I have rarely been as concerned about a development in language education as I was about the pedagogy promoted by the National Centre for Excellence for Language Pedagogy (NCELP), led by Emma Marsden and Rachel Hawkes. This was not because everything the NCELP advocated was wrong. Far from it. Explicit attention to phonics is important, vocabulary selection matters, grammar matters, frequency matters and careful sequencing and recycling are essential. My problem was never with these individual principles, several of which I have advocated myself for years, but with what happened when they are assembled into a pedagogy and, more importantly, with the model of language learning underpinning it.

At the heart of NCELP was a deceptively attractive idea: systematically teach the building blocks of language — phonics, vocabulary and grammar — carefully sequence them and provide sufficient practice, and learners will gradually become able to put them together and communicate. This sounds perfectly logical and makes for a wonderfully tidy curriculum. However, knowing the components of a language and being able to deploy them rapidly and accurately under the enormous processing pressure of real-time communication are two very different things! This is a distinction well established in the skill-acquisition literature (Anderson, 1982; DeKeyser, 1998, 2007).

Take a Year 8 learner attempting to say something relatively simple in French. They may need to retrieve vocabulary, select the correct verb form, apply agreement, organise words syntactically, retrieve accurate phonological representations and monitor what they are saying, whilst simultaneously thinking about the message they want to convey. For an experienced speaker this happens largely automatically; for a novice, whose working memory is already heavily taxed, it can be enormously demanding. Words and rules can eventually be assembled into fluent speech, but only once a great deal of the language involved has become sufficiently familiar and automatised for its retrieval to place relatively little burden on working memory.

This is why I found NCELP’s deliberate de-emphasis of unanalysed chunks particularly difficult to reconcile with what we know about fluent language use. Decades of research into formulaic language, from Pawley and Syder through Wray, Ellis, Conklin and Schmitt, Boers and others, have shown that fluent speakers do not generate every utterance from individual words by applying grammatical rules from scratch. Language is massively phraseological: we store and retrieve countless multi-word sequences, collocations and partially fixed constructions, dramatically reducing processing demands. For a novice learner, therefore, a familiar sequence such as je voudrais aller, ich möchte gern or me gusta mucho is not somehow pedagogically inferior because every component has not yet been fully analysed. Quite the opposite: it gives them something they can understand, retrieve and use whilst the underlying grammatical system develops.

This does not mean teaching endless prefabricated phrases which learners cannot manipulate. Chunks and highly familiar sentence patterns provide cognitive scaffolding from which greater generativity can develop. Once learners possess a sufficiently large repertoire of familiar language, we can draw their attention to patterns, substitute elements, contrast forms, explain rules and progressively help them recombine what they know. The choice is not between “chunks” and “grammar”; good language teaching exploits the interaction between the two.

The problem becomes even more obvious when we consider listening. Before meaning can be constructed, learners must segment a continuous acoustic stream, recognise phonemes and syllables, identify word boundaries, activate lexical representations, parse syntactic relationships and retain information in working memory. Research by Field, Cutler, Goh, Vandergrift and others has demonstrated how demanding these lower-level processes are, particularly for novice listeners. This is why I have argued for years that listening should be conceived primarily as modelling and processing practice rather than merely comprehension testing.

Explicit phonics can undoubtedly help, and NCELP was right to give phonics greater prominence. However, phonics is not listening, just as knowing vocabulary is not listening and knowing grammar is not listening. The learner has to integrate all of these things in fractions of a second whilst speech continues to arrive through their ears. Breaking language down into three neat “pillars” may make curriculum progression easier to document, but it risks giving a misleading picture of how communicative competence develops. The components matter enormously; what ultimately matters is whether learners can integrate them rapidly enough to understand and produce language in real time.

I had similar reservations about the prominence given to frequency in vocabulary selection. NCELP was right to challenge the often arbitrary vocabulary choices found in traditional topic-based textbooks, and corpus frequency data can be extremely useful. However, frequency is a useful criterion, not a curriculum. It tells us how often a lexical item occurs in a corpus but it does not tell us how useful that word will be to a thirteen-year-old, how learnable it is, which constructions it combines productively with, whether it facilitates meaningful interaction or whether pupils actually want to say it! A sensible curriculum balances frequency against communicative usefulness, generative potential, learnability, cultural relevance and learner interest.

This brings me to what I regard as the biggest weakness of the whole enterprise: motivation. We are not teaching miniature linguists who have voluntarily enrolled on a university language course; we are teaching children, most of whom have not chosen the language and many of whom already believe that languages are difficult. Research deriving from Self-Determination Theory (Deci & Ryan, 1985; Ryan & Deci, 2000), Bandura’s work on self-efficacy, Dörnyei’s work on L2 motivation and the literature on willingness to communicate all point to the importance of perceived competence, success and agency. Pupils need to experience themselves understanding and using language. They need those moments when something which previously sounded like meaningless noise suddenly makes sense, or when they discover they can sustain a short interaction without consciously constructing every sentence. Those experiences of competence are not fluffy motivational extras; they are central to whether adolescents decide language learning is something they can succeed at.

