English-language vowel changes before historic /r/

In English, many vowel shifts affect only vowels followed by in rhotic dialects, or vowels that were historically followed by  that has been elided in non-rhotic dialects. Most of them involve the merging of vowel distinctions and so fewer vowel phonemes occur before than in other positions of a word.

Overview
In rhotic dialects, is pronounced in most cases. In General American English (GA), is pronounced as an approximant  or  in most positions, but after some vowels, it is pronounced as r-coloring. In Scottish English, is traditionally pronounced as a flap  or trill, and there are no r-colored vowels.

In non-rhotic dialects like Received Pronunciation (RP), historic is elided at the end of a syllable, and if the preceding vowel is stressed, it undergoes compensatory lengthening or breaking (diphthongization). Thus, words that historically had often have long vowels or centering diphthongs ending in a schwa, or a diphthong followed by a schwa.
 * earth: GA, RP
 * here: GA, RP
 * fire: GA, RP

In most English dialects, there are vowel shifts that affect only vowels before or vowels that were historically followed by. Vowel shifts before historical fall into two categories: mergers and splits. Mergers are more common and so most English dialects have fewer vowel distinctions before historical than in other positions of a word.

In many North American dialects, there are ten or eleven stressed monophthongs; only five or six vowel (rarely seven) contrasts are possible before a preconsonantal and word-final (beer, bear, burr, bar, bore, bor, boor). Often, more contrasts exist if appears between vowels of different syllables. In some American dialects and in most native English dialects outside North America, for example, mirror and nearer do not rhyme, and some or all of marry, merry, and Mary are pronounced distinctly. (In North America, those distinctions are most likely to occur in New York City, Philadelphia, some of Eastern New England (including Boston), and in conservative Southern accents.) In many dialects, however, the number of contrasts in that position tends to be reduced, and the tendency seems to be towards further reduction. The difference in how the reductions have been manifested represents one of the greatest sources of cross-dialect variation.

Non-rhotic accents in many cases show mergers in the same positions as rhotic accents even though there is often no phoneme present. That results partly from mergers that occurred before the was lost and partly from later mergers of the centering diphthongs and long vowels that resulted from the loss of.

The phenomenon that occurs in many dialects of the United States is one of tense–lax neutralization in which the normal English distinction between tense and lax vowels is eliminated.

In some cases, the quality of a vowel before is different from the quality of the vowel elsewhere. For example, in some dialects of American English, the quality of the vowel in more typically does not occur except before, and it is somewhere in between the vowels of maw and mow. It is similar to the vowel of the latter word but without the glide.

It is important to note, however, that different mergers occur in different dialects. Generally, these correlate to accents with rhotic vowels, as opposed to non-rhoticity (as in most of British English) or fully pronounced /r/ (as in Scottish English).

Mergers before intervocalic R
Most North American English dialects merge the lax vowels with the tense vowels before and so "marry" and "merry" have the same vowel as "mare", "mirror" has the same vowel as "mere", "forest" has the same vowel as the stressed form of "for", and "hurry" has the same vowel as "stir" as well as that found in the second syllable of "letter". The mergers are typically resisted by non-rhotic North Americans and are largely absent in areas of the United States that are historically largely nonrhotic.

Hurry–furry merger
The hurry–furry merger occurs when the vowel before intervocalic  is merged with. That is particularly a feature in many dialects of North American English but not New York City English, Mid-Atlantic American English, older Southern American English, some speakers of Eastern New England English, and speakers of Southeastern New England English. Speakers with the merger pronounce hurry to rhyme with furry and turret to rhyme with stirit.

To occur, the merger requires the mergers to be in full effect, which is the case in nearly all English dialects worldwide, particularly outside the British Isles. However, in Scotland, hurry is a perfect rhyme of furry, but also the  mergers have never developed there, meaning that ,  and  can all still exist before both intervocalic and coda ; thus, fur, fern, and fir have distinct vowels:.

Dialects in England, Wales, and most others outside North America maintain the distinction between both sounds and so hurry and furry do not rhyme. However, in dialects without the foot–strut split, hurry has an entirely different vowel: (in a number of those dialects, a square–nurse merger is in effect instead).

