Phonemic inventory: how many phonemes occur in SNAE?

There is no need to review the inventory of SNAE consonant phonemes here: in most respects, this is identical with that of other varieties and is described elsewhere in this volume. Only one matter of phonemic contrast among consonants will be mentioned here: that involving the voiced and voiceless types of /w/, or /w/ and /hw/, as found in pairs like wear and where, weather and whether, wine and whine, or witch and which. While some conservative speakers of NAE maintain a distinction between these sounds, it has largely disappeared among younger speakers in most regions, so that it can be safely described as absent in SNAE.

Regional pronunciation differences in English are far more likely to involve vowels than consonants. In particular, there are four important variables of vowel contrast that distinguish major dialects of English, including SNAE. These are shown in Table 13.2.

The first line of Table 13.2 refers to the split of Middle English short /u/. This had affected southern English speech by the seventeenth century (Wells 1982: 197), early enough to be transplanted to North America, but never spread to northern England, where FOOT and STRUT still rhyme today. This variable therefore divides Britain into two dialect regions but unites the standard variety of British English, which is regionally rooted in the south-east of England, with SNAE, which has /u/ in FOOT but /ʌ/ in STRUT.

The remaining lines of Table 13.2 display important trans-Atlantic differences in phonemic inventory. The first involves another split, in this case of Middle English short /a/. Like the split of short /u/, this occurred in southern England, leaving the North unaffected. It seems to have occurred in two stages. First, in the seventeenth and early eighteenth centuries, short /a/ was lengthened to [a:] before voiceless fricatives and a few other environments (the BATH class), elsewhere remaining short and shifting forward to [æ] (the TRAP class; Wells 1982: 203–204). This aspect of the split did make it across the Atlantic, at least to some founding communities, though its subsequent history in American English is complicated and led to several of the dialect differences that will be discussed below. The second stage of the split was a backing of the lengthened vowel from [a:] to [ɑ:] in south-eastern England, producing the sound of the modern BATH class in Standard British English, with /ah/. This must have happened in the late eighteenth or early nineteenth century (Wells 1982: 234), early enough to be transplanted to Australia and New Zealand, but not to North America, where the lengthened vowel tended to be raised rather than backed. While some regions of North America retain a distinct, raised vowel in a much expanded version of the BATH class today, the split of short /a/ has collapsed in most NAE dialects. SNAE has a single phoneme, /æ/, in both TRAP and BATH words, with only subphonemic variation in phonetic quality (Kurath and McDavid 136; ANAE 173–174).

The last two lines of Table 13.2 involve mergers rather than splits. The first concerns the small remnant of Middle English long /ā/ that was not raised to /ey/ (FACE) in the Great English Vowel Shift, remaining in a low-central position. This includes the word father, along with lengthened /a/ before /-lm/ (almond, alms, balm, calm, palm, psalm) and a few other unusual words (ma, pa, rah, etc.). Its residual status made this PALM class prone to merger with neighboring vowels. In southern England it merged with the lengthened and retracted BATH class and, as a result of /r/ vocalization, with the START class. In North America, where BATH was not retracted and START generally retained its /r/, the tendency was instead for PALM to merge with LOT, the regular development of Middle English short /o/. LOT began to shift down and forward from its original mid-back position by the seventeenth century, reaching a low-central unrounded position, approximately [ɑ], in some southern English dialects as North America’s English-speaking colonies were being founded. Wells (1982: 245) suggests that it is not clear whether this happened in England or was an American innovation. It later moved back and re-rounded in standard British English, but the low-central unrounded vowel survives in much of North America, occupying the same phonetic range as southern British BATH. This accounts for the American tendency to transcribe LOT as /a/ or /ɑ/, based on its phonetic identity in many American dialects, rather than as /o/, based on its historical identity. In these dialects, LOT has merged with PALM, so that father rhymes with bother and balm and bomb are homophones (PEAS 141–142; ANAE 171). The main exception to this pattern is New England, as discussed below.