This was also one of the recurring concerns expressed by teachers about NCELP materials on social media and teacher blogs. The criticism was not simply that teachers disliked phonics, explicit grammar or systematic vocabulary instruction; many of the teachers criticising NCELP used all three. Concerns repeatedly centred on the atomisation of language, the nature of the activities, the relative lack of meaningful communication, the treatment of chunks and the dryness of some materials. Steve Smith, for example, whilst acknowledging the logic behind much of the approach, famously described some materials as “dull as ditchwater” and questioned the extent to which communication had become secondary. Other experienced practitioners raised similar concerns, although others responded positively to NCELP and that should also be acknowledged. The reaction was not unanimous, but neither was the resistance insignificant.

One of the most unhelpful tendencies in education is to assume that university researchers possess “the evidence” whereas teachers merely possess opinions, with resistance to research-informed initiatives consequently interpreted as conservatism. Teachers spend thousands of hours observing what happens when particular activities and sequences encounter actual children. That evidence is contextual and fallible, but so is much educational research when one attempts to generalise from controlled studies to whole-school curriculum design. Research should inform practice, whilst sustained classroom experience should interrogate research. When substantial numbers of experienced teachers independently raise similar concerns, policymakers should at least consider that they may be noticing something important.

This is why what is happening in Northern Ireland deserves careful scrutiny. Rachel Hawkes chaired the Modern Languages Working Group involved in developing the proposed curriculum, and the language surrounding the new direction has a distinctly familiar feel, with considerable emphasis on systematic linguistic construction, carefully specified knowledge and progression towards fluency. None of those things is inherently problematic; indeed, I strongly support systematic curriculum design. My concern is whether Northern Ireland is importing not merely sensible principles from NCELP but the same underlying conception of language learning which shaped the FAILED English initiative.

The context makes this particularly worrying because language education in Northern Ireland is already fragile. Language Trends Northern Ireland 2025 paints a deeply concerning picture of declining participation, with German GCSE entries having more than halved since 2002, a third of responding schools reporting that lower-attaining pupils are discouraged from taking a language and more than a third reporting that senior leaders may advise pupils against choosing one. In such an environment, the last thing language education needs is a curriculum which risks making the subject feel even more technical, analytical and linguistically atomised to the pupils we are desperately trying to retain.

There is something extremely seductive about a curriculum built around precisely specified phonics, vocabulary and grammar. Everything can be sequenced, mapped, inspected and assessed, and on paper the progression looks impeccable. The danger is that we confuse the tidiness of the curriculum with the messier reality of acquisition, because what the teacher has covered and what the learner can retrieve automatically under communicative pressure are emphatically not the same thing!!!

A pupil may have “covered” the perfect tense and still fail to recognise j’ai mangé in rapid speech. Another may know hundreds of high-frequency words yet struggle to understand a short conversation because their phonological representations are weak and they cannot segment the speech stream efficiently. Another may explain adjective agreement perfectly on a worksheet and still be incapable of applying it whilst retrieving vocabulary and planning a message. None of this is surprising if we understand the distinction between declarative knowledge and proceduralised competence, yet curriculum documents have a persistent tendency to equate what has been explicitly taught with what learners have acquired.

Northern Ireland therefore has an opportunity to take what was useful from NCELP without reproducing what, in my view, went badly wrong! – so badly wrong, that despite having invested several millions of pounds in it, the UK government didn’t renew their contract. Teach phonics systematically, select vocabulary intelligently, teach grammar explicitly, sequence learning carefully, use frequency data where they genuinely help and recycle relentlessly. However, put communicative competence at the centre of the architecture. Give learners large amounts of highly comprehensible listening and reading, exploit repeated language patterns and formulaic sequences, establish strong sound-form-meaning mappings, scaffold production heavily before expecting spontaneity, and teach grammar in a way which helps learners analyse and extend language they have already begun to internalise rather than expecting grammatical knowledge to magically transform itself into fluent speech.

Most importantly, engineer success early and often, because the central problem facing school language education is not that twelve-year-olds cannot explain grammatical terminology or manipulate isolated linguistic components accurately enough. It is that far too many pupils conclude, often very early, that languages are difficult, that they are not very good at them and that continuing with them would therefore be irrational. Once that self-perception has taken hold, another carefully sequenced grammar progression is unlikely to rescue them.

NCELP was an ambitious attempt to translate selected strands of second-language research into a coherent national pedagogy, and some of the principles it promoted were sound, some of its research was valuable and some teachers undoubtedly benefited from its resources. My criticism is not personal; it concerns the pedagogical architecture which emerged and, in particular, what I regard as an excessive faith in the systematic teaching and manipulation of linguistic components as the principal route towards communicative competence.

Northern Ireland now has something England did not have when NCELP was launched: hindsight. It can examine the research, look carefully at what happened in classrooms, listen seriously to the teachers who used the materials and ask whether the answer to an existing crisis of motivation and uptake is really to make language learning more systematically componential. If the result is that pupils know more about how the language is constructed but feel less able or less inclined to actually use it, we will have achieved precisely the opposite of what language education urgently needs, and given the precarious state of language uptake in Northern Ireland, that is a risk I simply do not think we can afford to take.