You are standing at a corner asking for directions, and you say the phrase bus stop. You say it carefully: bus, then stop. Two clean words, two clean /s/ sounds. The person understands you, but something in their face says you sound like a recording being read aloud.
Then the reverse happens. Someone says "that's a nice sight" and you hear "that's an ice sight". You did not mishear the consonants. You missed the one thing that marked the boundary between the words, because it was not a consonant at all. It was length.
Native speakers of American English do something at these joints that almost no course teaches directly, and it changes both how you sound and how much you understand.
The rule
When a word ends with the same sound that the next word begins with, English speakers do not say the sound twice: they say one consonant and hold it about twice as long.
So bus stop is /ˈbʌsːtɑp/, with a single /s/ that lasts roughly double. Black coat is one long held /k/. The same goes for this seat, big game, some more, half fixed, next time, hot toast, feel like, and one night. One articulation, stretched.
This matters in both directions. When you produce two separate consonants, you insert a gap that native speakers never make, and the phrase sounds slow and mechanical. When you listen, you are waiting for two consonants that will never come, so the boundary vanishes and words run together in your ear. Length is the boundary cue. Nothing else marks it.
Why it works this way
English is stress-timed. Stressed syllables land at roughly even intervals, and everything between them gets compressed. Saying a consonant twice costs an extra beat that the rhythm has no room for, so the language solved it by keeping one gesture and stretching its duration instead.
The physical mechanics split by consonant type.
Stops: /p t k b d ɡ/
Close once, hold the closure in silence, release once. For black coat, the back of your tongue rises to the soft palate for the /k/ of black, and it simply stays there. No air escapes. You hold that silence for about a beat, then release directly into the vowel of coat. There is exactly one puff of air in the whole phrase. If you hear two small explosions, you released twice.
That silent hold is the part learners skip. It feels like nothing is happening, so the instinct is to fill it. Resist that. The silence carries the information.
Fricatives, nasals and liquids: /s z f v m n l/
These are continuous, so there is nothing to hold silently. You just keep the friction or the hum running for about twice as long. In bus stop, the hiss never stops between the words; it lengthens. In some more, your lips stay closed and the nasal hum continues. In feel like, the tongue tip stays on the ridge behind your teeth. These long consonants are clearly audible, and they are the easiest place to start practicing.
What the length is doing for the listener
| Phrase | Long consonant | Sounds like this if you shorten it |
|---|---|---|
| bus stop | long /s/ | bus top |
| stop pushing | long held /p/ | stop ushing |
| life force | long /f/ | life orce |
| some money | long /m/ | so money |
| white tie | long held /t/ | why tie |
| a nice sight | long /s/ | an ice sight |
| ten nights | long /n/ | ten ights |
| cheap price | long held /p/ | chea price |
Read the middle column and the right column out loud one after the other. The only difference you are producing is duration, and yet the meaning moves.
The exceptions
1. Affricates never merge
/tʃ/ and /dʒ/ keep their identity. Which chair has two separate /tʃ/ sounds, and orange juice has two separate /dʒ/ sounds. You must articulate both. An affricate is a stop plus a release burst, and the burst cannot be stretched, so merging would delete a whole sound.
2. /h/ does not merge and does not lengthen
High heels keeps two distinct /h/ sounds. /h/ is just breath through an open throat; a long breath is not perceptible as a doubled consonant, so English restarts it.
3. Similar is not the same: that is assimilation, not gemination
In good boy, /d/ meets /b/. These are different sounds, so nothing lengthens. Instead the /d/ is left unreleased and the phrase comes out as a single held /b/ closure. Gemination means one sound made long; assimilation means one sound reshaped by its neighbor. Same feel in the mouth, different mechanism, and it is worth keeping them apart in your head.
4. Deliberate or contrastive speech overrides the rule
If you are correcting someone ("I said BUS stop, not bus STATION"), you are allowed to separate the words fully. The rule describes connected speech at normal speed, not emphatic speech.
5. Doubled letters inside a word are never long sounds
Letter, running, and summer have one short consonant each, and in American English the double t in letter is actually a quick flap /ˈlɛɾɚ/. The rule is about sound boundaries between words, not about spelling.
The drill
- Say bus stop and stretch the /s/ deliberately: "busssstop". Exaggerate first, then shorten until it feels natural. Answer: one hiss, never two.
- Say black coat in three steps: close the /k/, count one silent beat in your head, release into "oat". Answer: one puff of air only.
- Alternate a nice sight and an ice sight ten times. Record it and check that you can tell them apart on playback.
- Now test the exceptions: say which chair and high heels and make sure you hear two clear starts in each.
- Put it in a sentence at speed: "The big game is one night after the bus stop closes." Aim for the same rhythm, not slower.
Length is a grammatical signal in English, not decoration. Once you hold these consonants, your speech stops sounding word-by-word and your listening comprehension of fast speech improves at the same time. Take it further with sentence-level practice, and sharpen your ear on minimal pairs.