Professional Documents
Culture Documents
Peas PDF
Peas PDF
Peas PDF
net/publication/284754654
The Pronunciation of English in the Atlantic States: Based upon the Collections
of the Linguistic Atlas of the Eastern United States
CITATIONS READS
36 123
3 authors, including:
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Music and Poetry’s Great Leap (Forward?): The Private Aesthetic of the 20th Century View project
All content following this page was uploaded by Samuel Jay Keyser on 02 December 2015.
Review
Author(s): Samuel Jay Keyser
Review by: Samuel Jay Keyser
Source: Language, Vol. 39, No. 2 (Apr. - Jun., 1963), pp. 303-316
Published by: Linguistic Society of America
Stable URL: http://www.jstor.org/stable/411215
Accessed: 23-11-2015 22:11 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
Linguistic Society of America is collaborating with JSTOR to digitize, preserve and extend access to Language.
http://www.jstor.org
This content downloaded from 18.9.61.111 on Mon, 23 Nov 2015 22:11:18 UTC
All use subject to JSTOR Terms and Conditions
REVIEWS 303
The pronunciation of English in the Atlantic states: Based upon the collections
of the Linguistic Atlas of the Eastern United States. By HANS KURATHand
RAVENI. MCDAVIDJR. (Studies in American English, No. 3.) Pp. xi, 182, with
180 full-page maps. Ann Arbor: University of Michigan, 1961.
Reviewed by SAMUELJAY KEYSER, Massachusetts Institute of Technology
As its subtitle indicates, the Pronunciation of English in the Atlantic states
(PEAS) is based upon the collections of the Linguistic atlas of the Eastern United
States. It is not, however, simply a collation of published material. Recent studies
have been added, in particular those by Raven McDavid in Georgia, Florida,
New York State, and Canada, and his expansion of the late Guy S. Lowman's
preliminary study in South Carolina.1
One of the major difficulties facing any dialectologist is that of presenting his
material in an accessible manner. To accomplish this the authors have employed
'outlines', accompanied by a large number of 'synopses', to describe the 'Regional
dialects of cultivated speech' (Chapter 2). The outlines note the major features
of the areas delineated. The 'synopses' present in tabular form the most impor-
tant allophones of the vowel phonemes in sixty-six key words in the dialects of
seventy cultivated speakers.
Another device used is maps. Maps 4-33 show the regional variants or dia-
1This work was supported in part by the U. S. Army, the U. S. Navy (Office of Naval
Research), and the U. S. Air Force (Office of Scientific Research and Development Com-
mand), and in part by the National Science Foundation (Grant G-16526).
I wish to thank G. H. Matthews and P. M. Postal for their useful comments, Noam
Chomsky for several valuable suggestions, and Morris Halle for his astringent readings of
the manuscript, which have both clarified the argument and spared the reader many digres-
sions.
This content downloaded from 18.9.61.111 on Mon, 23 Nov 2015 22:11:18 UTC
All use subject to JSTOR Terms and Conditions
304 LANGUAGE,VOLUME39, NUMBER2 (1963)
phones for each of the vowel phonemes.2 These maps are augmented by Chapter
3, 'The regional and social dissemination of the diaphones of stressed vowels'.
Maps 34-58 present the distribution of the consonant /r/ and the incidence
of vowels before /r'a/. Chapter 4 deals with this aspect in detail, the authors
having decided to devote a separate chapter to it because of the extremely com-
plicated nature of the material and its interpretation. The remaining maps,
59-180, present the material discussed in Chapter 5, which deals with the 're-
gionally and socially varying incidence of vowel and consonant phonemes in the
vocabulary'. (This chapter is chiefly the work of McDavid; Kurath is responsible
for Chapters 1-4.)
The use of synopses, outlines, and simplified maps has effected a great economy
in presentation, and for anyone who has attempted to use, for example, the
Linguistic atlas of New England, the compactness and accessibility of the data in
this volume cannot be praised enough.
