English Contractions and Reduced Speech: the 60 Forms Natives Actually Use

Reduced speech is not slang, not sloppiness and not an accent. It is what English sounds like when it is spoken at conversational speed by anybody, anywhere, including newsreaders and professors. This page lists the forms in full, says what each one sounds like, and explains the two that cause the most damage — can versus can't, and the disappearing /h/.

If you learned English from a textbook, you learned a version of it that nobody speaks. Not a simplified version — a different one. Textbook English gives every word its own space. Real English runs words together, drops sounds out of the middle of them, and leaves you with a stream where you expected a sequence.

This is not a vocabulary problem and studying more words will not fix it. It is a decoding problem, and decoding is trainable.

The three things that happen to a word at speed

Reduction — an unstressed vowel collapses into a schwa, the neutral "uh" sound. For becomes /fər/, to becomes /tə/, can becomes /kən/. Roughly every function word in English does this.

Assimilation — a sound changes to match its neighbour. Did you becomes "didja", because /d/ + /j/ fuses into /dʒ/. Don't you becomes "doncha".

Elision — a sound disappears entirely. Next day becomes "nex' day". Old man becomes "ol' man". The /h/ in him, her, his vanishes after almost any word.

Everything below is one of those three.

The written contractions

These are the ones you have seen in print. They are the easy half.

The spoken reductions — never written, always said

This is the half that breaks listening. None of these appear in formal writing; all of them appear in almost every spoken sentence.

Verb + to

Modal + have — the single most misheard group, because have here sounds exactly like of, which is why natives themselves write "could of" by mistake.

Preposition and quantifier

Pronoun fusion — the question forms, where two or three words become one.

The disappearing /h/

In he, him, his, her and have, the /h/ drops whenever the word is unstressed — which is nearly always, because pronouns carry no new information. The remaining vowel then glues itself to the word in front.

This one is worth singling out because learners do not hear a reduced word — they hear a word that is simply not there, and then try to parse a sentence with a missing object.

T-flapping and the glottal stop

In American English, a /t/ between two vowels becomes a quick /d/-like tap.

• water → wa-der
• better → be-dder
• city → ci-ddy
• party → par-dy
• little → li-ddle
• get a → ge-da
• a lot of it → a lo-da-vit

Before a syllabic /n/, it disappears into a catch in the throat instead:

• button → bu'-'n
• mountain → moun-'n
• important → impor-'nt
• written → wri-'n

British English keeps the /t/ in water but glottalises it in other places — bottle as "bo'le", what as "wha'". Same mechanism, different distribution.

The two that cause the most damage

1. Can versus can't. Learners are taught to listen for the /t/. In natural American speech the /t/ in can't is usually unreleased — you cannot hear it. The real difference is vowel and stress:

• can is reduced and unstressed: /kən/, fast and flat. "I kən see it."
• can't keeps a full vowel and takes the stress: /kænt/. "I CAN'T see it."

So the rule is inverted from what it looks like on paper: if you clearly hear the word, it is probably can't. If it slips past almost unheard, it is can. Getting this backwards reverses the meaning of the sentence, which is why it is first on this list.

2. Of versus have. Both reduce to /əv/ or just /ə/. "I could've gone" and "I could of gone" are acoustically identical. There is no listening fix — only knowing that of cannot follow a modal, so it must be have.

How to train this

Reading a list gets you recognition. It does not get you the ear, because the ear needs the sound attached to the meaning, at speed, repeatedly.

Three things that work, in order of how much they cost you:

Narrow listening. Pick one speaker and stay with them for hours — one podcast host, one YouTuber, one actor. Reductions are partly individual; a single voice is a far smaller problem than "English" is, and once you have cracked one voice the next is easier.

Transcript-first, then audio. Read the line, then hear it, then hear it without reading. The gap between what you read and what you heard is exactly your reduction inventory, and it is different for everybody.

Shadowing. Say it along with the speaker, at their speed, copying the swallowing rather than correcting it. Producing a reduction is the fastest way to start hearing it, because your mouth learns the shape your ear is looking for.

What does not work: slowing the audio down. A reduction played at 0.75× turns back into the full form, so you practise hearing the thing you already understood.

Where Deep In fits

This is the problem Deep In was built for, so treat what follows as interested rather than neutral.

You paste a YouTube link — a podcast, an interview, a channel you already follow — and get a word-level transcript in English and in your own language, synced to the speech. When a phrase goes past as one sound, you tap it and Solomia, the AI in the app, tells you what the words were and why they came out that way: which reduction, what the full form is, whether it is regional.

It is English only, it assumes you already have a base, and after the first day it costs money — $18 a month for Basic, $29 for Deep. The first 24 hours need no account and no card, which is enough to find out whether this is your problem.

If it is not, the three techniques above work without us. The list on this page is the same list either way.

Related

• How to understand native English speakers — why the gap exists and what closes it
• How to learn English from YouTube videos — choosing material you will actually finish