

Independent reading aloud (reading familiar language without an immediate model, applying learned sound–spelling correspondences to new word combinations)

Importantly, this is a pedagogical progression, not a rigid sequence of isolated stages. Sound–spelling correspondences can be modelled from the outset, and auditory processing, phonological awareness and explicit phonics instruction can overlap and reinforce one another. The essential principle is that learners should not be expected to apply correspondences independently before they have had sufficient opportunities to perceive, recognise and practise them.

In conclusion: before expecting independent reading aloud, provide repeated opportunities to listen to, discriminate between, recognise and process the target language. Use activities such as Faulty Echo, aural sorting, syllable identification, matching audio to written text and identifying recurring sound–spelling patterns. Explicitly teach high-frequency grapheme–phoneme correspondences and provide supported decoding practice before gradually withdrawing auditory and visual scaffolding.

Research basis: Baddeley (2003); Ehri (2014); Share (1995).

2. Failing to teach learners how to segment written language into meaningful chunks

One of the defining characteristics of fluent reading aloud is the ability to process and articulate language in meaningful units rather than as a succession of isolated words. Skilled readers draw on familiar lexical sequences, syntactic structures and prosodic patterns to organise their reading. This reduces the need for laborious word-by-word processing and facilitates more natural rhythm, stress and intonation.

Chunking is therefore both a cognitive and a prosodic phenomenon. Familiar multiword sequences can be retrieved and processed relatively efficiently, while prosodic chunking helps readers organise speech into meaningful phrases. These two processes are related but not identical: a prosodic phrase may contain several lexical chunks.

Consider the following examples, where slashes indicate natural or pedagogically useful phrase boundaries:

• French: Le week-end dernier / je suis allé au cinéma / avec mes amis.
• Spanish: Cuando tengo tiempo / me gusta salir / con mis amigos.
• German: Am Wochenende / gehe ich oft / mit meinen Freunden / ins Kino.
• Italian: Quando ho tempo / mi piace andare al cinema / con i miei amici.

A novice reader may decode these sentences word by word, producing frequent pauses and unnatural stress patterns. A proficient reader is more likely to recognise familiar sequences and syntactic groupings, allowing articulation to proceed more smoothly. The exact prosodic boundaries will, of course, vary according to emphasis, speaking rate and context.

Why does this matter for fluency? When word recognition and access to familiar sequences become more automatic, fewer attentional resources are needed for local processing. This makes it easier to plan articulation, anticipate upcoming material and produce appropriate prosody. Skilled reading aloud thus involves both efficient recognition and the organisation of speech into meaningful units.

Instead: Teach learners to identify and rehearse meaningful phrases. Use visual chunking, teacher modelling, echo reading and repeated phrase-level practice. Gradually remove the visual boundaries and check whether learners can reproduce appropriate phrasing independently.

Research basis: LaBerge & Samuels (1974); Ellis (2002); Kuhn & Stahl (2003); Rasinski (2004).

3. Asking learners to decode too much unfamiliar language at once

Reading aloud is cognitively demanding because it requires the coordination of several processes, often under considerable time pressure. Novice learners must recognise written symbols, access or construct their phonological forms, retrieve relevant lexical information, organise articulation and produce speech with appropriate rhythm and intonation. When too much language is unfamiliar, these processes compete for limited attentional and working-memory resources.

A cognitive model of reading aloud

The following is a simplified, research-informed model of the processes involved:

1. Visual and orthographic processing

The learner identifies letters, letter combinations and familiar written word forms.

Example: Recognising the French sequence eau in beaucoup.

↓

2. Orthographic–phonological conversion and lexical access

The learner retrieves a familiar word’s pronunciation or constructs a plausible pronunciation using learned sound–spelling correspondences. Recognition of meaning may occur alongside these processes.

Example: Recognising that beaucoup is pronounced /bo.ku/, rather than pronouncing each letter separately.

↓

3. Phonological encoding and temporary maintenance

The learner activates and organises the sound sequence for production, temporarily maintaining information as necessary while preparing the utterance.

Example: Holding the phonological sequence of beaucoup de travail while preparing to articulate it.

↓

4. Prosodic planning and chunking

The learner organises the utterance into appropriate rhythmic and intonational units, drawing on syntactic knowledge, familiar sequences and the intended meaning.

Example: Grouping J’ai beaucoup de travail / aujourd’hui into meaningful phrases.

↓

5. Articulatory planning and execution

The learner prepares and coordinates the speech movements needed to produce the utterance.

Example: Producing the sentence with appropriate sound contrasts, liaison where required, rhythm and intonation.

↓

6. Auditory and somatosensory monitoring

The learner monitors their speech and may detect mismatches between the intended and actual pronunciation, making adjustments where possible.

Example: Noticing an incorrectly pronounced vowel and attempting to correct it.

Important: This model is a pedagogical simplification, not a strictly serial account. In skilled reading aloud, several processes overlap and operate interactively. Familiar words may be recognised and pronounced through relatively direct lexical pathways, while unfamiliar words require more effortful decoding.

Why unfamiliar language creates problems

When orthographic recognition and phonological decoding are not sufficiently automatic, they consume attentional resources that could otherwise support prosodic planning, articulation and monitoring.

The result may be hesitant reading, inaccurate pronunciation, inappropriate pauses and reduced fluency.

Instead: Reduce unnecessary processing demands by using familiar vocabulary, previously modelled sound–spelling patterns and short, highly patterned sentences. Introduce new decoding challenges selectively and progressively. As recognition and decoding become more automatic, gradually increase sentence length, syntactic complexity and the proportion of unfamiliar language.

The objective is not to eliminate challenge but to manage cognitive demands so that learners can practise the processes targeted by the activity without being overwhelmed by competing demands.

Research basis: Coltheart et al. (2001); Levelt, Roelofs & Meyer (1999); Baddeley (2003); Sweller (1988); LaBerge & Samuels (1974).

4.Using round-robin reading as the main form of practice

Having students take turns reading sentences or paragraphs aloud provides limited active reading time for each learner. It can also create unnecessary performance pressure, particularly among less confident readers.

Instead: Maximise simultaneous participation through choral reading, paired reading and repeated short rehearsals. Reserve individual performances for carefully selected opportunities, ideally after learners have had sufficient preparation.

Research basis: National Reading Panel (2000); Kuhn & Stahl (2003).

5. Correcting every pronunciation error immediately

Constant correction can interrupt the flow of reading, increase self-consciousness and divert attention from the overall task. Moreover, not all pronunciation errors have the same implications for intelligibility or learning.

Instead: Prioritise errors that affect intelligibility or reveal important gaps in sound–spelling knowledge. Use selective feedback, brief modelling and opportunities for immediate self-correction or another attempt. Where appropriate, delay feedback until the learner has completed a meaningful phrase.

Research basis: Lyster & Ranta (1997); Saito & Lyster (2012).

6. Mistaking successful supported reading for independent mastery

A student who reads a sentence accurately immediately after hearing it may simply be reproducing a recently activated phonological sequence. Successful performance under these conditions does not necessarily demonstrate durable decoding ability.

Instead: Revisit the language after a delay, remove the model and present familiar words in unfamiliar combinations. Check whether learners can decode accurately without immediate auditory support. This provides a more meaningful indication of whether sound–spelling correspondences have been consolidated and can be applied independently.

Research basis: Share (1995); Ehri (2014).

7. Neglecting stress, rhythm and intonation

Accurately pronouncing individual words does not automatically result in fluent, natural reading. Learners also need to develop prosodic control: appropriate stress, rhythm, phrasing and intonation.

Instead: Explicitly model sentence stress, rhythm, intonation and meaningful pauses. Use echo reading and repeated phrase-level practice to develop expressive, intelligible delivery. Encourage learners to imitate the prosodic features of whole utterances, not merely the pronunciation of individual words.

Research basis: Rasinski (2004); Kuhn & Stahl (2003).

8. Prioritising reading speed before accuracy and expression

Reading quickly is not necessarily reading fluently. When learners are encouraged to accelerate before they have developed sufficiently – as when we ask them to read aloud tongue twisters meant for native speakers – accurate and stable decoding skills, they may resort to guessing, omit words, distort pronunciation or sacrifice natural phrasing. Fluency involves the efficient coordination of accuracy, automaticity and prosody, not speed in isolation.

Instead: Establish accurate decoding and appropriate phrasing before introducing speed as an explicit performance target. Initially, learners should concentrate on recognising words reliably, articulating them correctly and grouping them into meaningful prosodic units. Once these processes are reasonably secure, increasing reading rate becomes a legitimate and valuable instructional objective.

Research basis: National Reading Panel (2000); Kuhn & Stahl (2003).

9. Failing to systematically develop reading automaticity through fluency-building activities

However, the fact that speed should not be prioritised prematurely does not mean that it should be neglected. Quite the opposite. Once learners have developed reasonably accurate decoding skills, teachers should deliberately create opportunities for them to read familiar language increasingly quickly and effortlessly.

Reading-aloud fluency does not develop simply through exposure to written texts or occasional opportunities to read aloud. Learners need a carefully staged sequence of activities that progressively develops accuracy, automaticity and prosodic control.

Importantly, automaticity is not simply the by-product of accurate reading: it is an instructional objective in its own right. Repeated, appropriately challenging practice can help learners reduce the time and attentional effort required to recognise, retrieve and articulate familiar language.

This is where purposeful EPI fluency-building activities such as timed repeated reading, paced reading and personal-best challenges become particularly valuable. They introduce an element of challenge, make progress visible and encourage learners to process familiar language more efficiently, without sacrificing accuracy or expression.

Instead: Organise reading-aloud instruction around a progression that initially prioritises accurate, supported decoding and subsequently incorporates deliberate fluency-building practice. Once learners can read the target material accurately, introduce repeated readings with gradually reduced completion times, paced reading and other carefully calibrated speed challenges.

The objective is not to make learners rush. It is to accelerate automaticity so that fluent reading becomes progressively less effortful.

A possible progression towards fluent reading aloud

Stage 1 – Phonological familiarisation

Learners listen to short phrases and sentences, focusing on pronunciation, stress and intonation before attempting to read them aloud.

Stage 2 – Sound–spelling mapping

Learners identify recurring grapheme–phoneme correspondences in familiar words and practise recognising these patterns in other lexical items.

Stage 3 – Supported decoding

Learners read short, familiar phrases with visual support, teacher modelling and echo reading.

Stage 4 – Chunked reading

Learners practise reading sentences divided into meaningful phrases, gradually developing more natural phrasing and prosody.