The basic orientation of the book, as is evident from the attention paid to
devices for presenting data, is extreme empiricism, an orientation the authors
share with many contemporary linguists. They see their primary task as that of
presenting data in the most convenient form; the reader gets no specific account
of the theoretical framework behind their study. They are, to some extent, sus-
picious of all contemporary schemes of phonemicization, even their own, and
they are careful to point out to the reader that their theoretically motivated
results are not irreversible, that the original data can always be recovered (2):
A phonic record of pronunciation is needed for all attempts at establishing the system of
phonemes and describing the phonic range of the several phonemes (i.e. their positional
and prosodic allophones or their free variations). The preservation of the phonic record is
all the more important since all schemes of phonemicization are to some extent arbitrary
in view of the present state of our knowledge of the segmentation of utterances. If the
basis of the adopted scheme of phonemicization is explicitly stated and each phoneme in
that system is described in phonic terms, the phonic data thus preserved can be interpreted
in terms of different systems of phonemicization.
It would be a misunderstanding to regard the study simply as a record of more
or less well-chosen facts. On a number of points the authors take positions that
are of the greatest interest for linguistic theory. Since the general tenor of the
study may obscure this fact, I shall examine a few of these points in detail.
I take the theoretical orientation of PEAS to consist essentially in the follow-
ing three conditions: (1) STRONG BIUNIQUENESS, (2) UNIFORMITY, and (3) MINI-
MIZATION. By strong biuniqueness I mean the constraint that allophones of a
given phoneme must be phonetically distinct from allophones of other phonemes.
By uniformity I mean the constraint that all phonetic diphthongs must be treated
exclusively as unit phonemes or else all phonetic diphthongs must be treated
exclusively as phonemic sequences. By minimization I mean the constraint that
phonemicizations in dialectology should be so formulated as to yield a minimum
of phonemic heteroglosses between dialects.
The authors have established a hierarchy among the features in which dialects
(and/or idiolects) of a language differ (2):
2 In the Preface (page v) Kurath says, 'The term diaphone as used here comprises all
regional and social variants of a phoneme or its allophones.'
This content downloaded from 18.9.61.111 on Mon, 23 Nov 2015 22:11:18 UTC
All use subject to JSTOR Terms and Conditions
REVIEWS 305
Phonemic, phonic, and incidental heteroglosses are of unequal value in determining the
degree of difference between dialects and in evaluating the relative importance of the
boundaries between speech areas. Of the three types, phonemic heteroglosses obviously
outrank the others.
It should be noted that this hierarchy is not explicitly justified in the book,
probably because it seemed self-evident to the authors, and one can only specu-
late what the justification would be. I do not question their hierarchy. I only
want to observe that the introduction of such a hierarchy is an important the-
oretical step which does not follow in any simple sense from the data and-what
is more important-that a number of other features of their description are
directly dependent on it.3
Given the primacy of phonemic heteroglosses, the authors have, I suggest,
reasoned in the following manner: since phonemic heteroglosses are indices of
the most important differences among dialects and since different dialects of a
given language must resemble each other (what other sense could there be to the
concept 'dialect'?), it is necessary to devise phonemicizations for the individual
dialects in such a way as to minimize the number of phonemic heteroglosses
among dialects. This is condition 3.
An especially clear illustration of the manner in which the authors utilize
that condition is provided by the treatment of /awi/ and /a_u/ in PEAS.4
As is well known, there are three competing phonemicizations of English phonetic
diphthongs: the binary regards them as sequences of two phonemes; the unitary
regards them as allophones of unit phonemes; and the mixed is a blend of the
two, regarding some phonetic diphthongs as phonemic diphthongs and others as
unit phonemes. The authors have not even considered the third alternative and
may well be in agreement with Hockett, who, on the grounds that IC levels are
mixed, rejects that analysis in which the vocalic nuclei of beat, boot, bait, etc. are
regarded as units and those in bite, bout, and Hoyt as sequences.5
Choosing, then, between a strictly binary and a strictly unitary phonemiciza-
tion, the authors reject the former (3):
Although admitting that a binary interpretation of diphthongal phones of American English
more satisfactory than any advanced so far may in time be devised, a unitary interpretation
seems more realistic and certainly more convenient for comparing the dialects of English
spoken in the Altantic States and establishing degrees of diversity of kinship between them
on a stated basis.
The reasons for their choice are further amplified in the following (4):
In most dialects diphthongal and monophthongal variants occur side by side under ob-
servable (hence statable) conditions, i.e., in certain positions or under certain prosodic
3 A similar hierarchy may be found in Uriel Weinreich, 'Is a structural dialectology
possible?', Word 10.392 (1954).