The last line of Table 13.2 also involves the contrastive relations of the LOT class, in this case with the THOUGHT class, sometimes referred to as long open /o/ and transcribed as /ɔ/ (the /oh/ symbol used here, like the /o/ for LOT, is consistent with the binary system of broad transcription mentioned above). Like PALM, THOUGHT is not the regular development of any single Middle English vowel phoneme, and its membership varies by dialect. For instance, as a result of a lengthening of Middle English /o/ before voiceless fricatives parallel to the development of Middle English /a/ described above, the THOUGHT class includes the CLOTH subclass in most American dialects but not in Standard British English, where CLOTH retains its original association with LOT. In North America, THOUGHT has therefore shown a similar tendency to merge with its neighbor, LOT, except where other phonetic developments have kept the two categories distinct (ANAE 123). These phonetic developments include some of the most salient regional differences in pronunciation, to be discussed below, but tend to occur in regions that are not the main source of SNAE (specifically, the Mid-Atlantic, the Inland North, and the South). Outside these areas, in SNAE, unrounding of THOUGHT and phonetic approximation to LOT has generally led to the “low-back merger”, making homophones of such pairs as cot and caught, sod and sawed, stock and stalk, don and dawn, and collar and caller. This merger is now complete in northern New England, the West and Canada, as well as in parts of the Midland and South, (ANAE 170). It is in progress in the remaining parts of the Midland and South – more advanced in some communities and social groups than in others – and may even be making inroads among younger, upwardly mobile speakers in the areas that have historically resisted it, as the pronunciation features that prevented it in the past become socially stigmatized. Nevertheless, the retention of the LOT–THOUGHT contrast among many Americans who would consider themselves – and not without reason – to be speakers of SNAE compels us to place parentheses around the “NO” in the last line of Table 13.2, to concede the persistence of dialect variation in this respect.

To summarize this section, the distinctive sound of SNAE is strongly influenced by three important features connected with variables of phonemic inventory. Most North Americans use /æ/ in BATH words, so that they have the same vowel as TRAP words; make no distinction between PALM and LOT words, because of the low-back to low-central, unrounded pronunciation of LOT; and, increasingly, also fail to distinguish LOT and THOUGHT words, using a lowered, unrounded vowel in THOUGHT that is too close to the vowel of LOT to support a stable phonemic contrast.

Systematic variation in phonemic incidence: words whose pronunciation varies by phonological rule

The most important and pervasive systematic variable of phonemic incidence in English is the occurrence of /r/ in “coda” position; that is, when it is not pre-vocalic, as in Wells’ keywords NEAR, SQUARE, START, NORTH, FORCE, CURE, and NURSE. The tendency to delete or “vocalize” post-vocalic /r/ in English began in restricted environments in the Middle English period, but did not become a more general feature of British English until the eighteenth or nineteenth century, too late to be implanted with the initial English-speaking settlement of North America (Wells 1982: 218). Nevertheless, in the nineteenth century “r-lessness” became a defining feature of Standard British English, whence it spread across the Atlantic as a prestige feature to several regions along the east coast of the United States. These included eastern New England, New York City, and parts of the South, but not the intervening southern Mid-Atlantic region around Philadelphia, or Canada (PEAS 171, Map 156; ANAE 48). The original colonial dialects of NAE, already carried inland by westward migration, remained unaffected. In the late twentieth century, the prestige of r-less pronunciation remained high in Britain, where it continued to conquer new territory, but was reversed in the United States, in favor of the more “American” sound of a fully constricted post-vocalic /r/. This prestige reversal was examined by the most famous sociolinguistic study ever undertaken, in New York City, where /r/ was being re-inserted by middle-aged and younger speakers interested in upward social mobility in the 1960s (Labov 1966, 1972a).

Today, r-lessness remains a variable feature along the Atlantic seaboard, heard more from older speakers with local social networks than from the young or globally oriented. Its recession is more or less complete in European-American Southern speech, especially in large cities like Atlanta and Houston, but it has persisted to a greater extent in African-American speech, as mentioned below. It also continues to be widespread in eastern New England, even in large cities like Boston and Providence, and survives in popular culture in the catch phrase used as a stereotype of Boston speech, “pahk the car in Hahvuhd Yahd”.

Even in communities that retain /r/ vocalization, however, it is usually variable. Its frequency responds both to stylistic factors, with less vocalization in formal, monitored speech, and to a range of phonological and other linguistic constraints. The most favorable environment for vocalization is the unstressed /ər/ of lettER words, as in letter, September, character, or spectacular, but especially in word-internal position, as in permission, Saturday, afternoon, or information. This context seems to “fly below the radar” of speakers concerned with moving away from the r-less speech they grew up with. Another vocalization context for otherwise /r/-constricting speakers is the presence of two r’s in a word, which promotes a kind of dissimilation in which one of them is vocalized or deleted. This applies to the first /r/ in words like corner, former, ordinary, and quarter, and to the second in words like mirror and terror, or when an agentive or comparative suffix <-er> is added to a stem ending in /r/, as in bearer or fairer. The most variable vocalization environments are the NEAR, SQUARE, START, NORTH, FORCE, and CURE sets that feature /r/ in their stressed syllables: these often have a restored /r/ in communities moving away from vocalization, but remain r-less in communities where this feature is more stable. Generally r-less speakers have the PALM vowel in the START class and the THOUGHT vowel in the NORTH and FORCE classes: Labov (1966, 1972a) investigates potential homophones like dock and dark and sauce and source in New York City. (Where NORTH and FORCE are different, as in traditional Boston speech, NORTH has [ɒ] and FORCE has an in-gliding variant of the GOAT vowel [ɔə].) The least favorable context for /r/ vocalization, and the first in which it is normally restored, is the syllabic /r/ or stressed /ɜ˞/ of NURSE words, like her, girl, bird, or first. Almost all Americans have a constricted /r/ in these words today, perhaps because a diphthongized variant of this vowel, pronounced as [ɜɪ], was the target of negative stereotypes of r-less dialects in the mid-twentieth century, as when New York cabbies were reputed to say “toity-toid street” for thirty-third street.