Stage 5 – Repeated reading and timed fluency practice

Learners rehearse short passages several times, aiming for progressively greater accuracy, automaticity and expressive delivery.

Once accuracy is established, introduce activities such as:

• Timed repeated reading: Students read the same short, familiar passage aloud several times, recording how long each attempt takes and trying to improve their previous performance without sacrificing pronunciation or prosody.
• Beat your personal best: Students repeat a familiar passage, attempting to reduce their completion time while maintaining an agreed accuracy criterion.
• Timed chunk reading: Students repeatedly read short, meaningful phrases, gradually increasing the rate at which they can recognise and articulate them accurately.

These activities make progress visible and introduce a motivational element of self-competition. The objective is not simply to speak faster but to reduce the attentional effort required to recognise, retrieve and articulate familiar language.

Stage 6 – Independent reading and paced reading

Learners read familiar texts without immediate modelling or visual chunking cues, demonstrating increasing control over decoding and prosody.

At this stage, paced reading can provide an additional challenge. Students read accessible passages within predetermined time limits, with targets gradually adjusted as their processing becomes more efficient.

For example, learners might read a short passage aloud at a comfortable rate, establish a baseline and then practise towards a modestly faster rate over subsequent attempts. Accuracy and natural phrasing remain essential.

The principle is similar to the timed reading approaches discussed by Chung and Nation (2006), although their research primarily concerns silent L2 reading speed rather than oral reading fluency.

Stage 7 – Delayed transfer

Learners read new combinations of familiar language after a delay, demonstrating that they can apply established decoding and prosodic skills beyond the original practice material.

Where appropriate, teachers can also compare reading accuracy, rate and prosody across different occasions to determine whether improvements are retained and transferable rather than restricted to a repeatedly rehearsed passage.

Why timed practice can help

The theoretical rationale is straightforward. As recognition and decoding processes become more automatic, they require fewer attentional resources, potentially freeing capacity for articulation, prosodic planning and comprehension.

Repeated, appropriately challenging practice can contribute to this automatisation. Timed activities add a measurable performance goal and may increase motivation, particularly when learners compete against their own previous performance rather than against classmates.

Nevertheless, faster reading is not automatically better reading. Excessive time pressure can encourage guessing, omissions and unnatural delivery. Timed practice should therefore be introduced only when learners can already read the material with reasonable accuracy, and improvements should be judged against accuracy and prosodic quality, not speed alone.

The key is to increase processing demands systematically while progressively reducing scaffolding, rather than simply asking learners to read longer or more difficult passages.

This progression is a pedagogical synthesis informed by research on automaticity, orthographic learning and guided oral reading, rather than a single experimentally validated instructional programme.

Research basis: LaBerge & Samuels (1974); National Reading Panel (2000); Kuhn & Stahl (2003); Ehri (2014); Chung & Nation (2006).

10. Failing to develop learners’ ability to self-monitor and self-correct their reading aloud

Reading aloud is not simply a matter of decoding written language and articulating it. It also involves monitoring one’s own speech, detecting discrepancies between intended and actual pronunciation, and making adjustments where necessary. Skilled readers can use auditory and articulatory feedback to regulate their performance, whereas novice learners may struggle to recognise their own errors, particularly when much of their attention is devoted to decoding.

This creates an important instructional challenge: learners cannot reliably correct pronunciation errors they are unable to perceive. Moreover, even when they notice that something sounds wrong, they may lack the phonological knowledge or articulatory control needed to correct it.

Self-monitoring therefore needs to be developed explicitly and progressively, rather than simply expecting learners to recognise and correct their mistakes independently.

Instead: Incorporate activities that gradually develop learners’ ability to evaluate and regulate their own reading aloud. For example:

• Listen, compare and correct: Learners record themselves reading a short, familiar passage, listen to an accurate model, identify discrepancies and attempt a second recording.
• Focused self-monitoring: Learners listen back to their reading with one specific target in mind, such as French nasal vowels, silent final consonants or appropriate phrasing, rather than attempting to monitor everything simultaneously.
• Best Recording (a variation of Paul Nation’s activity of the same name): Learners work in pairs, taking turns to read aloud a short, familiar passage while their partner records them. After each attempt, they listen to the recording together, identify aspects that could be improved (e.g. pronunciation, phrasing, stress or intonation), and rehearse before recording again. The process is repeated until the learner is satisfied that they have produced their best possible version. The partner provides focused feedback and helps the reader notice discrepancies they might otherwise overlook. The emphasis is on successive improvement through repeated performance, self-evaluation and peer support, rather than simply producing a flawless recording on the first attempt.

Initially, the teacher should provide substantial support through modelling, focused listening tasks and explicit feedback. As learners become more proficient, this support can be progressively withdrawn, encouraging greater independence in error detection and correction.

Importantly, self-monitoring should not become so intrusive that it disrupts reading fluency. The aim is to develop increasingly accurate internal monitoring, not to encourage learners to interrupt themselves after every word. During early practice, listening back to recordings may be more productive than attempting continuous self-correction while reading.

Ultimately, the goal is to help learners move from dependence on external correction towards greater autonomy in evaluating and improving their own oral reading performance.

Research basis: Levelt (1983); Levelt, Roelofs & Meyer (1999); Saito & Lyster (2012).

Concluding remarks

Effective reading-aloud instruction is not about getting students to pronounce a passage successfully on a particular occasion. It is about developing the cognitive processes that eventually make accurate, fluent and independent reading possible.

This requires a carefully sequenced progression:

Phonological familiarisation → Sound–spelling mapping → Chunking → Supported decoding → Guided oral practice → Selective feedback → Prosodic development → Automaticity → Independent transfer.

Purposeful fluency-building activities, including timed repeated reading and paced reading, can be embedded within this progression to make practice more engaging and encourage increasingly efficient processing.

From an Extensive Processing Instruction (EPI) perspective, reading aloud should therefore be embedded within a broader input-to-output sequence, rather than used as an isolated pronunciation activity.

One important qualification: much of the strongest evidence on oral reading fluency and orthographic learning comes from first-language literacy research. The implications for MFL instruction are plausible and pedagogically useful, but should not all be presented as findings directly demonstrated in secondary foreign-language classrooms. Likewise, findings on timed silent reading should not automatically be generalised to reading aloud.

The central principle remains:

Model before expecting performance. Scaffold before demanding independence. Consolidate before accelerating. And test transfer before assuming mastery.

Selected references

• Baddeley, A. (2003). Working memory: Looking back and looking forward. Nature Reviews Neuroscience, 4, 829–839.
• Chung, M., & Nation, I. S. P. (2006). The effect of a speed reading course. English Teaching, 61(4), 181–204.
• Coltheart, M., Rastle, K., Perry, C., Langdon, R., & Ziegler, J. (2001). DRC: A dual route cascaded model of visual word recognition and reading aloud. Psychological Review, 108(1), 204–256.
• Ehri, L. C. (2014). Orthographic mapping in the acquisition of sight word reading, spelling memory, and vocabulary learning. Scientific Studies of Reading, 18(1), 5–21.
• Ellis, N. C. (2002). Frequency effects in language processing: A review with implications for theories of implicit and explicit language acquisition. Studies in Second Language Acquisition, 24(2), 143–188.
• Kuhn, M. R., & Stahl, S. A. (2003). Fluency: A review of developmental and remedial practices. Journal of Educational Psychology, 95(1), 3–21.
• LaBerge, D., & Samuels, S. J. (1974). Toward a theory of automatic information processing in reading. Cognitive Psychology, 6(2), 293–323.
• Levelt, W. J. M., Roelofs, A., & Meyer, A. S. (1999). A theory of lexical access in speech production. Behavioral and Brain Sciences, 22(1), 1–38.
• Lyster, R., & Ranta, L. (1997). Corrective feedback and learner uptake. Studies in Second Language Acquisition, 19(1), 37–66.
• National Reading Panel. (2000). Teaching Children to Read. National Institute of Child Health and Human Development.
• Rasinski, T. V. (2004). Assessing Reading Fluency. Pacific Resources for Education and Learning.
• Saito, K., & Lyster, R. (2012). Effects of form-focused instruction and corrective feedback on L2 pronunciation development. Language Learning, 62(2), 595–633.
• Share, D. L. (1995). Phonological recoding and self-teaching: Sine qua non of reading acquisition. Cognition, 55(2), 151–218.
• Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257–285.

Introduction

Dictation is one of the oldest activities in the language teacher’s repertoire. Yet, despite its apparent simplicity, it is also one of the most cognitively demanding things we can ask a beginner language learner to do.

Consider what happens when we dictate even a relatively simple sentence to a Year 7 student. The learner must perceive a rapidly disappearing acoustic signal, segment it into words, recognise those words, hold what they have heard in working memory, retrieve their written forms and reproduce them accurately — while the next part of the sentence may already be arriving.

Thing is: for an experienced language learner, many of these processes have become relatively automatic. For a beginner, they haven’t, though.

This is why simply doing more dictation is not necessarily the best way to make students better at dictation. If we want learners eventually to transcribe connected speech successfully, we need to teach and scaffold the processes that successful dictation requires.

In this post, I will suggest a progression which moves from highly supported listening to independent transcription, following a simple principle rooted in what we know about how the brain works:

Hear and discriminate → segment → identify → reconstruct → complete → transcribe with support → transcribe independently.

The key thing here is: scaffolded, inclusive progression.

The cognitive bases of dictation fluency: from perception to grapho-motor execution

Before considering how we should scaffold dictation, it is useful to understand what successful dictation actually involves cognitively. The diagram below provides a simplified model of the journey from hearing the speech signal to producing the words on the page.

At its simplest, successful dictation requires learners to turn a continuous stream of sound into an accurate sequence of written words. First, they have to perceive the sounds accurately. They then have to work out where words and chunks begin and end, recognise those words, and use their knowledge of vocabulary, grammar and context to help predict and confirm what they are hearing. Next, they have to retrieve the correct spellings and physically write them down. Meanwhile, working memory has to keep hold of what has just been heard while all of this processing takes place, and learners need to monitor what they are writing to check that it makes sense and corresponds to the original input.

With experienced listeners and writers, many of these processes happen rapidly and relatively automatically. With Year 7 beginners, however, several of them are still slow, fragile and effortful!This is precisely why asking beginners simply to “listen and write what you hear” can impose such a heavy cognitive burden, overwhelming for many of your less confident learners.

The table below breaks the process down further. Of course, these processes do not operate as a perfectly neat sequence in the brain. Brain processes rarely do! Instead, they interact constantly and often operate in parallel. The model above is therefore best understood as a useful pedagogical representation of the major demands involved in dictation, rather than as a literal step-by-step account of cognition.

