IELTS Fluency and Coherence: The Scale Judges Which Pauses, Not How Many

The official Fluency and Coherence descriptors never count your pauses or your speed. They classify what you paused for, and one read-aloud test tells you which kind of pause you have.

Published on September 10, 2026

Direct answer

Fluency and coherence is one of the four Speaking criteria, and it measures whether you can keep going and stay understandable while you do it, not how fast you talk. The descriptors never count pauses; they classify them, and the line between band 6 and band 8 is whether you paused to plan what to say or to hunt for the words to say it with.

That distinction is written into the official scale, and once you can see it, most of the standard advice about this criterion turns out to be aimed at the wrong thing.

What the scale actually says

Here is the fluency and coherence column of the official IELTS Speaking band descriptors, reduced to the lines about hesitation.

Read the column from the bottom up and you can watch two separate things improve.

From band 4 to band 6, the change is quantity: you stop stopping. Band 6 is “able to keep going and demonstrates a willingness to produce long turns” — it still allows coherence to be lost “at times.”

From band 6 to band 9, the change is cause. Every band above 6 is written in terms of what the pause was for. Band 7 pauses “indicate problems with accessing appropriate language” but do not break coherence. Band 8 allows the same kind of pause “occasionally,” and requires that most hesitation be “content related.” Band 9 permits hesitation only to prepare content, “not to find words or grammar.”

Nobody at the top of that scale is silent. They pause to think about what, never about how.

The speed trap

Candidates who are told they lack fluency usually respond by speaking faster. Nothing in the criterion asks for that.

Speed appears nowhere in the fluency descriptors as something to increase, and a rapid speech rate is named in the pronunciation column as something that damages rhythm. Band 6 pronunciation, in full: “Chunking is generally appropriate, but rhythm may be affected by a lack of stress-timing and/or a rapid speech rate.”

So the instinct costs you twice. It cannot lift fluency, because fluency is not measured in words per minute, and it can lower pronunciation, because rushing is one of the two things the pronunciation descriptor names as a rhythm problem. Word endings are the first casualty: asked, worked, lists, months lose their final consonants when the delivery is rushed, and those endings carry tense and number. Our guide to consonant clusters covers the endings that go first.

Speed is only mentioned once, at band 5, and it is mentioned as a crutch: candidates who are “usually able to keep going” but rely “on repetition and self-correction to do so and/or on slow speech.” Slow speech is listed as a symptom, not as the fault to fix.

The complexity trade

Band 5 contains the most useful sentence in the whole column, and it is a warning about strategy rather than a description of a speaker:

More complex speech usually causes disfluency but simpler language may be produced fluently.

Band 5 describes exactly that trap in writing: more complex speech usually causes disfluency but simpler language may be produced fluently, so reaching for vocabulary you do not control costs you on one criterion to gain on another. The candidate who inserts a memorised idiom mid-sentence and then stalls for two seconds has traded a fluency point for a lexical one, and the trade rarely pays.

There is a version of this that does pay: vocabulary you have said out loud enough times that it arrives without a search. Anything else is a hesitation waiting for a place to happen.

The read-aloud test

You cannot tell from inside your own answer whether you paused for an idea or for a word. There is a two-take test that separates them.

1. Improvise. Take any Part 2 cue card, speak for two minutes with no notes, and record it.
2. Write the same answer out, then read it aloud into a second recording.

The second take removes language search entirely. The words are on the page; nothing is left to find. Whatever hesitation survives into the read-aloud take is delivery — breath, phrasing, unfamiliar words on the page. Whatever disappears was language access.

Read an answer you have already written out loud: if it is smooth on the page and broken when you improvise it, your delivery is fine and your language access is the problem. That means vocabulary and grammar practice, not pronunciation practice, and no amount of recording yourself reading will move it.

The reverse result is more common than candidates expect. If the read-aloud take is also broken — long gaps mid-phrase, endings dropped, a rate that swings between crawling and sprinting — then the problem is not that you lack words. It is that the delivery is not stable even when the words are handed to you, and that is the half you can measure and fix alone.

Label your pauses

Play the improvised take back once with a pen.

• Mark C where you were deciding what to say — the next example, whether to mention your brother, how to end.
• Mark L where you were looking for how to say it — the word, the tense, the preposition.

Bands 8 and 9 allow the C pauses and are written to catch the L ones. Count the L marks across three recordings and you have the actual size of your problem, which is usually four or five recurring spots rather than a general condition.

