A large share of fluent language is prefabricated: chunks, collocations, sentence frames. The disagreement is about what follows for teaching: notice chunks as they emerge (Thornbury), pre-select and flood them (Conti), or treat them as the whole syllabus (Lewis).

Where the sources disagree

  • Selection: Lewis gives no syllabus. Conti selects by communicative function → pattern. Thornbury combines a frequent-formulae core with emergent items.
  • Fluency mechanism: chunk storage (Pawley & Syder, Thornbury) vs process automatisation (Goodey’s comment, DeKeyser via Conti). These are complementary. Chunks are what gets automatised. (Claude)
  • Risk: “all chunks, no pineapple”: retrievable formulae without generative grammar.
  • Bottom-up grammar: Dellar and Selivan say most “grammar” errors are colligation/priming gaps; Thornbury (Error) and Pienemann via Conti say many are developmental. Both can be true: word-pattern errors (I am agree) look lexical, question-order errors look developmental. (Claude)
  • Chunks → rules: Selivan (citing N. Ellis 1996) says chunks bootstrap grammar; Walter (2012) says chunk teaching avoids grammar. Conti sequences both (chunks first, rules in Expansion).

Theory

Procedures

Research

More notes (by source)

Adaptive Learning in ELT (Philip Kerr)

Key terms (DELTA M1)

formulaic sequence · collocation · colligation · lexical chunk · lexical priming (Hoey) · grammaticalised lexis · lexical syllabus (Willis) · nativelike selection (Pawley & Syder) · premature lexicalisation · lexical coverage · involvement load (Laufer & Hulstijn)

Already in my vault

Open questions

  • How many chunks does an FCE candidate actually need, and which? (Check the English Vocabulary Profile.)
  • Can sentence builders work for IELTS Writing Task 2 without producing template-speak?