Cognitive process	What does the learner have to do?	French example	What might go wrong for a beginner?
1. Acoustic input	Receive a continuous and rapidly disappearing speech signal. Spoken language does not arrive with spaces between words.	The learner hears Je vais au cinéma avec mes amis as one continuous stream.	The sentence may simply sound like a blur of unfamiliar sounds.
2. Phonological perception	Identify the relevant speech sounds and other phonological information such as rhythm and intonation.	The learner needs to perceive cinéma accurately enough to recognise it.	An unstable phonological representation may cause the learner to mishear a familiar word or fail to recognise it altogether.
3. Segmentation	Work out where words and meaningful chunks begin and end in continuous speech.	The learner needs to perceive je vais / au cinéma / avec mes amis rather than one undifferentiated stream.	The learner may know every word on the page but fail to locate those words in connected speech.
4. Lexical access	Match what has been heard with words stored in long-term memory and activate their meaning.	Hearing /sinema/ activates the known lexical item cinéma.	Recognition may be too slow, or the learner may know cinéma visually but not recognise it quickly enough from its spoken form.
5. Grammatical and contextual prediction	Use grammar, meaning and familiar chunks to anticipate and confirm what is likely to come next.	After je vais au…, the learner may expect a place such as cinéma. After avec mes…, a noun such as amis becomes highly predictable.	Beginners have fewer automatised chunks and less grammatical knowledge available to constrain the possibilities.
6. Orthographic retrieval	Retrieve the correct written form of the words that have been recognised.	The learner recognises /sinema/ and retrieves cinéma, including the correct accent.	The learner may recognise the word perfectly but write sinema, cinema or another plausible sound-based spelling.
7. Grapho-motor execution	Turn the retrieved orthographic representation into handwriting or typed output.	The learner physically writes cinéma while retaining the next part of the sentence.	If handwriting or spelling is slow and effortful, it consumes resources that could otherwise be devoted to listening and remembering.
8. Working memory	Temporarily maintain and coordinate what has been heard while words are recognised, spellings retrieved and previous material written down.	While writing au cinéma, the learner may already need to retain avec mes amis long enough to transcribe it next.	Earlier words may disappear from memory while the learner concentrates on spelling or writing another word.
9. Monitoring and self-correction	Compare the developing written version with what was heard and with existing knowledge of vocabulary, grammar and spelling.	The learner writes avec mes ami, notices the mismatch and corrects it to avec mes amis.	So much attention may be devoted to getting words onto the page that little capacity remains for checking and correcting.

From controlled processing to fluent dictation

This model also helps explain what we mean by dictation fluency: fluent learner does not necessarily possess a fundamentally different set of skills from a beginner. The crucial difference is that many of these processes have become faster and more automatic, eliciting very little processing load in working memory.

Consider the word cinéma. A beginner might have to consciously analyse the sounds, search memory for the word, retrieve its meaning, remember its spelling and then think about how to write it. An experienced learner may move from hearing /sinema/ to activating cinéma — its pronunciation, meaning and spelling — extremely rapidly.

The same applies at sentence level. A more proficient learner hearing:

Je vais au cinéma avec mes…

is already using grammatical, lexical and contextual knowledge to anticipate what could plausibly follow. Processing is therefore not exclusively bottom-up. What learners already know helps them interpret what they are hearing. This has a major implication for teaching, the most important one is that full dictation requires the successful coordination of phonological perception, segmentation, lexical access, grammatical prediction, orthographic retrieval, working memory and grapho-motor execution.

It follows that repeatedly giving beginners full dictations and hoping that these processes improve is not necessarily the most efficient instructional approach.Instead, we can reduce the complexity of the final task, isolate or foreground some of its component processes, provide support where learners need it and gradually remove that support.

That gives us the rationale for the progression that follows:

Hear and discriminate → segment → identify → reconstruct → complete → transcribe with support → transcribe independently.

And this is where both my experience with weaker French L2-learners and research show that effective dictation instruction should begin.

1. Why full dictation is so difficult for beginners

Suppose a Year 7 French class hears:

Le week-end, je joue souvent au foot avec mes amis.

To write this sentence successfully, learners need to do much more than “know the vocabulary”.

They must identify where one word finishes and another begins. They need sufficiently stable phonological representations of souvent, avec, mes amis, etc. They must map the sounds they perceive onto words stored in long-term memory. They must retain several elements while writing. They must retrieve appropriate spelling conventions. They also need to use their emerging knowledge of French grammar and syntax to predict what is likely to come next.

In other words, dictation simultaneously places demands on speech perception, segmentation, phonological processing, lexical access, working memory, grammatical prediction and orthographic encoding.

That is quite a lot for an eleven-year-old who started learning French six weeks ago.

There is another important issue. The speech signal does not arrive neatly packaged as the sequence of words we see on the page. Native and proficient speakers perceive meaningful units within a continuous stream of sound because their brains have become extremely efficient at segmentation and lexical recognition. Beginners have not yet developed this ability. Consequently, when learners repeatedly fail at conventional dictation, the problem may not simply be that they “haven’t learnt the words”. They may know a word perfectly well when they see it and still fail to recognise it in connected speech.

This is why I would not normally begin dictation training with full dictation. Instead, I would build towards it carefully scaffolding the process. starting with the most important process: phonemic discrimination.

2. Start with discrimination, not transcription

At the beginning of the progression, I want the learner’s attention almost entirely on the sound.

A simple multiple-choice listening task works extremely well here.

For example, students might hear:

Je joue souvent au foot.

and choose between:

A. Je joue souvent au foot.
B. Je joue souvent au tennis.
C. Je joue parfois au foot.
D. Je ne joue jamais au foot.

There is virtually no writing involved. This matters because the learner is not simultaneously worrying about spelling, handwriting and remembering an entire sentence, more attentional capacity can be devoted to the all-important careful auditory discrimination.

The alternatives should not simply be random sentences. Good distractors should require learners to attend to acoustically or semantically important contrasts.This makes multiple choice an excellent entry point into dictation training, provided we see it not as a testing device but as a listening scaffold.

3. Move from general recognition to precise listening

Once learners can identify the overall sentence, we can increase the perceptual demand. One useful next step is Missing Detail.

Students see most of the sentence but one important element is missing. They listen and identify what belongs in the gap.

For example:

Le week-end, je joue ______ avec mes amis.

They hear:

Le week-end, je joue au tennis avec mes amis.

The task now requires more than recognising which sentence was spoken in that the learner must focus attention on a specific point in the speech stream and extract the relevant information. The writing burden remains small and manageable, however.

This is important in the early stages. We are increasing the listening challenge without unnecessarily increasing every other source of cognitive demand at the same time.

4. Teach learners to reconstruct what they hear

The next stage can introduce Sentence Puzzles. Learners hear a sentence and are given its constituent chunks in a scrambled order:

avec mes amis | je | le week-end | au foot | joue

Their task is to reconstruct:

Le week-end, je joue au foot avec mes amis.

This is a powerful bridge between recognition and transcription because the learner no longer simply selects an answer. They must hold the sentence in mind sufficiently well to reconstruct its sequence. At the same time, the words themselves are provided.

This means the activity removes much of the orthographic retrieval burden while retaining demands on listening, memory, word order and syntactic prediction. It also encourages learners to process language in chunks, rather than attempting to hold an entire sentence as a sequence of isolated words.

5. Teach them to find the boundaries in speech

One of the greatest obstacles beginner listeners face is segmentation. On the page:

Je vais au cinéma avec mes amis.

looks like seven clearly separated words.

In speech, learners do not hear seven convenient little boxes. Instaed, they encounter a continuous acoustic stream.

An activity such as Break the Flow can therefore explicitly train learners to identify boundaries between words or meaningful chunks. Students hear a sentence and indicate where they believe the divisions occur.

At first, I would use highly familiar language. The purpose is not to test whether learners know obscure vocabulary. It is to help them become better at finding familiar language inside connected speech.

This distinction is fundamental. Listening instruction should not always be about asking:

“Did you understand it?”

Sometimes it should be about teaching learners:

“How do you find the language you already know when somebody actually says it?”

6. Zoom in on individual words

We can then narrow the learner’s attention still further. In Spot the Word, students hear a sentence and identify or type a particular word they have detected.This may sound relatively undemanding, but it trains an important listening behaviour: selective lexical identification. Instead of trying to transcribe everything, learners practise locking onto individual lexical items within a larger stream of speech.

Why is it valuable? Because with beginner learners because full dictation can otherwise become an all-or-nothing experience. Miss the beginning of the sentence and panic sets in; while the learner is trying to reconstruct what has already disappeared, the rest of the sentence passes them by.

Training learners to recover individual words helps develop more resilient listening behaviour.

7. Remove some of the word, but not all of it

We can now begin increasing the orthographic demand. Activities involving Missing Vowels or Missing Consonants provide learners with incomplete written representations of the words they hear. Instead of writing a word entirely from memory, students use the acoustic signal to reconstruct its missing letters.

This provides a useful intermediate stage because learners must increasingly connect sound and spelling, but they are not yet required to retrieve the entire orthographic representation independently.

I would not necessarily regard Missing Vowels and Missing Consonants as two rigidly consecutive levels of difficulty. Their relative difficulty depends on the target language, the particular lexical items and the sound-spelling relationships involved. It is better to regard them as two complementary forms of partial transcription.

The pedagogical principle is more important than their precise ordering: don’t remove all the support at once. Fade it gradually.

8. Move to guided dictation

Only after learners have had substantial experience of recognising, segmenting, reconstructing and partially transcribing language would I move towards something resembling conventional dictation. But even here, there is an important intermediate stage: Guided Dictation.

Students might be given:

Le ______, je joue ______ au foot avec mes ______.

and hear:

Le week-end, je joue souvent au foot avec mes amis.

Or they might receive first letters, word lengths, selected words, chunk boundaries or other carefully chosen prompts.

The learner now has to transcribe substantial portions of the sentence, but the existing information reduces uncertainty and working-memory demands.

Crucially, guided dictation allows us to manipulate difficulty very precisely.

Early on, perhaps 70 per cent of the sentence is provided.Later, 50 per cent.Then 20 per cent.Eventually, nothing.

This is scaffolding in the proper sense of the word: support is provided because learners need it, and progressively withdrawn as they become capable of performing independently.

9. Finally, full dictation

Full dictation should be the destination, not necessarily the starting point.By the time learners reach this stage, they have repeatedly practised many of its component processes:

discriminating sounds → identifying details → recognising words → reconstructing sequences → segmenting connected speech → mapping sounds onto spelling → retaining chunks → transcribing.

