The Lexical Approach

Language isn't grammar plus a bag of words — it's prefab chunks ("would you mind", "take a chance", "in the long run") that fluent speakers grab whole. The Lexical Approach says: train the chunks, and grammar comes along for the ride.

What it is

The Lexical Approach is a way of understanding language — and teaching it — that puts multi-word chunks at the center instead of single words and grammar rules. It was popularized by Michael Lewis in his 1993 book The Lexical Approach and the follow-up Implementing the Lexical Approach (1997). Lewis's famous one-liner: "Language consists of grammaticalised lexis, not lexicalised grammar." Translation from academic-speak: you don't build sentences by taking grammar rules and plugging words in. You mostly recombine ready-made pieces you've already met, and grammar is the pattern that emerges from those pieces.

Those ready-made pieces have a few names, but the key category is the collocation — words that habitually travel together. You make a decision, you don't do a decision. You take a heavy shower in Spanish-influenced English? No — a heavy rain, a strong coffee, a fast food. Nothing about the grammar tells you which adjective is "allowed"; it's just what natives say. Beyond collocations, chunks include fixed expressions ("by the way", "as a matter of fact"), semi-fixed frames ("I'd like to __", "the thing is __"), and institutionalised utterances ("Could you pass the salt?").

Here's the gym framing: most courses train you like a bodybuilder isolating one muscle at a time — vocab list on Monday, verb conjugations on Tuesday — then act shocked when you can't move fluidly in a real conversation. The Lexical Approach trains compound movements. A chunk like "would you mind if I" is a single rep that delivers vocabulary, grammar, word order, and a social function all at once. You're not memorizing the word "mind" plus the conditional plus polite-request rules. You're grooving one pattern your mouth and ear can fire automatically.

This matters because fluency is largely about retrieval speed, not knowledge. Working memory is tiny — you can't assemble a sentence word-by-word from rules in real time without sounding like a stalling robot. Chunks bypass the assembly line: you pull a whole pre-built phrase off the shelf. That's why a native uses "I have no idea" instantly while a textbook learner is still parsing "I + have + negation + indefinite article + noun."

The evidence

The chunk-first view isn't a fringe opinion — it's mainstream corpus linguistics and second-language acquisition research.

Corpus studies (analyses of huge databases of real language) show that a startling proportion of natural speech and writing is formulaic — built from recurring multi-word units rather than freshly assembled. Researchers like Alison Wray (Formulaic Language and the Lexicon, 2002) argue that formulaic sequences are stored and retrieved as single wholes, which is exactly why fluent speech is fast and idiomatic. Estimates of how much language is formulaic vary widely by method, but the consistent finding is: a lot of it, far more than the "rules + words" model assumes.

John Sinclair, whose COBUILD corpus project reshaped dictionary-making, proposed the "idiom principle": speakers operate mostly by selecting ready-made phrases, only falling back on the slower "open-choice" (word-by-word) principle when needed. This is the linguistic backbone under Lewis's approach.

On the acquisition side, Paul Nation — the leading authority on vocabulary learning — stresses that knowing a word means knowing its collocations and patterns of use, not just its dictionary gloss. His work on vocabulary size and the value of meeting words in rich context dovetails with the lexical view: words live in company.

Crucially, this all fits the input-first worldview. Stephen Krashen's Comprehensible Input explains where chunks come from: you don't consciously memorize most of them, you absorb them from large amounts of understandable language. You meet "the thing is" a hundred times across podcasts and shows, and one day it's just yours. Lewis himself leaned on this — he argued that learners acquire chunks largely by noticing them in input, not by drilling rules.

The honest caveat: Lewis's teaching recommendations were sometimes light on hard experimental proof, and "just notice the chunks" can be vague. But the underlying descriptive claim — that language is heavily chunked and that collocational knowledge is central to fluency — is one of the better-supported ideas in the field. No "fluent in 30 days" promises here: building a deep stock of chunks is the work of months and years of reps. But it's the right work.

How to actually use it

You don't "do" the Lexical Approach as a separate method — you bolt a chunk-hunting habit onto an input-first routine. Here's the training plan.

1. Get your reps in input first. Chunks come from volume. Do your daily listening and reading (see Building Your Daily Routine and Finding Comprehensible Input). You can't mine ore from a mountain you never visit.

1. Notice, then capture phrases — never lone words. When something jumps out, save the whole chunk, not the bare word. Not "decision" — save "make a difficult decision". Not "mind" — save "would you mind if". This is the single biggest habit shift. Your flashcards and notes should look like phrases, because that's how the language actually ships.

1. Mine sentences, not vocabulary lists. This is literally Sentence Mining: pull a full sentence you understood from your input, drop it into your SRS, and review it as a unit. You're memorizing the neighborhood, not the lone house.

1. Feed chunks into spaced repetition. Put your collocations and frames into Anki or another SRS. Test yourself on recognizing and producing the whole chunk, ideally with audio. Fighting the Forgetting Curve with phrases beats fighting it with isolated words, because each rep teaches grammar and usage for free.

1. Train the patterns out loud. Once a chunk is familiar, shadow it (see Shadowing) so it lives in your mouth, not just your eyes. Glossika-style sentence reps do the same job — see The Glossika Method. The goal is automaticity: the phrase fires before you can think.

1. Don't cram grammar rules — let them crystallize. When you've internalized "I'd like to __", "I'd love to _", "I'd rather __", you've acquired the conditional without a single rule sheet. Grammar becomes the pattern you feel across your chunks. That's Grammar: Acquiring Intuition in action.

1. Use a collocation tool when you're unsure. Writing something and not sure if a pairing is natural? Check a corpus or collocation dictionary instead of guessing from grammar.

Languy's rule of thumb: if you'd be tempted to memorize a word alone, you're leaving gains on the table. Grab the chunk it came with.

Resources

• The Lexical Approach and Implementing the Lexical Approach by Michael Lewis — the foundational books, written for teachers but readable.
• Oxford Collocations Dictionary for Students of English — the classic reference for "which words go together." Many languages have equivalents; search "[your language] collocations dictionary."
• Online corpora — search "COCA corpus" (Corpus of Contemporary American English) or "Sketch Engine" to see real chunks in context across millions of examples.
• Anki↗ — for storing mined chunks; see our Anki: The Complete Guide.
• Language Reactor↗ — captures phrases straight from Netflix/YouTube subtitles; see Language Reactor & Immersion Tools.
• LingQ — built around saving and reviewing words in context rather than in isolation; see LingQ.
• Glossika — drills full sentences as chunked reps; see The Glossika Method.
• For English specifically: search "Michael McCarthy Felicity O'Dell English Collocations in Use" — excellent graded chunk practice.

Related

• Sentence Mining
• Vocabulary Acquisition
• Grammar: Acquiring Intuition
• Comprehensible Input (Krashen)
• Spaced Repetition (SRS)
• The Glossika Method
• Grammar-Translation (and Why It Fails)
• The Forgetting Curve & Memory