Most other systematic variables of phonemic incidence in English involve conditioned mergers, or neutralizations of phonemic contrast in particular phonological contexts. A whole set of these neutralizations is connected with the variation in /r/ just discussed, but involves inter-vocalic rather than post-vocalic /r/ (see PEAS 123–127; Gregg 1957b). It is a general property of English phonology that an /r/ in the coda of a syllable (post-vocalic /r/) limits the range of vowels that can occur before it. In particular, short vowels (/i, e, æ, o, ʌ, u/, or those of KIT, DRESS, TRAP, LOT, STRUT, and FOOT) are generally not licensed in this position. If we think of inter-vocalic /r/ in words like very, carrot, orange, and hurry as ambi-syllabic, at once closing the preceding syllable (coda position) and starting the next (onset position), we can see how the incidence of vowels in the first syllables of such words will be constrained by the variable presence of coda /r/. In r-less dialects, coda /r/ is not present, so that the full range of vowels can occur. In these dialects we therefore hear the vowel of DRESS in very, that of TRAP in carrot, that of LOT in orange, and that of STRUT in hurry. In the /r/-retaining dialects most commonly associated with SNAE, by contrast, we hear compromise vowels in these contexts that represent neutralizations of contrast between long and short vowels, which match the set of vowel qualities that occur before final coda /r/, in SQUARE, FORCE, and NURSE. The DRESS and TRAP vowels merge with FACE; the LOT vowel with GOAT; and the STRUT vowel with NURSE. Thus, Mary, merry, and marry all sound more or less like merry; coral sounds like choral; and hurry has the vowel of her.

Another common neutralization of vowel contrast affects the vowels /uw/ and /iw/ after the alveolar stops /t, d, n/. In most English dialects, this contrast has survived after labial and velar consonants (boot versus beauty; coop versus cube), where /iw/ is distinguished by a palatal glide before the vowel, [ju]. After liquids and /s/, however, the contrast has now been lost, so that /uw/ now occurs in place of /iw/ in words like pollution and super. In NAE, this neutralization is generally extended to instances of /iw/ after /t, d, n/, as in Tuesday, student, duty, and news, though some conservative speakers retain a palatal glide in at least some of these words (PEAS 113, Map 33), especially in the American South (ANAE 55) and parts of Canada.

Another important systematic difference in phonemic incidence concerns the consonant /t/, which is partially voiced, or “flapped”, when it occurs after a stressed vowel or vowel-liquid sequence and before an unstressed vowel or syllabic sonorant, as in city, party, faulty, daughter, or battle; flapping after /l/ is variable. Medial /d/ also has a slightly weakened articulation in these contexts, so that for many speakers the /t-d/ contrast is neutralized in pairs like atom and Adam; coated and coded; rater and raider; metal and medal (or meddle); diluted and deluded; etc. Flapping can also occur two syllables after the stress, as in charity, monitor, or penalty, but is more variable in this position. Following a stressed vowel and /n/, as in twenty or winter, the /t/ is often deleted altogether, so that winter and winner sound the same; preceding a syllabic /n/, as in button or Latin, it is replaced in most dialects with a glottal stop. Flapping in NAE is now completely standard, to the point where its absence, especially in the core environment after stressed vowels or vowel-/r/ sequences (city and party), is considered pompous or affected.

The foregoing discussion of systematic variation in phonemic incidence can be summarized by noting that SNAE retains coda or post-vocalic /r/, with a consequent reduction in the number of vowel contrasts before intervocalic /r/, so that hairy, berry, and carry all have the vowel of berry, and forest and worry have the vowels of four and were respectively; has also lost the contrast between /uw/ and /iw/ after /t, d, n/, so that due and dew sound like do; and replaces /t/ in post-tonic, inter-vocalic contexts with a sound that, for most speakers, is identical with /d/, so that seated and seeded, or bitter and bidder, are homophones.