General American has a three-way merger between the first vowels in hurry and furry and the unstressed vowel in letters. In Received Pronunciation, all of them have different sounds (, and, respectively), and some minimal pairs exist between unstressed  and , such as foreword  vs. forward. In General American, they collapse to, but in phonemic transcription, they can still be differentiated as and  to facilitate comparisons with other accents. General American also often lacks a proper opposition between and, which makes minimal pairs such as unorthodoxy and an orthodoxy variably homophonous as. See the strut–comma merger for more information.

In New Zealand English, there is a consistent contrast between hurry and furry, but the unstressed is lengthened to  (phonetically ) in many positions, particularly in formal or slow speech and especially when it is spelled $⟨er⟩$. Thus, boarded and bordered might be distinguished as and, which is homophonous in Australian English as  and distinguished in Received Pronunciation as  and , based on the length and the rounding of. The shift was caused by a complete phonemic merger of and, a weak vowel merger that was generalized to all environments.

Mary–marry–merry merger
One notable merger of vowels before is the Mary–marry–merry merger, a merging of the vowels  (as in the name Carrie or the word marry) and  (as in Kerry or merry) with the historical  (as in Cary or Mary) whenever they are realized before intervocalic. No contrast exists before a final or preconsonantal, where merged with  and  with  (see nurse mergers) centuries ago. The merger is fairly widespread and is complete or nearly complete in most varieties of North American English, but it is rare in other varieties of English. The following variants are common in North America:
 * The full Mary–marry–merry merger (also known, in this context, as the three-way merger) is found throughout much of the United States (particularly the Western and Central United States) and in all of Canada except Montreal. This is found in about 57% of American English speakers, according to a 2003 dialect survey. The merger is highlighted in the song Merry Go 'Round, whose central wordplay revolves around "Mary", "marry", and "merry" having the exact same pronunciation in the singer's accent.
 * No merger, also known as a three-way contrast, exists in North America primarily in the Northeastern United States and is most clearly documented in the accents of Philadelphia, New Jersey, New York City, Rhode Island, and Boston. In the Philadelphia accent, the three-way contrast is preserved, but merry tends to be merged with Murray (see merry–Murray merger below). The three-way contrast is found in about 17% of American English speakers overall.
 * The Mary–marry merger is found alone with 16% of American English speakers overall, with the highest concentration in New England, especially New Hampshire.
 * The Mary–merry merger is found alone among 9% of American English speakers overall, concentrated in the American South, especially Louisiana where it is the most common variant, and the Southern part of the Mid-Atlantic region. It is also found among Anglophones in Montreal.
 * The merry–marry merger is found alone rarely, with about 1% of American English speakers.

In accents without the merger, Mary has the a sound of mare, marry has the "short a" sound of mat, and merry has the "short e" sound of met. In modern Received Pronunciation, they are pronounced as, , and ; in Australian English, as , , and ; in New York City English, as , , and ; and in Philadelphia English, the same as New York City except merry is. There is plenty of variance in the distribution of the merger, with expatriate communities of those speakers being formed all over the country.

The Mary–merry merger is possible in New Zealand, and the quality of the merged vowel is then (similar to  in General American). However, in New Zealand, the vowel in Mary often merges with the vowel  instead (see near–square merger), which before intervocalic  may then merge with  and so Mary (phonemically ) can be  or  instead. In all of those cases, there is a clear distinction between Mary and merry (regardless of how both are pronounced) and marry (with the  vowel) on the other.

Merry–Murray merger
The merry–Murray merger is a merger of and  before, with the resulting vowel being. It is common in the Philadelphia accent, which does not usually have the marry–merry merger; its "short a", as in marry and its SQUARE vowel remain distinct unmerged classes before. Therefore, merry and Murray are both pronounced as, but marry and Mary  are distinct from this merged pair (and each other).

