Plan a curriculum from communicative functions, then the high-frequency patterns that realise them, then contexts. Teach grammar through lexis; explain rules only after the chunks have been processed.
Claim
Curriculum design order: communicative functions (his “Majestic 12” by end of KS3) → high-frequency patterns and chunks → contexts. “Grammar and Lexis are too closely intertwined to be considered as distinct from one another.”
Mechanism / theory
- Structural Deficit Hypothesis (Bentin, Deutsch & Liberman 1990; Scarborough 1991): syntactic-processing deficits impair comprehension.
- Communicative functions (Finocchiaro & Brumfit 1983); Nunan (2015) on tasks.
- Memory is context-dependent, so recycle across contexts for transfer.
Classroom procedure
- Choose functions → key patterns/chunks.
- Model with sentence builders (substitution tables read aloud; translation on mini-whiteboards).
- Pop-up grammar.
- “Death by chunks” receptive work: read-aloud games, narrow reading/listening, sentence puzzles, dictation at ~90–95% comprehensible input.
- Structured production: translation games, info-gap, creative tasks.
- Inductive grammar → low-stakes checks → fluency/spontaneity.
Limits & critique
- Cuts content from ~6 to ~3 units a year. Defensible only if recycling really delivers.
- A teacher with 50 min/week couldn’t make it fit (comment; unanswered).
- Rejects the “decontextualised chunks” charge, but sentence builders are decontextualised at the modelling stage. Context comes later. (Claude)
- Transfer caveat: written for UK secondary MFL (French/Spanish/German) with ~2–3 lessons a week and minimal out-of-class exposure. Tashkent EFL learners get far more ambient English (YouTube, games), so input-scarcity arguments apply less strongly. (Claude)
Apply in my context (Claude)
- A sentence builder is a substitution table, which is old (Palmer, Hornby). What’s new is the volume of receptive processing around it.
- For KET Writing Part 2 and PET emails: one sentence builder per function (inviting, apologising, suggesting) beats a model text. Learners need to recombine, not copy.