This book provides a delightful survey of the global variety of pronunciation and usage of English as an educated standard. It focuses on the phonetics of the various Englishes, especially on the vowels, where so much of the variability resides, and on differences in usage, lexical and syntactic as well as orthographic. Although this small volume necessarily deals with most topics briefly, it includes a wealth of detail.

T&H's observations are largely accurate, but, even though the second author is American, they have something of a tin ear for the English of the United States (USEng). Consider the following examples illustrating the differential possibilities of coreferential pronoun deletion in English English (EngEng) and USEng (p. 79). The EngEng examples This shirt has two buttons off and What kind is that tree with flowers round? are said to correspond to USEng This shirt has two buttons off it and What kind is that tree with flowers round it? More expected USEng would be This shirt has two buttons off of it and What kind of tree is that one with flowers around it? Among other points of difference here, EngEng and USEng differ remarkably in the use of round as a preposition; this point is missed in another set of examples:

In fact, in normal USEng usage the word round, again, would be expressed around; moreover, the authors do not attempt to reflect the different conventions for quotation marks in the two styles. Similarly strange examples recur throughout the discussion of USEng.

The book begins with the RP (Received Pronunciation) accent of EngEng, then proceeds to other (British and colonial) English varieties of the language: EngEng other than RP, Australian (AusEng), New Zealand (NZEng), South African (SAfEng), and Welsh English (WEng) (all in chap. 2); North American (NAmEng), including USEng and Canadian English (CanEng) (chaps. 3 and 4); ScotEng, Northern Irish English (NIrEng), and Southern Irish English (SIrEng) (chap. 5); West Indian English, including standard Jamaican English (JamEng), and, in a departure from the avowed topic of the book, a brief discussion of English-based pidgins and creoles (chap, 6); “Lesser-known Englishes,” English as spoken in small settlements of long standing such as Bermuda, Pitcairn Island, and the Falklands (chap. 7); and second-language varieties of English where these are institutionalized, as in South Asia and the Philippines (chap. 8).

As might be expected from this organizational scheme, the description of RP establishes a baseline for comparison. Using RP as the accent of orientation is useful in that many readers will be familiar with it – and it has been described elsewhere in detail (cf. Jones 1926, Jones 1997, Upton et al. 2001) – but it is an unfortunate choice in that RP is one of the most highly evolved of the present-day standard varieties of English, and thus comparisons are often a bit strained, at least from a historical point of view. Also, T&H tend to cite the phonemes of the various kinds of English using their RP phonetic values, so that, for example, the diphthong of bout is given as

no matter how it is pronounced, whether

in NZEng (p. 17),

in SAfEng (29), or

in WEng (31), except that it is represented as /au/ for ScotEng (92), JamEng (109), WAfEng (125), EAfEng (129), Singaporean English (136), and Philippine English (139). The motivation behind this treatment seems to be that if there is no autonomous phoneme

in the accent in question, the representation is /au/ rather than

. This makes little sense if

is a phonemic unit, as the discussion throughout would have it; but if

is a sequence of phonemes, then the elements surely should be represented in terms of what they sound like, for example, as

for NZEng

, or as

for SAfEng

], rather than in terms of what sounds an RP speaker would substitute.

This RP-centered perspective finds American English confusing and even bizarre. For example, “Foreign learners may find the distribution of

in USEng confusing and hard to learn” (37). This imagined confusion has to do with the fact that in the variety of USEng in question,

rather than

is found before a syllable-closing (i) voiceless fricative

or (ii) voiced velar

. The only difference between the American pattern and the RP described by Jones 1926 (e.g., soft

, with a less common variant

), is the extension of tensing to the environment before the velar nasal – long, song, strong – and sometimes before the voiced velar stop: frog, log. The unsolved problem here is to explain exceptions such as cog (always with the lax vowel) and dog (always with the tense vowel). (Unless, of course, there is complete merger of

as in much of USEng.) This may be imagined to be confusing to the foreign learner simply because it is counter to the tendency in RP of recent generations to suppress

in favor of

in this environment; cf. Jones 1997, which has for RP only

for soft.

By comparison, a far greater challenge for the foreign learner is in learning the RP distribution of

. How does one master alternations like

but

? Why is it

but

(the fish),

but

? This differential tensing seems to represent a sound change in progress, but it is slow or frozen progress, at least within RP; there is no change between Jones 1926 and Jones 1997. It might also be noted that the environment for tensing to

is almost exactly the same as the American and Jones 1926 environment for tensing to

, namely before syllable-closing

; compare RP after, path, pass and USEng often, cloth, cross. Confusing the matter further is the fact that in other varieties of EngEng, the distribution of

is quite different (12). It might be interesting to chart the progress of this sound change in what has become known as Estuary English, a near-RP variety widely spoken among educated people in southern England.

