IELTS Speaking band 7 pronunciation: what the examiner is actually listening for

Pronunciation is one of the four equally weighted criteria in IELTS Speaking, and it is the one candidates prepare for least, usually because the advice available is either "practise more" or "get a native accent". Neither is what the rubric says.

This guide is about band 7 specifically: what separates it from band 6, what separates it from band 8, and how to check where you currently sit using your own recordings rather than a guess.

How the descriptor is actually structured

The public band descriptors for pronunciation do not give band 7 its own list of features. Bands 6 and 8 are described positively, and band 7 is defined as having all the positive features of band 6 plus some, but not all, of the positive features of band 8. Band 5 works the same way between bands 4 and 6.

That structure is the single most useful thing to know, because it turns a vague target into arithmetic. Paraphrasing the two anchor bands:

Band 6 — a range of pronunciation features is used, but with mixed control; some effective use, not sustained; generally understandable throughout, though mispronounced words or sounds reduce clarity at times.

Band 8 — a wide range of pronunciation features; flexible use sustained throughout with only occasional lapses; easy to understand throughout; first-language accent has minimal effect on intelligibility.

Read the official wording on the IELTS website before your test — it is published and free — but the shape is what matters here. Band 7 means: you clearly do everything at band 6, and you have started to sustain some of band 8. Not all of it. Some of it, visibly.

Two consequences fall straight out of this.

Accent is not the target. Band 8, the higher band, is the one that mentions accent, and only to say its effect on intelligibility should be minimal. Nothing anywhere in the scale asks you to sound like a native speaker. Intelligibility is the axis; accent is only relevant where it damages intelligibility. That is the same principle behind accent-tolerant scoring.

Consistency is worth more than peak quality. "Sustained" and "not sustained" is the pivot between the bands. A candidate who produces beautiful intonation in a rehearsed Part 2 answer and goes flat in Part 3 is describing band 6, not band 7. Which means your practice should optimise for holding up across fourteen minutes, not for a perfect thirty seconds.

The features, decoded

"Pronunciation features" is examiner shorthand. The standard teaching breakdown — not a verbatim rubric list, but the one examiners and trainers work from — has five layers.

1. Individual sounds

The meaning-changing contrasts: /ɪ/ vs /iː/ (ship/sheep), /θ/ vs /s/ (think/sink), /v/ vs /w/ (vet/wet), final consonants that carry grammar (walk/walked).

Mixed control looks like: the contrast is there sometimes, usually on words you have practised, and collapses on new vocabulary or under pressure.

Effective use looks like: the contrast holds on words you have never rehearsed.

2. Word stress

Which syllable in a multi-syllable word gets the beat: pho-TO-gra-pher, not PHO-to-gra-pher.

This one is undervalued by candidates and heavily weighted by listeners. A word with the stress on the wrong syllable is often simply not recognised, even when every phoneme in it is correct — misplacing the stress in comfortable or develop costs you more than a slightly off vowel anywhere in the sentence.

3. Sentence stress and rhythm

Which words in the sentence get the beat, and the reduction of everything else to weak forms. English is stress-timed: content words are strong, function words reduce to a schwa. Give every word equal weight and the examiner hears a candidate who has not internalised English rhythm — and has to work harder to follow you. This is the layer covered in detail in the schwa.

4. Intonation

Pitch movement across a phrase: falling for a completed statement, rising for a genuine question or an unfinished list, and the wider excursions that mark contrast and attitude.

Flat delivery is the single most common thing capping otherwise strong candidates at band 6. It is also the feature most improved by a nervous candidate simply slowing down and breathing.

5. Chunking and pausing

Grouping words into meaningful phrases with pauses between them, rather than pausing mid-phrase to search for a word.

The distinction the examiner makes is where you pause, not how often. A pause between thought groups is a pronunciation feature. A pause between an adjective and its noun is a hesitation.

The self-check

The point of the descriptors is that they are checkable. Record a two-minute Part 2 answer on any topic you have not rehearsed, then go through this table. Each row has something you can look at, not just listen for.

If you record the answer in sayit, most of these have a number attached: per-word clarity verdicts and the specific phoneme substitutions for the first row, a stress score and coverage for the second, pace against a target for the third, and a pitch contour with a monotone flag for the fourth. Chunking is the one you still have to judge by ear — look at where the gaps fall in the transcript.