Lexical variation in phonemic incidence: words whose pronunciation varies in phonologically irregular ways

Some variables of phonemic incidence are truly idiosyncratic. A classic example from American dialectology is the fricative in the word greasy, which varies between /s/ in the North and /z/ in the Midland and South of the Eastern United States (PEAS 176–177, Map 171). The irregular nature of this variation is demonstrated by its absence in phonologically similar words like easy, teasing, fleecy, or increasing: variation between /s/ and /z/ is clearly a property of the word greasy, not a phonological rule affecting intervocalic /s/. The same could be said about the word vase, which varies in both its vowel and final consonant, rhyming alternately with face, phase, or spas (PEAS 177); this pattern is not observed in similar words like base or raise, etc. A lack of systematicity has made this type of variation less interesting to phonologists, but not to the general public. Many of the most frequently cited examples of dialect variation involve pronunciations of particular words. Further examples from PEAS include the vowels of DRESS, KIT, or FACE in again (131); of TRAP, PALM or LOT in aunt (135); of DRESS or TRAP in catch (139); of FLEECE or KIT in creek (148); of FLEECE or PRICE in either and neither (149); of GOOSE or FOOT in roof (154) and root (155); or of LOT or THOUGHT in several words, including sausage, water, wash, fog, long, and on (162–164); or the voicing of the fricative in without (176). Nevertheless, though such words may excite the interest of casual observers, it is difficult to give a general account of them, since each tends to display a unique regional distribution.

Somewhat less idiosyncratic are several sets of words of varying size that display more or less regular differences between SNAE and Standard British English. One of them, already mentioned above, is the CLOTH set: words that, in British English, feature short /o/ before voiceless fricatives, as in coffee, lost, or boss. In North American dialects that distinguish LOT and THOUGHT, most of these words have the vowel of THOUGHT (though there are exceptions, generally involving less frequent words). The same is true of most words that feature this vowel before /g/ or /ŋ/, like dog, log, song, and wrong, though phonemic incidence in this subset is even more variable, as mentioned above. For most people, cog and gong, for example, have the LOT vowel, as does the new word blog, while hog varies by region.

Another loosely cohesive set of British English–NAE differences involves reduction or deletion of unstressed vowels, for instance in the set of Latinate words that end in -tary or -tory, such as secretary, military, preparatory, or mandatory. The penultimate vowel of these words is usually deleted or reduced in Britain but preserved in North America. NAE also retains unstressed vowels in words like medicine, police, and the names of berries (blackberry, raspberry, strawberry, etc.), as well as in place names like Birmingham and Manchester and names beginning with Saint. When we add to this the evidence of other distinctively British shortenings in words like forehead and waistcoat, NAE appears the more generally conservative dialect in this respect.

Within North America, some of the differences between Canadian and American English also involve variation in phonemic incidence, with variable adherence to British norms in Canada. Several of these involve the PRICE vowel, /ay/. Words in -ile, such as fertile, futile, hostile, missile, mobile, and sterile, have a reduced vowel in the second syllable, like that of noble, in American English but a full PRICE vowel, like that of profile, in Britain and Canada (Avis 1956: 46). On the other hand, Americans tend to have /ay/ in -ine words, like genuine, and in the Latin prefixes anti-, semi-, and multi-, where /i/ (KIT) is heard for the first set and /iy/ (FLEECE) in the second in Britain and, variably, in Canada (Avis 1956: 47). Verbs that begin with the prefix di-, like digress, direct, dissect, and diverge, along with their nominal forms, also vary between the PRICE and KIT vowels on either side of the Atlantic as well as within NAE, though in the case of vitamin, North Americans are united in using /ay/ in contrast to British /i/. Similarly irregular is a set of words that contain the Latin prefix pro-, such as the nouns process, produce, and progress: these tend to have the LOT vowel in the United States but vary between LOT and GOAT in Canada (Avis 1956: 45).

The cases of lexically governed phonemic incidence discussed so far have involved sets of a few dozen words at most, but there is one case that involves not dozens but hundreds or thousands of words. This is the set of “foreign (a)” words, discussed in Boberg (2010: 137–140) and first studied systematically in earlier work cited therein: words borrowed from other languages in which the stressed vowel is spelled with the letter <a>. Because English <a> has several phonemic values, these words can have the vowel of FACE, like potato, of TRAP, like tobacco, or of PALM, like spa. Most recent borrowings get either TRAP or PALM, but national dialects of English have different systems for deciding which vowel goes in which words. British English bases its choice mostly on vowel length. Since /æ/ is a short vowel, it is used before voiceless consonants, which tend to shorten preceding vowels, especially when they are spelled with double letters. Thus, /æ/ is heard in British English in macho, mafia, pasta and Picasso; exceptions are taco, which is variable, perhaps due to American influence, and Iraq, which usually has /ah/. A following voiced consonant encourages the preceding vowel to be treated as long: long /ah/ is preferred in the stressed syllables of avocado, Colorado, drama, façade, lager, lava, llama, pajamas, Pakistani, panorama, plaza, Slavic, and soprano; one exception to this is lasagna (or lasagne), which usually has /æ/. In American English, vowel length is far less important than the foreign status of the words, which demands the use of /ah/ rather than /æ/, perhaps on the model of Spanish, the most familiar “foreign” language in many parts of the United States. All of the foreign (a) words just listed have /ah/ in American English, except Pakistani, panorama, and soprano, which normally have /æ/, and Colorado, Iraq, and pajamas, which vary between /ah/ and /æ/. Canadians have a third pattern all their own, in which most of these words, at least traditionally, had /æ/, though some (like façade, lasagna, lager, macho, and mafia) have now begun to switch over to /ah/, apparently under American influence. Even most younger Canadians today, however, continue to use /æ/ in avocado, Colorado, drama, Iraq, lava, pajamas, Pakistani, panorama, pasta, Picasso, plaza, Slavic, and soprano, a list that includes several words (avocado, drama, lava, and Slavic) in which both Britons and Americans agree on /ah/. The Canadian preference for /æ/ as the default vowel for these words likely has its origins in the conventional Canadian understanding that where Britons say /ah/ in BATH words, Canadians say /æ/; this correspondence was simply transferred to foreign words as well, so that if British English had /ah/ in avocado, drama, lava, or Slavic, this should be rendered as /æ/ in Canada.