Rather than treating the many varieties of English from the point of view of how they differ from RP, how much more straightforward it would be to take a historical approach, deriving all varieties of modern English from a common ground – one that is surprisingly accessible, all the more so owing to the data presented in this rich small volume.

The discussion of

above anticipates the reconstruction proposed. The

in those examples results from the contextual tensing of a vowel that when not tensed results in RP

. This reconstructed lax vowel is *a. It remains [a] in the two most conservative accents reported by T&H, ScotEng (92) and JamEng (113). (It is interesting to note, as reported by Upton et al. 2001:xii, that RP

appears to be reverting to [a], a reversal of a sound change; cf. the reversal of tensing in RP of

in soft, etc., discussed above.) The following is the reconstructable vowel system: six lax vowels

(as in bit, bet, bat, pot, butt, put); five tense monophthongs

(as in beet, made, yacht, boat, boot); three front-gliding diphthongs

(as in weight, bite, boy); and four back-gliding diphthongs

(as in cute, bout, bought, know).

There are many lexically specific departures from this basic system in different varieties of English. Examples include the development of

from *ö as in the adverb just (which for many speakers of USEng is distinct from the adjective just) or from *i, as in sister, which in some varieties of Southern USEng does not rhyme with kissed her (Hill 1962:21).

Within standard Englishes, *ê and *ei and *ô and *ou seem to remain distinct only in some varieties of WEng: made [me:d] < *mêd vs. maid

< *meid and nose [no:z] < *nôz vs. knows

< *nouz (p. 32). Note that *ei and *ou fell together with *ê and *ô in conservative ScotEng and JamEng, where mid vowel monophthongs remain (92, 113), but in the other Englishes, *ê and *ô diphthongized, falling together with *ei and *ou. T&H phonemicize the WEng contrasts as

(32), resulting in /ei/ and /ou/ as orphans, there being otherwise no WEng phonemes /e/, /o/. This shows a hazard of the procrustean treatment of English monophthongal tense vowel nuclei as vowel + semivowel.

The reconstructable system of reduced syllabics is of less interest because differences in this system do not figure in differentiating different contemporary Englishes, except in minor details. This system centers on a reduced vowel or feature that I represent with the nonphonetic symbol

, which also covers the syllabic equivalents of the various resonant consonants:

as in bottom, button, bottle, better, yellow, many. (Thus, I do not use

to represent this sound/feature/phenomenon. Final

in many, fifty is the syllabic equivalent of /j/, i.e., a weak [i] or, in RP,

, not a sequence

.

derives from something

from did not; there were never intermediate derivational stages

.)

It seems fair also to reconstruct variation in just which vowels are subject to reduction, just as in contemporary English; cf. missile, evil, record, python, dictionary: EngEng

; USEng

. The only serious systemic changes seem to be the loss of non-prevocalic

in favor of simple

in non-rhotic varieties of English; the differential treatment of

as high or not, with the consequent merger, it seems, of the pronunciation of candid and candied for such speakers; and, as in AusEng and much of EngEng, the phonetic change of

to front rounded (a new

) in contrast with back rounded vocalized

(a new

). T&H note that in NZEng, word-final

(their

) has the quality of

: butter

, a cup

(23); though interesting, this does not seem to represent any systemic change. The importance of a reconstruction such as this is not in the detailed phonetic values of the entities reconstructed but rather in providing an accent-neutral basis for the discussion of the variants of English.

The reconstruction of a non-back rounded vowel *ö underlying RP

is perhaps surprising, but this vowel is rounded in SIrEng (101), “retains some lip-rounding” in conservative JamEng (113), and is involved in the morphophonemic alternation with *iu, as in punish/punitive, study/studious, consumption/consume, which indicates the involvement of both frontness and rounding. Furthermore, the tense equivalent (

in RP) is often rounded, as in NZEng (24), SAfEng (29), and WEng (31).

*u also underlies RP

but also RP

, as in put, bush, cushion. The fact that the

of southern or abundance alternates with

in south, abound points to its being from *u. Most examples of *u have merged with *ö. But in the Midlands and the north of England (12), it is the opposite: *ö has merged with *u, resulting in

from both sources. (Northern EngEng

is often remarkably fronted. Is this a holdover from a feature of *ö?) Laxing of *û has provided most of the extant examples of

, as in book, took (spook remaining unchanged).