Now the scaffolds disappear and learners hear:

Le week-end, je joue souvent au foot avec mes amis.

and write:

Le week-end, je joue souvent au foot avec mes amis.

What appears to be a very simple classroom activity is actually the culmination of a carefully engineered instructional sequence. And because learners have been prepared for it, full dictation is far less likely to become the familiar scenario in which confident students succeed, weaker listeners write three words and everyone waits for the teacher to reveal the answer.

10. A few important principles

Several principles should underpin the progression.

First, use familiar language. Especially at beginner level, dictation training should predominantly consolidate language that has already been extensively modelled and processed. If the vocabulary, grammar, pronunciation and spelling are all new, it becomes very difficult to know what exactly the task is developing.

Second, keep the language short initially. Sentence-level listening is ideal because it allows repeated, intensive processing without overwhelming working memory.

Third, allow repetition strategically. The objective is not to simulate an examination every time students listen. During learning, replaying the input allows learners to test hypotheses, notice previously missed information and gradually construct a more accurate representation.

Fourth, vary what learners have to listen for. Sometimes the target should be a phonological contrast; sometimes a word boundary; sometimes a grammatical morpheme; sometimes lexical meaning; sometimes spelling.

Fifth, make difficulty progressive. Don’t move from “listen and tick” directly to “write everything you hear”.

And finally, don’t confuse dictation practice with dictation testing.

If every dictation is scored, high-stakes or performed under strict test conditions, we risk turning an extremely useful instructional technique into yet another assessment event. During the learning phase, errors should provide information:

What didn’t I hear? What did I mis-segment? Which sound-spelling correspondence fooled me? Which word did I know on paper but fail to recognise aurally?

Having this metacognitive reflection at the end of any such activity is for me the icing on the cake. Having done my PhD in metacognition I can vouch for the importance of the lightbulb moments that this self-questioning often sparks.

11. The Language Gym Dictation Trainer: putting the progression into practice

This thinking lies behind the Dictation Trainer on The Language Gym, the latest addition to the http://www.language-gym.com’s game room. Rather than presenting beginner learners immediately with an empty box and asking them to transcribe complete sentences, the trainer takes them through a series of increasingly demanding listening and reconstruction activities:

Multiple Choice → Missing Detail → Sentence Puzzle → Break the Flow → Spot the Word → Missing Vowels / Missing Consonants → Guided Dictation → Dictation

Each activity removes a little more support.

Multiple Choice develops careful auditory discrimination without requiring transcription.

Missing Detail directs attention towards specific information in the speech stream.

Sentence Puzzle requires learners to reconstruct the sequence they have heard while providing the language itself.

Break the Flow focuses explicitly on one of the most important bottom-up listening skills: segmenting continuous speech into meaningful units.

Spot the Word develops selective lexical recognition within connected speech.

Missing Vowels and Missing Consonants strengthen the mapping between phonological and orthographic representations while still providing substantial visual support.

Guided Dictation removes much of that support and requires increasingly independent transcription.

Finally, Dictation removes the scaffold altogether.

The important point is that these are not simply nine different ways of making dictation more entertaining.

They embody a learning progression.

The learner moves from:

recognition → focused identification → reconstruction → segmentation → partial transcription → supported transcription → independent transcription.

And that progression mirrors the cognitive model discussed earlier. Rather than asking the novice learner to coordinate all the processes involved in dictation successfully from the outset, we manipulate the task so that particular processes can be practised with reduced demands elsewhere.

In other words, the Dictation Trainer operationalises a principle that, according to cognitive load theory, process-based instruction and, most importanty common sense should underpin much of beginner language teaching of any complex skills, including dictation:

Don’t repeatedly test the complex performance you eventually want. Break that performance down, practise its component processes, scaffold their integration, and gradually remove the support.

By the time a Year 7 learner reaches the final Dictation activity, we are no longer simply hoping that they can write what they hear. Rather,we have taught them how to get there.

Introduction

Over the years, through workshops, conferences, social media and conversations with teachers, I have noticed that the very same questions about EPI keep coming back. They are important questions because they concern what ultimately matters most: how we move from supported performance to independence, from knowledge to fluency, and from classroom practice to genuine communication!

In this post, I address six of the questions I am asked most frequently:

1. How do you teach writing in EPI, and how do you move from micro-writing to extended writing?
2. How do we move students from sentence builders and scaffolding to genuinely independent speaking and writing?
3. How do we get students to speak spontaneously rather than simply reproducing memorised sentences?
4. What is the best way to correct errors without damaging fluency, confidence or motivation?
5. How do we recycle language effectively across units so that students actually remember it months later?
6. How do I embed culture in an EPI curriculum?

Although these may appear to be six separate questions, they are actually connected by one fundamental principle: we cannot expect learners to perform complex language tasks successfully unless we systematically develop and practise the processes that underpin them. Whether we are talking about writing, speaking, error correction, retrieval, recycling or culture, the objective is ultimately the same: to help learners develop a repertoire of language that is durable, rapidly retrievable, flexible and usable in real communication!

And, as always, the journey from input to independence matters just as much as the destination!

1. How do you teach writing in EPI, and how do you move from micro-writing to extended writing?

In EPI, writing is taught as a process, rather than simply practised as a product. Too often, students are given a title, a few bullet points and perhaps a vocabulary sheet and are expected to produce 90 or 150 words. Yet extended writing requires the learner to perform numerous demanding processes simultaneously. L2 writing research has long conceptualised writing as the coordination of multiple cognitive and linguistic processes, including planning, formulation, lexical retrieval and monitoring (Kellogg, 1996; Kormos, 2012; Manchón, 2011). If those processes have not been adequately trained, simply asking students to write more extended pieces will not necessarily make them better writers!

The model I use identifies eight processes involved in successful writing: understanding the task, generating ideas, retrieving language, building sentences, applying grammar, organising and linking ideas, checking whilst writing, and improving after writing. Each needs to be explicitly developed.

This is where micro-writing is particularly useful. Instead of practising all eight processes simultaneously, we can design short tasks which foreground one or two:

• Understanding the task: 1. give students three possible responses to an exam bullet point and ask which actually addresses it and why. 2. Give them a title such as The pros and cons of driving a car and a set of sentences some of which are pertinent and some of which aren’t. The task: crossing out the irrelevant ones, or ranking them in descending order of relevance
• Generating ideas: 1. give a prompt such as What did you do last weekend? and allow 45 seconds to generate five possible ideas before writing anything. 2. Give them the same prompt and sentence stems to complete.
• Retrieving language: 1. Give five English chunks or visual prompts and ask students to retrieve the target-language equivalents from memory. 2. Stage a No snakes no ladders with mini whiteboards
• Building sentences: 1. provide three familiar chunks and challenge students to combine them into one well-formed sentence. 2. Give them a sentence and then ask them to extend them with prompts such as when, how, where, who with
• Applying grammar: 1.Give a sentence in the present and ask students to rewrite it accurately in the past; 2. Give them incorrect sentence to rewrite accurately
• Organising and linking ideas: 1. Give three short sentences and ask students to connect them into a coherent mini-paragraph. 2. Stage a jigsaw reading task
• Checking whilst writing: 1. Students write three sentences but pause after each to run through a short checking routine, for example checking the verb, agreement and word order. 2. Students write one sentence at a time, pausing after each to check meaning, accuracy and cohesion, identify and repair any problems, then continue writing.
• Improving after writing: 1. Give students their own four-sentence text back and ask them to upgrade two sentences by adding detail, complexity or greater lexical precision. 2. Ask them to keep a personalised list of their recurrent mistakes and ask them to use it before handing it a written piece.

These individual processes sit within five broader principles which shape how I approach writing instruction in EPI.

1. Write little and often. Replace occasional marathon writing sessions with frequent, low-stakes writing opportunities. Writing fluency is built through regular retrieval, sentence-building, manipulation and quick-response tasks rather than through constantly producing extended pieces. A student who writes four or five carefully focused sentences several times a week may get far more useful writing practice than one who produces a long piece every few weeks! This emphasis on repeated retrieval and increasingly fluent access to language is consistent with research on retrieval practice and skill development (DeKeyser, 2007; Nakata, 2017).

2. Train the processes, not just the product. Successful writing depends on processes such as idea generation, lexical retrieval, grammatical and orthographic encoding, sentence assembly and monitoring. These processes need to be practised explicitly and systematically, rather than assuming that they will somehow develop simply because students write longer texts (Kellogg, 1996; Kormos, 2012).

3. Prime before you write. Activate the language before asking students to write. Listening, reading, retrieval and speaking tasks can reactivate relevant vocabulary, chunks, structures and ideas, making them more readily accessible when students begin writing. Research on pre-task planning and pre-learning similarly suggests that making relevant linguistic resources more accessible can support subsequent L2 production (Ellis, 2005; Webb, 2009). We should not repeatedly send learners to the blank page linguistically “cold”!

4. Build automaticity through rapid response. Activities such as mini-whiteboard work, retrieval tasks, translation sprints, Do Nows and plenaries can develop greater speed and fluency in accessing language. Skill-acquisition accounts of L2 learning emphasise the importance of extensive practice in moving from effortful processing towards increasingly rapid and automatised performance (DeKeyser, 2007; Segalowitz, 2010). Knowing a word or structure is not enough if retrieving it consumes so much cognitive capacity that little remains for constructing and monitoring the sentence!

5. Use extended writing sparingly but purposefully. Extended writing certainly has a place, but primarily to integrate, demonstrate and diagnose writing skills, not as the dominant form of writing practice. If the underlying processes are weak, repeatedly asking students to produce 90 or 150 words simply gives them repeated practice at struggling with those same processes.

So how do we eventually get to extended writing?

The transition involves progressively integrating the processes that have previously been practised in smaller combinations. Students move from individual sentences to sentence expansion and manipulation, from sentences to linked sentence chains, from sentence chains to short paragraphs, and eventually from paragraphs to extended texts. At the same time, scaffolding is gradually faded and retrieval demands increase.

This is important motivationally too! A struggling learner confronted with a blank page and “Write 150 words” can experience failure before writing has even begun. Micro-writing creates frequent, achievable successes, whilst allowing us to diagnose exactly where the problem lies.

The principle is simple: write little and often, train the processes, prime before writing, build automaticity, progressively integrate the processes and only then expect learners to orchestrate them independently in extended writing!

2. How do we move students from sentence builders and scaffolding to genuinely independent speaking and writing?

One criticism occasionally levelled at EPI is that sentence builders create dependency. They can, but only if they are used incorrectly. A sentence builder is a temporary scaffold, not the destination. The purpose of scaffolding is ultimately to make itself redundant!

