many of the Scotch-Irish by way of the Shenandoah Valley and Kentucky.

The most valuable contribution of Krapp to our knowledge of American English is certainly the second volume, which is devoted to a detailed study of the history of our pronunciation. Unintentionally 'phonetic' spellings in the New England town records and the rarer Virginia documents, the rimes of the poets of the revolutionary period, and the statements of American orthoepists (Webster, and others) are used to determine the earlier pronunciation and the fluctuations and changes in pronunciation.

In general, Krapp's judgement can be trusted. But one can hardly agree to his interpretation of certain 'phonetic' spellings, which I shall point out here. (a) The spelling Fleg for Flag and the reverse spelling salf for self, etc. (95) in the New England town records are probably occasioned by a closer pronunciation of [e], which still distinguishes eastern New England, rather than being 'merely orthographic' substitutions, as Krapp hesitatingly suggests. (b) The spellings reast for rest (1695), and cheast for chest (1681) probably do not indicate [tSi st] and [rist], but [test] and [rest], possibly with a somewhat lengthened vowel (104). (c) The spelling of the vowels in booshel, boolet (Watertown Records, 1651) probably means no more than a wellrounded [u] rather than [u] (150).

On the whole, Krapp has too much confidence in the purity of rimes. He does not strictly abide by the sound principle (II 5) that rimes 'may serve a useful purpose as providing confirmatory evidence for points in pronunciation which are more or less established by other evidence', which he sets up. This is apparent in the following cases: (a) Rimes like head: maid, tread: shade (88) do not prove that the vowel in head and tread was long, and the long vowels of daid, haid in dialect stories (89) are not 'popular survivals' of a long vowel, but arise from the wide-spread Southern tendency to pronounce all originally slack short vowels in monosyllables long and rather close, i.e. bed [bed], bid [bid], mud [mad], etc. (cf. fraish for fresh in Tennessee mountain dialect (94), and aige for edge (105) in Georgia speech). (b) Krapp's opinion that the vowel [1] as in sit was formerly pronounced 'with a higher value, approximately as [i]' (113), as rimes like sin: seen 'occur not infrequently', is flatly contradicted by the numerous e-spellings in the town records, as in hender for hinder (115), as well as by the general tendency in the development of this and other slack vowels. (c) Rimes like state meet, brain: unseen, and tree: lay,

which Krapp includes without comment among the flame: beam rimes containing mid-front-open and mid-front-close vowels (128), rudely shake one's confidence in the purity of the rimes of Barlow and Freneau and their age, for meet, unseen, and tree certainly had [i], which is here rimed with [e] or [e]. And the vowels in Freneau's imperfect rimes like mourn urn probably were not much different from what they are now in the East, and not [o] and [u], respectively (140), for the vowel in urn had certainly been unrounded to [A] long before Freneau's time. (d) Learn: man was surely not 'a perfect rime according to New England pronunciation of the time' (220), i.e. 1786-7, but rather [lɛən, læǝn]: [mæn].

As regards Krapp's interpretation of the statements of the older orthoepists, one can hardly agree with him on all points. (a) He does Webster an injustice when he calls the eminent lexicographer's statement that 'a in cart has its short in carry' a 'physical absurdity', adding: 'unless one pronounces carry as [kari], a pronunciation which Webster certainly did not have in mind' (87). Why not [kærı] and [kært]? Even if Webster did not advocate this pronunciation of cart, it was still current in New England in his day, as one may gather from Dwight's rimes far: war: spare: bare (83), to mention only one source. To be sure, Webster is not a careful observer. He is too prone to consider his personal pronunciation as the best, and to defend it either by an appeal to anti-British feeling or, in a pinch (122) by referring to British authorities. He is of prime importance because of his vast influence rather than his judicious observations. Krapp's estimation. of Webster wavers (cf. 87, 136). (b) The universal preference of all the earlier dictionaries for [a] in haunt (78) is more probably due to New England bias than to Walker's dictionary. (c) There seems to be no reason for assuming that Alexander had in mind 'approximately [a] or [ɔ]' when he wrote in his Columbian Dictionary (Boston, 1800) that 'the sound of a in bare, though long, is really distinct from the sound of a in bate .. the aperture is greater in sounding a, in bare, and approximates near to the apperture of sounding ă, in ball' (108). He probably meant [æ], which is still current in New England, the Western Reserve, and elsewhere (see Kenyon, American Pronunciation 93), and is much more widely used than Krapp's statement (111) implies.

Krapp's descriptions of the articulation of sounds and of changes in articulation are not always carefully worded, and in several cases rather faulty. (a) One is surprised to read that the point of the tongue is

'pressed against the back of the lower teeth' in '[A] and all other vowels' (167). (b) The statement (218) that the consonantal r is produced by 'lifting the point of the tongue so high as to cause an actual contact between the tongue and the palate' is misleading. (c) The explanation (184) of the dialect forms yarb, yarn for herb, earn is rather cryptic: "The syllabic character of the r seems here to cause first a diphthongal vowel preceding it, the first element of which becomes consonantal through shifting of the stress to the second element'. If the r was preceded by a vowel it could not be syllabic; moreover, the author does not concede (167) the existence of a syllabic r even for the present day. The y in yarb, yarn is, in origin, rather a glide between vowels in phrases like the herbs, he earns. (d) It would be better to speak of the change from [dj] (instead of [d]) to [dz] in duty (201), and from [tj,tç] (instead of [t]) to [t] in tune (234). (e) The restitution of [dj] for [dz] in obedient, etc., on the one hand, and the preservation of [tS], from older [tj], in nature, etc., on the other, cannot be explained by 'the influence of orthography' (201), for in that case we should also expect the restitution of [tj]. The cause is a phonetic one: [tS] is so different from [tj] that [ne tjər] for [ne-t§r] sounded too much like an affectation, while [obi djant] for [obi'dzant] was only a slight departure. (f) The glide sound [j] in the formerly fashionable, but now dialectal, pronunciation of garden as [gjardn] and card as [kjard] did not arise after [ær] had become [ar] (208), for, aside from purely phonetic objections, dialectal Southern girl [gjæl] and probably also cow [kjæu] never had the back-vowel [a], and thus should not have the glide.

