Lewis: The Lexical Approach
Summary: Michael Lewis’s challenge to grammar-first language teaching — arguing that language consists of “grammaticalized lexis” (chunks, collocations, and formulaic sequences) rather than “lexicalized grammar” (structures filled with vocabulary), and that fluency comes from a large mental lexicon of ready-made phrases, not mastery of rules.
Sources: Physical book clipnotes/Library_Quotes_Batch_6.md
Source pages: Library Quotes Analysis: Batch 6 (Books 65-83 Selection)
Last updated: 2026-05-06
Language as chunks
“Language as Chunks: We don’t speak in words, we speak in ‘pre-fabricated’ phrases.” “Language consists of grammaticalized lexis, not lexicalized grammar.” “Fluency is the result of a large ‘mental lexicon’ of ready-made expressions.”
Lewis’s central claim inverts the standard model of language production: rather than building sentences from rules plus words, fluent speakers retrieve pre-packaged multi-word units — “as a matter of fact,” “I was wondering if,” “it goes without saying” — and stitch them together. These chunks are not irregular exceptions to be memorized despite the rules; they are the primary building material. Compare pinker-selected-works on Words and Rules: Pinker’s dual-mechanism theory (rules + stored forms) is compatible with Lewis, but Lewis extends the stored-forms claim much further into what looked like rule-governed territory.
Collocation as core knowledge
“Collocation: The natural company that words keep (e.g., ‘heavy rain’ vs. ‘strong wind’).” “The Learner, who needs collocations more than they need conjugation tables.” “The grammar-lexis dichotomy is a false one; they are two sides of the same coin.”
Collocation is the most practical application of the chunk insight: native speakers do not say “strong rain” because “rain” collocates with “heavy,” not “strong” — but this is not a grammatical rule, it is a lexical fact. Language learners who study grammar but not collocation produce technically correct but unidiomatic output. The knowledge that makes speech sound native is overwhelmingly collocational. Compare crystal-how-language-works on the native speaker’s implicit rule knowledge: collocation is the most invisible component of that competence.
Grammar rules that native speakers don’t think about
“Why do we spend years teaching grammar rules that native speakers never think about?” “Can we learn a language faster by focusing on ‘strings’ rather than ‘rules’?”
The pedagogical challenge Lewis poses to ELT: the communicative efficiency of chunk-based learning vs. rule-based learning. A learner who has internalized “I was wondering if you could possibly…” as a unit can produce polite requests without analyzing the past progressive, the conditional, or the modal. The analysis, Lewis argues, is not how native speakers produce the utterance anyway. Compare language-testing-and-identity: what we test affects what we teach; a lexical approach implies lexically-based assessment.
Observe-Hypothesize-Experiment
“Observe-Hypothesize-Experiment: The new cycle of linguistic discovery.”
Lewis’s proposed learning cycle: encounter chunks in natural data (observe), form a hypothesis about their patterns (hypothesize), test the hypothesis in production (experiment). This replaces the traditional Present-Practice-Produce sequence with a more inductive, corpus-driven approach. Compare langacker-cognitive-grammar: both approaches treat grammatical knowledge as emerging from exposure to patterns rather than from the application of abstract rules.