4 Since I shall be discussing both unitary and binary interpretations throughout, to
avoid confusion I adopt the following notation: letters (between slant lines) that are joined
by a link or tieline represent unit phonemes; otherwise, two adjacent symbols between
diagonals represent binary sequences of phonemes.
6 Given condition 2 above, however, it is not surprising that the authors pay no attention
to the third alternative and probably do not consider it viable. - For Hockett's complete
argument see Manual of phonology160 ff. For a criticism of his position see Chomsky, IJAL
23.229 ff. (1957).
This content downloaded from 18.9.61.111 on Mon, 23 Nov 2015 22:11:18 UTC
All use subject to JSTOR Terms and Conditions
306 LANGUAGE, VOLUME 39, NUMBER 2 (1963)
conditions. For instance, a speaker may have [I] in bit, [n] in put, [ae]in bat and [a] in lot
but [19] in crib, [ua] in good, [se] in half and [a-] in rod.Y
phonic. This decision may be based upon their realism principle, though one
could also argue that the decision follows equally well from the convenience
principle. Like realism, however, convenience is nowhere defined explicitly. I
suspect that it is essentially embodied in condition 3 above. A binary interpreta-
tion meeting the conditions of uniformity and strong biuniqueness would neces-
sarily lead to an increase in the number of phonemic sequences postulated for the
Atlantic States dialects. An increase in the number of phonemic sequences,
however, carries with it a proliferation of phonemic heteroglosses (as will be
seen in detail below). But this proliferation directly counters condition 3, that
phonemic heteroglosses be minimized. The unitary interpretation, which mini-
mizes phonemic heteroglosses, is therefore termed more convenient.
We find that the choice of the unitary interpretation has allowed the authors
to assert that the major dialects spoken in the Eastern States all share the free
vowels /a vi/ and /a wu/. This assertion amounts to the complete elimination of
phonemic heteroglosses in the major dialects for /awi/ and /a u/, and it
attributes the differences among the dialects to the less crucial phonic hetero-
glosses. The cost of organizing the data so that this assertion holds, however, is
5a Because of typographical limitations, no distinction will be made between full-size
and superior phonetic symbols. Thus the superior [o] in this passage is reproduced here as
full-size [a].
This content downloaded from 18.9.61.111 on Mon, 23 Nov 2015 22:11:18 UTC
All use subject to JSTOR Terms and Conditions
REVIEWS 307
This content downloaded from 18.9.61.111 on Mon, 23 Nov 2015 22:11:18 UTC
All use subject to JSTOR Terms and Conditions
308 LANGUAGE, VOLUME 39, NUMBER 2 (1963)
This content downloaded from 18.9.61.111 on Mon, 23 Nov 2015 22:11:18 UTC
All use subject to JSTOR Terms and Conditions
REVIEWS 309
This content downloaded from 18.9.61.111 on Mon, 23 Nov 2015 22:11:18 UTC
All use subject to JSTOR Terms and Conditions
310 LANGUAGE, VOLUME 39, NUMBER 2 (1963)
weaker than the authors', since it requires fewer constraints and treats minimiza-
tion not as a condition but as an interesting empirical question. As will be seen
below, the question is in fact answered positively.
Every system of phonemic analysis must account for the relationship between
the abstract phonemes and the phonic substance (allophones or phones). In the
authors' theoretical framework this account is embodied in the entries in the
synoptic tables of Chapter 2. In the following I shall make use of a different
convention. Instead of being shown as entries in the synoptic tables, the allo-
phonic variants will be given by means of statements. For example, the material
collected for the Charleston dialect (Synopsis 136 above) is accounted for by the
following statement or rule:
Rule 2. a -* ae before u.
