What native speakers actually say — and whether it scores

What a corpus is, and why it changes the picture

A corpus is a very large collection of real English — conversations, lectures, newspapers, academic articles — that can be searched and counted. Instead of asking a grammarian what is correct, you ask several million words of actual usage what is frequent.

The results are often uncomfortable for textbooks. Structures given a whole chapter turn out to be vanishingly rare in speech; structures that are barely mentioned turn out to be everywhere.

Some things the counting shows

Speech and writing are not the same grammar

This is the central finding, and everything else follows from it. Conversation and academic prose differ so systematically that they are almost different systems.

Learners are typically taught one grammar and then use it in both places. That is where the marks go — a candidate who writes an essay in conversational grammar sounds informal, and a candidate who speaks in academic grammar sounds memorised. Both are penalised, and both are the same mistake.

The passive is much rarer than learners think — except where it is essential

Corpus counts show the passive is uncommon in conversation and concentrated in academic and scientific prose, where the agent is unknown or irrelevant. So the useful rule is not use the passive or avoid the passive — it is use it where there is nobody to name, which is exactly the IELTS Task 1 process diagram.

Progressive tenses are rarer than the textbook implies

Conrad and Biber's count is blunt: "In general, we use simple verbs much more than progressive verbs." Learners drilled heavily on the continuous tend to overuse it, and the simple tenses are quietly doing most of the work in real speech. But the rule you were given for when to avoid it turns out to be the wrong rule entirely — see below, because this is the most useful thing in the book.

Very few words do most of the work

A small set of extremely common verbs — get, go, make, take, have, do — accounts for a huge share of everything anyone says, mostly through phrasal and collocational combinations. Which means fluency lives in combinations, not in rare words — and that is precisely what the Lexical Resource criterion means when it asks for collocation and natural usage.

The stative-verb rule you were taught is not the rule English follows

This is the single most valuable page in Real Grammar, and it is worth reading slowly, because almost every learner has been taught the opposite.

The textbook rule: non-action (stative) verbs are not used in the progressive. Do you want some more cake? — never are you wanting. Clean, memorable, and it is not what the corpus finds.

Look at the first list. Hoping, wondering, feeling — these are exactly the verbs a textbook calls stative and forbids. And I was just wondering if you could help is not an error. It is one of the most natural sentences in spoken English.

So what is the real rule?

It is not about action versus non-action at all. The corpus gives two conditions, and a verb takes the progressive when it meets both:

1. The subject actively controls the action or state.
2. The action or state lasts over an extended period of time.

That is why the two lists cut across the action/non-action distinction so untidily. You control wondering, hoping and listening, and each one takes time — so they are progressive constantly. You do not control knowing, wanting or hearing; those simply happen to you, so they stay simple. And shut or swallow is over in an instant, so it stays simple too, even though you control it completely.

We had the textbook version of this rule on this page ourselves until we went back to the corpus. That is the point of the book, and Conrad and Biber say so in their own introduction: "Corpus research has shown that we often are not consciously aware of the most common language choices we make." The rules we remember are not always the rules we follow.

Register: the same sentence, right and wrong

Appropriacy is inside the criteria — every major rubric asks whether the language suits the task. Corpus work makes it concrete, because it shows the two registers really are different. It is also the authors' whole reason for writing:

Appropriateness, not just correctness is the whole distinction a band descriptor is built on, and it is the one thing a grammar checker cannot give you. A sentence can be flawless and still be the wrong sentence for the room.

Both sentences are perfectly good English. Each is wrong in the other room — and notice that the wrongness is not a grammar error, which is why candidates cannot see it in their own work.

What real spoken grammar actually contains

Corpus recordings of fluent conversation are full of things learners were taught to avoid — and they are not errors. They are the grammar of speech.

• Ellipsis. Been there? Want one? Full sentences are not required in conversation, and insisting on them sounds formal.
• Vague language. Sort of, and things like that, or whatever. Extremely frequent, and a legitimate part of natural speech.
• Discourse markers. Well, anyway, I mean, you know. They organise speech and every fluent speaker uses them constantly.
• Repetition and self-correction. Real speech restarts, rephrases and repairs. Doing it smoothly is a sign of control, not of weakness.
• Short clauses joined by and and but. Speech does not subordinate as densely as writing does — and trying to speak in written subordination is what makes candidates sound rehearsed.

How to use this

1. Learn the two registers as two registers. When you learn a phrase, learn where it lives. Practising written phrases for a speaking test is the most common way to sound memorised.
2. Trust natural usage over remembered rules. If a construction feels stilted, it probably is, and the rubric agrees — appropriacy is scored.
3. Learn combinations, not words. The high-frequency verbs and the collocations they live in do most of the work in real English, and Lexical Resource rewards exactly that.
4. Use the passive where there is no actor, and rarely otherwise.
5. Listen to unscripted English — real conversation, not textbook dialogues. That is the corpus, and it is free.

For the test, and for real English

For the test: match the register. Speak like a person, write like a writer, and let the descriptors reward you for the difference — appropriacy is a criterion in every major rubric, and this is the cheapest way to satisfy it.

For real English: the corpus is simply describing the language you are trying to join. Everything in it — the vagueness, the discourse markers, the small set of verbs doing enormous work — is what you will actually need in a meeting, a seminar or a conversation, long after the exam.

That is the happy part of this particular book. It is one of the few places where preparing properly for the test and learning the real language are not in tension at all — because the examiner is asking you to sound like someone who speaks English, and so is everybody else.