2.

From Heterogloss to Dialect Boundary

2.1 Introduction

The data secured by systematic selective sampling and presented cartographically as described above furnish the basis for outlining the dialectal structure of the area surveyed.

As a first step in working out the areal structure, only items exhibiting fairly clear-cut dissemination patterns are taken into consideration. The expectation is that the items whose variants appear to be disseminated in a complicated or haphazard fashion can best be dealt with after the skeletal structure of the area has been set up tentatively.

In handling the items one by one, attention is focused on the lines that separate the variants, on the heteroglossic lines. These delimit the areas occupied by the variants, on which attention is concentrated in dealing with features peculiar to the several subareas.

When the heteroglossic lines (heteroglosses, for short) drawn for the items provided by the sample are assembled on maps, one finds that in some parts of the area they run in bundles of various sizes— close-knit or spaced. These bundles show the location of major and minor dialect boundaries and thus indicate the dialectal structure of the total area.

Every systematic sampling survey has revealed this situation, definitely disposing of the earlier notion that heteroglossic lines are more or less evenly spaced [so G. Paris 1888; perhaps also H. Schuchardt 1928 (< 1900): 166–88]. That perhaps no two hetero-glosses run precisely the same course is not in conflict with the observation that they often do form bundles.

To evaluate the relative importance of the dialect boundaries suggested by the bundles, it is not enough to count the heteroglosses composing the bundles. The heteroglosses must be evaluated from the structural point of view before a sound decision can be reached. For this reason, heteroglosses of different kinds should be assembled separately, so that they can be evaluated by groups. Each set will contribute some evidence for subdividing the area; taken together they will furnish the basis for a generalized scheme designed to exhibit the dialectal structure of the area. This procedure is laborious, and to some extent arbitrary. But there is no other method that will do justice to the evidence yielded by the sampling survey.

From the structural point of view, heteroglosses fall into three major classes: phonological, morphological-syntactic, and lexical. In turn, phonological and morphological-syntactic heteroglosses are either structural or nonstructural. Finally, the nonstructural (subphonemic) heterophones are of two kinds: incidental and phonic.

Listed in the order in which they are discussed below, the classes of heteroglosses and their subdivisions are as follows:

I. Lexical

II. Morphological-syntactic: (1) structural, (2) incidental

III. Phonological: (1) structural (phonemic); (2) nonstructural (subphonemic): (a) incidental, (b) phonic

2.2 Lexical Heteroglosses: the Upper South, the Eastern States, the Midwest

The lexical heteroglosses can be more easily handled than the morphological and the phonological. Since the lexicon of a language, though not lacking systematization, is not as rigidly structured as the morphology and the phonology, all heterolexes yielded by the survey can be given the same rank and assembled on a single map.

The procedure is as follows. On an outline map on which every community investigated is set off from its neighbors by lines so as to provide a “honeycomb grid,” the course of each heterolex is entered segment by segment. In the end, the number of heterolexes running between any two communities in the area is recorded on the grid. As a result, the grouping of heteroglosses in bundles of various sizes is brought into relief.

This method is illustrated by E. J. Dearden’s investigation of the regional vocabulary in an area comprising the states of Maryland, Virginia, and North Carolina [Dearden 1943]. In this study, all of the varying lexical items recorded for the Linguistic Atlas of the Eastern United States that exhibit fairly clearly defined regional disseminations are included. The number of heteroglossic lines running between any two communities varies from zero to twenty-seven (see Figure 1). The most prominent bundles form a fairly continuous, though jagged, contour running from the Blue Ridge in Virginia southward into the piedmont of North Carolina. Less prominent bundles separate the Eastern from the Western Shore of Maryland, tidewater from piedmont Virginia, and eastern Virginia from eastern North Carolina. These bundles of heteroglosses form the primary and the secondary dialect boundaries in this extensive area insofar as the lexicon is concerned.