Phonetic realization: what is the phonetic quality of each phoneme?

Even variables of phonemic incidence that involve large sets of words, like the foreign (a) class just discussed, are limited in their role as indicators of dialect difference by their frequency of occurrence: while speakers may react strongly to unfamiliar or different pronunciations of words like vase, roof, greasy, or pasta when they hear them, the likelihood of any one of these words occurring in ordinary discourse is fairly small. Far more likely is that any substantial quantity of speech will include several examples of the more common vowel phonemes. Given their high frequency in discourse, as well as their systematic and regular application, variables of phonetic quality must therefore play a leading role in allowing speakers to project their own regional or social identities, as well as to perceive and assess the identities of others whose speech they hear.

There is a great deal of regional variation in the phonetic value of phonemes across North America, as recorded by the ANAE, or by Thomas (2001). Analysis of this variation will be reserved for the next section. Here, in Table 13.3, we offer a general statement of the approximate phonetic quality of the vowels of SNAE. For other, analogous descriptions, see, inter alia, Wells (1982: 121–122), Kretzschmar (2004: 263–264) or Ladefoged (2006: 39); an earlier equivalent appears in Bloomfield (1933: 91). Where substantial inter-speaker variation occurs even within SNAE, two phonetic symbols appear. Allophonic variation due to phonetic context is assumed rather than explicitly indicated, so that the values in Table 13.3 indicate the main quality of each vowel, rather than the total range of its allophones.

New England: Boston and Providence

Though New England is often thought of as a unified region in historical and cultural terms, it embraces several distinct dialect areas (PEAS, Map 2; Boberg 2001). The LOT-THOUGHT variable divides it into a northern half, from Vermont to Maine, including Boston, where this distinction has been lost, and a southern half, from Connecticut to Rhode Island, including Providence, where it is maintained. Bisecting this division is a line separating eastern New England, including Boston and Providence, which is traditionally r-less, from western New England, including Springfield and Hartford, which remained r-full. Northeastern New England, including Boston, traditionally resisted the merger of PALM and LOT, which has affected the rest of North America, because LOT was merged instead with THOUGHT. In traditional Boston speech, some members of the BATH class were identified with the PALM class, as in British English, rather than with TRAP, though this pattern is now recessive. Northeastern New England also held out against the merger of NORTH and FORCE found in SNAE, but this, too, is fading with time (Laferrière 1979). The quality of the back up-gliding vowels of GOOSE, GOAT, and MOUTH tends to be conservative across New England, with less centralization and fronting than occurs farther south. Table 13.4 lists some vowel qualities typical of traditional Boston speech, which can be compared with those given for SNAE in Table 13.3.

The Mid-Atlantic: New York City, Philadelphia, and Baltimore

The Mid-Atlantic region between New England and the South is also bisected by the /r/ line, with New York City and region, to the north, being traditionally r-less and Philadelphia and Baltimore, to the south, being the major exception to this pattern along the east coast. In terms of Labov’s pivot points, however, the Mid-Atlantic region is more unified than New England: its northern and southern sections share a common vowel system in which phonemic distinctions are maintained in both corners of the vowel space, with minor differences in lexical distribution. The low-back merger has been resisted by shifting THOUGHT (including the CLOTH subset) up to mid-back position, where it becomes a diphthong with a central in-glide, in the range between [ɔə] and [ʊə]. Here it is easily distinguished from the [ɑ] of LOT, but merges, in New York, with NORTH/FORCE (sauce = source). Labov (1966) showed that the height of the diphthong nucleus was an important sociolinguistic variable in New York, with higher qualities receiving a negative evaluation even from New Yorkers themselves. In the low-front quadrant, the TRAP-BATH split displays a parallel development: an upward shift of the tense vowel, BATH, along the front periphery of the vowel space. Its quality ranges from [ɛə] to [ɪə], with a parallel social evaluation to that of raised THOUGHT, and a parallel tendency to merge with SQUARE in New York (bad = bared). TRAP remains in the low-front position, at [æ]. As in most of North America, PALM is merged with LOT; /r/ vocalization in New York adds the START class to this set (dark = dock). These developments are summarized in Table 13.5.

