Which Parts of Your Accent Actually Cost You, and Which Are Free

You do not need a native accent, and by now most people have heard that. What almost nobody tells you is the more useful half: some pronunciation errors genuinely break comprehension, and others cost you nothing. Fixing the wrong ones burns months. This is how to tell them apart.

Intelligibility and nativeness are two different targets

Aiming to sound native means matching thousands of small habits: vowel colour, rhythm, the exact moment a consonant is released, the melody of a question. Aiming to be intelligible means something narrower. It means a listener never has to stop and rebuild your sentence.

These targets overlap far less than people assume. You can carry an obvious accent and be perfectly easy to follow. You can also flatten your vowels toward something American and still lose a listener every few sentences because you are putting the stress in the wrong place. The second speaker sounds “better” in a five-second clip and is harder to work with in a meeting.

The practical consequence: the question is never “how foreign do I sound”. It is “how much work is the listener doing”.

There is a second reason the distinction matters, which is that listeners adapt and vowels do not need fixing for them to. A few minutes into a conversation your accent is far less of an obstacle than it was in your first sentence, because the person you are talking to has recalibrated to your voice. That adaptation works on the parts of your accent that are consistent. It does not rescue you from a misplaced stress, because a stress error does not look like an accent to the listener. It looks like a different word.

Word stress is the expensive error

If you fix one thing, fix this. English listeners use stress to identify the word itself, not just to add emphasis. Say a word with the stress on the wrong syllable and a native listener will often not recognise a word they know perfectly well.

Part of the reason is that English hollows out its unstressed syllables. The vowel in an unstressed syllable collapses toward a schwa, so the shape a listener is scanning for is not “five clear syllables” but “one strong beat with mush around it”. Move the beat and you have handed them a different shape.

Word	Stress	Common misplacement
deVELopment	second	DEvelopment
phoTOGraphy	second	PHOtography
comFORtable	second	comforTAble
deTERmine	second	DEtermine
necESSary	second	necesSARY

Korean does not use stress this way, so there is nothing in your first language pushing you toward the right syllable. It has to be learned per word, which is tedious but finite: the stress is printed in the IPA on every dictionary entry you have ever looked at, and most people have simply never read that part.

The consonant contrasts that carry real cost

Not all substitutions are equal. A substitution is expensive when it turns one real English word into another real English word, and cheap when the result is not a word at all.

• /r/ and /l/. Expensive. Right and light, collect and correct, glass and grass. Both members exist, both are common, and context often permits either.
• /p/ and /f/. Expensive. Pour and four, cup and cuff, copy and coffee.
• /b/ and /v/. Expensive. Berry and very, boat and vote, curb and curve.
• /θ/ as /s/ or /t/. Usually cheap. Sink and think are both words, but they almost never fit the same sentence. Listeners repair this one without noticing.
• Final /z/ devoiced to /s/. Cheap. Nobody has ever misunderstood “he sees” because the final consonant was crisp.

This is a ranking by consequence, not by difficulty. The th sound gets an enormous share of classroom attention and sits near the bottom of the list, while /p/ and /f/ get almost none and sit near the top.

The test is easy to apply to your own speech. Take the substitution you make and ask whether the result is still an English word. If it is, you have a real problem worth hours. If it is not, the listener’s ear will patch it and you can leave it alone. Almost every ranking in this article falls out of that single question, and it works for contrasts I have not listed.

The extra vowel is the one nobody warns you about

Korean syllables do not permit most of the consonant clusters English piles up, so a vowel gets inserted to break them apart. The word grows syllables it does not have.

asked is one syllable, /æskt/. Said with inserted vowels it becomes three or four, and a listener scanning for a one-beat word does not find it. The same thing happens to worlds, texts, clothes, months.

This one deserves attention because it is invisible from the inside. You are producing every sound in the word, in the right order, and it still fails. Nothing feels wrong. The fix is not a new sound but a subtraction: get the consonants to run together with nothing between them, even if the result feels rushed and unfinished. It is supposed to feel rushed. That is what the word sounds like.