At the beginning of an instructional sequence, students need substantial support because the language is unfamiliar and retrieval from long-term memory is not yet possible. Sentence builders reduce cognitive load, make useful patterns transparent and allow learners to experience successful comprehension and production very early on. This is compatible with cognitive-load accounts of instruction, which emphasise the importance of reducing unnecessary demands on limited working-memory resources during initial learning (Sweller, Ayres & Kalyuga, 2011).

However, there is a crucial stage between using a sentence builder and dispensing with it: retrieval practice.This is where, in my experience, quite a few teachers who have adopted the MARS EARS cycle move too quickly. They spend considerable time modelling the language and carrying out receptive-processing activities, but not nearly enough time on the repeated retrieval required to make that language readily available without the scaffold. Then, when the sentence builder disappears, students struggle, and the conclusion is sometimes that they have become “dependent” on it. Often, the real problem is simply that retrieval has not been practised sufficiently! Research on retrieval practice shows that effortful retrieval can strengthen subsequent retention more effectively than additional exposure alone (Karpicke & Roediger, 2008), whilst L2 research has demonstrated benefits from repeated retrieval for vocabulary learning (Nakata, 2017).

Fading should therefore be gradual. Students might initially work with the complete sentence builder, then with sections covered, then with increasingly minimal prompts, then perhaps with English cues or images, and eventually with no linguistic support whatsoever. Activities such as delayed copying, sentence reconstruction, translation, retrieval grids, oral retrieval from visual cues and writing from memory can progressively increase retrieval strength.

However, independence requires more than recalling the original sentences. Students must also manipulate and recombine the language. Change the subject. Change the tense. Add a reason. Substitute an adjective. Combine two familiar structures. Respond to an unexpected question. Use a familiar chunk in an unfamiliar context.The criterion for removing scaffolding should therefore not be time, as in “we have used this sentence builder for three lessons”. It should be evidence that learners can retrieve and manipulate the target language successfully.

This is why the retrieval work in MARS EARS matters so much. Repeated retrieval, spaced over time and under increasingly demanding conditions, is what converts supported performance into increasingly autonomous performance (Nakata, 2015, 2017).

The journey is essentially:

scaffolded processing → retrieval → faded retrieval → manipulation → recombination → autonomous use

The goal was never independence from the language in the sentence builder. We want that language firmly embedded in long-term memory! What we want is independence from the sentence builder itself.

3. How do we get students to speak spontaneously rather than simply reproducing memorised sentences?

First, we need to rethink what “spontaneous speaking” actually means. Spontaneity does not mean generating language from nowhere. Fluent speakers draw constantly on familiar lexical chunks, collocations, syntactic patterns and communicative routines stored in long-term memory. Research on formulaic language has repeatedly highlighted the role played by prefabricated sequences in fluent language production (Pawley & Syder, 1983; Wray, 2002; Wood, 2010). You cannot spontaneously retrieve language which you have never acquired!

The real objective is therefore to develop a sufficiently large and automatised repertoire which learners can retrieve, adapt and recombine rapidly in response to an unpredictable communicative need. Research into L2 fluency similarly stresses the importance of automaticity and speed of access to linguistic knowledge (Segalowitz, 2010).

This is why speaking instruction, like writing instruction, needs to be process-based. Simply telling students to “have a conversation” does not train the processes which successful conversation requires.

I typically build towards communicative interaction progressively. Early on, chunking aloud helps establish accurate and fluent articulatory routines. For example, students can repeatedly produce a difficult phrase such as je voudrais aller or ich würde gerne as a single phonological unit rather than assembling it word by word. Research on formulaic sequences provides a strong rationale for building readily accessible multiword units, and intervention research has found benefits of explicit formulaic-sequence instruction for aspects of oral fluency.

We then need substantial oral and written interactional retrieval practice. Students might see a visual cue and retrieve a question, hear a question and rapidly retrieve an appropriate response, or complete a written interaction in which each answer determines the next question. The aim is to make high-frequency interactional language increasingly accessible and rapidly retrievable.

Next come mini role-plays. These are very short exchanges with a limited communicative objective. Students might buy a train ticket, decline an invitation politely, arrange when and where to meet, order something in a café or complain about a hotel room. These provide genuine interaction without overwhelming working memory.

We then progressively increase unpredictability through communicative tasks such as information gaps, spot-the-difference activities, problem-solving, surveys, decision-making tasks or conversations in which students do not know in advance exactly what their partner will say. Task-based research has long emphasised the importance of meaning-focused interaction and increasingly demanding conditions for L2 production (Ellis, 2003; Skehan, 1998).

Throughout this sequence, however, we must train the micro-skills of speaking: asking and answering questions, turn-taking, initiating and closing an interaction, reacting, hesitating naturally, requesting clarification, repairing communication, circumlocuting, agreeing and disagreeing, using fillers and buying oneself processing time.

These behaviours also need routinisation. For example, repeated two-minute conversations with changing partners can require students to use expressions equivalent to “Really?”, “Why?”, “What about you?”, clarification requests and follow-up questions until these interactional routines become increasingly automatic.

This is why I do not accept the opposition between memorisation and spontaneity. Memorisation of useful language is necessary but insufficient. The progression should be:

chunking → retrieval → interactional retrieval → mini role-plays → increasingly unpredictable communication

At the same time, we systematically train and routinise the micro-skills which hold conversation together.

Spontaneity is not where speaking instruction begins. It is what well-sequenced, process-based speaking instruction gradually makes possible!

4. What is the best way to correct errors without damaging fluency, confidence or motivation?

The first question I ask is: what is the purpose of the activity?

Not every error needs correcting, and certainly not every error needs correcting immediately. Error correction should be driven by the learning objective and by where students are in the instructional sequence. Research on corrective feedback strongly supports its potential value for L2 development, whilst also showing that its effects depend on factors including feedback type and instructional context (Li, 2010; Lyster & Saito, 2010; Lyster, Saito & Sato, 2013).

During highly controlled work, particularly when students are consolidating a new pattern, accuracy matters greatly. Errors which become repeatedly rehearsed risk becoming entrenched. At this stage, relatively immediate feedback can therefore be useful, particularly when it directs attention to the precise feature students are trying to acquire. Indeed, Lyster and Saito’s classroom meta-analysis found significant and durable benefits of oral corrective feedback, although the effects varied according to the feedback provided.

However, the situation changes when the primary objective is fluency.If a learner is trying to sustain an interaction and we interrupt every few seconds to correct gender, adjective agreement or a verb ending, we force attention away from meaning and towards form. We increase cognitive load, disrupt the flow of communication and potentially make the learner increasingly reluctant to speak.

So correction needs to be selective, purposeful and timed intelligently!

I tend to prioritise errors which impede communication, relate directly to recently taught high-value language, occur repeatedly, or are likely to become habitual. Many other errors can be left temporarily untouched.

Feedback can also be delayed. During a speaking task, for example, the teacher can note recurring errors and address them afterwards with the whole class. This preserves communicative flow whilst still creating an opportunity for learning.

The same principle applies to writing. Covering a student’s page in corrections is rarely productive. Working memory and attention are limited. A learner is far more likely to benefit from focused feedback on two or three recurring or important problems than from twenty corrections which they barely process. Research on written corrective feedback similarly highlights the importance of what learners actually do with the feedback they receive, rather than viewing teacher correction as an end in itself (Ferris, 2010; Storch & Wigglesworth, 2010).

Most importantly, correction should generate further processing. Simply showing students the correct form is often insufficient. Ask them to retrieve it again, discriminate between correct and incorrect versions, reconstruct the sentence or use the corrected pattern in another context.

Error correction should therefore not be conceived as the teacher’s attempt to eliminate every mistake. Errors are inevitable in language development.

The objective is to decide which errors matter now, when attention should be drawn to them, and what students should subsequently do with that feedback.

Correction is most effective when it becomes another opportunity for learning, rather than merely an indication that something was wrong!

5. How do we recycle language effectively across units so that students actually remember it months later?

One of the biggest problems in language curricula is that we confuse having taught something with students having learnt it. A structure appears in Unit 2, students practise it, take a test and everybody moves on. By Unit 6, much of it has disappeared.

That is not surprising. Forgetting is normal. A curriculum designed for durable learning must therefore engineer systematic recycling across units. Research on distributed practice and L2 vocabulary learning supports revisiting material across spaced encounters rather than concentrating all practice into a single period (Nakata, 2015; Nakata, Suzuki & He, 2023).

The first mechanism is interleaving. Rather than practising only the language belonging to the current unit, deliberately mix previous material with current learning. If students are studying holidays, for example, a retrieval task might combine where I went with previously learnt language relating to weather, transport, food, opinions and descriptions of people. Students now have to select the appropriate language rather than simply retrieve items from a predictable single set. L2 research has found benefits of interleaving for the longer-term retention of grammatical distinctions and foreign-language verb conjugation (Nakata & Suzuki, 2019; Pan et al., 2019).

Do Now tasks are ideal for this! The first five minutes of a lesson can retrieve language not merely from yesterday but from last week, last month and the previous term. An Exit Ticket can similarly include one item from today’s learning and one or two strategically selected items from earlier units. The important principle here is not the label attached to the activity, but the use of spaced, effortful retrieval.

We also need cumulative receptive and productive tasks. A listening text in Unit 5 should not contain only Unit 5 language. It can deliberately incorporate high-value material from Units 1 to 4. The same applies to reading. Productively, cumulative speaking and writing tasks should require students to mobilise language from several previous contexts rather than merely reproduce the latest unit.

Curriculum sequencing itself can facilitate this. Semantic proximity between adjacent units creates natural opportunities for recycling. A unit on daily routine can lead naturally into school life; school into free time; free time into health; holidays into travel, weather and accommodation. Previous language remains communicatively useful rather than disappearing because the textbook has suddenly moved to an unrelated topic.

Other useful techniques include planned carry-over language, deliberately identifying high-value chunks and structures which must recur in subsequent units; cumulative retrieval grids; interleaved translation; cumulative quizzes; old-language/new-context activities; and requiring students to reuse previously learnt structures with new vocabulary.

The important point is that recycling cannot be left to chance. When planning Unit 6, ask not only “What new language am I teaching?” but also “What from Units 1 to 5 am I deliberately bringing back, and in what form?”

A retrieval-driven curriculum therefore behaves less like a sequence of sealed boxes and more like an expanding network. New learning is constantly connected to old learning!