One way of distinguishing New York City and Philadelphia, apart from vocalization of /r/, is in the lexical distribution of tense and lax vowels (ANAE 173). While we have used Wells’ keywords to represent the tense vowels, the membership of these sets in NAE dialects is larger than in British English, for which the keywords were designed. The split of Middle English short /a/ in the Mid-Atlantic region, in particular, has received a great deal of scholarly attention because of the complexity of the conditioning factors that determine which vowels are tense, like BATH (e.g., Labov 1972a: 72–75; ANAE 173). In New York, the tensing environment was extended from the British BATH set, before voiceless fricatives, to vowels before voiced stops (cab, bad, badge, and bag) and front nasals (ham, band), though several nonphonetic constraints create exceptions to this rule. Philadelphia has tensing in a smaller range of environments and with more exceptions: among voiced stops, only /d/ causes tensing and the single word sad is a notable exception even to this. In the back vowels, the distribution of words like chocolate, laundry, and sausage between the /o/ and /oh/ or LOT and THOUGHT classes also shows regional variation. Particularly noteworthy is the preposition on, which rhymes with don in New York (as in the North generally) but with dawn in Philadelphia (as in the Midland and South; ANAE 189).

The Inland North: Chicago, Detroit, Cleveland, and Buffalo

The Inland North extends along the southern shores of the Great Lakes, from Milwaukee, Wisconsin, through Chicago, Detroit, Toledo, Cleveland, and Buffalo to Rochester, New York. Here, TRAP and BATH are a single phoneme, which has undergone the same phonetic development as BATH in the Mid-Atlantic territory, raising to mid-front position, approximately [ɛɐ] or [eə]. This has left room in the low-front quadrant, still occupied by TRAP in the Mid-Atlantic vowel system, for the forward shift of LOT (merged with PALM) to [a], which maintains its contrast with THOUGHT along a front-back dimension. The raising of TRAP-BATH and fronting of LOT-PALM are the initial and most striking components of the Northern Cities Vowel Shift (NCS, Labov 1991: 15–17; ANAE 187–191), which also involves several consequent developments in the short vowel subsystem. Fronting of LOT-PALM allows THOUGHT to unround and move down to [ɑ], which in turn makes room for STRUT to approach the lower-mid back region of the vowel space; this frees up the lower-mid central space, into which DRESS is retracted; finally, KIT is lowered into the lower-mid front space once occupied by DRESS. The ANAE uses the NCS to define its Inland Northern dialect region (204), though recent research by McCarthy (2011) suggests that its initiating stages, the raising of /æ/ and the fronting of /o-ah/, are no longer active changes, at least in her sample of Chicago speakers. The effects of the NCS are illustrated in Table 13.6, though it should be noted that it represents extreme targets for vowel shifting that are not reached by all speakers or in all contexts; they are intended to indicate the direction of shift.

The Inland North was initially settled mainly from New England, with which it shares several general northern characteristics, such as the conservative treatment of GOOSE, GOAT, and MOUTH referred to above, with relatively little movement away from the rear periphery of the vowel space. As a result, MOUTH [ɑʊ] is articulated further back in the Inland North than PRICE [aɪ], the opposite of what we find farther south.

The South: Richmond, Charlotte, Atlanta, Nashville, Dallas, and Houston

In the Southern United States, an entirely different set of vowel shifts, known as the Southern Shift (Labov 1991: 25; ANAE 242–254), has developed in response to the most frequently cited element of Southern phonology, the monopthongization of /ay/, the vowel of PRICE. (In this case Wells’ choice of keyword is unfortunate, since many Southerners do not monopthongize /ay/ before voiceless consonants; glide deletion happens most frequently in open syllables and before voiced consonants, so that prize would be a more appropriate keyword.) The realization of /ay/ as [a:] created a hole at the bottom of the subsystem of front up-gliding vowels that has pulled the nucleus of /ey/ (FACE) down toward low-central position, [ɐɪ], with /iy/ (FLEECE) following it downward in a chain shift. In their shifted positions, these long vowels have switched places with their short counterparts, KIT and DRESS, which have become tense, inward-gliding diphthongs, with nuclei higher and fronter than those of the originally long vowels. The third short front vowel, TRAP-BATH, is also tensed and diphthongal, or even triphthongal, with an upward then downward contour, especially in the BATH subset, which may parallel the lengthening of this class in other dialects.