Two patterns show up almost every time. L pauses cluster before the same few topics — your job, your studies, the thing you do at weekends — because those are the answers that need vocabulary you rarely use in English. And they cluster mid-sentence, which is exactly what band 7 and band 5 both name: the mid-sentence search is the signature of a language problem, while a pause between sentences reads as planning.

Fix the L pauses

For each L pause, say the same idea again with the vocabulary you already control. The goal is one version of the sentence you can produce without a mid-sentence search, not a faster version of the broken one.

Then say that version out loud until it is automatic. Not memorised as a script — the examiner will notice — but rehearsed at the level of the phrase, so the words for your own job and your own city arrive without a hunt. Those phrases repeat across almost every Part 1 and Part 2 answer you will ever give.

Pause placement is the other half. Put your pauses where a comma would go if you were writing the answer, and nowhere else. Pauses at meaning boundaries read as control; pauses in the middle of a phrase read as a search. The same skill is called chunking in the pronunciation column, which is why it is the one habit that pays on two criteria at once — connected speech covers what happens inside a chunk, and English rhythm covers the beat that holds the chunks together.

Check the last thirty seconds

Re-record the whole long turn and check the last thirty seconds, where content runs out and both hesitation and word endings usually collapse at the same time.

This is not a coincidence. When you are searching for something to say, attention leaves the delivery, and the two criteria fail together: hesitation rises on fluency, endings vanish on pronunciation. Preparing more content for Part 2 is therefore also delivery practice, which is the same conclusion the IELTS pronunciation guide reaches from the other direction.

What Phonema measures, and what it does not

Phonema works one sentence at a time. You give it the sentence — typed, pasted, or captured with the camera — and alongside the word-by-word clarity score it returns a delivery score with one line saying what moved it: how many pauses longer than half a second fell inside the sentence, or that your rate was slow and what it was in words per minute. Those numbers come from where each sound actually landed in your recording, not from an estimate, and it runs on the phone, so the audio never leaves it.

That makes it the instrument for the read-aloud half of the test above, taken a sentence at a time rather than a two-minute turn at a time: record the long turn on your phone’s voice recorder for the C-and-L labelling, and bring the sentences that broke into Phonema. You supply the text, so language search is out of the picture by construction, and what is left is the delivery: whether you can say a prepared line at a steady rate, with the pauses at the boundaries rather than in the middle, without losing the ends of words.

Be clear about the limit. Because you supply the text, it cannot know whether you paused to think — there was nothing to think about — and it cannot judge coherence, discourse markers, or whether your answer developed the topic. Those are the parts of this criterion that need a listener or an examiner, not a measurement. If you want the score to mean something across weeks, how to measure pronunciation progress explains what has to stay constant between recordings.

Where this stops

Fluency and coherence is a quarter of your Speaking score, and half of it — the coherence half — is about how an answer is built rather than how it is delivered. Discourse markers, topic development, the shape of a Part 3 answer, being interrupted and recovering: none of that is visible in a recording of a sentence you wrote.

For that part, my other app, OralPrep, simulates the full three-part test and scores answers against all four criteria, including whether your answers are developed enough to be worth a band. Phonema does the narrower job: the delivery of the words themselves, measured the same way every time, so you can tell whether this week’s take was steadier than last week’s.

Common questions

What is fluency and coherence in IELTS Speaking? Fluency and coherence is one of the four Speaking criteria, and it measures whether you can keep going and stay understandable while you do it, not how fast you talk.

Do pauses lower your IELTS Speaking score? The descriptors never count pauses; they classify them, and the line between band 6 and band 8 is whether you paused to plan what to say or to hunt for the words to say it with.

Does speaking faster improve IELTS fluency? Speed appears nowhere in the fluency descriptors as something to increase, and a rapid speech rate is named in the pronunciation column as something that damages rhythm.

Is it better to use simple language and stay fluent? Band 5 describes exactly that trap in writing: more complex speech usually causes disfluency but simpler language may be produced fluently, so reaching for vocabulary you do not control costs you on one criterion to gain on another.

How can I tell if my hesitation is a language problem or a delivery problem? Read an answer you have already written out loud: if it is smooth on the page and broken when you improvise it, your delivery is fine and your language access is the problem.

Do fillers like “um” count against you? They are hesitation like any other, and the descriptors care about what caused them. A filler while you choose your next example is content-related; a filler in the middle of a clause while you search for a verb is the mid-sentence pattern named at bands 5 and 7.

Band descriptors quoted from the official IELTS Speaking band descriptors; see also IELTS scoring in detail.

Related sounds

• t — as in tea
• d — as in day
• s — as in see
• θ — as in think