Linking and Reduction: How Fast English Speech Works

Native English sounds fast mainly because words link into each other and unstressed words shrink, not because speakers talk at a dramatically higher speed. "What do you want to do?" is spoken as something close to "Whaddaya wanna do?" Every change in that sentence follows a pattern: weak vowels become schwa, t between vowels becomes a soft tap, and short words fuse. Once you know the patterns, speech that sounded like a blur breaks back into words.

This guide covers the main types of linking, the sound changes at word boundaries, the common reduced phrases, and drills for listening and speaking. The reference accent is General American, with British differences noted.

Why written words disappear in speech

English sentences have a rhythm built on stressed content words (nouns, main verbs, adjectives, adverbs) and unstressed function words (articles, pronouns, prepositions, auxiliaries). Stressed words get full vowels and clear consonants. Unstressed words are squeezed into the gaps between them, and squeezing produces every change below. In "I NEED to GO to the BANK," the three stressed words carry the message; the others shrink to tə, tə and thə.

The foundation of all reduction is the weak vowel, schwa, covered in our guide to the schwa sound. This article starts where that one ends: what happens between words.

Type 1: consonant to vowel

When a word ends in a consonant and the next starts with a vowel, the consonant is pronounced as if it began the next word. There is no pause and no glottal catch.

This is why "an ice cream" and "a nice cream" can sound identical, and why a few English words changed historically: "a napron" became "an apron" because speakers misjudged where the n belonged.

In General American, a final r always links, because r is always pronounced: "far away," "for ever." In non-rhotic accents such as standard southern British English, the r is silent at the end of "far" alone but reappears before a vowel ("far away" with an r), and some speakers even insert an r where none is written ("law and order" as "law-r-and order"). This inserted r is common in British speech but not typical of General American.

Type 2: vowel to vowel

When one word ends in a vowel sound and the next begins with one, English inserts a light glide so the two vowels do not crash together. Which glide depends on the first vowel.

You do not need to add these glides deliberately; they appear naturally when you move straight from one vowel to the next without stopping. The mistake to avoid is inserting a glottal stop, a tiny catch in the throat, which makes speech sound clipped.

Type 3: same or similar consonants

When one word ends and the next begins with the same consonant, English does not pronounce it twice. It holds a single, slightly longer consonant.

• bad dog: one long d, not bad-dog with a release in between
• big game: one long g
• stop pushing: one held p
• some money: one long m
• with this: one th
• gas station: one s

The same happens with similar pairs like t and d: in "hot dog," the t is not released before the d.

Type 4: sounds that change (assimilation)

Some sounds adjust to their neighbors. The most important changes for listening are these.

y after t, d, s and z

When a word ending in t, d, s or z is followed by you, your or year, the two sounds fuse into a new one. This is the source of many famous reductions.

n before p, b, m, k and g

An n takes the shape of the next consonant's position. Before p, b or m, it moves to the lips and becomes m; before k or g, it moves back and becomes ng.

• ten people: tem people
• in bed: im bed
• green bag: greem bag
• in case: ing case
• one game: wung game

Native speakers do not hear these changes as errors, and neither should you. If you say n carefully in every position, you will still be understood.

Type 5: sounds that disappear (elision)

t and d between consonants

When t or d sits between two other consonants, usually across a word boundary, it is often dropped.

This also explains why past-tense endings seem to vanish: "I asked Mary" and "I ask Mary" can sound almost the same, and listeners rely on context. For the reverse skill of producing those clusters clearly, see our guide to consonant clusters.

h in unstressed pronouns

The h of he, him, his, her and have is usually dropped when the word is unstressed and follows another word. The remaining vowel then links to the previous consonant.

• tell him: TEL-im
• ask her: AS-ker
• Did he go?: DID-ee go
• What's his name?: WHAT-siz name
• should have: SHOOD-əv, often just SHOOD-ə

At the start of a sentence, or when stressed, the h stays: "HE did it, not me."

Type 6: the American flap T

In General American, a t (and d) between vowel sounds, when the following vowel is unstressed, becomes a flap: a single quick tap of the tongue that sounds like a soft d. It happens inside words and across word boundaries.

The flap does not occur when the t starts a stressed syllable: in atTACK and reTURN the t is a full, aspirated t. British English generally keeps a clear t in water and better, or replaces it with a glottal stop in casual speech.

Also in American English, t after n is often dropped between vowels, so twenty can sound like TWEN-ee, internet like IN-er-net and wanted like WAN-id. Before a syllabic n, t becomes a glottal stop: button, mountain and written have a catch in the throat instead of a released t.

Common reduced phrases

These fused forms are extremely frequent in conversation, films and songs. Most are not written in formal English, but you need to recognize all of them.

One grammatical detail: wanna generally cannot replace want to when the word after want is the understood subject of the next verb. "Who do you want to see?" can become "Who do you wanna see?", but "Who do you want to win?" normally keeps want to, because the person you want is the one who wins.

Sentence-level examples

Here is how several changes combine in everyday sentences:

Drills for listening

Dictation with a twist. Choose a clip of ten to twenty seconds from a podcast or show with a transcript. Listen and write exactly what you hear, including fragments. Compare with the transcript and label each gap with the change that caused it: linking, flap, fusion, dropped sound or weak form. Over time you will see that a handful of labels account for most errors.

Spot the fusion. Listen specifically for did you, would you, don't you and got you, which are among the most frequent fusions. Tally each one you catch.

Gonna or going to. In any clip, note every future going to and whether it was reduced. You will find that the full form appears mainly when speakers are careful or emphatic.

Drills for speaking

Backward linking. Take a phrase like "pick it up." Say "up," then "tup," then "ki-tup," then "pi-ki-tup." Building from the end forces the consonants onto the next vowel.

Shadowing. Play a short clip and talk over it in real time, copying rhythm and pitch rather than individual words. Linking happens automatically when you keep pace.

Minimal effort sentences. Say "I want to go to the store" three ways: every word full, with weak forms only (I want tə go tə thə store), and fully casual (I wanna go tə thə store). Notice that the middle version is already natural enough for any situation.

Flap practice. Read these with a quick d-like tap: water, better, city, pretty, bottle, a lot of, get it, put it on, what about it.

Connected speech builds on clear single words and correct word stress, so if a phrase keeps breaking down, check the individual words first. Our guides to word stress and minimal pairs cover those layers, and each word of the day gives you a fresh word to link into a sentence of your own. In WordFren, each word pairs its audio with an example sentence, so you can hear it alone, say it, then try it inside a phrase.