Fluency (real-time production without undue pausing) depends less on prioritising meaning over form than on a large store of memorised, quickly retrievable chunks: Palmer’s ‘memorize perfectly the largest number of common and useful word-groups’.
Claim
Thornbury disputes Ellis & Barkhuizen (2005, following Skehan 1998) that fluency arises when learners prioritise meaning over form. He argues it rests on holistically stored sequences (Pawley & Syder 1983: nativelike selection and nativelike fluency). Brumfit (1984) used “fluency” methodologically; cognitive SLA defines it temporally.
Six steps (conditional on the claim being right)
- Clarify what a “word group” is.
- Select common and useful ones.
- Set a target for the largest practicable number.
- Decide the criteria for “perfect” memorisation.
- Teach memorisation strategies.
- Design activation tasks that replicate real use without undue pausing.
Measures
Pause frequency, mean length of run, pause placement (at clause boundaries vs mid-clause), L1 pausing transfer (Chambers 1997).
Limits & critique
- Goodey (comment): fluency = automatisation of processes, not recall; chunk recall can itself slow speech.
- Sylvester (comment): memorised dialogues don’t transfer to new interlocutors; personal relevance aids retention.
- Thornbury concedes an “ear” may be aptitude.
Apply in my context (Claude)
- IELTS Fluency & Coherence: examiners respond to pause placement more than raw speed. Train mid-clause pauses out by rehearsing chunks to automaticity, e.g. 4/3/2 (Maurice 1983; popularised by Nation) on Part 2.
- Combine with Conti’s routinisation stage: MARS-EARS - the EPI instructional sequence is in effect an implementation of steps 2–6.