The analysis of present pronunciations is not always correct. (a) Regarding the vowel sound in firm Krapp says: "The writer is convinced, however, that ordinarily in the speech of the Americans who may be said to pronounce their r's, the reverted vowel is followed by a slight consonantal frictional element which can be designated only as [r]' (167); and, accordingly, he transcribes firm as [form]. This point will bear experimental investigation. In my opinion the great majority of Americans, whether they turn up the tip of the tongue or not (perhaps all except the recent immigrants from Scotland and Ireland, and the [AI]-New Yorkers), have only three sounds in firm, hurt, fern, etc., thus Eastern [fam, hзt, fan] and Western [fum, hit, fin]. At any rate, the upward movement of the tongue is no more noticeable than in the vowels of beam and boom, and less so than in the vowels of aim and home, which Krapp prefers to write as simple vowels. (b) Midwesterners will be surprised to read that the lengthened [ɔ] in soft,

moss, etc., 'is not found in

American English, except occasionally in rustic dialects' (30); and that hurrah ending in [a] is only 'sometimes heard, especially in the pronunciation of women', the 'most common' pronunciation being [hu'rɔ] (47).

Whatever shortcomings Krapp's treatment of American pronunciation and its history may have, arise in a large measure from its character as a pioneer work; and these deficiencies are greatly outweighed by the unquestioned merits of this first comprehensive treatment of American pronunciation. Many an excellent observation will be found by the reader. We must content ourselves here with pointing out a few of them: (a) 'Academic and dictionary authority is strangely opposed to the pronunciation of [u']', instead of [ju], in lute, rule, new, duty, 'and it is often characterized as a mark of vulgarity or illiterate speech, in spite of the fact that it is and has long been widely current in the speech of persons of undoubted cultivation and education', such as Noah Webster and W. D. Whitney (155). We need to be reminded of this. (b) The treatment of the perplexing fluctuations in the pronunciation of the vowels of fir, fur, fern, learn toward the end of the eighteenth century from locality to locality and even from person to person-as in loss, log, wash today-is very helpful (177 ff.). (c) The influence of Negro talk on polite Southern speech is rightly discounted: "The negroes omitted their r's because they heard no r's in the speech of their white superiors' (226). (d) We must welcome Krapp's sound statement that 'the questions of direct and determining influence of British upon American speech [i.e. in matters of pronunciation] since the period of migration must always be stated with many qualifications and limitations' (33), and that 'there is no evidence to indicate that any single pronunciation which has become general in America has become so through imitative influence of British pronunciation' (80). Purveyors of imported pronunciations would do well to realize this fact..

HANS KURATH

Grammar of Early Welsh; Part I Phonology. (Supplement Philologica, Vol. II) Pp. 178. By Jos. BAUDIŠ. New York: Oxford University Press, American Branch, 1924. For the Philological Society, London, England.

It is now nearly twenty years since Strachan's Introduction to Early Welsh, a work which has served well the present generation of Keltic

1 The reviewer wishes to explain that though published in 1924 the work came into his hands for review in Language early in 1927.

scholars, was published. Strachan's book was both written and issued under very special conditions, and it was inevitable that it should be superseded earlier than it would have been had its author lived to give it its final form and see it through the press. Besides, much important work has been done in the field of Keltic linguistics since 1909. Not to mention a large number of discussions of special topics, many of them of considerable length and importance, including at least two of great merit by Professor Baudiš himself, which have appeared in the periodicals or separately, we have had new standard works in Pedersen's great Vergleichende Grammatik and Morris-Jones' Welsh Grammar, both of which Baudiš has naturally had constantly before him. His own Grammar of Early Welsh, the completion of which will be eagerly awaited, will, some defects notwithstanding, certainly rank with these. His purpose, as described in his own words, was to provide students of Early Welsh with a grammar 'serviceable and useful both in study and reference', and this purpose he may justly claim to have accomplished as regards the phonology with great success, a success qualified mainly by the somewhat unfortunate method of presentation of his materials. Like Strachan-or rather, perhaps, like Strachan's editor, for there can be little doubt that had Strachan written the preface to his Introduction himself he would have left no room for question-Baudiš has not defined the term 'Early Welsh'. Clearly neither he nor his predecessor use it in the same sense as Morris-Jones; and in his text Baudiš (again with Strachan) regularly writes of Old Welsh, but occasionally uses the term Early Welsh (e.g. on pp. 27, 144) but without stating what distinction, if any, he makes between OW and EW.

Baudiš' book is essentially a descriptive grammar, and to most students of Early Welsh, whose interests in Welsh at all are mainly if not solely linguistic, it will be for that reason disappointing and often less useful than Morris-Jones's book, with all its shortcomings. It is true that there is a certain amount of comparison with MW and NW, and also some haphazard comparison-at least the present writer has not been able to discover any underlying system in accordance with which it is provided with other Keltic or IE tongues, which, however, is often hidden away in footnotes or in paragraphs printed in smaller type, one or two of which have the appearance of having been added as if by an afterthought (as §30 a, on p. 21, which unnecessarily repeats two items just given on p. 16). The extremely concise method of

2 This criticism does not apply to the useful summary account appearing on pp. 167ff.