IELTS speaking band descriptors: 4 criteria, band by band

A candidate walks out of the speaking test sure it went well. The conversation flowed, the examiner nodded, the ideas were interesting. The result comes back: 6.5. Same as last time.

The reason is almost never the ideas. IELTS speaking is scored on four things, each worth exactly a quarter of your mark, and not one of them is what you said. The examiner is grading how you said it. Most speaking advice skips this entirely and tells you to "speak more confidently," which is like telling someone stuck at 6.5 to "just be better."

This article does the opposite. Here are the four band descriptors examiners use, what separates band 6 from band 7 in each, and how to find the one that's quietly capping you.

The four criteria, and why your ideas aren't one of them

Your speaking band is built from four equally weighted criteria. Each is worth 25%:

• Fluency and Coherence (FC): can you keep talking and stay on track?
• Lexical Resource (LR): the range and fit of your vocabulary.
• Grammatical Range and Accuracy (GRA): your sentence structures and how often they're correct.
• Pronunciation: how easy you are to understand.

Notice what's absent. There's no criterion for intelligence, originality, or having the "right" opinion. You can argue that ice cream cures sadness and still hit band 8, as long as you say it fluently, with precise words and clean grammar, clearly enough to follow. Content is the vehicle, not the score.

Two of these names should look familiar. If you've studied Writing, you've already met Lexical Resource and Grammatical Range and Accuracy — they're the same two criteria, judged a little differently when you're speaking instead of editing on paper. Fluency and Coherence is the spoken cousin of Writing's Coherence and Cohesion: both care about flow and linking, but one listens for hesitation while the other reads your paragraphs. Pronunciation is the only criterion with no writing equivalent, and Task Achievement, the writing criterion for answering the question, has no speaking equivalent at all.

How four scores become one band

The examiner gives you a band for each of the four, then averages them and rounds to the nearest half band. That's it. No criterion is weighted more than another.

Each criterion is given a whole band, since the official notes require a candidate to "fully fit the positive features of the descriptor at a particular level". So say you score 7 for fluency, 6 for vocabulary, 7 for grammar and 7 for pronunciation. The average is 6.75, which rounds up to band 7.0. The weak criterion cost you nothing, because the other three carried it.

Now flip one axis. Strong fluency, strong vocabulary, strong grammar, but pronunciation that's hard to follow:

7 + 7 + 7 + 5 = 26, averaged to 6.5

One weak criterion just erased an otherwise band-7 performance. This is the maths that traps people at 6.5: they pour months into the three things they're already decent at and never touch the fourth. The same averaging runs your overall score across all four skills, one level up. The lesson repeats: your lowest axis sets your ceiling.

So the useful question isn't "how do I get better at speaking." It's "which of my four is lowest, and what does the next band up look like there." Let's take them one at a time.

Fluency and coherence: keep going, stay on track

Fluency is whether you can keep speaking without grinding to a halt. Coherence is whether your words connect into something a listener can follow. The examiner is listening for hesitation, repetition, self-correction, and the linking words you use to signal where you're going.

The public Band Descriptors draw a clear line here. At band 6 you can keep going and produce long turns, but coherence slips at times: repetition, a self-correction, a hesitation while you hunt for the next word. Your connectives and discourse markers get used, just not always in the right place. At band 7 you keep those long turns going without noticeable effort, and you handle that same set of linking words with more flexibility.

Hear the difference on a typical part 1 question:

Band 6: "I like my city because... because it has many things. And the food is good. And, yeah, there are parks and... people are friendly."

Band 7: "What I love about my city is how much there is to do. The food scene is great, and on top of that there's green space everywhere, so weekends never get boring."

The second answer isn't smarter. It moves. "What I love about" opens it, "on top of that" adds, "so" draws a consequence, and nothing stalls. The band-6 version leans on "and" as a crutch and stops to restart twice. If your fillers are all and, um, because, that's the repetition the descriptor flags.

Lexical resource: the same criterion, spoken out loud

Vocabulary in speaking rewards the same precision and collocation it rewards in writing, with one twist. Spoken LR also wants natural paraphrase and idiomatic language. At band 6 your vocabulary is wide enough to discuss topics at length and keep your meaning clear, even when some word choices don't quite fit, and your paraphrasing generally gets there. Band 7 brings in less common and idiomatic vocabulary with a feel for style and collocation, and the paraphrase lands cleanly rather than approximately.

Band 6: "Traffic is a very big problem and the government must make it better."

Band 7: "Traffic congestion is a real headache, and honestly the city needs to get to grips with it before it gets worse."

"A real headache" and "get to grips with" are idiomatic, and "congestion" is the precise word a band-7 speaker reaches for without thinking. Here's the twist worth remembering: idioms that would hurt you in an academic essay help you in speaking. The register is different. The collocation and precision habits, though, transfer straight across, so the collocation and word-form drills in the writing LR guide work for both.

What doesn't transfer is the memorised showpiece. "Every coin has two sides," bolted onto a part-3 answer, reads as rehearsed, and examiners hear it forty times a day.

