Guides · 11 min read

Why can't people understand my English? Six causes, and how to find out which one is yours

Fluent English that people keep asking you to repeat is almost never a vocabulary problem. Here are the six things that actually break intelligibility, ranked by how much damage they do, with a way to test each one.

You can hold a technical conversation in writing. You read comfortably. Your grammar is better than most native speakers'. And still, several times a week, someone says "sorry?" or repeats your sentence back with a question mark on it.

This is one of the most demoralising places to be in a language, partly because the usual advice — "practise more", "immerse yourself", "get an accent coach" — does not tell you what is actually wrong. Something specific is happening in your speech, and it is almost certainly on a short list.

Here is that list, ordered by how much damage each item does, with a way to test yourself on each.

First: it is probably not your accent

The word "accent" gets used for two completely different things, and conflating them is why so much advice is useless.

Accent is the systematic colour of your speech: the exact quality of your vowels, your rhythm, your voice. Nigerian English, Indian English, Scottish English, Singaporean English and Texan English are all accents. Listeners adjust to an unfamiliar accent within about a minute of conversation, and then it costs them nothing.

Intelligibility is whether the listener recovers the words you intended. That is a different variable, and it is the one that breaks.

The evidence that they are separate is everywhere: speakers with very strong regional accents who are perfectly easy to follow, and speakers with faint accents whom nobody can understand. If "reduce your accent" were the answer, the first group would not exist.

So the useful question is not "how do I sound less foreign" but "which specific thing is costing my listener the word". There are six candidates.

1. A meaning-changing sound collapse

The heaviest hitter. Somewhere in your speech, two English sounds that distinguish real words have merged into one.

The usual suspects, and the words that expose them:

CollapseSounds likeCommon for
/θ/ to /s/ or /t/think becomes sink or tinkmost languages
/v/ to /w/vet becomes wetHindi, Urdu
/r/ to /l/rice becomes liceJapanese, Korean
/ɪ/ to /iː/ship becomes sheepSpanish, Arabic, Hindi
/b/ to /v/boat becomes voteSpanish
final consonants droppedcat, cap, can mergeMandarin, Vietnamese

This one does the most damage per occurrence, because it does not degrade the word — it replaces it. Your listener does not hear a slightly odd think; they hear sink, a real English word, and try to make it fit. By the time they have failed, you are three sentences further on.

The test. Say these pairs to someone and have them write down which one they heard, without seeing your face: think/sink, vet/wet, rice/lice, ship/sheep, cat/cap. Any pair they get wrong more than once is a collapse. The method for fixing them is in minimal pairs practice; the specific guides are TH, v and w and word endings.

2. Word stress on the wrong syllable

Underrated, and often the actual cause when sounds are fine.

English listeners use the stressed syllable as the primary key for looking a word up. Move the stress and the lookup fails, even when every individual sound is correct. comFORtable instead of COMfortable, DEvelop instead of deVELop, PHOtographer instead of phoTOgrapher — a listener hearing the wrong stress pattern often does not recognise the word at all, and cannot even tell you which word they missed.

This is why "I said it perfectly and they still didn't get it" happens. You did say the sounds perfectly. The beat was in the wrong place.

The test. Record yourself saying ten long words you use often at work. Check each stress placement in a dictionary. Words borrowed from your own language, and words you learned by reading rather than hearing, are the likely offenders. Word stress and intonation has the patterns.

3. Flat rhythm from unreduced vowels

If you pronounce every vowel fully, every syllable gets equal weight, and your listener loses the map of which words carry the meaning.

English is stress-timed. Content words are strong and long; function words shrink to a schwa. When that alternation is missing, the listener has to process every word with equal attention — which is exhausting and slow, and produces exactly the "I have to concentrate really hard to follow you" reaction that people are too polite to say out loud.

This one rarely causes an outright misunderstanding. It causes fatigue, and fatigue causes people to stop listening carefully, which then causes misunderstandings downstream.

The test. Say "I'd like to speak to the manager about the problem" at normal speed and count your strong beats. There should be about four. If there are twelve, you are not reducing. The schwa is the fix.

4. Speaking too fast for your own control

Counterintuitive, and very common among fluent speakers.

Speed is not the problem in itself — native speakers are fast. The problem is running faster than the rate at which you can maintain your sounds and your stress placement. Past that point, precision collapses first and speed keeps going.

There is usually a psychological driver: people speak fast when they are anxious about taking up time, or to demonstrate fluency, or because a rehearsed sentence pours out at once. The result is a listener who catches the first three words and reconstructs the rest.

The test. Record a minute of speech and compare your words per minute against a normal conversational target. If your rate is high and your sound accuracy is poor on the same recording, but both improve when you deliberately slow down, speed is your bottleneck rather than articulation.

5. Pausing in the wrong places

Related to speed, and it goes wrong in a specific way.

English groups words into thought groups with pauses between them. If your pauses fall inside those groups — between an adjective and its noun, between a verb and its object — because you are searching for the next word, the listener's parser breaks. They were assembling a phrase and the phrase stopped.

The number of pauses matters much less than their position. Plenty of slow, heavily-pausing speakers are perfectly easy to follow because their pauses fall at the joints.