In passing, it should be noted that this procedure also brings into relief the subareas of relative uniform word usage, such as the Piedmont of Virginia and western North Carolina.

The dialect boundaries presented by Kurath in his Word Geography of the Eastern United States [1949] result from this procedure, although it cannot be claimed that the method was applied with statistical precision (see Figure 2).

A few of the heterolexes that set the Midland off from the North and from the South are presented in Figure 3.

This method of establishing dialect boundaries has also been used by A. L. Davis in his Word Atlas of the Great Lakes Region [1948]. Relying upon twenty-five heterolexes exhibiting a fairly clear geographic dissemination, he discovered a prominent bundle of lines that runs in a westerly direction a short distance south of the Great Lakes (see Figure 4). This bundle reflects the westward extension of the dialect boundary that separates the North from the Midland in Pennsylvania. To the north of it the New England variants of the twenty-five lexical items are current.

The method described above is applicable to the data secured in any sampling survey and has been used with good results in several other investigations of the regional vocabulary of American English: J. D. Hawkins 1935; C. T. Hankey 1960; A. H. Marckwardt 1957; G. R. Wood 1963.

To my knowledge, the only European study of regional vocabulary in which an actual count of heteroglosses was used to establish speech boundaries is that of W. Wenzel [1930].

For the general purpose of determining major and minor speech boundaries on the basis of regional vocabulary all heterolexes are accorded the same value, so that the relative importance of such boundaries rests solely on the number of heterolexes in any given bundle. There can be no valid objection to this procedure.

FIGURE 1: Word Boundaries in the Upper South

Cities: B(altimore, C(harlottesville, F(ayetteville, G(reensboro, L(ynchburg, N(orfolk, R(ichmond, W(ashington.

Adapted from E. J. Dearden Dialect Areas of the South Atlantic States.

On the other hand, weighting the heterolexes from a certain point of view, as with reference to their expansive, stable, or recessive character, or with regard to the “ranges” of the lexicon, e.g., the vocabulary of the farm and the home, of business and politics, of literature, of science, etc., has its proper function. Such refinements in method often yield valuable insight into the sociocultural dynamics of a speech area.

FIGURE 2: The Speech Areas of the Eastern States

FIGURE 3: A. Northern Words

1. buttry ‘pantry’ [Kurath 1949: fig. 51]

2. spider ‘frying pan’ [ibid.: fig. 68] darning needle ‘dragon fly’ [ibid.: fig. 141]

4. pail ‘bucket’ [ibid.: fig. 66]

B. Southern Words

5. low ‘moo’ [Kurath 1949: fig. 96]

6. lightwood /laitəd / ‘kindling’ [ibid.: fig. 29]

7. corn house ‘corn crib’ [ibid.: fig. 57]

2.3 Types of Phonological Heteroglosses

The handling of phonological and morphological heteroglosses yielded by selective sampling is much more complicated. Structural heterophones and heteromorphs must be kept apart from the non-structural.

Phonological heteroglosses fall into three distinct classes, which owing to their relative importance in the structure of the language, must be separately handled.

Phonemic heteroglosses exhibited in the data are obviously of the greatest importance, since by their very nature they occur in more or less numerous morphemes, of which only a few instances are apt to be included in the questionnaire. Separating phonemic heteroglosses from other phonological variants presupposes a phonemic analysis of at least the major dialect types current in the area, so that the shared phonemic units can be distinguished from the regionally restricted ones.

FIGURE 4: Bundles of Heterolexes in the Great Lakes Area

Cities: C(hicago, Cleveland, Co(lumbus, D(etroit, Indianapolis, Milwaukee, P(eoria.

Adapted from A. L. Davis A Word Atlas of the Great Lakes Region.

When two or more plans of phonemicizing the data seem feasible, preference is given to the one that facilitates the comparison of the dialects. Needless to say, the plan chosen should be described and the inventory of shared and of regionally or socially restricted phonemes set up. This procedure has been applied, and strictly adhered to, in dealing with the phonological data recorded in the Eastern United States [Kurath and McDavid 1961].