Mirror–nearer and mergers
The mergers of and  (as in mirror and nearer, or Sirius and serious, respectively) and  occur in North American English as a part of pre- laxing, together with the Mary–merry and horse–hoarse mergers. The phonetic outcome of the first merger is either a lax vowel, or a somewhat raised vowel that approaches the monophthongal allophone of : , often diphthongal as. In the case of the merger, it tends to approach the monophthongal variant of :.

The mirror–nearer merger is absent from traditional, local, or non-standard accents of the Southern and Eastern United States, where nearer is pronounced with a tense monophthong or a centering diphthong  (phonemicized as  or, depending on whether the accent is rhotic or not), whereas mirror has a lax monophthong.

In the case of the first merger, only a handful of minimal pairs (e.g., cirrus–serous and Sirius–serious) illustrate the contrast, in addition to morphologically distinct pairs (e.g., spirit–spearit), all of which are rendered homophonous by the merger. Indeed, the number of the words containing is itself low. No minimal pairs exist for the merger, due to the extreme scarcity of the  sequence in dialects of English with the foot–strut split. Furthermore, the hurry–furry merger that occurs in most varieties of North American English results in a merger of with, removing almost any trace of the historical  vowel in this position. Instead, it is a simple replacement of one phoneme with another, so that the word tour is perceived to contain the  vowel, rather than the  vowel. However, this change may not hold where morpheme boundaries apply; allowing a qualitative distinction to be maintained between the stressed vowels in tourist (a fairly close back monophthong of variable height) on the one hand, and two-wrist  (a fully close monophthong in free variation with a narrow closing diphthong) on the other (cf. traditional RP ). The same applies to the mirror–nearer merger, which laxes the vowel in clearing but not in key ring, cf. RP. Certain words are pronounced as if they contained a morpheme boundary before, notably hero and zero.

Some words originally containing the sequence are merged with either  (see cure–force merger) or, more rarely,  (see cure–nurse merger) instead of  +.

The mirror–nearer and mergers are not to be confused with the fleece–near and goose–cure mergers that occur in some non-rhotic dialects before a sounded  and which do not involve the lax vowels  and.

Merger of and  before vowels
Words with a stressed before intervocalic  in Received Pronunciation are treated differently in different varieties of North American English. As shown in the table below, in Canadian English, all of them are pronounced with, as in cord. In the accents of Philadelphia, southern New Jersey, and the Carolinas (and traditionally throughout the whole South), those words are pronounced by some with, as in card and so merge with historic prevocalic in words like starry. In New York City, Long Island, and the nearby parts of New Jersey, those words are pronounced with like in Received Pronunciation. However, the sound is met with change to and so still merges with the historic prevocalic  in starry.

On the other hand, the traditional Eastern New England accents (especially around Boston), the words are pronounced with, but the cot–caught merger still applies elsewhere. In that regard, it is the same as Canadian, rather than Received Pronunciation. Most of the rest of the United States (marked "General American" in the table), however, has a distinctive mixed system. Most words are pronounced as in Canada, the five words in the left-hand column are typically pronounced with, all common words ending in an unstressed full vowel.

In accents with the horse–hoarse merger, also includes the historic  in words such as glory and force. When an accent also features the cot–caught merger, is typically analyzed as  to avoid postulating a separate  phoneme that occurs only before. Therefore, both cord and glory are considered to contain the phoneme in California, Canada, and elsewhere. Therefore, in accents with the horse–hoarse merger, and  are different analyses of the same word cord, and there may be little to no difference in the realization of the vowel.

In the varieties of Scottish English with the cot–caught merger, the vowel is pronounced towards the of caught and north. It remains distinct from the of force and goat because of the lack of the horse–hoarse merger.

Even in the American East Coast without the split (Boston, New York City, Rhode Island, Philadelphia and some of the coastal South), some of the words in the original short-o class often show influence from other American dialects and end up with anyway. For instance, some speakers from the Northeast pronounce Florida, orange, and horrible with but foreign and origin with. The list of words affected differs from dialect to dialect and occasionally from speaker to speaker, which is an example of sound change by lexical diffusion.

