What Are Chunks, and Why Native Speakers Don't Translate

1. What a chunk is: bigger than a word, smaller than a sentence

Open almost any English coursebook and you'll find two units of learning: a vocabulary list and a set of grammar rules. Behind that pairing sits an assumption nobody states out loud — that speaking is computation. Think of a meaning, select the words, apply the rules, produce the sentence.

Linguists have been pointing at the same problem with that picture since the 1980s: when native speakers actually talk, the units they work with are much larger than words. A great deal of the time they aren't computing at all. They're retrieving.

A chunk is one of those retrievable strings — stored whole in the mental lexicon, pulled out without being rebuilt word by word. It might be three words or seven; it might be a complete sentence or a frame with a gap in it. Length isn't the test. Being remembered as one thing is the test.

Take Could you give me a hand with this?. As words, that's eight units, each of which has to be chosen, ordered and inflected correctly — miss one and the sentence stalls. What a native speaker stores is three pieces: Could you, give me a hand, with this. The middle one matters most: give me a hand has nothing to do with hands, so there is no way to derive it. It can only be learned whole.

2. How much of native speech isn't built on the spot?

This isn't an impression — there are corpus numbers. Erman and Warren (2000) estimated that prefabricated expressions account for around 58.6% of spoken English and 52.3% of written English. More than half of what a native speaker says was not assembled in the moment.

The first systematic argument for this came from Pawley and Syder (1983), who posed two puzzles that are still quoted today. Why do native speakers reliably pick the one natural phrasing out of the many that would be perfectly grammatical (nativelike selection)? And how do they produce long stretches of connected speech with so few pauses (nativelike fluency)?

The corpus linguist John Sinclair (1991) turned this into two coexisting principles: the idiom principle and the open-choice principle. Open choice is the model we all know — combine words freely according to the rules. The idiom principle is retrieval of semi-fixed phrases already stored. Sinclair's argument was that in real language use the idiom principle is the default, and open choice is the fallback for when no ready-made phrase will do.

3. Why more vocabulary doesn't close the gap

A lot of learners share one experience: solid reading scores, a few thousand words banked, and still a stall the moment they have to speak. That isn't a discipline problem. It's a structural bottleneck caused by learning at the wrong unit size.

Working at word level means running the whole pipeline for every sentence: reach for the meaning in your first language, translate the words, apply the grammar, speak. All of that has to finish inside the two or three seconds a conversation gives you, and any step going wrong sends you back to the start. Working at chunk level lets you skip most of the middle.

There's processing evidence too. Conklin and Schmitt (2008) compared formulaic sequences against matched non-formulaic strings and found that both native and non-native readers processed the formulaic ones significantly faster. The advantage isn't only that chunks come to mind — they're cheaper to handle once they do.

On the teaching side, Boers et al. (2006) ran one group through a course with an explicit focus on formulaic sequences and another through conventional instruction; blind raters judged the formulaic group noticeably more fluent in speech. Worth noting: that study had 32 participants, which is small enough that it shouldn't be treated as settled. But its direction agrees with the corpus and processing evidence above.

4. What chunks look like: four common types

Chunks aren't one single thing. Nattinger and DeCarrico (1992) and Wray (2002) proposed various taxonomies; in practice four categories cover most of what a learner needs:

Sentence frames give the fastest return for speaking. Learn I was wondering if you could ... once and you have the opening for dozens of different requests — you've acquired a whole family of sentences in a single item.

5. How to actually practise chunks

The method is usually traced to Michael Lewis's Lexical Approach (1993) and its most-quoted line: language is "grammaticalised lexis, not lexicalised grammar". In daily practice it comes down to four steps:

1. 1Choose chunks by situation, not from a word list. Start from the settings you'll actually be in — ordering food, an interview, small talk with colleagues. Chunks from one situation reinforce each other, which beats memorising scattered items.
2. 2Say the whole thing out loud; don't break it up. Half a chunk's value is its rhythm. give me a hand is one rhythmic unit; pronounced as four separate words it loses the very cue that makes it retrievable as a whole.
3. 3Use shadowing to get the intonation. Hear a line, say it back immediately. What you're copying isn't only the sounds — it's the stress placement and the linking. This is the step that moves a chunk from *recognised* to *sayable*.
4. 4Review on a spacing schedule. Chunk memory decays like any other. Coming back after a day and again after three beats one thirty-minute session.

Steps two through four are what ChunksEnglish is built to do: native audio on every chunk, shadowing with recording playback, and spaced review. The method doesn't depend on any app, though — a notebook and YouTube will get you there, just with more friction.

6. Three common misconceptions

"Chunks are just sentence patterns"

Traditional pattern drills are a structure plus a list of substitutable words, and the point is the grammatical structure. Chunks are about habitual co-occurrence — which words native speakers actually put together. heavy rain is grammatically unremarkable, but it's a chunk, because nobody says strong rain.

"Learning chunks will make me sound robotic"

The opposite. Because the fixed part is automatic, your attention is free for the small part that carries what you actually mean. Cognitive resources are finite; the effort you don't spend assembling sentences is effort you can spend on content.

"This is only for beginners"

The chunk gap is more visible at higher levels, not less. Beginners are limited by vocabulary. Intermediate and advanced learners are usually the ones who can say it but still don't sound native — which is exactly Pawley and Syder's nativelike selection problem, and chunks are the only thing that fixes it.

7. Where to start

The most effective opening move is to pick one situation you'll genuinely be in this month and take its chunks as a set, rather than collecting them at random. Each list below gives every chunk with its meaning and three examples in context:

• Restaurant English: 31 chunks — menu to check
• Hotel English: 20 chunks — booking, check-in, problems, check-out
• Complimenting in English: 40 chunks — work, appearance, ideas, effort
• Dating English: 40 chunks — first hello to saying how you feel

Take one situation, drill it until the lines arrive without thinking, then move on. That finishes far better than opening four situations and learning ten items in each.