Treat any app's numbers, including these, as a training signal that tracks direction rather than an official band prediction. No tool outside the test is examiner-calibrated, and anyone claiming otherwise is selling something.

The band 6 to band 7 gap, in practice

Across candidates who are stuck just below 7, the gap is almost always two specific things rather than a general weakness.

One recurring meaning-changing sound. Not a scattering of imperfect sounds — one substitution that shows up dozens of times because it lives in high-frequency words. TH becoming /s/ or /t/. /v/ becoming /w/. Final consonants disappearing. It reads to the examiner as "mispronunciation reduces clarity at times" precisely because it recurs.

Flat intonation on long answers. Candidates who sound lively in Part 1 often go monotone in Part 2, because a two-minute solo turn is cognitively expensive and prosody is the first thing to be dropped. This is the "not sustained" clause in the band 6 descriptor, almost verbatim.

Fixing either one moves you. Fixing both is usually the whole band.

A two-week plan

This assumes you already have your vocabulary and fluency work under way, and are adding pronunciation to it. Roughly thirty minutes a day.

Days 1–3 — diagnose. Record three unrehearsed Part 2 answers on different topics. Do not prepare them; you want a true baseline. Go through the self-check table for each. Write down the one sound that recurs most and whether your intonation flattens after the first thirty seconds. Everything else is noise for now.

Days 4–9 — one sound, one pattern, per day. Half the session on your recurring sound: isolated words, then minimal pairs, then sentences at speed. The relevant guide for the common cases is TH, v vs w, final consonants or initial clusters. Half the session on prosody: read a paragraph aloud with the content words deliberately louder and longer and the function words deliberately reduced, then say a one-minute answer trying to keep the same shape.

Days 10–13 — full answers, scored. Two timed Part 2 answers per day plus a Part 3 style follow-up discussion, reviewed against the same table. The goal here is not improvement in any single feature; it is sustaining what you fixed in days 4–9 across a full-length answer. That is literally the difference between the two bands.

Day 14 — rest. Do one short answer to stay warm and stop. Fatigue costs more than one more practice session gains.

If you want the full four-criteria version of this plan, including fluency, lexis and grammar, it is in IELTS speaking practice at home.

Examiner realities worth knowing

Your accent is not being marked. It is only relevant through intelligibility. Candidates with strong Indian, Nigerian, Spanish or Chinese accents get band 8 and 9 every day.

Speaking faster does not help. Fluency is scored separately, and it is about coherence and hesitation, not speed. Speeding up usually costs you chunking and intonation at the same time. Slowing down slightly and pausing in the right places generally raises both scores.

The examiner hears you for the whole test. Pronunciation is assessed across Parts 1 to 3, and the "sustained" language in the descriptors means a strong Part 2 does not rescue a weak Part 3. Practise the long, unscripted parts.

Memorised answers hurt you. They are audibly different — the prosody flattens into a recitation — and examiners are trained to notice. They also collapse the moment a follow-up question goes off script.

Questions people actually ask

Can I get band 7 with a strong regional accent? Yes. The descriptors are about intelligibility and the range of features you use. Accent enters the scale only at band 8, and only as "minimal effect on intelligibility".

Is band 8 pronunciation a different skill? It is the same features, sustained. Band 8 language is about wide range, flexible, throughout and easy to understand. If band 7 is "you clearly do this, most of the time", band 8 is "you do this without it costing you anything".

How much can I move in two weeks? Prosody — rhythm, intonation, chunking — can move noticeably in two weeks, because it is a habit rather than a motor skill. A new consonant contrast usually needs longer to hold up under exam pressure, which is why the plan above spends its time on one sound rather than five.

Does practising with an app count? For the measurable features, yes, and it is faster than self-assessment because it removes the problem of your ear not hearing what your mouth did. For the parts that need a human — reading your hesitation, probing a weak answer — book at least one mock with a qualified examiner or teacher. The trade-off is set out in AI feedback vs a human coach.

Where to go next

The companion guide with the full four-criteria plan is IELTS speaking practice at home; the TOEFL equivalent, which is scored differently, is TOEFL speaking practice. For the prosody half of the work, start with word stress and intonation.

To run the self-check on a real answer, open sayit in your browser — free to start, no install and no card — record a two-minute unrehearsed Part 2 answer, and go through the table above against the numbers rather than your impression of how it went.