/aʊr/–/aʊər/ merger
The Middle English merger of the vowels with the spellings $⟨our⟩$ and $⟨ower⟩$ affects all modern varieties of English and causes words like sour and hour, which originally had one syllable, to have two syllables and so to rhyme with power. In accents that lack the merger, sour has one syllable, and power has two syllables. Similar mergers also occur in which hire gains a syllable and so makes it pronounced like higher, and coir gains a syllable and so makes it pronounced like coyer.

Card–cord merger
The card–cord merger, or merger, is a merger of Early Modern English  with, which results in the homophony of pairs like card/cord, barn/born and far/for. It is roughly similar to the father–bother merger but before r. The merger is found in some Caribbean English accents, in some West Country accents in England, and in some accents of Southern American English. Areas of the United States in which the merger is most common include Central Texas, Utah, and St. Louis, but it is not dominant anywhere and is rapidly disappearing. Rhotic dialects with the card–cord merger are some of the only ones without the horse–hoarse merger; this correlation is well-documented in the United States.

merger
In Modern English, the reflexes of Early Modern English and  are highly susceptible to phonemic mergers with other vowels. Words belonging to that class are most commonly spelled with oor, our, ure, or eur. Examples include poor, tour, cure, Europe (words such as moor ultimately from Old English ō words). Wells refers to the class as the words after the keyword of the lexical set to which he assigns them.

In traditional Received Pronunciation and General American, words are pronounced with Received Pronunciation  ( before a vowel) and General American. However, those pronunciations are being replaced by other pronunciations in many accents.

In Southern England, words are often pronounced with  and so moor is often pronounced, tour , and poor. The traditional form is much more common in Northern England. A similar merger is encountered in many varieties of American English, whose prevailing pronunciations are and ⁓, depending on whether or not the accent is rhotic. For many speakers of American English, the historical merges with  after palatal consonants, as in "cure", "sure", "pure", and "mature", and merges with  in other environments such as in "poor" and "moor".

In Australian and New Zealand English, the centering diphthong has practically disappeared and is replaced in some words by  (a sequence of two separate monophthongs) and in others by  (a long monophthong). The outcome that occurs in a particular word is not always predictable although, for example, pure, cure, and tour may rhyme with fewer and have, and poor, moor, and sure rhyme with for and paw and have.

merger
In East Anglia, a  merger in which words like fury merge to the sound of furry is common, especially after palatal and palatoalveolar consonants and so sure is often pronounced, which is also a common single-word merger in American English in which the word sure is often. Also, yod-dropping may apply, which yields pronunciations such as for pure. Other pronunciations in the accents that merge cure and fir include pure,  curious,  bureau and  mural.

– merger
Varieties of Southern American English, Midland American English and High Tider English may merge words like fire and far or tired and tarred towards of the second words:. That results in a tire–tar merger, but tower is kept distinct.

–– merger
Some accents of southern British English, including many types of Received Pronunciation and in Norwich, have mergers of the vowels in words like tire, tar (which already merged with, as in palm), and tower. Thus, the triphthong of tower merges with the  of tire (both surface as diphthongal ) or with the  of tar. Some speakers merge all three sounds and so tower, tire, and tar are all pronounced.

Horse–hoarse merger
The horse–hoarse merger, or  merger, is the merger of the vowels and  before historic, which makes word pairs like horse–hoarse, for–four, war–wore, or–oar, morning–mourning pronounced the same. Historically, the class belonged to the  phoneme (as in contemporary Received Pronunciation lot), but the  class was  (as in Scottish English go), which is similar to the contrast between the short lax  and the long tense  in German.

The merger now occurs in most varieties of English. Accents that have resisted the merger include most Scottish and Caribbean accents as well as some African American, Southern American, Indian, Irish, older Maine, South Wales (excluding Cardiff), some Northern English (Lancashire, Yorkshire), and West Midlands accents.

In the non-rhotic British accents that make the distinction, is typically merged with, while the sound of  varies. The areas of Wales that make the distinction merge it with the monophthongal variety of : (those accents lack the toe–tow merger), but in the West Midlands, it corresponds to  + :  or a separate  phoneme:. The words belonging to each set vary to an extent region to region, for example from Port Talbot tend to use, instead of the traditional , in forceps, fortress, important and importance.

