It is two criteria wearing one name
Fluency is about flow: can you keep going without breaking down. Coherence is about sense: can a listener follow where your ideas are going. They are marked together, and most candidates only train the first one.
That is why people who speak quickly and smoothly still get a 6. Smooth is not the same as easy to follow.
What each band actually sounds like
The official descriptors are qualitative. Here is what they come down to in practice.
| Band | Fluency | Coherence |
|---|---|---|
| 5 | Keeps going, but leans on repeating and re-starting to do it. Noticeable slowing. | Ideas are there but the links are basic or mechanical. Connectives get over-used. |
| 6 | Willing to speak at length, but flow breaks now and then — repetition, self-correction, hesitation. | Mostly followable. Some links are used wrongly. Ideas can wander. |
| 7 | Speaks at length without obvious effort. Hesitation happens, but it does not derail the answer. | Ideas are organised and connected with some flexibility, not just firstly / secondly. |
| 8 | Fluent throughout. Occasional self-correction only. Hesitation is about what to say, not how to say it. | Topics are developed properly — each point goes somewhere. |
| 9 | Rare repetition or correction. Any pause is thinking, never searching. | Fully coherent, ideas fully developed. |
The two kinds of hesitation
Because of that line, not all pauses are equal — and this is where most candidates lose marks without knowing it.
| Kind | What it sounds like | How it is treated |
|---|---|---|
| Content hesitation — choosing an idea | A pause at a clause boundary: "I usually study at home … because the library is too far." | Normal at every band, including 8 and 9. |
| Language hesitation — hunting for a word | A pause inside a phrase, after a, the, to, is: "I study at the … um … the place near my house." | This is the hesitation that holds you at 6 or 7. |
The two silences can be exactly the same length. The listener hears them completely differently, because one lands where a thought ends and the other lands where the grammar has already committed you to a word you cannot find.
This is also why trying to eliminate "um" does not work: the sound is not the problem, the position is. There is more on this in how to stop hesitating.
What we measure, and what it lines up with
When Aflo scores a spoken answer, we measure how fast you speak while actually speaking, how many words you get between pauses, how much of the answer is silence, and where each pause lands. Our measurements agree with human raters at about ρ = 0.8.
| What we measure | Stronger | Weaker |
|---|---|---|
| Words between pauses (length of run) | 7+ words | under 4 |
| Long pauses (½ sec or more), per minute | 4 or fewer | more than 12 |
| Pause ratio — share of the answer that is silence | under 30% | over 45% |
| Speaking rate while speaking | about 2.8–3.3 words/sec | below ~1.5 |
Two warnings about that table. Speed is not the goal — 2.5 words per second is perfectly fluent, and rushing causes the stumbles that produce mid-phrase pauses. And a short answer is not a slow one: brief answers look worse on some of these numbers automatically, which is why no single figure should ever decide a band.
The half nobody trains: coherence
Coherence is not a list of linking words. In fact, reaching for a lot of them is a band 5–6 signal, not a 7+ one: mechanical firstly, secondly, moreover, in conclusion on a two-minute answer marks you as someone using a memorised frame.
What raises coherence is that your ideas actually go somewhere:
- Answer the question first, then develop. Do not warm up for twenty seconds.
- Give a reason for every claim. Because is the highest-value word in the test. It forces development and it fills time honestly.
- Then give one concrete example. A specific, real detail is worth more than three general sentences.
- Do not start a new idea until the last one has finished. Wandering is what "loses coherence" means in the descriptors.
That is the whole structure: point → reason → example. It sounds organised, it produces long runs of speech, and it gives you something to say — which removes the language hesitation that was capping your fluency in the first place. The two halves of the criterion fix each other.
What actually moves you up a band
| If you are at… | The change that moves you |
|---|---|
| 5 → 6 | Stop restarting sentences. Finish the sentence you began, even if it is imperfect. Say more per turn. |
| 6 → 7 | Stop losing the thread. Use point → reason → example so every answer has a shape, and drop the mechanical connectives. |
| 7 → 8 | Move your hesitation. Simplify your sentences so you stop hunting for words — pause to choose an idea, not to find vocabulary. |
Notice that none of these is "speak faster" or "use more linking words". Both of those are how people usually try to raise this criterion, and neither is what it measures.
You cannot hear this in yourself
While you are speaking, your attention is on the idea, so you do not notice your own hesitation — and afterwards you remember whether it felt smooth, which is a different thing. Record a two-minute answer and listen back with one question: did I stop inside my phrases, or between them?
Aflo listens to your answer, measures your pauses and pace, and shows you where fluency dropped and whether your ideas were developed — so you can see which half of the criterion is actually holding your band down.
Frequently asked questions
What is the difference between band 7 and band 8 in Fluency and Coherence?
Where your hesitation comes from. At band 7 you may still stop to find a word. At band 8 you only stop to decide what to say — hesitation is content-related, not language-related. That is the actual line in the descriptors, and it means the fix is to simplify your sentences so you stop hunting for vocabulary, not to speak faster.
Do linking words raise my Fluency and Coherence score?
Not by themselves, and over-using them lowers it. Mechanical firstly / secondly / moreover on a short answer reads as a memorised frame, which is a band 5-6 signal. Coherence is about ideas that develop and go somewhere: answer, reason, example.
Does speaking faster improve fluency?
No. Speed is not what the criterion measures. Around 2.5 words per second while you are actually speaking is perfectly fluent. Rushing causes stumbles and restarts, which produce exactly the mid-phrase hesitation that holds your band down.
Are the numbers in this article official IELTS criteria?
No. IELTS publishes qualitative band descriptors and no acoustic thresholds at all. The figures here are Aflo's own measurements, which we validated against human ratings at about ρ = 0.8. Use them as a mirror on your speech, and be sceptical of anyone presenting numbers like these as official.