*iu survives in USEng as

in what Trager (1972:41) calls an “old-fashioned American pronunciation,” such that rude *riud

and rood *rûd

remain distinct for some speakers. (I find it intriguing that the OED uses iu for this vowel nucleus in specifying pronunciation.) In most accents, the center of syllabicity is in the second component of this diphthong, giving rise to an intermediate *jû. Then, in most accents, the *j is lost within the syllable after liquids and palatals – lute, chews, yew; cf. loot, choose, you. (Note the different syllable division in value, where *j is retained.) This loss of *j is extended in most varieties of NAmEng to the environment of a preceding dental or alveolar so that *iu has merged with *û, as in tune, toon; dune, doom; super, soup (44). There seems to be an unexplained exception with the archaic word thews, which in my experience retains *j after

; compare enthusiasm, which, within this accent, follows the pattern of losing the *j. There are also unexplained exceptions like sure, sugar in which *sj have merged into

, the regular pattern in other environments, as in issue, fissure. There is nothing novel about this account, but it needs to be stated because all too many observers coming at the description of the treatment of *iu in USEng from an RP perspective get hopelessly confused. (I've had an EngEng speaker express to me the expectation that Americans should pronounce cute as coot because Americans pronounce dew the same as do.)

The reconstruction of

for the first vowel of law, daughter, August should not be too surprising, given its typical spellings. It is retained as

in what Kurath & McDavid (1961:22) refer to as the “Upcountry” of the U.S. South.

is a point of considerable instability in the system. In many accents, it merges with

– in much of USEng (37, 44), in Eastern New England English (46), CanEng (48), ScotEng (92), and to some extent in NIrEng (99). In New York City English (not in upstate New York English, which is like general inland northeastern USEng),

and the tense

which merges with it diphthongize to

or

, as in off

(47), showing what might be thought of as a polarity reversal from the postulated source

.

*â is marginal, mainly “foreign”: do re mi fa sol la ti do; yacht; obbligato; guava; Guatemala (compare Alabama). The “foreign” flavor of this vowel does not mean that it is of recent origin. Obbligato is documented in English since 1724, guava since 1555, fa and la since 1325 (dates from the OED). Much of the development of the present-day varieties of English has to do with the expansion of the role of *â and changes in its systemic status. Father may have had *â from early on, but compare rather, lather, variably with /â/ or /a/. Monosyllables in -al- such as calm, palm, half, calf may have developed /â/ from *a (variably in half, calf ). The tensing of *a before tautosyllabic *r, as in star, part, is exceptionless but of uncertain antiquity in the system.

The tense vowels remain monophthongal in ScotEng and in JamEng. In other accents they are subject to various degrees of diphthongization. They may be further distinguished from corresponding lax vowels by having greater duration, as in RP.

In some accents, duration is a matter of syllable shape and morphological boundary phenomena. This is particularly true of ScotEng, in which vowels are noticeably longer before

, and word-finally, and vowels so lengthened in word-final position retain that length even in the presence of an inflectional suffix, resulting in contrasts between short wade, toad, mood and long weighed, towed, mooed (93). A similar condition exists in USEng, in which vowels are lengthened before postvocalic voiced obstruents, resulting in the now famous pair writer : rider, in which the medial consonants reduce to identical taps and the contrast between the words is revealed by the fact that rider has a longer first syllable than does writer. A personal example illustrating this feature of USEng as compared with RP comes from a recent experience in Australia, where there were two colleagues named Bob. We jokingly referred to them as “short-vowel Bob” (the Englishman) and “long-vowel Bob” (the American), and some of us went so far as to refer to English Bob as

and American Bob as [ba:b].

There is serious confusion in this book regarding the treatment of foreign words in NAmEng: “Many words felt to be ‘foreign’ have

in USEng corresponding to the

in RP. Thus Milan is

in RP but may be

[sic; should be

or better

– there is no */a/ in T&H's vowel inventory for USEng, and the first vowel is either tense or reduced] in NAmEng, and Datsun in USEng is

[same vowel representation mistake], as if it were spelled Dotsun” (37). Similarly, “‘foreign’ words spelled with o tend to have /ou/ in NAmEng corresponding to

in EngEng” (37), with examples “Bogota EngEng

, NAmEng

and Carlos EngEng

, NAmEng

.” T&H treat NAmEng as the innovator here, but the sol-fa syllables cited above, in the language since 1325, provide the model of English tense *â for “foreign” /a/, as in fa and la, and of tense *ô for “foreign” /o/, as in the name of the do note. It is further worth remembering that phonetically in NAmEng,

: has been lost in favor of *â, not the other way around. Thus

can be seen as rather conservative pronunciations for words treated as “foreign” while

are more nativized. (Part of the difference in treatment of foreign words doubtless has to do with the difference in the RP and USEng phonetic treatments of *ô. In Mexico City I heard an Englishman order

from a street vendor, I think unsuccessfully. An American who orders

gets served correctly with only slight hesitation. Both the Englishman and the American are aiming at Spanish un taco

.)

