Connected Speech: Linking, Elision, Assimilation, Weak Forms

In natural English the words do not stay separate. Sounds link across the gaps between them, some disappear, some change to match their neighbours, and the small grammar words shrink. Next please is said /nɛks pliːz/, with the t gone entirely. That set of changes is called connected speech, and it is the main reason spoken English sounds so different from the written version.

None of it is careless. These are the ordinary productions of careful speakers at normal speed, and they follow patterns you can learn. Type any phrase into the pronunciation Studio to hear how it sounds run together.

The four things that happen

Connected speech is not one process. It is four, and they can all apply to the same phrase.

Linking joins the end of one word to the start of the next, so there is no silence between them.

Elision removes a sound entirely, usually a /t/ or /d/ caught between other consonants.

Assimilation changes a sound so that it matches its neighbour more closely, which saves the mouth a movement.

Weak forms reduce common grammar words to a shorter, vaguer version, usually built around the schwa.

Taken together they explain most of the gap between what you read and what you hear. They also explain why the same sentence sounds slower and clearer when someone repeats it for you: repeating it removes most of these processes and restores the citation forms.

Linking

English does not put a gap between words. The sounds run straight into each other, and where a word ends in a consonant and the next begins with a vowel, the consonant effectively moves across.

That last column is deliberately re-syllabified to show what your ear actually receives. This is why a phrase you know perfectly well in writing can be unrecognisable when spoken: the boundaries are not where you expect them.

Linking r and intrusive r

In non-rhotic accents, including standard British English, a written r at the end of a word is silent, so far is /fɑː/. But when the next word starts with a vowel, that r comes back to do the linking job: far away is /fɑː rəˈweɪ/.

The same thing then happens where there is no written r at all. Law and order is commonly said /lɔː rənd ˈɔːdə/, and media event as /ˈmiːdiə rɪˈvɛnt/. This is called intrusive r, it is completely standard in ordinary British speech, and it is a natural consequence of the linking pattern rather than an error. Some careful speakers avoid it in formal contexts.

Glide linking

Where the first word ends in a vowel and the next begins with one, English inserts a small /j/ or /w/ to bridge them.

After the vowels /iː/, /ɪ/, /eɪ/, /aɪ/ and /ɔɪ/, a /j/ appears: she is becomes /ʃiː jɪz/, my own becomes /maɪ jəʊn/.

After /uː/, /ʊ/, /əʊ/ and /aʊ/, a /w/ appears: go on becomes /ɡəʊ wɒn/, do it becomes /duː wɪt/, how often becomes /haʊ wˈɒfən/.

You almost certainly do this already if you speak English at all fluently. Naming it is useful mainly for listening, since it explains sounds arriving that no spelling accounts for.

Elision

Elision is a sound dropping out. English does it most often with /t/ and /d/ when they are trapped between two other consonants.

The rule is narrow and reliable: a /t/ or /d/ at the end of a consonant cluster drops when the next word starts with a consonant. It stays when the next word starts with a vowel, which is why last night loses its t but last order keeps it.

Whole vowels disappear too, especially the schwa in an unstressed syllable next to a liquid consonant. Camera becomes /ˈkæmrə/, interesting becomes /ˈɪntrəstɪŋ/, comfortable becomes /ˈkʌmftəbəl/, police is often /pliːs/ in fast speech. Those first three are so established that dictionaries give them as the main form, which is why they appear in the difficult words guide as syllable-count traps.

The /h/ in weak pronouns goes the same way. Tell him becomes /ˈtɛl ɪm/, ask her becomes /ˈɑːsk ə/, give them becomes /ˈɡɪv əm/. The h survives when the word is stressed or starts a sentence.

Assimilation

Assimilation is a sound changing to become more like its neighbour. The commonest kind in English works backwards: a sound takes on the place of articulation of the sound that follows it.

Say ten bottles quickly and honestly, and you will find your lips are already closed for the b while you are still making the n. That is all assimilation is: the mouth starting the next movement early.

Coalescence

A special case, and the one that produces the most recognisable effects. When /t/, /d/, /s/ or /z/ meets a following /j/, the two merge into a single new sound.

This is the same process that produced nature /ˈneɪtʃə/ and soldier /ˈsəʊldʒə/ inside single words, centuries ago. It is still running, now across word boundaries.

Weak forms

Around forty English words have two pronunciations: a full one used when the word is stressed or said alone, and a weak one used everywhere else. The weak form is usually built on a schwa. Because these are the most frequent words in the language, weak forms account for a huge share of what makes spoken English hard to follow.

There is one reliable exception. A weak-form word takes its strong form at the end of a clause: What are you looking at? ends with /æt/, not /ət/. The same applies when the word is being contrasted: I said to him, not from him.

Schwa does so much of this work that the schwa guide is the natural companion to this section, and it covers the vowel itself in far more detail.