Grammatical range and accuracy: clean sentences, with no eraser

GRA looks at two things: the range of structures you attempt, and how often they come out correct. Writing gives you time to fix a clause. Speaking doesn't, which is why this axis trips up people whose grammar looks fine on paper.

At band 6 you use a mix of short, simple sentences and complex ones, but without much flexibility, and the complex structures bring frequent errors, even if they rarely block understanding. Band 7 uses a wider range of structures more flexibly, and produces error-free sentences frequently. That last phrase, error-free sentences, is the target the descriptor actually names. Not perfect grammar. Stretches where nothing breaks.

Band 6: "When I was child, I live in small town and it was very quiet."

Band 7: "When I was a child, we lived in a small town, which was so quiet that not much ever happened."

Same memory, but the band-7 version runs a relative clause into a result clause and keeps the articles and tense intact. The band-6 version drops "a" twice and slips into the present. If you can produce three or four clean complex sentences in a row, under pressure, you're showing the examiner band 7. The drills that build this are the same ones from the writing GRA guide, just rehearsed aloud.

Pronunciation: the criterion with no band-7 checklist

Pronunciation has a quirk worth knowing, because it changes how you climb. The public descriptor gives a detailed feature list only for the even bands: 2, 4, 6, 8. The odd bands aren't left out, but they don't get a list of their own. Band 7 is defined by reference instead: all the positive features of band 6, plus some, but not all, of what band 8 asks for.

In plain terms: there's no band-7 checklist to tick off. You reach 7 by doing everything band 6 asks and starting to do what band 8 asks. So the bands you study are 6 and 8.

Band 6 uses a range of pronunciation features, but with mixed control, so you stay generally understandable while the odd mispronounced word or sound still costs you clarity. Band 8 sustains a wide range of those features and is easy to understand throughout, with a first-language accent that has minimal effect on intelligibility. Read that last line twice. Accent is not the problem. A strong Japanese, Hindi, or Vietnamese accent can score band 8, because the test is intelligibility, not sounding British.

The features under assessment are word stress, sentence stress, intonation, rhythm, and individual sounds. Word stress alone can change meaning: a REcord (noun) and to reCORD (verb) are the same letters with the stress in different places. Put it on the wrong syllable often enough and the listener works harder, and "works harder to understand you" is exactly what drops you from 8 toward 6. The dedicated guide to IELTS pronunciation takes each feature band by band, with a five-minute drill for the one costing you most.

Find your weakest axis before you drill anything

You now have four band-6-versus-7 lines. The move is not to grind all four. It's to find the one that's lowest and aim there, because that's the axis setting your ceiling.

Do this once. Record yourself answering a part-2 cue card for two minutes, then play it back four times, listening for one criterion each pass. Did you stall and restart (FC)? Reach for vague words like good, big, thing (LR)? Stay in the present tense the whole way (GRA)? Lose the listener on certain words (Pronunciation)? The axis that's clearly weakest is your missing 0.5.

That diagnose-then-target approach is the whole idea behind IELTS Drills: feedback is tagged to a single Band Descriptor, so when your weak axis is grammar, you drill grammar for five minutes instead of doing another full mock and hoping. The four-axis self-diagnosis here was written for writing, but the logic is identical for speaking — name the axis, then drill it alone.

Frequently asked questions

Are the speaking and writing band descriptors the same? Two of the four overlap. Lexical Resource and Grammatical Range and Accuracy appear in both, judged a bit differently aloud. Fluency and Coherence is speaking's version of writing's Coherence and Cohesion. Pronunciation is speaking-only, and Task Achievement is writing-only.

Does my accent lower my pronunciation score? No. The band 8 descriptor explicitly allows a first-language accent that has minimal effect on intelligibility. You're scored on how easy you are to understand, not on how native you sound.

Is grammar or vocabulary more important for speaking? Neither. All four criteria carry an equal 25%. Improving your lowest one moves your band more than polishing a criterion you're already strong in.

Do examiners score my opinions or ideas? No. Content isn't assessed. A simple, honest answer delivered fluently and accurately outscores a clever one delivered in broken sentences.

How is the speaking band rounded? It's the average of the four criteria, rounded to the nearest half band. Each criterion gets a whole band, so scores of 7, 6, 7, 7 average to 6.75 and round to 7.0. A single 5 among three 7s averages to 6.5.

Can one weak criterion cap my whole speaking score? Yes. Because the four are averaged, your lowest axis pulls everything toward it. Three 7s and a 5 is a 6.5, not a 7.

The bottom line

IELTS speaking isn't graded on what you think. It's graded on four criteria, weighted equally, and your lowest one decides your band. Fluency and coherence wants flow without stalling. Lexical resource wants precise, idiomatic words that fit. Grammar wants stretches of error-free complex sentences. Pronunciation wants you easy to follow, accent and all.

Learn the band-6-to-7 line in each, find the one that's lowest, and spend your five minutes a day there. That's the difference between practising harder and practising the 0.5 you're actually missing.