The tensing of the vowel of can't deserves special mention. It is somehow separate from the tensing in path, pass, and gets its own isophone in T&H's schematic chart of the relationships of major varieties of English (5). A form that was tensed prior to the Great Vowel Shift is found in the U.S. South and South Midlands,

(Cassidy 1985:524). In RP the form shows post-Great Vowel Shift tensing with backing,

. In most USEng, the vowel of can't shows tensing, usually with raising,

, though this is part of a wider phenomenon of variable tensing of

, the so-called Northern Cities Chain Shift, such that, for example, for some speakers bad

and sad

do not rhyme, and Ann

and Ian

are hard to distinguish (45). In all three accents can't remains distinct from cant

.

Vowels plus *r have an intricate but rather straightforward development.

Other than in ScotEng, *e and *i have merged with *ö before tautosyllabic *r, such that the vowels of fern, bird, hurt that remain distinct in ScotEng (92) are not distinct in other accents.

Vowels tense before tautosyllabic *r. The result of the tensing of *ör is a tense monophthong which will later de-rhotacize along with the rest of the rhotic vowel nuclei in the ancestor of the non-rhotic accents. The tensing of the remaining lax vowels plus *r results in

, the latter monophthongizing except in the U.S. Upcountry South. The subsequent development of the vowels of start, fort is thus different from their counterparts not before *r.

Next comes the “breaking” (to use the term from grammars of Old English) of a non-low vowel before a tautosyllabic *r, as in peer

, pear

, pore

(35). Low

remain monophthongal in this environment: star

, port

. (Throughout I use [r] for the English sound, whatever the details of its phonetic manifestation. T&H sometimes use the turned r symbol for this sound.)

In the inland north of USEng, vowels tense (and break) before /r/ even when /r/ is not syllable-final. Thus *er becomes

.

also tenses (and and breaks) to

, since

is the tense of

(cf. the above discussion of can't). Thus Mary, merry, marry merge, and hurry rhymes with furry. (The vowel of merry has a slightly different history in Philadelphia English, where it merges with the vowel of hurry and furry.) Note that starry

has a vowel that tensed (and backed) prior to this development so its vowel nucleus does not fall together with that of marry.

Then comes the sound change that has done the most to restructure the underlying vowel system: de-rhotacization, the loss of tautosyllabic postvocalic *r. For most non-rhotic accents, the loss of *r is suspended when the following word begins in a vowel (“linking r”), and non-etymological *r may be inserted between words to break up vowel clusters (“intrusive r”). SAfEng has evolved one step further, losing *r in this position absolutely (29). De-rhotacization defines the “English” varieties of English discussed in chap. 2 (see above) as well as Eastern New England USEng and several varieties of coastal Southern USEng. It is also a feature of many other varieties of English that lack a straightforward genetic connection with any specific accent of (British) English, such as South Asian English and creolized African and Caribbean varieties of English.

The vowel systems that result from de-rhotacization are remarkably rich and remarkably different from the proposed underlying system above, especially when an additional sound change, “smoothing” (11), occurs. Smoothing converts triphthongs to monophthongs, e.g., tower

. For a Southeastern accent of USEng, the following series of contrasts is reported: fear

, fare

, fire

, far

, for

, four

, sure

(Hill 1962:111). (In this accent tower remains disyllabic.)

It should be emphasized that the other varieties that remain rhotic – most of NAmEng, ScotEng, IrEng, JamEng, EngEng in the southwest of England – do not form a natural class; they simply have not undergone this catastrophic sound change.

The development of *i in the Antipodes is worthy of mention. AusEng has a rather high vowel for *i, almost [i] (17). NZEng has

for this vowel (23). Not surprisingly, this difference has become a focus of popular culture. In graffiti seen at the beach at Bondi, perhaps during the 2000 Olympics, a Kiwi wrote Australia sux (‘sucks’), and another hand added, New Zealand nil.

This book is a fourth edition and as such should be fairly free of typographical errors, but it suffers from an excessive number of them. In discussing the difference in the use of capitalization after a colon in British and American usage, the supposedly contrasting examples both have a capital letter after the colon (84). Bown appears for bowl (88).

seem to be reversed in describing a possible SIrEng rendition of the contrast between tin and thin (102).

should be

on p. 115. The like themselves should be They like themselves in West African English (126).

Overall, however, this book is a wonderful reference for anyone interested in varieties of English. I had more fun with this book than with just about anything else I've read on this topic.