Ultimately, the test of recycling is not whether students recognise something when we show it to them. It is whether, months later, they can retrieve and deploy it rapidly without being told where it originally came from.

6. How do I embed culture in an EPI curriculum?

When people discuss embedding culture in MFL, the conversation often begins with content: festivals, food, monuments, famous people, films, music, literature and historical events. All of these can enrich a curriculum. However, if the overarching aim of EPI is to develop language fluency and successful communication, I would start somewhere else.

Hence, my first question would be:What culture-related issues could impede successful communication or interaction in the target-language country or countries?

Culture is not simply something we know about. Culture influences what people do with language! This is central to the notion of intercultural communicative competence, in which successful language use involves not merely linguistic knowledge but understanding how language, relationships, social practices and cultural meanings interact (Byram, 1997; Kramsch, 1993).

Take forms of address. A student may know the grammar of tu and vous, du and Sie, or tú and usted, but successful interaction requires more than grammar. Who uses which form with whom? Under what circumstances? Who normally initiates the move towards greater informality? A linguistically accurate choice can still be socially inappropriate. Research into interlanguage pragmatics has repeatedly demonstrated that grammatical competence alone does not guarantee pragmatically appropriate L2 performance (Kasper & Rose, 2002; Taguchi, 2015).

The same applies to requests. A learner may translate “Give me a coffee” perfectly accurately at sentence level but fail to understand how requests are conventionally softened in the target culture. So, when teaching ordering in a café, I would model not merely the vocabulary for drinks but culturally plausible openings, requests, responses, thanks and leave-taking.

Consider greetings. Teaching bonjour or hola as lexical equivalents of “hello” is insufficient. When are they used? Does one greet staff when entering a small shop? What happens when joining a group? When might handshakes, kisses or simply verbal greetings be expected? The language and the cultural script belong together.

The same principle applies to disagreement. Students can learn je ne suis pas d’accord, but authentic interaction requires sensitivity to how directly disagreement is normally expressed in different contexts. We can give students several ways of disagreeing, ask them to decide which might be appropriate in different situations and relationships, and then use them in mini-dialogues and role-plays.

Or take invitations and refusals. Students can role-play inviting someone out, but the cultural learning lies partly in how one accepts, declines, gives a reason, proposes an alternative and closes the exchange without sounding abrupt.

Culture can therefore be embedded naturally into EPI sequences through models, listening texts, dialogues, interactional retrieval, role-plays and communicative tasks. If students are learning requests, teach culturally appropriate requesting. If they are learning opinions, teach culturally plausible agreement and disagreement. If they are learning transactional language, teach the interactional routines surrounding the transaction.

Of course, literature, history, music, geography, festivals and the diversity of target-language societies remain valuable. But cultural learning should not become a succession of interesting facts detached from what students can actually do with the language.

From an EPI perspective, one of its most important purposes is to develop pragmatic and intercultural competence, helping learners understand not merely what to say, but what to say, to whom, when, how and why!

Because grammatical accuracy alone does not guarantee successful communication. You can produce a perfectly grammatical sentence and still get the interaction completely wrong!

Selected bibliography

Byram, M. (1997). Teaching and Assessing Intercultural Communicative Competence. Multilingual Matters.

DeKeyser, R. M. (Ed.). (2007). Practice in a Second Language: Perspectives from Applied Linguistics and Cognitive Psychology. Cambridge University Press.

Ellis, R. (2003). Task-Based Language Learning and Teaching. Oxford University Press.

Ellis, R. (Ed.). (2005). Planning and Task Performance in a Second Language. John Benjamins.

Ferris, D. R. (2010). Second language writing research and written corrective feedback in SLA: Intersections and practical applications. Studies in Second Language Acquisition, 32(2), 181–201.

Karpicke, J. D., & Roediger, H. L. (2008). The critical importance of retrieval for learning. Science, 319, 966–968.

Kasper, G., & Rose, K. R. (2002). Pragmatic Development in a Second Language. Blackwell.

Kellogg, R. T. (1996). A model of working memory in writing. In C. M. Levy & S. Ransdell (Eds.), The Science of Writing. Lawrence Erlbaum.

Kormos, J. (2012). The role of individual differences in L2 writing. Journal of Second Language Writing, 21(4), 390–403.

Kramsch, C. (1993). Context and Culture in Language Teaching. Oxford University Press.

Li, S. (2010). The effectiveness of corrective feedback in SLA: A meta-analysis. Language Learning, 60(2), 309–365.

Lyster, R., & Saito, K. (2010). Oral feedback in classroom SLA: A meta-analysis. Studies in Second Language Acquisition, 32(2), 265–302.

Lyster, R., Saito, K., & Sato, M. (2013). Oral corrective feedback in second language classrooms. Language Teaching, 46(1), 1–40.

Manchón, R. M. (Ed.). (2011). Learning-to-Write and Writing-to-Learn in an Additional Language. John Benjamins.

Nakata, T. (2015). Effects of expanding and equal spacing on second language vocabulary learning. Studies in Second Language Acquisition, 37(4), 677–711.

X. (2023). Costs and benefits of spacing for second language vocabulary learning. Language Learning, 73(3), 799–834.

Pawley, A., & Syder, F. H. (1983). Two puzzles for linguistic theory: Nativelike selection and nativelike fluency. In J. C. Richards & R. W. Schmidt (Eds.), Language and Communication. Longman.

Segalowitz, N. (2010). Cognitive Bases of Second Language Fluency. Routledge.

Skehan, P. (1998). A Cognitive Approach to Language Learning. Oxford University Press.

Storch, N., & Wigglesworth, G. (2010). Learners’ processing, uptake, and retention of corrective feedback on writing. Studies in Second Language Acquisition, 32(2), 303–334.

Taguchi, N. (2015). “Contextually” speaking: A survey of pragmatic learning abroad, in class, and online. System, 48, 3–20.

Webb, S. (2009). The effects of pre-learning vocabulary on reading comprehension and writing. Canadian Modern Language Review, 65(3), 441–470.

Wood, D. (2010). Formulaic Language and Second Language Speech Fluency. Continuum.

Here is the list of the upcoming research-informed workshops for those who intend to pursue the EPI accreditation, want to learn a bit more about EPI or simply widen there repertoire of engaging activities across all four language skills. The links below will direct you to the University of Bath’s http://www.networkforlearning.org.uk where you will be able to enrol on your elected courses.

Becoming an EPI Teacher: A Deep Dive into Dr Conti’s Extensive Processing Instruction  Gianfranco Conti | 9 & 10 November         Find out more                                    15 & 16 February           Find out more ___________________________________

Phonics in MFL Gianfranco Conti | 17 November          Find out more                                  23 February           Find out more ____________________________________

Strategies for Effective Grammar Instruction Gianfranco Conti | 18 November            Find out more                                   24 February            Find out more __________________________________

Curriculum Design & Lesson Planning in MFL Gianfranco Conti | 23 & 24 November        Find out more _________________________________

Implementing EPI at KS4 for the new MFL GCSEs Gianfranco Conti | 2 December             Find out more __________________________________

Implementing EPI for Learners at KS2 Gianfranco Conti | 19 January    Find out more __________________________________

Becoming an Accredited EPI Teacher Program Gianfranco Conti | various   Find out more _______________________________________

Becoming an Accredited EPI Teacher – Final Assessment Tutorial Gianfranco Conti | tbc                     Find out more

Introduction

I have written extensively in my books and on this blog about the more obvious fundamentals of vocabulary teaching: retrieval practice, recycling, chunking, spacing and interleaving.

So, let us be clear: this is not another post about those things.

This post is about some of the less obvious reasons why vocabulary often does not stick or, just as importantly, why words that pupils apparently “know” are not available quickly enough when they need them most, especially in GCSE listening, speaking and writing.

1. Vocabulary needs an explicit input-to-output journey

A major weakness in vocabulary teaching is not that words are never recycled. The recently published MFL textbooks do recycle important vocabulary a few times. The bigger problem is that there is often no systematic relationship between the vocabulary encountered in input and the vocabulary subsequently expected in output. This is a major flaw that I have noticed in every single one of the MFL books published by ‘big’ publishers in the last five years. Especially in the new GCSE books.

By an input-to-output sequence, I mean something very simple: pupils first establish the meaning of a carefully selected core of language; they then encounter and process that same language repeatedly through listening and reading; once it has become sufficiently familiar, they retrieve and manipulate it with support before eventually using it more independently in speaking and writing. In other words, there is a substantive phase in which the target vocabulary is taught, an even more substantive input phase flooded with that vocabulary, and finally a pushed-output phase in which the target vocabulary initially taught is practised through interactional speaking and writing.

Boers (2021) is particularly useful here. His research synthesis discusses enhancing vocabulary learning through input, the movement from input to output, retrieval and practice rather than assuming that exposure alone will result in productive command. Research comparing input-only conditions with input followed by learning tasks has likewise found substantially greater vocabulary gains when learners have to do something further with the language rather than merely encounter it (Duong, Montero Perez, Desmet & Peters, 2021).

Why does this matter for GCSE? Because pupils can recognise vocabulary during classroom reading yet fail to retrieve the same language during speaking or writing! Repeated experiences of “I knew it but couldn’t use it” are also hardly conducive to self-efficacy (Bandura, 1997; Mills, Pajares & Herron, 2007).

The fix:

1. Decide on your high-value core first.
2. Pre-teach it.
3. Flood subsequent listening and reading with it.
4. Move through structured retrieval.
5. Only then expect freer production.

This continuity is central to MARS EARS, PIRCO (see image below) and the Sentence Builders approach: the language pupils are eventually expected to use is systematically prepared beforehand rather than appearing almost by accident at the production stage.

2. Do not make new vocabulary and new grammar compete for attention

This follows directly from point 1. A common curriculum sequence introduces a substantial amount of unfamiliar vocabulary and, at almost exactly the same time, asks pupils to understand a new grammatical rule.

That is asking working memory to deal with two major sources of novelty simultaneously, with potentially disastrous consequences for the less able and/or focused students.

Cognitive Load Theory reminds us that working-memory capacity is limited when processing unfamiliar information (Sweller, Ayres & Kalyuga, 2011). In second-language learning, the problem is particularly acute because even apparently simple sentences may require learners simultaneously to decode unfamiliar words, hold their meanings in memory, notice morphology and word order, and work out the underlying message.

The result can be superficial learning of both vocabulary and grammar, which later produces the familiar Year 11 complaint: “But we covered this.” Perhaps. But covering something and establishing a usable representation in long-term memory are very different things.

