Complex sentences in IELTS: what Grammatical Range means and what it never meant

The Grammatical Range and Accuracy descriptors draw a line that candidates keep reading backwards. Band 6: uses a mix of simple and complex sentence forms, with errors that rarely impede communication. Band 7: uses a variety of complex structures, with frequent error-free sentences. Read carelessly, this says "write longer sentences". What it actually says is two separate things — show more than one kind of complexity, and keep a meaningful share of sentences perfectly clean — and the second condition is the one that decides most papers. How the other three criteria draw their own band 7 lines is set out in what band 7 actually requires, criterion by criterion.

The error-rate arithmetic

"Produces frequent error-free sentences" is the operative band 7 phrase. A 270-word essay runs 13–16 sentences; "frequent" in examiner practice means roughly half of them carry no error at all — not small errors, none. Now consider the trade every candidate makes when they inflate sentences: a 35-word sentence with three clauses has three chances to break. Two long broken sentences do more band damage than six short clean ones, because they attack the exact phrase the band 7 judgement rests on.

The practical rule: default to one idea per sentence, and deploy complexity where the logic demands it — a condition, a contrast, a cause. Complexity that mirrors thought reads as range; complexity added for length reads as risk, and prices itself accordingly.

Five structures that cover "variety"

The descriptor says variety of complex structures — plural. These five are learnable in a week each, cover essentially every logical move a Task 2 essay makes, and are exactly what examiners can identify quickly:

1. Concession: Although outright bans rarely work, they do signal official disapproval. The workhorse of argument essays; one per essay, minimum.
2. Real and unreal conditionals: If advertising to children were restricted, pressure on parents would ease. The unreal form (were / would) is a band 7 marker precisely because band 6 essays avoid it.
3. Relative clauses, including reduced ones: Policies aimed at commuters (reduced) alongside policies that target commuters — the reduced form is quieter and rarer, which is why it registers.
4. Noun clauses as subject or object: What matters more than the fine itself is whether drivers believe they will be caught. One of these per essay lifts the range judgement on its own.
5. Purpose and result: so that, such … that, in order to — Fares were cut so that low-income workers could reach the city centre.

An essay using four of these once each shows more range than an essay using although five times — variety is counted across types, not instances.

What this looks like in revision

Diagnose by sentence, not by essay. Take your last three essays and mark every sentence clean / flawed. Most candidates find a pattern within twenty sentences: it is nearly always two or three error types on repeat — articles, subject–verb agreement across a long subject, comma splices, tense drift in conditionals. Fixing your top two error types converts several flawed sentences per essay into clean ones, which moves the "frequent error-free" ratio directly.

Build complexity from clean simplicity, not the reverse. Write the paragraph in short clean sentences first, then join the two pairs whose logic connects — a concession here, a relative clause there. This ordering keeps accuracy anchored while range is added. Candidates who draft long and then "fix errors" reliably miss half of them; errors survive editing much better than they survive not being written.

Rehearse structures, not sentences. Memorised whole sentences get noticed and discounted; memorised frames (Although X, Y, What X fails to consider is Y) are simply grammar you own. Write five fresh examples of each frame across a week — different topics each time — and the frame stops needing attention, which frees it for exam conditions.

Band 7 grammar is a portfolio decision: a floor of short sentences you never miss, four or five structure types deployed where the argument calls for them, and your two personal error types hunted to extinction. None of that requires rare grammar — it requires knowing precisely which sentences of yours break, and refusing to write those. The vocabulary criterion rewards the same discipline, which is why big words do not raise your Lexical Resource band either.