The distinction was once present in the speech of southern England, the NORTH vowel being sounded as and the FORCE vowel as the centring diphthong. For many speakers, however, as noted by Henry Sweet, this contrast had by 1890 become constricted to word-final positions if the following word began with a consonant (so 'horse' and 'hoarse' had thus become homophonous, but not 'morceau' and 'more so'). In his 1918 Outline of English Phonetics, Daniel Jones described the distinction as optional, but he still considered it to be frequently heard in 1962; the two vowels are differentiated in the first (1884–1928) and second (1989) editions of the Oxford English Dictionary with the caveat that in most varieties of southern British pronunciation the two had become identical;  no distinction is drawn in the third edition, as well as in most modern British dictionaries (Chambers being a notable exception). John C. Wells wrote in 2002 that the distinction had become obsolete in RP.

In the United States, the merger is now widespread everywhere but is quite recent in some parts of the country. For example, fieldwork performed in the 1930s by Kurath and McDavid showed the contrast to be robustly present in the speech of the entire Atlantic coast, as well as Vermont, northern and western New York State, Virginia, central and southern West Virginia, and North Carolina. However, by the 1990s, surveys showed those areas had completely or almost completely undergone the merger. Even in areas in which the distinction is still made, the acoustic difference between the of horse and the  of hoarse was found to be rather small for many speakers. Some American speakers retain the original length distinction but merge the quality. Therefore, hoarse is pronounced longer than horse.

In the 2006 study, most white participants in only these American cities still resisted the merger: Wilmington, North Carolina; Mobile, Alabama; and Portland, Maine. A 2013 study of Portland, however, found the merger to have been established "at all age levels". In the 2006 study, even St. Louis, Missouri, which traditionally maintained the horse–hoarse distinction so strongly that it instead merged card and cord, showed that only 50% of the participants still maintained the distinction. The same pattern (a horse–hoarse distinction and a card–cord merger) also exists in a minority of speakers in Texas and Utah. New Orleans prominently shows much variability regarding the merger, including some speakers with no merger at all. Black Americans are rapidly undergoing the merger but are also less likely to do so than white Americans, with a little over half of the 2006 study's black participants maintaining the distinction nationwide.

In some Indian, Welsh, and Southern American dialects, the distinction between and  may be maintained through the presence or absence of, with horse being  and hoarse being.

The two groups of words merged by the rule are called the lexical sets (including horse) and  (including hoarse) by Wells (1982).

In dialects that maintain the distinction between the two phonemes, is indicated almost exclusively by the spellings or, aur and ar (when preceded by /w/), as in horse, aural, war, while  is generally indicated by the spellings oar, ore, our and oor, as in hoarse, wore, four, door.

However, can also sometimes occur in words with the or spelling. This is usually in one or more of the following circumstances: However, it does not occur in all words that fit the above criteria. The following table lists some words irregularly with the sound, rather than, with the cases that make them so and regular  words by comparison. Note that in non-standard accents many words can shift their pronunciation without changing diaphonemes due to lexical diffusion.
 * When the vowel immediately follows a labial consonant,, as force itself.
 * In past participles in -orn with corresponding past tense forms in -ore, as in torn, or words made from ones with the vowel.
 * When the /r/ is followed by a vowel within the same morpheme, as in words like glory and flora.

merger
The  merger or cheer–chair merger is the merger of the Early Modern English sequences and, as well as the  between them, and is found in some accents of Modern English. Many speakers in New Zealand merge them towards the vowel, but some speakers in East Anglia and South Carolina merge them towards the  vowel. The merger is widespread in Caribbean English, including Jamaican English.

mergers
Common in a vast majority of modern English dialects worldwide is the merger of as many as five Early Modern English vowels (,, , , and ) into when followed by an  before a consonant or at the end of a syllable. Thus, the vowels in words like, , and are the same in almost all modern accents of English. John C. Wells briefly calls it the  merger. When another vowel follows, these are often distinct; contrast the vowels in merry, hurry, weary, mirror, and furry (see the Mary–marry–merry merger, mirror–nearer merger, and hurry–furry merger for details). The major exceptions to most of the Nurse mergers are Scottish English and older Irish English, which also do not have mergers of vowels before following another vowel. What Scottish and older Irish English have in common is rhoticity without r-colored vowels, meaning that /r/ is used at the end of a syllable.