The exam consequences are obvious: if pupils have never had enough processing capacity to consolidate the lexical material properly, they will struggle to access it under pressure. If language lessons repeatedly generate overload and failure, pupils’ sense of competence can suffer too — important when self-efficacy is strongly related to persistence and achievement (Bandura, 1997; Mills et al., 2007).

The fix:

1. Establish the new lexical material first through highly comprehensible listening and reading.
2. Once much of that language has become familiar, use it as the known linguistic material through which pupils notice and work on the new grammatical pattern.
3. Only then move into heavier manipulation and freer use.

This is precisely why MARS EARS and PIRCO delay heavier grammatical manipulation until learners have built sufficient familiarity with the lexical material: reduce unnecessary novelty first, then ask pupils to analyse and manipulate.

3. Vocabulary needs to be taught through listening in context

This is one of the most important points in this post. Vocabulary learning in school is still heavily associated with print: vocabulary lists, knowledge organisers, textbook pages, flashcards and written retrieval.

Meanwhile, listening has historically been far more likely to be tested than systematically taught (Graham, 2017; Graham & Macaro, 2008). More recent work by Graham and Zhang has demonstrated that vocabulary can indeed be deliberately taught through listening and that different forms of explicit lexical support during oral input can promote vocabulary learning (Zhang & Graham, 2020; Graham & Zhang, 2024). Steven Smith and I have also advocated for substantive teaching of vocabulary to occur in every single input-to-output pedagogical cycle.

The consequence of a print-heavy diet is that a pupil can know perfectly well what a word looks like while possessing a far less stable representation of what it sounds like.

Listening also gives learners very little time. Native spoken-word processing begins extraordinarily quickly: lexical candidates can start being activated within roughly the first 100–150 milliseconds of speech input, and spoken-word recognition proceeds incrementally as the acoustic signal unfolds. For an L2 learner whose phonological representation of a word is weak, this is obviously brutal!

Matthews and Cheng (2015) found that the ability to recognise high-frequency words in speech correlated very strongly with listening comprehension; recognition of vocabulary at one frequency band alone accounted for 52% of the variance in listening scores in their sample. Masrai (2022), working with Arabic-speaking learners, similarly found a strong relationship between phonological vocabulary knowledge and listening comprehension.

So this is not solved merely by doing more phonics, reading aloud or dictation. Phonics matters because it helps stabilise sound–spelling representations, but pupils also need to meet vocabulary repeatedly in meaningful connected speech.

The fix:

1. Teach vocabulary through short, highly comprehensible listening contexts from the beginning.
2. Recycle the same lexical material through connected speech, not isolated word audio.
3. Gradually increase processing speed and complexity.
4. Keep sound–meaning mapping central throughout.

This is exactly why we created The Language Gym Listening Trainer, Dictation Trainer and Audio Boxing: to make pupils process familiar vocabulary repeatedly through the ear, in context and under progressively greater time pressure — not merely recognise it on a vocabulary sheet.

4. Knowing the meaning is not the same as knowing the word

A pupil sees chien and knows that it means dog. Good, but that does not demonstrate full lexical knowledge.Nation’s influential framework distinguishes knowledge of spoken form, written form, word parts, meaning, associations, grammatical behaviour, collocations and constraints on use (Nation, 2001). Schmitt (2010, 2014) likewise stresses that lexical knowledge has both breadth and depth.

Breadth asks how many words do I know? Depth asks how well do I know each one?That distinction matters enormously in the current GCSE climate because a finite prescribed vocabulary list can encourage the seductive idea that learning means accumulating form–meaning pairings. However, a pupil may know the basic meaning and still:

1. mishear the word;
2. spell it incorrectly;
3. place it incorrectly in a sentence;
4. fail to inflect it;
5. fail to recognise a common derivative;
6. fail to combine it with the right words;
7. or simply retrieve it too slowly.

Any of those can affect examination performance. From the learner’s perspective, repeatedly “knowing” vocabulary during revision but being unable to mobilise it successfully in an assessment is confusing and demoralising.

The fix: for your most important vocabulary, deliberately practise different dimensions of word knowledge. Sometimes identify the word aurally; sometimes reconstruct the spelling; sometimes manipulate the morphology; sometimes retrieve the surrounding chunk; sometimes place it in a new sentence.

This multidimensional treatment is built into the Sentence Builders books and Language Gym workouts: important lexical material is not simply presented and translated; it is repeatedly processed through different operations.

5. Pupils usually need many more encounters than neat CPD soundbites suggest

There is no magic number of encounters after which a word suddenly becomes “learned”. That needs saying clearly.Webb (2007) demonstrated continued growth across repeated encounters and, crucially, showed that different dimensions of vocabulary knowledge develop at different rates. Uchihara, Webb and Yanagisawa’s (2019) meta-analysis confirmed the general relationship between frequency of encounter and incidental vocabulary learning.

The problem is that research findings are sometimes translated into suspiciously tidy CPD claims such as “six encounters” or “eight encounters”. However, Six or eight encounters may produce some learning under certain conditions, but that is very different from saying they will produce a durable, multidimensional lexical representation capable of rapid retrieval months later.

This distinction matters because neat low encounter numbers are extremely convenient for overloaded schemes of learning and compact textbooks. If hundreds of items have to be “covered”, reassuring teachers that a handful of encounters will do makes the curriculum fit the available pages. Thing is tthe learner’s memory does not care how many pages the publisher has.

A textbook study by Bergström and colleagues, for example, found that target vocabulary in the materials they examined was not recycled sufficiently to support vocabulary development — a useful warning against equating textbook appearance with genuine learning opportunities.

The fix:

1. Stop asking “How many times have I taught this?”
2. Start auditing how many meaningful encounters your core vocabulary actually receives across a unit, term and year.
3. Distinguish basic familiarity from durable, fast, productive knowledge.

This is why our Sentence Builders materials and Language Gym resources deliberately recycle a relatively constrained lexical core intensively. I would rather pupils know 100 high-value items extremely well than “cover” 300 that disappear three weeks later.

6. Spacing is not enough: number, quality, modality and phonological stability all matter

Everyone has heard about spacing, distributed practice and interleaving. Important? Absolutely. But spacing weak learning does not magically turn it into strong learning.Four other variables matter enormously:

1. How many encounters pupils receive.
2. What they actually do during those encounters.
3. Which modalities they process the word through.
4. How stable the phonological representation becomes.

First, number matters, as we have just seen (Webb, 2007; Uchihara et al., 2019).

Second, quality matters. Depth-of-processing work and vocabulary-specific research suggest that encounters requiring richer semantic and formal processing are generally more memorable than superficial exposure (Craik & Lockhart, 1972; Laufer & Hulstijn, 2001).

Third, modality matters, although teachers should resist simplistic claims that “more modes are always better”. Research on multimodal vocabulary learning is mixed, partly because extra visual information can help under some conditions and create extra cognitive load under others. The important point for MFL is much more basic: vocabulary required for both listening and reading cannot sensibly be practised almost exclusively in print.

Fourth, the phonological representation has to stabilise. A pupil who is uncertain what a word actually sounds like is going to struggle to recognise it rapidly when embedded in connected speech, however confidently they can recognise the spelling.

GCSE exposes all four weaknesses. The listening paper does not reward vocabulary that exists only orthographically; spontaneous speaking does not reward vocabulary that requires four seconds of conscious search.

The fix: plan the learning history of high-value vocabulary across:

1. sound and print;
2. receptive and productive processing;
3. recognition and retrieval;
4. shallow and deeper processing;
5. repeated encounters over sufficient time.

This is very much the design principle behind PIRCO, MARS EARS, the Sentence Builders books and the Language Gym: repeated encounters, yes — but repeated encounters that actually do different cognitive work.

7. Eventually vocabulary has to become fast

This is one of the most neglected distinctions in school vocabulary teaching. A pupil can retrieve parce que after four seconds. Technically, they know it. The provlem is that in real-time listening or speaking, four seconds is an eternity.

Fluent language processing depends partly on automaticity: frequently required linguistic material has to become accessible with progressively less conscious attention (DeKeyser, 2007; Segalowitz, 2010).

This matters because working memory is limited. If a learner has to consciously search for the meaning of every second word, very little capacity remains for understanding the message. Likewise, in speaking, if almost every lexical item requires deliberate retrieval, the learner has fewer attentional resources available for syntax, morphology, pronunciation and planning the next part of the utterance.

The listening data discussed earlier make the point starkly. Matthews and Cheng’s learners’ ability to recognise spoken high-frequency vocabulary was very strongly associated with overall listening comprehension.

GCSE listening is therefore partly a race against processing time. Knowing vocabulary is not enough; learners need to access it quickly.

The fix:

1. Establish the vocabulary accurately first.
2. Practise increasingly rapid recognition and retrieval through familiar language.
3. Introduce time pressure gradually.
4. Keep success rates high enough that speed practice strengthens rather than destabilises knowledge.

This is one reason Language Gym games such as Audio Boxing and the vocabulary/listening trainers use repeated, increasingly rapid responses. The target is not frantic speed. The target is automatic access.

8. Small structural words deserve far more attention

Vocabulary teaching naturally privileges content words such as environment, cinema, holidays, football, and technology. They are salient, carry obvious semantic meaning and look good on vocabulary lists.

However, language is held together by large numbers of high-frequency function and structure words: articles, pronouns, auxiliaries, prepositions, conjunctions, particles, negators and similar items. 55% of every text we read or listen to consists of such words.

Field (2008) demonstrated how L2 listeners tend to rely heavily on content words and can underprocess function words. The latter are particularly difficult because they are often short, unstressed and phonologically reduced in normal connected speech.

This matters because the pupil can recognise all the impressive content vocabulary and still completely misunderstand the message if they:

1. miss the negative;
2. miss the pronoun;
3. miss the tense-bearing auxiliary;
4. miss the preposition;
5. miss the conjunction expressing contrast.

The whole interpretation may then change! This has obvious implications for GCSE listening and reading papers, but also for speaking and writing accuracy. The irony is that many of these words are among the highest-frequency items in the language, yet they often receive less deliberate practice precisely because teachers assume they are easy.

The fix: when planning a unit, identify not merely the topic vocabulary but the small structural words and chunks that pupils need to process automatically. Give them deliberate listening and retrieval practice too.

This is why Sentence Builders do not consist merely of lists of nouns and adjectives. The recurring small structures surrounding the content words are part of what pupils repeatedly process and retrieve.

9. Not every word needs the same level of mastery

Vocabulary lists often imply that every item deserves equal treatment. However, research distinguishes clearly between receptive and productive vocabulary knowledge (Nation, 2001; Webb, 2008). Learners normally understand substantially more vocabulary than they can spontaneously produce. That is not a defect; it is a normal feature of lexical development.