Labov’s description of the Southern Shift also includes a second component involving a parallel forward shift of the long back-upgliding vowels of GOOSE, GOAT, and MOUTH (Labov 1991: 25). The last of these is part of the Southern strategy for avoiding the low-back merger: as MOUTH shifts forward from [ɑʊ] to [æʊ], THOUGHT develops a back-upglide, so that it adopts the [ɑʊ] quality that MOUTH has in the North. It is thereby differentiated from LOT, which remains in low-back position and is lengthened but monophthongal, with a nuclear quality that often overlaps that of THOUGHT (ANAE 127, 254). Two short back vowels, /u/ (FOOT) and /ʌ/ (STRUT), are also lengthened and strongly centralized in Southern speech, while the up-glide of /oy/ (CHOICE) is either shortened, as in boy ([bɔə]), or deleted, as in boil ([bɔːl]). Together, these ten vowel shifts combine to create what is known in popular culture as a “Southern drawl”: their effects are summarized in Table 13.7.

The full set of shifts shown in Table 13.7 is only found among some speakers in the South and in some subregions more that others. Monophthongization of /ay/, which appears to be the initiating development of the Southern Shift, displays the widest spatial distribution, being found over most of what most people consider to be the South in a broader cultural sense: from Texas to Virginia, and from Kentucky, on the Ohio River, down to Mississippi and Alabama, on the Gulf of Mexico (ANAE 131). Glide deletion before a smaller class of liquid and nasal consonants, in words like tile, tire, and time, is variably found in an even larger region, reaching across the Ohio River into parts of the southern Midland, from southern Illinois across to Philadelphia; the ANAE therefore specifies glide deletion before obstruents and word-finally, as in tide and tie, as the diagnostic criterion for the South. Within the region established by this criterion, there are two subregions where the Southern Shift is particularly advanced, both in inland rather than coastal areas: one in North Texas, from Lubbock to Dallas; the other in the Appalachian region, including eastern Tennessee, western North and South Carolina, and northern Georgia and Alabama.

In the remainder of the South, including older coastal enclaves like Ocracoke Island, Charleston, Savannah, and New Orleans, the Southern Shift is less consistently present, displaying only a subset of its elements, or sociolinguistic variation within communities (for Ocracoke, see Wolfram and Schilling-Estes 1997; for Charleston, see Baranowski 2007). It is almost entirely absent from areas subject to strong non-Southern influence, including Washington, DC, on the northern edge of the South, and central and southern Florida: Orlando, Tampa, and Miami are not Southern cities in the linguistic sense. Moreover, unlike the Northern Cities Shift, which is most advanced in the major cities of the Great Lakes region, the Southern Shift is associated instead with smaller towns and rural areas, where it is identified with traditional Southern culture: Thomas (1997) documents this urban–rural split in Texas. Younger speakers in the largest urban centers of the South – especially Dallas, Houston, and Atlanta – often lack most or all components of the Southern Shift, which appears to be receding over time (ANAE 253). While back vowel fronting remains a vigorous change, supported by parallel developments in other regions, the front part of the Southern Shift is now subject to negative social evaluation and therefore rejected by young, urban speakers, particularly women; Fridland (2001) reports this development in Memphis, Tennessee, and Dodsworth and Kohn (2012) confirm it in Raleigh, North Carolina.

The Midland: Pittsburgh, Columbus, Cincinnati, Indianapolis and St. Louis

The existence and geographic extent of a Midland dialect region has been the subject of considerable debate among students of American dialect. Kurath proposed a broad Midland region between his North and South, extending westward from Philadelphia into the Appalachian Mountains and southern Midwest (PEAS, Map 2). Subsequent analyses, including that of the ANAE, have treated the Midland as a transition zone, characterized more by a gradual recession of Southern features as one moves north than by unique features of its own. Thomas (2010) examines the North–Midland boundary in Ohio; Habick (1993) reports on aspects of the Southern Shift heard in central Illinois; and Marckwardt (1957) and Frazer (1978) show the transitional nature of Midland speech across the entire North–Central region. The ANAE finds strong fronting of back up-gliding vowels across the Midland, as well as a tendency to merge PALM-LOT and THOUGHT, already complete in the Pittsburgh area. Unlike the Inland North, whose cities display a uniform development of the NCS, Midland cities, like Pittsburgh (Johnstone and Kiesling 2008), Cincinnati (Boberg and Strassel 2000), and St. Louis (Murray 1986), are characterized by somewhat greater diversity. Pittsburgh, for instance, displays a monophthongization of /aw/ (downtown stereotypically becomes dahntahn); Cincinnati has its own, simpler version of the Mid-Atlantic tensing and raising of /æh/ (BATH); while St. Louis has a unique system of back vowels before /r/, with NORTH distinguished from FORCE but merged with START (horse and hoarse are different, but born and barn are the same).

