Phonological history of English open back vowels

The phonology of the open back vowels of the English language has undergone changes both overall and with regional variations, through Old and Middle English to the present. The sounds heard in modern English were significantly influenced by the Great Vowel Shift, as well as more recent developments in some dialects such as the cot–caught merger.

Old and Middle English
In the Old English vowel system, the vowels in the open back area were unrounded:. There were also rounded back vowels of mid-height:. The corresponding spellings were $⟨a⟩$ and $⟨o⟩$, with the length distinctions not normally marked; in modern editions of Old English texts, the long vowels are often written $⟨ā⟩$, $⟨ō⟩$.

As the Old English (OE) system developed into that of Middle English (ME), the OE short vowel  merged with the fronted  to become a more central ME. Meanwhile, the OE long vowel was rounded and raised to ME. OE short remained relatively unchanged, becoming a short ME vowel regarded as, while OE long  became ME  (a higher vowel than ). Alternative developments were also possible; see English historical vowel correspondences for details.

Later, ME open syllable lengthening caused the short vowel to be normally changed to  in open syllables. Remaining instances of the short vowel also tended to become lower. Hence in Late Middle English (around 1400) the following open back vowels were present, distinguished by length:
 * , spelt $⟨o⟩$, as in dog, god
 * , often spelt $⟨oa⟩$, or $⟨o⟩$ before consonant+vowel or certain consonant pairs, as in boat, whole, old

16th-century changes
By 1600, the following changes had occurred:
 * The long vowel of boat had been raised to  as a result of the Great Vowel Shift.
 * The diphthong found in words such as cause, law, all, salt, psalm, half, change, chamber, dance had become an open back monophthong  or.
 * At this time, the short in dog was lowered to

There were thus two open back monophthongs:
 * as in lot
 * or as in cause

and one open back diphthong:
 * as in low

17th-century changes
By 1700, the following further developments had taken place:
 * The diphthong of soul was raised to, and then monophthongized to , merging with boat (see toe–tow merger). Before , this vowel further merged with  due to the horse–hoarse merger except in some varieties, as currently seen in Irish English, Scottish English and African American Vernacular English.
 * Short was retracted and rounded to . The shift was suppressed before a velar consonant, as in quack, twang, wag, wax, and also was suppressed by analogy in swam (the irregular past tense of swim). The change of  to  did not occur in Mid-Ulster English.
 * had begun to partake in lengthening and raising before a nonprevocalic voiceless fricative or /r/. That resulted in words like broth, cost and off having instead of, and was the start of the  split (see further below).
 * In words such as change and chamber, the pronunciation was gradually replaced in the standard language by a variant with, derived from Middle English . That explains the contemporary pronunciation of these words with.
 * However, when preceded, as in laugh and half,  was shifted to  instead, derived from Middle English.
 * An unrounded back vowel developed, found in certain classes of words that had previously had, like start, father and palm.

That left the standard form of the language with four open back vowels:
 * in lot and want.
 * in cloth and cost.
 * in start, father and palm.
 * in tor, cause, and corn.

Later changes
From the 18th century on, the following changes have occurred:
 * The three-way distinction between, , and was simplified in one of two ways:
 * In General American and old-fashioned RP, was raised to, merging with the vowel in  (the cloth-thought merger).
 * In many accents of England, the lengthening of the set was undone, restoring the short pronunciation . This became standard RP by the mid-20th century.
 * In General American, the lot vowel has become unrounded and merged into (the father–bother merger).

This leaves RP with three back vowels:
 * in lot, want, cloth, and cost.
 * in tor, cause, and corn.
 * in start, father, and palm.

and General American with two:
 * in lot, want, start, father, and palm.
 * in tor, cause, corn, cloth and cost.

Unrounded
In a few varieties of English, the vowel in lot is unrounded, pronounced toward []. This is found in the following dialects:
 * Most of Irish English
 * Much of the Caribbean
 * Norwich
 * The West Country and the West Midlands of England
 * Most of North American English
 * Excluding northeastern New England English and Western Pennsylvania English, where it is typically raised toward, merging with the vowel in thought.

There's also evidence for it in South East England as early as the late 16th century and as late as the 19th century.

Linguists disagree as to whether the unrounding of the lot vowel occurred independently in North America (probably occurring around the end of the 17th century) or was imported from certain types of speech current in Britain at that time.

In such accents outside of North America, lot typically is pronounced as, therefore being kept distinct from the vowel in palm, pronounced or. However, the major exception to this is North American English, where the vowel is lengthened to merge with the vowel in palm, as described below. This merger is called the  merger or more commonly the father–bother merger. (See further below.)

Father–bother merger
The father–bother merger is a phonemic merger of the lexical sets and. It represents unrounded lot, as detailed above, taken a step further. On top of being unrounded, the length distinction between the vowel in lot and bother and the vowel in palm and father is lost, so that the two groups merge. This causes father and bother to become rhymes.

This occurs in the great majority of North American accents; of the North American dialects that have unrounded lot, the only notable exception to the merger is New York City English, where the opposition with the -type vowel is somewhat tenuous.

Examples of possible homophones resulting from the merger include Khan and con as well as Saab and sob.

While the accents in northeastern New England, such as the Boston accent, also remain unmerged, lot remains rounded and merges instead with cloth and thought.

split
The  split is the result of a late 17th-century sound change that lengthened to  before voiceless fricatives, and also before  in the words gone and sometimes on. It was ultimately raised and merged with of words like thought, although in some accents that vowel is actually open. This means that is not a separate vowel; rather, it means "either  or, depending on the accent". The sound change is most consistent in the last syllable of a word, and much less so elsewhere (see below). Some words that entered the language later, especially when used more in writing than speech, are exempt from the lengthening, e.g. joss and Goth with the short vowel. Similar changes took place in words with $⟨a⟩$; see trap–bath split and /æ/-tensing.