The curriculum-design mistake is assuming that every item on a GCSE list needs to become equally retrievable in speaking and writing. Trying to automatise everything can actually result in automatizing nothing, because practice gets spread too thinly.

Some words are so useful that pupils should be able to retrieve them rapidly and accurately across many contexts. Others need to be readily recognised in reading and listening but may not justify the same investment in productive practice. A third group may be useful extension vocabulary for some learners.

This distinction is particularly important where curriculum time is limited. Every five minutes spent repeatedly retrieving a very low-value productive item is five minutes unavailable for a word or chunk pupils may need constantly.

The fix: classify vocabulary explicitly into:

1. productive core;
2. productive extension;
3. receptive recognition.

Then teach accordingly.

This principle runs throughout the Sentence Builders approach and our Language Gym resources: prioritisation matters. The aim is not to pretend every item is equally important but to ensure that the highest-value language receives the heaviest processing and retrieval load.

10. Retrieval practice is excellent. A culture of constant judgement is not necessarily excellent.

I have advocated retrieval practice for years. The evidence supporting it is strong: retrieving information from memory improves later retention more effectively than simply restudying it (Roediger & Karpicke, 2006; Karpicke & Roediger, 2008). In my opinion, though, there is a difference between retrieval as learning and retrieval as constant judgement, and that distinction is getting blurred.

In many schools, retrieval has become the compulsory Do Now at the beginning of every lesson, often with marks attached. For the pupil who repeatedly succeeds, this is absolutely fine. But what about the pupil who repeatedly scores:

1. Monday: 3/10.
2. Wednesday: 4/10.
3. Friday: 2/10.

And this happens in the first five minutes of the lesson. You have just begun every lesson by reminding that pupil that they are failing.

Agarwal and colleagues (2014) found that low-stakes retrieval can actually reduce test anxiety, but the operative words are low stakes. The motivational consequences are different when every retrieval episode becomes a scored judgement.

This also risks widening attainment gaps when the bulk of vocabulary learning is effectively delegated to pupils at home. Organised learners with strong routines and support arrive prepared, retrieve successfully and become more confident. The pupils who do not study effectively arrive behind, fail the quiz and fall further behind.

A Matthew-effect dynamic can then emerge. If the same learners spend three years beginning language lessons by being shown that they are poor at vocabulary, departments should not be surprised when some of them do not opt for GCSE.

The fix:

1. Keep retrieval frequent.
2. Make much more of it in-class.
3. Keep it predominantly low-stakes.
4. Engineer high-success practice.
5. Give immediate corrective feedback.
6. Use formal vocabulary tests diagnostically rather than turning every lesson into a mini judgement day.

The Language Gym trainers were designed precisely to generate huge amounts of individual retrieval without requiring every error to become another public or formal mark.

11. Relevance matters as much as frequency – if not more!

This point is especially important in England under the reformed GCSE.

The increased use of frequency evidence in selecting examination vocabulary has a perfectly sensible rationale. High-frequency vocabulary has high utility and gives learners greater coverage across texts and contexts. However, frequency is a linguistic statistic. It is not a theory of motivation. A word can be highly frequent in a corpus yet have little immediate relevance to the life of the children you teach! Conversely, something that pupil desperately wants to say may not sit high enough on somebody’s frequency list.

Research informed by Self-Determination Theory reminds us that perceived competence, autonomy and meaningfulness can influence engagement and persistence (Deci & Ryan, 2000). Dörnyei’s work on L2 motivation likewise shows the importance of connecting language learning to learners’ developing identities and goals (Dörnyei, 2009).

That matters in England because the challenge is not merely achieving good GCSE grades; it is persuading enough pupils to take the GCSE in the first place.

A curriculum driven exclusively by frequency risks becoming highly efficient statistically while being emotionally and communicatively remote from the learner. That is not an argument against frequency; it is an argument against making frequency the sole criterion.

The fix:

1. Retain a carefully selected high-frequency lexical spine.
2. Leave controlled space for language pupils actually want to understand and say.
3. Use relevance as a motivational complement to frequency, not a replacement for it.

This is something we have always tried to do in the Sentence Builders books: combine high-frequency reusable language with content that gives pupils something worth communicating.

12. Do not mistake a vocabulary list for a vocabulary curriculum

This final point brings everything together: a vocabulary list tells you what pupils may need to know. It tells you almost nothing about how well they will know it. A list cannot tell you:

1. whether pupils can recognise the words in connected speech;
2. whether their phonological representations are stable;
3. whether they can retrieve them rapidly;
4. whether they know their grammatical behaviour;
5. whether they have met them enough times;
6. whether the encounters involved sufficiently deep processing;
7. whether they can use them productively;
8. whether they will remember them six months later.

That distinction matters particularly under an examination system built around prescribed vocabulary. There is an understandable temptation to equate curriculum quality with coverage: “We have taught 400 words this year.”

That figure is almost meaningless without knowing what pupils can actually do with those words.

Nation (2001), Schmitt (2010) and Webb and Nation (2017) all point in different ways towards the same central conclusion: lexical knowledge develops incrementally and along multiple dimensions.

The real unit of curriculum planning therefore should not simply be the word introduced. It should be the learning history of the word.

The fix: for the highest-value vocabulary, ask four questions repeatedly:

1. How often have pupils processed it?
2. How deeply have they processed it?
3. Through which modalities?
4. How quickly can they now access it?

That is also, ultimately, the rationale behind MARS EARS, PIRCO, Sentence Builders and The Language Gym. They are not simply different activities; they are attempts to engineer a coherent learning journey so that language does not merely get taught but becomes available.

The bottom line

Why do words fail to stick? Usually because:

1. they were not processed sufficiently often;
2. they were not processed sufficiently deeply;
3. they were not processed through enough useful modalities;
4. their phonological representations remained unstable;
5. they were recognised but not retrieved enough;
6. they were learned but never made fast.

Why do pupils sometimes “know” a word but fail to recognise it in listening? Because knowing its written form and meaning does not guarantee a sufficiently robust phonological representation or sufficiently rapid aural lexical access.

Why can pupils succeed on vocabulary tests but fail in GCSE speaking and writing? Because successful isolated retrieval does not automatically guarantee fluent access while working memory is simultaneously dealing with grammar, meaning and message construction.

And why can a curriculum apparently cover huge amounts of vocabulary without producing confident linguists? Because coverage is not mastery.

So I would stop asking “Have they learned the vocabulary?” and ask instead:

1. How deeply do they know it?
2. How long will they retain it?
3. Can they recognise it in connected speech?
4. Can they retrieve it accurately?
5. Can they access it fast enough when they actually need it?

References

Agarwal, P. K., D’Antonio, L., Roediger, H. L., McDermott, K. B. & McDaniel, M. A. (2014). Classroom-based programs of retrieval practice reduce middle school and high school students’ test anxiety. Journal of Applied Research in Memory and Cognition, 3, 131–139.

Bandura, A. (1997). Self-Efficacy: The Exercise of Control. Freeman.

Boers, F. (2021). Evaluating Second Language Vocabulary and Grammar Instruction: A Synthesis of the Research on Teaching Words, Phrases, and Patterns. Routledge.

Craik, F. I. M. & Lockhart, R. S. (1972). Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior, 11, 671–684.

Deci, E. L. & Ryan, R. M. (2000). The ‘what’ and ‘why’ of goal pursuits: Human needs and the self-determination of behaviour. Psychological Inquiry, 11(4), 227–268.

DeKeyser, R. M. (Ed.). (2007). Practice in a Second Language: Perspectives from Applied Linguistics and Cognitive Psychology. Cambridge University Press.

Dörnyei, Z. (2009). The Psychology of Second Language Acquisition. Oxford University Press.

Duong, T. M., Montero Perez, M., Desmet, P. & Peters, E. (2021). Learning vocabulary in spoken input- and output-based tasks. TASK.

Field, J. (2008). Bricks or mortar: Which parts of the input does a second language listener rely on? TESOL Quarterly, 42(3), 411–432.

Graham, S. (2017). Research into practice: Listening strategies in an instructed classroom setting. Language Teaching, 50(1).

Graham, S. & Zhang, P. A. (2024). Learning vocabulary through listening: The role of strategy use and linguistic proficiency. AILA Review, 37(2), 241–265.

Karpicke, J. D. & Roediger, H. L. (2008). Repeated retrieval during learning is the key to long-term retention. Journal of Memory and Language, 57, 151–162.

Laufer, B. & Hulstijn, J. (2001). Incidental vocabulary acquisition in a second language: The construct of task-induced involvement. Applied Linguistics, 22(1), 1–26.

Masrai, A. (2022). The relationship between two measures of L2 phonological vocabulary knowledge and L2 listening comprehension. TESOL Journal.

Matthews, J. & Cheng, J. (2015). Recognition of high-frequency words from speech as a predictor of L2 listening comprehension. System, 52, 1–13.

Mills, N., Pajares, F. & Herron, C. (2007). Self-efficacy of college intermediate French students: Relation to achievement and motivation. The Modern Language Journal, 91(3), 417–442.

Nation, I. S. P. (2001). Learning Vocabulary in Another Language. Cambridge University Press.

Qian, D. D. (2002). Investigating the relationship between vocabulary knowledge and academic reading performance. Language Learning, 52(3), 513–536.

Roediger, H. L. & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249–255.

Schmitt, N. (2010). Researching Vocabulary: A Vocabulary Research Manual. Palgrave Macmillan.

Schmitt, N. (2014). Size and depth of vocabulary knowledge: What the research shows. Language Learning, 64(4), 913–951.

Segalowitz, N. (2010). Cognitive Bases of Second Language Fluency. Routledge.

Sweller, J., Ayres, P. & Kalyuga, S. (2011). Cognitive Load Theory. Springer.

Uchihara, T., Webb, S. & Yanagisawa, A. (2019). The effects of repetition on incidental vocabulary learning: A meta-analysis of correlational studies. Language Learning, 69(3), 559–599.

Webb, S. (2007). The effects of repetition on vocabulary knowledge. Applied Linguistics, 28(1), 46–65.

Webb, S. (2008). Receptive and productive vocabulary sizes of L2 learners. Studies in Second Language Acquisition, 30(1), 79–95.

Webb, S. & Nation, P. (2017). How Vocabulary Is Learned. Oxford University Press.

Zhang, P. A. & Graham, S. (2020). Vocabulary learning through listening: Comparing L2 explanations, teacher codeswitching, contrastive focus-on-form and incidental learning. Language Teaching Research, 24(6).