The West: Denver, Phoenix, Seattle, San Francisco, and Los Angeles

Little can be said about the West beyond what was said about SNAE above: there is almost nothing to distinguish them. The double merger of PALM, LOT, and THOUGHT is complete throughout the region: Reed (1952: 186–187) reported its progress in Washington State two generations ago. The West also has a single /æ/ vowel with raising only before nasals (in band and ham) and a more moderate fronting of back up-gliding vowels than is found in the Midland or South. Along the West’s eastern edge, several cities have a transitional status between the West and other regions. The largest of these is Minneapolis-St. Paul, which has the low-back merger of the West but the general raising of a unified TRAP-BATH vowel characteristic of the Inland North, as well as a typically northern resistance to the shifting of long up-gliding vowels, so that FACE and GOAT have an almost monophthongal quality, [e:] and [o:]. Much of the Great Plains region, including cities like Des Moines, Omaha, Kansas City, and Tulsa, displays a mixture of Midland and Western features, to the extent that these can be distinguished.

Canada: Vancouver, Edmonton, Calgary, Toronto, Ottawa, and Montreal

Like the West, most of Canada features a type of English that is difficult to distinguish from SNAE. The double low-back merger of PALM, LOT, and THOUGHT is complete across the country and TRAP and BATH are a single phoneme, /æ/, with raising only before nasals (Boberg 2010: 125–130). Canada shares the fronting of GOOSE with much of the United States but has comparatively little centralization of GOAT (Boberg 2010: 144). Nevertheless, two phonetic variables do distinguish Canadian English from neighboring American dialects, in addition to the retentions of British phonemic incidence and a unique foreign (a) pattern, discussed above. In Ontario, the most important of these is the Canadian Shift, a vowel shift that involves an opposite development of TRAP and LOT to that found across the border in the American Inland North (Boberg 2000). First identified by Clarke, Elms, and Youssef (1995) and confirmed as a change in progress by later work (ANAE 216–224; Boberg 2010: 230), the Canadian Shift involves a retraction of TRAP into the low-central position left empty by the low-back merger. As TRAP moves back, DRESS moves down toward the low-front quadrant, pulling KIT down behind it. Retracted Canadian TRAP in Ontario has the same phonetic quality as fronted LOT across the border in south-eastern Michigan and western New York: a Detroit or Buffalo pronunciation of solid might be misunderstood as salad in Toronto, and a Toronto pronunciation of black might be mistaken for block in Detroit or Buffalo.

The stark cross-border difference found around the Great Lakes gradually weakens as one moves west, until it all but disappears on the Prairies and the Pacific coast. Here, a common low-back merger prevents the Canadian Shift from being as distinctive as it is further east; Kennedy and Grama (2102), in fact, report a similar development in California. Instead, another feature, Canadian Raising, provides a more subtle degree of difference. First described in Ontario English by Joos (1942) and in Vancouver by Gregg (1957a), and later studied more extensively by Chambers (1973) and Boberg (2010: 149–151; 204–205), Canadian Raising produces non-low nuclei in the low diphthongs /aw/ and /ay/ (MOUTH and PRICE) before voiceless obstruents. Thus, cow, proud, tie, and tide have low nuclei, [aʊ] and [aɪ], whereas the nuclei of house, doubt, tight, and spice are raised to lower-mid position, ranging from [ɛʊ] to [ʌʊ] for /aw/ and from [ɐɪ] to [ɜɪ] for /ay/. Raising of /ay/ in pre-voiceless environments has also been noted in a number of American dialects: most famously on Martha’s Vineyard, Massachusetts, by Labov (1963), but also across much of the northern United States (ANAE 205–206). Raising of /aw/, by contrast, is more uniquely Canadian, and has inspired the most common American stereotype of Canadian speech, which has Canadians saying oot and a boot for out and about (like most stereotypes, this one is an exaggeration; in Western Canada, where the raised vowel is further back than in Ontario, a more accurate re-spelling would be oat and a boat). Boberg (2010: 156) demonstrates that a combination of Canadian Raising of /aw/ and retraction of /æ/ in the Canadian Shift separates most young speakers of Canadian English from their American peers, some of whom display moderate versions of one or the other feature but not both. The phonetic effects of the Canadian Shift and Canadian Raising are indicated in Table 13.8.

Boberg (2010) finds that the type of Canadian speech portrayed in Table 13.8 is particularly dominant across western and central Canada, from British Columbia to Ontario; it is also heard among some ethnic groups, particularly British-Canadians, in Montreal and to an increasing extent in Atlantic Canada, especially among younger, upwardly mobile people. Older, more locally oriented people in eastern Canada tend to speak a wider variety of local dialects, which limited space prevents us from discussing here: from traditional enclaves in the Ottawa Valley of eastern Ontario and several parts of the Maritime provinces to the highly distinctive dialects of Newfoundland, established by early nineteenth century settlement from south-western England and south-eastern Ireland (see Clarke (2004, 2010) for a description of Newfoundland pronunciation, which includes a low-central, unrounded vowel for LOT and a mid-back, rounded vowel in STRUT, in contrast to their usual qualities in mainland Canadian English).