Cot–Caught Merger: Which American Vowel System Is Your Ear On?

Look up caught in Cambridge's US dictionary. The transcription is /kɑːt/. Now look up cot on the same site: /kɑːt/. The same. Scroll down that caught page to the Cambridge Academic Content Dictionary block, still the same URL, and it gives you /kɔt/ against cot at /kɑt/. Different.

One publisher, one page, two answers. You did not mislearn this vowel. The cot–caught merger has split American English into two vowel systems, the reference book prints both, and nobody told you which one you were being taught.

Here is which regions use which, how to find out which system your own ear was trained on, and why the answer for your own mouth is probably to stop making the distinction at all.

What the cot–caught merger actually is

Say Don and dawn out loud, then stock and stalk. If each pair came out as one word said twice, you have the merger. If your tongue moved back and your lips rounded a little on the second word, you don't.

Dictionaries print two symbols for these vowels. /ɑ/ is the vowel of cot, hot, box, the one phoneticians call LOT. /ɔ/ is the vowel of caught, thought, law, the one called THOUGHT. Linguists call their collapse the low back merger, and you will not hear anyone say that out loud.

This is not the same thing as the father–bother merger, which people mix it up with constantly. That one collapsed the vowel of spa into the vowel of spot, it happened to most of America long before this, and it is the reason American dictionaries print /ɑ/ for cot where British ones print /ɒ/. British English keeps all three apart: cot is /kɒt/ and caught is /kɔːt/, which is covered in more detail in what actually separates British and American pronunciation.

Half of what? The number everyone repeats wrong

Search this topic and you will be told, on page after page, that half of Americans merge the two. The number is real. It is describing something else.

What the Atlas of North American English says is that the merger "does not define any one dialect region but embraces several of the regions to be defined in Chapter 11, including roughly half of the geographic territory that the Atlas covers." Territory. Not people. The territory inside that line is Canada, the West, Eastern New England and western Pennsylvania, which is a great deal of map and not a great many Americans. The regions that hold out hardest are the Inland North and the corridor from New York City down through Philadelphia to Baltimore, and those are full of people.

A population figure does exist. The 2003 Harvard Dialect Survey asked 11,050 people whether they pronounce cot and caught the same, and 39.07% said yes against 60.93% who said no. Two things to know before you quote it: anyone could fill it in, so it is not a random sample of the country, and it asked people to judge their own speech rather than recording it. That last point turns out to matter more than it sounds like it should.

The cot–caught merger map, read city by city

New York City returned 0% "same" judgements in the Atlas survey. Canada returned 87%. Everything else sits between those two.

The Atlas ranks its regions on a scale where 0 is a complete merger and 4 is a complete distinction. New York City has no row of its own, because the Atlas scores it separately, at the 0% above:

The four regions in the middle of that table are the honest answer to "is it just a western thing". They are not: the merger arose independently in Eastern New England, in western Pennsylvania and in the West, three places with nothing to do with each other, and the South, the Midland and the North sit in a broad band where between 16% and 50% of speakers hear the two words as the same.

Distance is a bad guide to any of this. Boston merges. Providence, fifty miles down the road, does not, and the Atlas goes out of its way to correct the older dialect literature on the point, noting that earlier surveys "were wrong in attributing the low back merger to Providence." All six Providence speakers in the survey kept the two vowels apart. If you are in the Bay Area or Los Angeles, expect a mess: the Penn atlas describes the merger as "still in progress in the West", with "considerable variation" in exactly those cities.

Why Boston and Seattle both merge and sound nothing alike

Boston says cot and caught identically, and a Bostonian saying them sounds nothing like someone from Seattle saying them. The merged vowel is not the same vowel.

Wiktionary marks cot as /kɒt/ under the label Boston and /kɑt/ for General American, and Boston never had the father–bother merger, so cart keeps a vowel of its own there.

The South is stranger again. The Atlas reports that there, "/o/ and /oh/ both show the same low back rounded nucleus, but /oh/ is distinguished from /o/ by the presence of a back upglide." Southern caught is not a different vowel from cot. It is the same vowel with a glide stuck on the end, closer to "caw-ut". The two regions that resist hardest do so by opposite moves: the Inland North pushes cot forward, and the Mid-Atlantic lifts caught upward, giving New York City the widest gap between the two of anywhere measured.

A merger is one variable. It is not an accent.

So should you merge? Yes, and here is what it costs

Merge. If you are learning American English and you have a choice, take the one vowel.

Nobody has run the study that would settle that, so take what follows as an argument rather than a finding. The nearest thing to a verdict in print comes from Levis and Challis, whose open textbook for pronunciation teachers says the merger "almost never causes misunderstandings, so it is fine for language learners to use whichever vowel is most comfortable for them." That licenses picking one system rather than straddling both.

The contrast lives in the lexicon: you learn it word by word or you do not learn it at all. No rule tells you which words take /ɔ/. English spelling never encoded it, because the THOUGHT class was assembled out of at least eight unrelated historical sources, and the Atlas notes in a footnote that the vowel "is frequently ignored in phonics texts, partly as the result of the fact that there is no single spelling pattern that serves as a cue for readers." Native children are not taught it either. So a learner who drills the contrast and then applies it word by word from guesswork gets it right some of the time and wrong the rest, and the result is an accent that belongs to no city in North America.

