IELTS Writing Band Score 7 Samples: The Criterion That Caps All Eleven of Ours

Every IELTS site publishes a band 7 sample answer. One answer tells you almost nothing, because you cannot tell which of its features put it at 7 and which are incidental.

Our sample library currently holds 25 published answers, and eleven of them came back estimated at band 7. Looking at eleven together says something one cannot. All eleven share exactly the same profile on three of the four criteria, and the criterion holding them at 7 is the same one every time. These are estimates from a model, not marks from a qualified examiner, so read this as a pattern in one scoring engine's output rather than as an examiner's verdict.

How four criteria become one band

Writing is marked on four criteria of equal weight: Task Response (Task Achievement in Task 1), Coherence and Cohesion, Lexical Resource and Grammatical Range and Accuracy. The four are averaged and the result is rounded to the nearest half band. The band score calculation post sets out the rounding rules, which matter more than candidates expect.

The arithmetic has a consequence worth stating before the data. An overall 7 can be reached from several different criterion sets, so two answers both labelled band 7 can have very different strengths. That is why a single sample is a poor teacher.

The eleven answers, side by side

Here is every answer in the library that was estimated at overall band 7, with the four criterion estimates behind it.

Three columns do not vary at all. Lexical Resource is 6 in every one of the eleven. Grammatical Range and Accuracy is 7 in every one. Task Response is 8 in every one. The only criterion that moves is Coherence and Cohesion, at 7 on six answers and 6 on five.

You can read the full answers, with the sentences each estimate was based on, in the band 7 samples library.

The criterion that caps them: Lexical Resource

Eleven out of eleven at Lexical Resource 6 is the finding. If vocabulary moved to 7 on any of these answers, and nothing else changed, the average would cross into 7.5 territory on several of them.

So the obvious question is what the engine actually objected to. Across the eleven answers it flagged 42 separate vocabulary problems. Sorted by how many of the eleven answers contained at least one of each kind:

Every single one of the eleven carries at least one word formation, collocation or spelling error. Range and repetition appear in eight, so narrow vocabulary is a real part of the picture, but it is not the universal part. Accuracy is.

That inverts the usual advice. The common prescription for a 6 in Lexical Resource is to learn more advanced words. The notes on these eleven answers point somewhere much duller: use the words you already have in the right form.

The actual errors

These are real sentences from the published answers, with what was flagged and the fix.

Read that list again and notice what is absent. Not one of them is a missing rare word. Every one is a word the writer already knew, used in the wrong form or with the wrong partner. The public descriptor at band 7 asks for "a sufficient range of vocabulary to allow some flexibility and precision" and permits "occasional errors in word choice, spelling and/or word formation". At band 6 the wording is "attempts to use less common vocabulary but with some inaccuracy". The distance between those two bullets is accuracy, not size.

What the band 8 answers do differently

The library holds two answers estimated at overall band 8. Their profiles are worth putting next to the eleven above.

Coherence and Cohesion is 7 in both, the same as the better half of the band 7 group. Grammatical Range and Accuracy is 7 in both, identical to all eleven. The criterion that actually moved is Lexical Resource, from 6 to 8. If you want that comparison worked through on sentences rather than on a table, the gap between 7 and 8 post does it line by line.

One number in this data is not about the essays

Task Response came back 8 on all eleven answers, and 9 on both band 8 answers. No ordinary answer in the library has ever had Task Response estimated below 8.

That is not credible as a description of eleven different essays by eleven different standards. It is a known weakness in this scoring engine: three criteria discriminate across a useful range and Task Response is compressed into the top two bands. We publish it rather than hiding it, because a page that quotes its own numbers has to say which of them to trust. The engine has never been compared against a qualified human examiner, so every band on this page is an estimate.

The practical reading is this. Treat the Lexical Resource, Coherence and Cohesion and Grammar estimates as a signal worth acting on, since they vary across answers in ways the evidence notes explain. Treat a high Task Response as unproven. For what Task Response actually asks for, band 7 in IELTS writing sets out the criterion by criterion difference from 6.5.

How to use a band 7 sample properly

Reading a sample answer and admiring it transfers nothing. Three things do.

1. Read the criterion estimates before the essay. Knowing the answer is strong on task coverage and weak on vocabulary tells you what to look for while you read.
2. Find the flagged sentences in the answer. Our Task 2 sample answer with the band explained walks one answer this way, criterion by criterion.
3. Write the fix yourself. Copy out three of the error sentences from the table above and correct them without looking. If you can repair someone else's word formation errors, you are closer to catching your own.

Do this today

Take your most recent Task 2 answer and proofread it for one thing only: word form and partner words. Nothing else.

1. Underline every preposition that follows a verb or an adjective. Check each pair, since depend on, interested in and result in are fixed and not guessable.
2. Circle every noun you have made plural. Research, information, advice, equipment and knowledge take no plural in English.
3. Check every word longer than eight letters for spelling.

On the eleven answers above, that pass alone addresses the problems flagged in all of them. When you have made the corrections, put the before and after versions through Bandly and compare the four estimates. They are model estimates rather than examiner marks, so what matters is whether the vocabulary number moves when you fix nothing but form.