Connected speech practice

When English words are spoken in isolation, each one sounds clear and complete. When they are spoken in a sentence, they merge, reduce, and reshape each other. This is connected speech — the natural process that makes fluent English sound like a continuous stream rather than a sequence of separate words. It is the main reason listening to English feels harder than reading it: the words you learned individually do not sound the same when they are strung together.

Connected speech is not sloppy or informal. It is the standard, correct way to speak English at any speed above very careful dictation. A newsreader does it. A university lecturer does it. Learning to produce it is the difference between sounding like a textbook and sounding like a speaker. Learning to recognise it is the difference between understanding English and not understanding why you studied it.

Linking is when the last sound of one word joins the first sound of the next, so the word boundary disappears and the two words sound like one. English has three types:

Consonant to vowel is the most common and most important. When a word ends with a consonant and the next word begins with a vowel, the consonant jumps across: an apple becomes a·napple, turn it off becomes tur·ni·toff. The consonant that belongs to the first word is pronounced as if it belongs to the second.

Vowel to vowel requires a bridge. When two vowels sit next to each other at a word boundary, a small glide appears between them. After /aɪ/, /eɪ/, /ɔɪ/ and other front vowels, a /j/ glide appears: I am → I·yam. After /oʊ/, /uː/ and other back vowels, a /w/ glide appears: go away → go·waway. This is not an added sound — it is the natural transition between the two mouth positions.

Vowel to consonant is the simplest — the consonant simply begins where the vowel ends, with no gap. she asked → she·waskt. This one happens automatically for most speakers.

Elision is the dropping of a sound in connected speech. English tolerates elision more than most languages, and it follows patterns — the dropped sound is almost always one that would create a difficult consonant cluster. Knowing the patterns makes fast speech less mysterious.

The most common elision is /t/ after /s/ or /k/. In next day, the /t/ after /ks/ drops because saying /kst/ followed by /d/ is too many consonants. In just between, the /t/ after /st/ drops for the same reason. In best friend, the /t/ after /st/ drops.

Another common pattern is unstressed syllable reduction. When a syllable between two stressed syllables is unstressed, it often loses its vowel entirely. This is why comfortable becomes kumf-tuh-buhl and temperature becomes tem-pruh-chuh. The written syllable is there in the spelling; the spoken syllable is not.

Elision is more common in fast, casual speech and less common in slow, careful speech. It is not wrong to pronounce every sound — it just sounds slower and more deliberate than a native speaker's normal pace. The skill is recognising elision in others, not necessarily producing it yourself at full speed.

Assimilation is when one sound changes to become more like a neighbouring sound. The change is usually anticipatory — a sound changes to match the sound coming next, because the mouth starts moving toward the next position before it gets there.

The most common assimilation in English is /n/ becoming /m/ before /b/ or /p/. In ten bikes, the /n/ becomes /m/ because /m/ is made with the lips in the same position as /b/. Saying /tenbikes/ requires the tongue to move to the ridge for /n/ and then immediately to the lips for /b/ — too much movement for fast speech, so the tongue stays down and the lips do both.

The other major assimilation is /d/ or /t/ plus /j/ merging into /dʒ/ or /tʃ/. would you becomes wuj-you. got you becomes go-chyou. did you becomes dij-you. This is so regular that it is essentially obligatory in fast speech — failing to do it makes you sound as though you are reading each word separately.

In real speech, linking, elision and assimilation happen simultaneously and continuously. Take the sentence Would you like to come to the party?. In careful speech it is seven separate words. In connected speech it becomes something like:

/wʊdʒu laɪktə kʌmtə ðə pɑːti/ — "would you" assimilates, "like to" elides the /k/ and links, "come to" links across the boundary. Every word boundary has been reshaped.

This is not fast speech or lazy speech — it is normal speech. Understanding it requires training your ear to hear word boundaries where the sounds do not stop. The best way to train this is to produce it yourself: once you can say connected speech naturally, you can hear it naturally.

In non-rhotic accents (British RP, Australian, New Zealand), the letter r is not pronounced at the end of a word unless a vowel follows. When a vowel does follow, the r appears:

Linking r: far away → far·raway. The r is in the spelling and appears because a vowel follows. car engine → car·rengine. four apples → fou·rapples.

