Word endings for Mandarin and Vietnamese speakers: why 'cat', 'cap' and 'can' merge

Ask a Mandarin or Vietnamese speaker which English sounds they find hard and you will usually hear about /r/, /l/ and TH. Ask an English listener where the misunderstandings actually come from and you get a different answer: the ends of words.

Cat, cap, can and cab differ only in their final consonant. So do walk and walked, play and plays, life and live. When word endings go quiet, all of those collapse into each other — and unlike a slightly-off vowel, a missing ending removes information the listener has no way to recover.

What your mouth is used to

Mandarin syllables are strictly limited at the end. A Mandarin syllable can finish with a vowel, with /n/, or with /ŋ/ — and that is essentially the whole list. There is no /t/, /d/, /k/, /p/, /s/, /z/, /f/ or /v/ in final position, and no clusters at all.

Vietnamese allows more final consonants, but they are unreleased: the mouth closes for the /t/ or /k/ and simply stops, with no audible burst of air afterwards. To a Vietnamese ear the consonant is fully present. To an English ear, an unreleased final stop is very quiet and easily missed, especially over a phone line.

Cantonese sits between the two: final /p/, /t/, /k/ exist but are also unreleased.

So there are really two separate problems, and it is worth knowing which one is yours:

• Deletion — the consonant is not produced at all. Cat comes out as ca. Common for Mandarin speakers.
• No release — the consonant is produced but never audibly finished. Cat comes out as ca[t] with the t silent. Common for Vietnamese and Cantonese speakers.

Deletion needs a new gesture. No-release needs a small extra puff of air. The second is much faster to fix, and many people who assume they have the first actually have the second.

What gets lost

English packs an unusual amount of meaning into its final consonants.

The grammatical endings matter more than they look. "I walk to the office yesterday" is not a small pronunciation slip; it reads as a grammar error to the listener, because the only thing marking the tense was the /t/ at the end of walked. Speakers with excellent written grammar are routinely judged as having poor grammar for this reason alone.

The lexical contrasts matter in a different way. Listeners repair cat from context surprisingly often, which is why this problem can persist for years without obvious feedback. It becomes visible in exactly the situations that matter most: names, numbers, addresses, technical terms, and phone calls.

The method: hold the ending

The instruction that works is not "pronounce the final consonant". It is hold it a beat longer than feels natural, and let it finish.

For final stops (/p/, /t/, /k/, /b/, /d/, /g/), the fix is release. Say cat, and after your tongue touches the ridge behind your teeth, let a small puff of air escape. Put a hand in front of your mouth: you should feel it. Then say ca with no ending and feel nothing. That difference is what your listener is using.

For final fricatives and the grammatical endings (/s/, /z/, /f/, /v/, /θ/, /ð/), the fix is duration. These sounds are continuous, so they can simply be held: "catsss", "livvv". Over-hold them during practice; it settles.

The one rule that matters: never add a vowel. Cat must not become cat-uh, walked must not become walk-ed. An added vowel creates a syllable that is not there and is a worse error than the one you are fixing, because it makes the word unrecognisable rather than ambiguous. If you need time, lengthen the consonant, not the word.

The -ed and -s rules

Both endings change shape depending on the sound before them, and both are far more regular than they look.

-ed has three forms:

Only the third case adds a syllable. Walked is one syllable, not two. This is the single most common over-correction: a speaker who has been told to pronounce endings starts saying "walk-ed", "help-ed", which is a new error.

-s has three forms as well:

The pairs that expose it

Say each pair and check that the two members sound genuinely different, not just different in your head.

The four-way sets are the honest test. Most speakers can produce one final consonant when they concentrate on it; producing four distinct ones in a row, without preparation, is the skill that transfers to conversation.

The drill

Block 1 — release practice (1 minute). Hand in front of your mouth. stop · cat · back · cup · hit · book · lock · put. You should feel the puff on every one. No vowel afterwards.

Block 2 — grammar endings (2 minutes). Say the base word and the inflected word back to back, so the difference is audible to you: walk / walked · help / helped · want / wanted · play / plays · box / boxes · love / loved.

Block 3 — sentences (2 minutes). These are loaded with endings on purpose.

• Record: "She walked to the shop and bought six books."
• Record: "The cat sat on the black mat last night."
• Record: "He wants the reports that she finished last week."

The third sentence is the workplace version: wants, reports, finished, last week — four endings in nine words, exactly the density of an ordinary status update.

Reading it off a recording

This is a hard problem to self-diagnose, because a missing sound is a negative — there is nothing to hear. You cannot notice the absence of something your first language never expected to be there.

What makes it visible is a comparison against the target. When you record one of the sentences above in sayit, each word gets a clarity verdict, and opening a flagged word shows the target pronunciation in IPA against the sounds that were actually detected. A dropped ending appears as a missing segment at the end of the word — a deletion rather than a substitution — which is a different picture from getting the sound wrong, and points at a different drill.

Two further signals are worth watching. When the same kind of ending disappears across several words in one take, it shows up in the focus areas summary, which is how you tell a systematic pattern from one careless word. And because the analysis works from the sounds rather than from a guessed transcript, it does not quietly insert the -ed that the sentence grammar implies. Dictation apps do exactly that, which is why they are useless for this particular problem — the mechanism is explained in the autocorrect problem.

Questions people actually ask

Do native speakers always release final consonants? Not always. In fast connected speech, a final stop before another consonant is often unreleased — "that boy" rarely has an audible /t/. But native speakers release reliably at the end of a phrase, before a pause, and before a vowel, and they never drop the grammatical /s/ and /d/. Practise the released version until it is automatic; the natural reductions will come back on their own.

Is this an accent issue? No. This is the clearest example of the opposite. Vowel colour, rhythm and voice quality are accent, and there is no reason to change them. A missing /t/ on walked deletes the past tense, and no listener can recover it. Fix the endings; keep the accent.

Why do people understand me anyway? Because English listeners repair heavily from context, especially in familiar situations. The gap opens up when context is thin: unfamiliar names, numbers, addresses, technical vocabulary, and anything on the phone. The fact that you get away with it in ordinary conversation is exactly why the habit survives.

How long does it take? Releasing final stops in isolated words takes days. Keeping the endings inside fast sentences, and especially remembering -ed and -s under pressure, takes a month or two of daily sentence work. The good news is there is no new sound to learn — you can already make /t/, /s/ and /d/ at the start of words. You are moving an existing gesture to a new position.

Where to go next

The mirror-image problem, where speakers add a vowel rather than dropping a consonant, is covered in consonant clusters for Spanish and Portuguese speakers. If your endings are landing but people still ask you to repeat yourself, the next layer up is rhythm and melody: word stress and intonation and the schwa.

To find out whether your endings are actually arriving, record one sentence free in your browser — "He wants the reports that she finished last week" — and look at which sounds were detected at the end of each word.