Words and names with historic are spelled   as in earn, earth or pearl and include the function words her and were, in  which have stayed distinct (see both the meet–meat and pane–pain mergers). The relevant words and names with historic are  in a stressed syllable, historic  are spelled as a stressed, and  is any  or. The diaphoneme originates from unstressed vowels before  and was not otherwise distinct.

Scottish English and rural Irish English dialects both use sequences of a vowel then not r-colored vowels, and both lack the foot–strut split; which result in comparable developments. However, the actual realizations of the retained Nurse vowels vary. Also, while most of Scottish English has some distinction, more prestigious/ younger Irish English realizes the Nurse merger as. The table below summarizes the overall differences: In Scottish English, mid front and  are merged into, paralleling the mid back vowel horse–hoarse merger, which Scottish English lacks. The vowel in fir is usually distinct, but is liable to merge than  because their non-rhoticized equivalents  and  belong to the same phoneme; this parallels the hurry–furry merger. All EME became, which included before. The (letter),  (term) and  (fur) vowels are fully distinct from each other.

For rural and very conservative Irish English, (in whirl) merges entirely with  (in earl), sometimes merging again with. The merged merges again with  after labials and coronal plosives (including  and  becoming  and ) in many common words, but this no longer productive.

merger
Some older Southern American English varieties and some of England's West Country dialects have a partial merger of. They generally pronounce as, which rhymes near with a  word like sir or fur (compare general English realisations of cue and coo). Words such as beard are then pronounced as. Usual word pairs like beer and burr are still distinguished as and. However, is dropped after a consonant cluster (as in queer) or a palato-alveolar consonant (as in cheer), likely because of phonotactic constraints, which then results in a merger with :,.

There is evidence that the African American Vernacular English in Memphis, Tennessee, merges both and  with  and so here and hair are both pronounced the same as the strong pronunciation of her.

merger
The  merger (words like perk being pronounced like pork) involves the merger of with  and occurs in broadest Geordie.

Some words (roughly those spelled with a) have a distinct  vowel in broad Geordie. Therefore, the merger involves only some of the words corresponding to historical in Received Pronunciation.

merger
The  merger, or fair–fur merger, is a merger of with  that occurs in some accents like Scouse, various other dialects within historic Lancashire, Teeside, Hull, the newer Dublin, and the Belfast accents.

Scouse, the accent of Liverpool and the Merseyside area, is the dialect with which the merger is most stereotypically associated. The most common realization in modern Scouse is [eː], but [ɛː] and [ɪː] are also possible. It is also found in many neighbouring regions of historic Lancashire, such as Bolton, Wigan and Blackburn, where the quality is generally a more central [ɜː]~[ɵː]. Shorrocks (1999) reports that in the dialect of Bolton, Greater Manchester, the two sets are generally merged to /ɵ:/, but some words such as first have a short /ɵ/.

The merger can also be found among some speakers in the Teeside conurbation and the Humberside (Hull - East Riding of Yorkshire - North East Lincolnshire) area with a quality intermediate between [ɛː] and [ɜː].

Thorne (2003) reports that the merger also occurs in Birmingham, remarking the merger as being "another principally northern characteristic". Interestingly enough, Tennant (1982) reports as being pronounced as /eə/ - which would lead  and  as being pronounced the opposite way of their RP pronunciation.

The merger is found in some varieties of African American Vernacular English and is pronounced : "A recent development reported for some AAE (in Memphis, but likely found elsewhere)." This is exemplified in Chingy's song "Right Thurr", in which the merger is spelled in the title.

Labov (1994) also reports such a merger in some western parts of the United States "with a high degree of r constriction".