Why this makes listening so hard

Most learners are taught words one at a time, in their citation form, and then meet English in which almost none of those forms appear. The vocabulary is known and the sentence is still unintelligible.

Three specific things break listening.

The word boundaries are in the wrong place. Pick it up arrives as three syllables that divide as pi-ki-tup. You are listening for a gap before it and there is no gap anywhere.

The words you rely on for grammar are the quietest ones. Articles, prepositions, auxiliaries and pronouns are exactly the words that reduce to a schwa. He could have gone can arrive as /i kʊdəv ɡɒn/, where three of the four words have lost most of their identity.

Sounds you are listening for are not there. If you have learned that next ends in a t, and the t is gone, your ear rejects the word and spends time looking for it while the sentence moves on.

The fix is not more vocabulary. It is learning the connected forms of the two hundred or so most frequent words, which is a much smaller job. Once would you arrives as /ˈwʊdʒu/ without surprising you, a large amount of ordinary speech becomes transparent at once.

A useful exercise is to take a short clip of natural speech, transcribe what you actually hear rather than what you know is there, then compare with the script. The gaps in your transcription are a precise map of which processes you have not internalised yet.

Does connected speech differ between accents?

The four processes are shared by every major variety of English. What differs is the detail.

Linking r and intrusive r are features of non-rhotic accents, which includes standard British, Australian and much of the English of southern England. Rhotic accents such as General American pronounce every written r anyway, so linking r has nothing to do and intrusive r does not arise.

Elision of t and d happens everywhere, and it is if anything more extensive in casual American speech.

Tapping is an American speciality. Get up and a lot of produce a fast tap in American English, so get up sounds like /ɡɛɾʌp/. British English generally keeps a full /t/, though many British speakers use a glottal stop instead.

Weak forms are shared, with small differences. Been has a weak form /bɪn/ in British English that is the normal form in American English.

Coalescence is universal, though British speakers use it slightly more freely than American speakers in formal registers.

The practical point is that connected speech is not an optional extra of one accent. Every fluent speaker of every variety does all of this. Choose the accent you are aiming for, then learn its version.

Careful speech and conversational speech

None of this is compulsory. English exists on a scale, and speakers move along it depending on the situation.

All three are correct English. The first would sound strange in a café and appropriate in a courtroom. A competent speaker moves between them without thinking, and the skill worth building is recognising all three when you hear them, not producing the fastest one.

For learners, the priority order is clear: understand the reduced forms first, because you will meet them constantly, and produce them second, once your individual sounds are reliable.

Practice sentences

Say each of these twice: once word by word, then once at conversational speed. The second version should sound noticeably different.

1. What do you want to do?
2. I should have asked her about it.
3. He kept quiet last night.
4. Can you give them a hand?
5. That cup of tea would be nice.
6. It’s the most common mistake I see.

How to build it into your own speech

Producing connected speech is a later skill than recognising it, and trying to force it early usually makes things worse rather than better. A blurred sentence is harder to follow than a slightly formal one.

A sensible order looks like this.

1. Get the individual sounds reliable. If /θ/ and /s/ are still merging, linking words together will bury the problem rather than fix it.
2. Get word stress right. Connected speech is built on the stressed and unstressed pattern. Without it, there is nothing for the weak forms to be weak against, and the word stress guide covers this first step.
3. Add weak forms, a few at a time. Start with to, and, of, for and a. These five appear in almost every sentence and reducing them alone changes the rhythm noticeably.
4. Add consonant-to-vowel linking. Practise phrases like turn off, pick it up, not at all until there is no gap in the middle.
5. Add elision last. Dropping sounds is the most advanced of the four, because dropping the wrong one produces a genuinely unclear word rather than a natural one.

Shadowing is the most efficient drill for all of this. Play a short recording, say it along with the speaker at the same time and speed, and repeat the same twenty seconds until you match. Doing that with the same clip ten times teaches more than doing it with ten different clips once.

Two habits to avoid. Do not aim for the fastest possible version, since native speakers vary their speed constantly and the fast form is not the target. And do not practise connected speech from a written transcript alone: the whole point is that the writing does not show what happens, so you need audio.

What this guide does not cover

Connected speech is about what happens to sounds at the joins. A separate system decides which whole words get emphasised in the first place, and that changes meaning rather than just texture. I didn’t say he stole it means six different things depending on which word you stress. That is covered in the sentence stress guide.

The two interact, since unstressed words are exactly the ones that reduce, but they answer different questions. Sentence stress asks which words matter. Connected speech asks what happens to the ones that do not.

Where consonants pile up at a join, the difficulty is physical rather than a matter of pattern, and the consonant clusters guide covers how English handles those runs. To hear any phrase on this page, type it into the Studio and play it in either accent.