Intrusive r: The same sound appears even when there is no r in the spelling, between a word ending in a non-close vowel and a word beginning with a vowel: the idea·r·of it, law·r·and order, France·r·and England. This is standard in British English, though some speakers avoid it in formal contexts.

In rhotic accents (General American, Irish, Scottish), the r is always pronounced, so this distinction does not apply.

Connected speech does not just change how words link — it changes what the words sound like. Function words reduce to weak forms: to becomes /tə/, of becomes /əv/, and becomes /ən/, can becomes /kən/. These reductions are so consistent that the strong forms sound marked and emphatic in context.

The weak form is also the form that links. cup of tea is not three separate words — it is /kʌpəvtiː/, a single linked chain where "of" has reduced to /əv/ and the /v/ links onto "tea". This is why learning weak forms and learning linking are the same skill: the reduction creates the link.

Weak forms revert to strong forms when the word is at the end of a sentence, when it is contrasted, or when it is emphasised. What are you looking AT? — full /æt/ because it ends the sentence. I said OF, not FOR — full forms because they are contrasted.

Say these phrases at normal speed, focusing on how the words connect. Play each one, repeat it immediately, and then try it without the audio.

Shadowing. Play a sentence from a podcast or video and say it at the same time, matching the rhythm, stress and linking exactly. Do not try to understand every word — focus on matching the flow. Shadowing trains your mouth to make connected-speech changes automatically.

Slow it down, then speed it up. Take a phrase like not at all. Say it slowly: /nɒt/ /æt/ /ɔːl/. Then faster: /nɒtætɔːl/. Then at full speed: /nɒtætɔːl/. The links appear naturally once you speed up. The sounds are not new — they are what happens when familiar sounds meet.

Record and compare. Record yourself saying a phrase, then play it next to the model audio. If your version sounds choppy — distinct words with gaps between them — you need more linking. If it sounds smooth but the words are wrong, you need more accuracy on the individual sounds first.

Chain sentences. Take three or four short sentences and read them back to back without pausing. This forces linking across sentence boundaries and builds the habit of continuous speech. Start with sentences that end and begin with similar sounds — the links will be easier and the exercise less frustrating.

Minimal pairs for assimilation. Say ten bikes and ten times side by side. In the first, /n/ becomes /m/ before /b/. In the second, /n/ stays /n/ before /t/. This builds awareness of when assimilation happens and when it does not.

Pausing at every word boundary. Saying each word as a separate unit with a micro-pause between them. This is the most common marker of non-native speech. The fix is to identify the last consonant of each word and the first vowel of the next, and link them without stopping.

Over-pronouncing function words. Giving full, clear pronunciation to to , of , and and other function words. This creates extra syllables, breaks the rhythm, and makes the sentence harder to follow. Reduce them to their weak forms.

Adding pauses that do not exist. Saying I am going to the shops with pauses after each word. In connected speech, "I am going to" is one smooth chain: /aɪm ɡɒnə/. The word boundaries are there in the grammar but not in the sound.

Not recognising assimilation. Hearing would you as two separate words when a native speaker says /wʊdʒu/. If you cannot hear the /dʒ/, you will have trouble understanding fast speech where this assimilation is constant.

Linking words · Weak form words · Elision words · Assimilation words · Connected speech phrases

Individual words worth a slow look: not, and, to, of, can, would, got, did.

Connected speech — overview — the theory behind linking, elision and assimilation. The schwa — the vowel that powers every weak form. Sentence stress — which syllables stay strong and which get reduced. Word stress — which syllable in each word is stressed.

Read slowly first. Pronounce each syllable deliberately before trying to blend them at full speed. Rushing hides the vowel and consonant shapes your ear needs to learn.

Record yourself. Most phones have a voice memo app — record a word, then compare it to the playback here. Hearing the gap between your version and the model accent makes targeted practice far easier.

Focus on one accent. Jumping between American, British and Australian in the same session creates confusion. Pick the accent you encounter most and stick with it until the patterns feel natural.

Every word above is a button. Set the accent and voice you want on the main screen, then come back — the guides use whatever you chose. To hear two accents against each other, use Compare; to test whether you can actually hear a contrast, use the listening quiz.