Both expensive errors also break your listening

Here is the part that turns this from a speaking problem into a general one. Whatever shape you store a word in is the shape you listen for.

Store determine with the stress on the first syllable and you are scanning incoming speech for DEtermine. When a native speaker says deTERmine at conversational speed, the word you are waiting for never arrives, and you experience that as “they spoke too fast”. Store asked as three syllables and a one-beat /æskt/ slides past unrecognised. People conclude their listening is weak and buy a listening course, when the actual fault is in how the words were filed.

That is also why fixing these two categories pays twice. Every hour spent putting the stress in the right place improves your production and your comprehension at the same time, which is not true of vowel colour or intonation work. If your listening has plateaued while your vocabulary keeps growing, this is the first place to look.

What you can safely keep

Here is the part that gets left out of pronunciation advice, so it is worth stating plainly. The following are audible, they mark you as a second-language speaker, and they cost you approximately nothing:

• Vowel colour. Your /æ/ not being quite an American /æ/. Listeners normalise vowel space within a sentence or two.
• Rhythm that is more even than a native speaker’s. It sounds a little metronomic. It is still perfectly parseable as long as the stressed syllable in each word is the right one.
• Intonation that is flatter than the local norm. This affects how warm or interested you sound, which is a real thing, but it is a social question and not a comprehension one.
• A recognisable regional flavour of any kind. English has no single reference accent, and listeners are used to enormous variation.

There is an exception worth naming. If your work depends on being read as local rather than merely understood, some of the decorative layer stops being decorative. Broadcasting and acting are the obvious cases. For everyone else, including everyone taking a speaking exam, the graders are trained on intelligibility and fluency rather than nativeness, and the decorative layer stays decorative.

Spending a year on any of these is a year not spent on stress and clusters. That is the actual cost of chasing nativeness: not failure, but the opportunity cost.

A triage table

Feature	Cost to the listener	Priority
Wrong syllable stressed	High — the word is not recognised	Fix first
Vowels inserted into clusters	High — the word gains syllables	Fix first
/r/–/l/, /p/–/f/, /b/–/v/	Medium to high — real word becomes another real word	Fix next
Vowel length distinctions	Medium — ship and sheep, full and fool	Fix next
/θ/ substitutions	Low — context repairs it	Optional
Vowel colour, rhythm, intonation	Low — marks you, does not block you	Leave it

Nothing in the bottom two rows is wrong to work on. It is just that the top two rows pay several times more per hour.

How to practise the expensive ones

The two high-cost categories need different treatment, which is why “practise pronunciation” as a single instruction rarely works.

1. For stress, learn it with the word. The moment you meet a new word is the moment to notice which syllable carries the beat. Adding it later means relearning the word, and the wrong stress will already be in muscle memory.
2. For clusters, slow the vowels and not the consonants. Most people slow down by stretching everything evenly, which makes the inserted vowel more prominent. Stretch the vowel you are supposed to have and let the consonants stay tight.
3. Record and compare, do not just listen. Listening to a model and thinking “yes, that is what I said” is the standard trap. Record yourself and play both back to back.
4. Read the IPA, not just the spelling. The stress mark is right there. /dɪˈtɜːrmɪn/ tells you where the beat goes before you have heard the word once.

That last one is the cheapest habit in the whole list and the most skipped. Vocaby carries IPA and audio on every one of its 29,000+ words, so the stress mark and the model recording sit side by side, and FSRS spaced repetition brings the word back before the pronunciation has faded along with the meaning. You can browse the entries on the Vocaby word pages too.

If you want the technique layer underneath this triage, how to improve English pronunciation covers the drills themselves.

The short version

Your accent is not a defect to be removed. Parts of it are, however, doing real damage, and they are a short and specific list: the syllable you stress, and the vowels you add where English does not have them. Everything else is decoration, and you can keep as much of it as you like.

Learn the stress with the word, not after it. Get Vocaby on the App Store and hear each one as you learn it.