The Winchester dialect shares Rule 1 with Charleston and Rule 2 with New
Bern. The allophonic rules accounting for Winchester, then, are:
If we postulate (as we have for Charleston and New Bern) that the diphthongs
in Winchester are represented by the sequences ai and au, Rule 1 will operate to
produce v in twice and out. If we next apply Rule 2, it will operate to produce
e in down since, after Rule 1 has been applied, it is only in this word that a now
occurs before u. Notice, however, that in order to achieve the Winchester output
from Rule 1 and Rule 2, an important restriction has been placed upon the rules:
Rule 1 must precede Rule 2. One important consequence of this restriction is that
it produces rules which are, in an obvious way, simpler than rules requiring no
order. For example, if we were to account for the Winchester data by rules un-
restricted as to order, we would require, beside Rule 1, the following counterpart
to Rule 2:
This content downloaded from 18.9.61.111 on Mon, 23 Nov 2015 22:11:18 UTC
All use subject to JSTOR Terms and Conditions
REVIEWS 311
This content downloaded from 18.9.61.111 on Mon, 23 Nov 2015 22:11:18 UTC
All use subject to JSTOR Terms and Conditions
312 LANGUAGE,VOLUME39, NUMBER2 (1963)
This content downloaded from 18.9.61.111 on Mon, 23 Nov 2015 22:11:18 UTC
All use subject to JSTOR Terms and Conditions
REVIEWS 313
course, that the authors knew what they were looking for. If, on the other hand,
their material was gathered without any specific theoretical framework in mind,
then whether the facts collected can accommodate future phonemicizations is
not a matter of preservation of the phonic record so much as a matter of pure
chance.
I have tried to show that by adopting a unitary interpretation of diphthongal
phones, the authors have been forced to pay a high price by increasing the
phonemic register in idiolects more economically treated by a binary interpreta-
tion, so that, in their effort to minimize phonemic heteroglosses, they have al-
lowed their description of one dialect to influence that of another.
These inadequacies may be overcome by two fundamental departures from
the theoretical framework in PEAS. The first is to abandon strong biuniqueness,
uniformity, and minimization as theoretical requirements in phonemic analysis.
The second is to impose a specific order on the application of allophonic rules.
By making use of a binary interpretation and by abandoning strong biuniqueness
and uniformity as a-priori theoretical constraints upon that interpretation, we
have been able to effect economy where PEAS has not and still arrive at intui-
tively satisfying empirical results, namely the minimizing of phonemic hetero-
glosses in the Atlantic States.1l By ordering our rules, we have been able to
simplify our description of a given dialect. We have seen, moreover, that through
a comparison of rules and their respective orderings from one dialect to another,
the possibility now arises of accounting for the formation of new dialect groups
through the geographical dissemination of rules.
In the course of PEAS the authors have had occasion to make certain histori-
cal statements. A few of these are questionable and seem to call for discussion.
The following statement appears on page 134:
The type of /jalo/, with the vowel of hollow, which is historically derived from /jela/, oc-
curs chiefly in Tidewater Virginia folk speech.
The derivation of [a] from [ae] in /jala/ for yellow is rather doubtful.'3 The
OE preform is geolu. The occurrence of spellings like 3olowe from the 12th cen-
tury onward (cf. OED) suggests early stress shift in this word.'4 The o spelling in
3olowe is, however, ambiguous, since it may represent rounded o in the West.'6
But dialectal forms with [jola]in Yorkshire, Lancashire, and Leicester'6 definitely
point to an early o, which could easily have been unrounded to a, i.e. [a]. (The
OED records a forms from the 14th century onward.) A similar suggestion is
I1 It should be noted that the phonemic analysis presented here corresponds precisely to
that which Hockett considered indefensible (see fn. 5 and the corresponding text above for
Hockett's position).
18 I am indebted to Helge Kokeritz of Yale University for bringing this error to my
attention.
14 With this can be compared NE show from OE sctawian, assuming stress shift
to sceawsan. NE shew derives regularly from sctawian without shift of stress.
'5See Richard Jordan, Handbuch der mittelenglischen Grammatik?65, ?66 (Heidelberg,
1934).
16 See Joseph Wright, English dialect grammar, Index (Oxford, 1905).
This content downloaded from 18.9.61.111 on Mon, 23 Nov 2015 22:11:18 UTC
All use subject to JSTOR Terms and Conditions
314 LANGUAGE, VOLUME 39, NUMBER 2 (1963)
This content downloaded from 18.9.61.111 on Mon, 23 Nov 2015 22:11:18 UTC
All use subject to JSTOR Terms and Conditions
REVIEWS 315
This content downloaded from 18.9.61.111 on Mon, 23 Nov 2015 22:11:18 UTC
All use subject to JSTOR Terms and Conditions
316 LANGUAGE,VOLUME39, NUMBER2 (1963)
This content downloaded from 18.9.61.111 on Mon, 23 Nov 2015 22:11:18 UTC
View publication stats
All use subject to JSTOR Terms and Conditions