The cot–caught merger, discussed below, has removed the distinction in some dialects.

As a result of the lengthening and raising, in the above-mentioned accents cross rhymes with sauce, and soft and cloth also have the vowel. Accents affected by this change include American English accents that lack the cot-caught merger and, formerly, RP, although today words of this group almost always have short in RP.

The lengthening and raising generally happened before the fricatives, and. In American English, the raising was extended to the environment before velars and, and sometimes before  as well, giving pronunciations like  for long,  for dog and  for chocolate.

In the varieties of American English that have the lot–cloth split, the lot vowel is usually symbolized as, often called the "short o" for historical reasons, as the corresponding RP vowel is still short (and it contrasts with  as in father and start). The thought vowel is usually transcribed as and it is often called the "open o". Its actual phonetic realization may be open, whereas the lot vowel may be realized as central. Some words vary as to which vowel they have. For example, words that end in -og like frog, hog, fog, log, bog etc. have in some accents and  in others.

There are also significant complexities in the pronunciation of written o occurring before one of the triggering phonemes in a non-final syllable. In other cases, however, the use of the open o as opposed to the short o is largely predictable. Just like with /æ/-tensing and the trap–bath split, there seems to be an open-syllable constraint. Namely, the change did not affect words with /ɑ/ in open syllables unless they were closely derived from words with in closed syllables. Hence occurs in crossing, crosser, crosses because it occurs in cross. In contrast, possible, jostle, impostor, profit, Gothic, and boggle all have. However, there are still exceptions in words like Boston and foster. A further list of words is mentioned in the table below:

Some words may vary depending on the speaker like (coffee, offer, donkey, soggy, boondoggle, etc. with either or ). Meanwhile, other words vary by region. For example, the word on, which in Northern American English dialects without the cot-caught merger is pronounced, rhyming with don, but in Midland and Southern American English without the merger is pronounced , rhyming with dawn. The isogloss for this difference, termed the ON line, lies between New York City and Philadelphia on the East Coast and runs West as far as speakers without the merger can be found.

Cot–caught merger
The cot–caught merger (also known as the low back merger or the lexical set merger) is a phonemic merger occurring in many accents of English, where the vowel sound in words like cot, nod, and stock (the vowel), has merged with that of caught, gnawed, and stalk (the  vowel). For example, with the merger, cot and caught become perfect homophones.

merger
The  merger is a merger of the English vowels of  and   that has been reported in Geordie since the late 20th century, with a quality around [oː]. The merger is more common among younger female speakers.

The merger also exists among older speakers in Bradford English with a quality around [ɔː], but younger speakers are more likely to resist the merger by fronting the vowel.

split
In some London accents of English, the vowel in words such as thought, force, and north, which merged earlier on in these varieties of English, undergoes a conditional split based on syllable structure: closed syllables have a higher vowel quality such as (possibly even  in broad Cockney varieties), and open syllables have a lower vowel quality  or a centering diphthong.

Originally-open syllables with an inflectional suffix (such as bored) retain the lower vowel quality, creating minimal pairs such as bored vs. board.

In broad Geordie, some words (roughly, those spelled with a, as in walk and talk) have  (which phonetically is the long counterpart of  ) instead of the standard. Those are the traditional dialect forms which are being replaced with the standard. is therefore not necessarily a distinct phoneme in the vowel system of Geordie, also because it occurs as an allophone of before voiced consonants.

Distribution of /ɑː/
The distribution of the vowel transcribed with ⟨ɑː⟩ in broad IPA varies greatly among dialects. It corresponds to, , and (when not prevocalic within the same word)  and even  in other dialects:
 * In non-rhotic dialects spoken outside of North America, corresponds mostly to  in General American and so is most often spelled $⟨ar⟩$. In dialects with the trap–bath split (such as Received Pronunciation, New Zealand English and South African English), it also corresponds to GA, which means that it can also be spelled $⟨a⟩$ before voiceless fricatives. In those dialects,  and  are separate phonemes.
 * In native words, in most non-rhotic speech of North America corresponds to both  in GA (RP ) and  in RP, as those dialects feature the father–bother merger.
 * North American English features the father–bother merger, where often corresponds to  in RP. Only New York City English and New England English distinguish between the two, and with an unrounded  vowel. Modern-day New York City English also has rounded  for reflexes of, which ironically, came from an unrounded vowel.
 * The is the cot–caught merger usually occurs in addition to the father–bother merger. This applies to almost all of Canadian English and many varieties of American English. The result is usually, the vowel, which is used for  as well. Some dialects will have  as the merged vowel, not ; these include Standard Canadian English or Pittsburgh English.
 * The caught-cot merger without the father–bother merger is found in Scottish English and most of New England English. Uniquely, Scottish English merges the vowel into the  vowel, rather than the other way around.
 * In loanwords and names, the open central unrounded vowel in a source language is regularly approximated with  in North America and  in RP. However, in the case of mid back rounded vowels spelled $⟨o⟩$, the usual North American approximation is, not  (in RP, it can be either  or ). However, when the vowel is both stressed and word-final, the only possibilities in RP are  in the first case and  in the latter case, mirroring GA.

For the sake of simplicity, instances of an unrounded vowel (phonetically ) that do not merge with / are excluded from the table below. For this reason, the traditional Norfolk dialect is included but the contemporary one, nor the Cardiff dialect, are not.

Fronted /oʊ/
In many dialects of English, the vowel has undergone fronting. The exact phonetic value varies. Dialects with the fronted include Received Pronunciation; Southern, Midland, and Mid-Atlantic American English; and Australian English. This fronting does not generally occur before, a relatively retracted consonant.