The comprehension cost of merging is close to nothing, and there is a real number behind that. Of 24 recorded natural misunderstandings that turned on these two vowels, collected by Labov and reported in the Atlas, 23 were on the part of two-phoneme speakers rather than one-phoneme speakers. The people who get caught out are the ones who keep the two vowels apart. If you merge, you resolve by context, the way a merged American does, and context does the work almost every time.

Almost. Four cases genuinely collide, and cot versus caught is not one of them, because one is a noun and the other is a past participle and they never compete for the same slot. The ones that do:

• Don and Dawn. Two first names, same sentence position, nothing to disambiguate them. "Don's here" and "Dawn's here" are a real problem, which is why the Atlas used the pair to test for the merger in the first place.
• collar and caller. Both nouns, and both turn up on a phone call.
• stock and stalk. Both nouns and both verbs, and both live in business English.
• "at dawn" and "at Don's". A time and a place, in the same slot, after the same preposition.

Nobody will think less of you for merging. The Atlas is explicit that the merger "takes place below the level of social awareness and is normally not the focus of sociolinguistic evaluation." Americans do not hear it as an accent. They mostly do not hear it at all.

One honest gap: nobody has measured whether merging makes you misspell stalk for stock. It is plausible and it is not established, so keep reading carefully and do not count it against merging.

Test your own ear: start with sock and talk, not Don and dawn

The usual advice is to say cot and caught and listen, and Wikipedia runs several minimal pairs together in one sentence and leaves you to it. Both ask you to hear a difference, which is the exact thing you cannot do if you merge.

The merger does not arrive everywhere in a word at once. The Atlas found it "is favored in syllables closed by nasal consonants" and that it is "more advanced before nasals than in any other environment," while the consonant that holds out longest is /k/. In the Indianapolis and Columbus sample, 50% of responses before /k/ were still completely distinct.

So the pairs are not equivalent, and no page on the internet orders them:

1. Don / dawn (before /n/) merges first. A speaker who keeps the contrast everywhere else may still merge here.
2. hot / caught (before /t/) is next.
3. dollar / caller (before /l/).
4. sock / talk (before /k/) holds out longest. This is your diagnostic.

Test yourself on sock and talk. If you get one answer on Don/dawn and the opposite on sock/talk, you have not made a mistake. You have found the seam.

Then test your ear and your mouth separately, because they come apart. Record yourself saying all four pairs and listen back a day later, or have someone else listen. Introspection is unreliable here for a specific reason: this change runs below the level of awareness, so asking yourself what you just said is asking the wrong instrument.

The harder half: your ear has to handle both systems

Across the whole Atlas sample, about 105 speakers were further along with the merger in what they could hear than in what they said, against 51 the other way round. Outside the merged region the ratio is 3 to 1 in favor of hearing.

That finding reframes the whole problem for a learner. Every article on this topic treats the merger as a question about your mouth, and you can settle that question in an afternoon by deciding to merge. Your ear does not get the same luxury. The person across the table may be producing a distinction that their own ear no longer reliably registers, or keeping two vowels apart that you now hear as one, and you have to follow either of them. You do not get to pick one system to understand.

This lands hardest on one particular reader: the one who learned from British materials. That ear is tuned for a rounded /ɒ/ in hot, box and dog, so an American hot already sounds wrong, and now there are two American answers rather than one. Three systems, and the coursebook showed you two. If American fast speech is also swallowing whole syllables on you, that is a separate problem with its own rules.

The words Americans themselves argue about

Frog is the sharpest example. Wiktionary lists both /fɹɔɡ/ and /fɹɑɡ/ for it under the non-cot-caught label, which means unmerged Americans, the ones who supposedly have a clean two-vowel system, disagree with each other about which vowel the word takes.

Frog is not alone. A group of short-o words drifted into the THOUGHT class before voiceless fricatives and back nasals, and the drift was patchy. The Atlas singles out dog as "the most commonly found shifted" and reports "extensive dialect variation in other words before /g/: fog, log, hog, frog." Wiktionary tags a merged and an unmerged variant for each of the words below, which are where Americans divide, roughly in order of how often you will meet them: on, dog, off, long, gone, soft, cloth, coffee, loss, song, chocolate.

No rule covers them, and the reason is worth knowing. The shift spread word by word rather than by a sound law, so uncommon words like Goth and toff were skipped entirely, and longer words like hospital, toggle and soggy stayed put. Spelling will not save you here any more than it will with the silent letters English keeps for decoration.

If you merge, all of this evaporates.

The map is thirty years old and the merger kept moving

The Atlas fieldwork was done by telephone in the 1990s, and the merger has been moving the whole time since. The Atlas measured the drift itself: on that 0-to-4 scale, the merger advances 0.43 units for every 25 years of age, with significant effects in Eastern New England, the West, the Midland and the South. "Only in Canada is the merger well enough established to show no correlation with age."

Two studies since have watched the line move. Aaron Dinkin's "Weakening resistance: Progress toward the low back merger in New York State" looked at Upstate New York, deep inside the Inland North, and found the region trending toward the merger in apparent time. And in 2025, Lelia Glass published work in American Speech on young Asian Americans in Atlanta, who show "the greatest overlap between /ɑ/ and /ɔ/" among the speakers she studied, and who are at the front of this change in the South, where the Atlas had found almost no complete merger at all.

So the table above is a floor. If the person you are talking to is under thirty, shade every "distinct" region toward merged.

Next American voice you hear, skip cot and caught. Listen for sock and talk, and then for Don and dawn, and notice whether the same person gives you the same answer on both. That gap tells you more about where they grew up than the vowel itself does.