The test. Read back a transcript of your own speech and mark where you actually paused. Pauses at commas and clause boundaries are fine. Pauses in the middle of a noun phrase are the problem. Connected speech covers how phrases hold together.

6. Monotone delivery

Pitch does grammatical work in English. A falling contour ends a statement. A rising one marks a genuine question or an unfinished list. A wide excursion signals contrast or new information.

Flatten all of that and the listener loses the punctuation. They cannot tell where your sentences end, whether you have finished, whether you are asking or telling. It is also frequently misread as boredom or coldness, which is a social cost on top of the comprehension one.

Monotone delivery is often a confidence artefact rather than a skill gap: people flatten out when speaking a second language in a high-stakes setting. It usually returns the moment the speaker relaxes, which is a useful diagnostic in itself.

The test. Record a two-minute answer and look at the pitch contour, or just listen for whether every phrase has the same shape. If the last three words of every sentence sound identical in shape, you are flat.

Finding out which one is actually yours

Reading a list of six causes does not tell you which is operating in your speech, and self-assessment on this is genuinely unreliable — your ear was trained by your first language, so it filters out exactly the distinctions you are failing to make. Your own successful attempts and your failures can sound identical to you.

Three ways to get an outside read, in rough order of accessibility:

A patient listener. Have someone mark, in real time, the moment they lose you — not what was wrong, just when. Then look at what you were saying at each mark. Patterns emerge quickly. This costs nothing and works well, but people find it socially awkward to do honestly.

A teacher. A qualified pronunciation teacher will find your pattern in one session. It is the fastest route and the most expensive one, and it is worth it if you can do it once even if you cannot do it weekly. What a human catches that software does not is covered in AI feedback versus a human coach.

A tool that inspects sounds, not words. This is what sayit is for. You read a sentence aloud, and instead of a transcript you get a per-word clarity verdict, the target pronunciation in IPA beside the sounds actually detected, a stress score, a pace reading against a target, and the pitch contour across the phrase. Those map onto the six causes above almost one to one: substitution patterns for cause 1, stress score for causes 2 and 3, pace for cause 4, and the contour for cause 6. Repeated problems across a whole take appear as a focus areas list, which is what turns "I sometimes get TH wrong" into "TH accounted for most of the unclear words in this recording".

One important caveat about tooling, and the reason ordinary dictation apps cannot do this job: standard speech recognition is built to guess the most likely sentence, so it silently repairs your errors. Say "I sink so" and it prints "I think so". The error that is costing you is exactly the one it hides. That mechanism is explained in the autocorrect problem.

What to do once you know

Fix one thing at a time, in this order:

  1. 1.The meaning-changing sound, if you have one. Highest damage per occurrence, and usually the fastest visible win.
  2. 2.Word stress on your frequent words. Cheap to fix — it is knowledge, not motor skill. Look up thirty words you say every week and correct the ones you have been getting wrong.
  3. 3.Vowel reduction and rhythm. A habit change, roughly a week to shift the default and a month to stop reverting under pressure.
  4. 4.Speed and pausing. Mostly a decision rather than a skill, though a nervous decision is hard to hold.
  5. 5.Intonation. Slowest to move deliberately, and often the thing that improves on its own once the four above stop consuming your attention.

Working on all five at once is the standard mistake, and it produces slow progress on everything. One at a time, with a way to check, is faster even though it feels slower.

Questions people actually ask

Should I get accent reduction training? Be careful with the framing. Programmes that target intelligibility — meaning-changing sounds, stress, rhythm — are useful. Programmes that promise to erase your accent are selling something you do not need and probably cannot achieve, and adult second-language accent is famously resistant regardless of effort. Judge a programme by whether it can name the specific contrast it will fix.

Is it my vocabulary? Almost never, if this is happening with words you use confidently. Vocabulary problems feel different: you know you did not have the word. This problem feels like you said the right thing and it did not arrive.

Why is it worse on the phone? Phone audio removes high frequencies, which is exactly where the fricatives live — /s/, /f/, /θ/, and the endings that carry grammar. Listeners also lose your lip movements, which they use more than they realise. If you are only misunderstood on calls, cause 1 or a dropped-endings problem is the likely culprit.

Why do some people understand me and others don't? Familiarity. Colleagues who hear your accent daily have adapted; strangers have not. If people who know you understand you perfectly and strangers do not, that is a strong signal that the issue is real but recoverable from context — which is exactly what a meaning-changing sound collapse looks like.

Does it get better just from living in an English-speaking country? Listening improves a great deal. Pronunciation often plateaus, because you get very little corrective feedback: people are polite, and they repair what you said from context rather than telling you what they heard. Which is precisely why an explicit diagnosis is worth more than another year of immersion.

Where to go next

If you already suspect a specific sound, go straight to it: TH, v and w, word endings, initial clusters. If you suspect rhythm, start with the schwa and connected speech. For a structured plan across all of it, how to improve English pronunciation is the overview.

If you do not know which one is yours — which is the honest answer for most people — get a diagnosis before you pick a drill. Open sayit in your browser, read one sentence aloud, and look at which words were flagged and why. It is free to start, there is no install and no card, and it takes about ten seconds.

Free in your browser

Hear exactly which sounds to fix.

Say one sentence and get sound-by-sound feedback in seconds. No install, no card.