Phonological variants that do not involve the phonemic structure are of two kinds, phonic and incidental.

Shared phonemes may differ markedly in their phonic realization, as the checked vowels / ae /of man and / Λ / of sun, or the free vowels / o / of no and / e / of day in American English. Such purely phonic heteroglosses, striking and pervasive as they may be, are of lower rank than phonemic ones, and must be handled as a separate set.

The incidence of shared phonemes in the lexicon, i.e., in shared words, may differ from area to area or from one social level to another. Thus some varieties of American English have the free vowel /u /of do, fool in room, roof, root, others the checked vowel / U / of full; some have voiced / z / in greasy, others voiceless / s /. Such incidental heterophones, though easily observed and by no means negligible, should be given a rank lower than the phonic ones. They affect neither the phonemic system nor the phonic realization of the phonemes.

2.31 Phonemic Heteroglosses

In the Eastern United States, phonemic, phonic, and incidental heteroglosses are only in partial agreement.

Phonemic heteroglosses set off a number of subareas on the Atlantic Slope (see Figure 5):

(1) The semivowel / /, as in ear, care, door, poor, is confined to four subareas: eastern New England, metropolitan New York, eastern Virginia, and South Carolina-Georgia.

(2) The same subareas have contrastive vowels in such words as heart ≠ hot, lark ≠ lock. Elsewhere postvocalic /r / survives in heart, lark, differentiating them from hot, lock.

(3) Checked /e / contrasting with free / o /, as in whole ≠ hole, road ≠ rode, is largely confined to New England.

(4) Merging in /ס /of the checked vowel of cot, stock and the free vowel of law occurs in eastern New England and in western Pennsylvania.

FIGURE 5: Phonemic Heteroglosses on the Atlantic Slope

1. The semivowel / /, as in ear, care, door, poor [Kurath-McDavid 1961: map 156]

2. The free vowel / a ~ /, as in far, heart (contrasting with the checked vowel /a /, as in hot) [ibid.: map 46]

3. law, caught, stalk and cot, stock have the same vowel, i.e. the free vowel / D / [ibid.: map 15]

FIGURE 6: Southern Diaphones

1. The phoneme / Λ / of sun, love articulated as a raised [Λ] or as a diphthongal [Λ^ә] (Kurath-McDavid: map 10)

2. The /æ / of ashes, half sounded as diphthongal [æ^ɛ] (ibid.: map 14)

3. The / u / of two, moon fronted to [ u ] (ibid.: map 17)

4. The / ɔ / of salt, law articulated as [ɒɔ], with progressive rounding (ibid.: map 23)

5. The /ai / of high, nine sounded as[a^ɛ ~ a^ә ~ a^•] (ibid.: map 26)

6. The /au / of down, mountain realized as [æ[] (ibid.: map 28)

2.32 Phonic Heteroglosses

The course of some of the phonic heteroglosses agrees to some extent with that of phonemic ones.

Thus (1) unconstricted or slightly constricted / 3 /, as in thirty, is largely confined to the / / areas [Kurath-McDavid: Map 25].

(2) The unconstricted diphthongal articulation [3I] of / 3 /, as in thirty, appears only in the / / areas of metropolitan New York and South Carolina-Georgia [ibid.: Map 25].

(3) Striking positional allophones of / ai / and / au /, as in nine ~ night and down ~ out, are restricted to the / /areas of Virginia and South Carolina [ibid.: Maps 26–29].

Other phonic heteroglosses follow an entirely different course. As a loose bundle running in a westerly direction south of the Pennsylvania line, they set the South and parts of the South Midland off from the North Midland. They agree in a general way with a prominent bundle of lexical heteroglosses. In the Lower South they agree rather markedly with phonemic heteroglosses. See Figure 6 and compare Figures 5 and 1.

FIGURE 7: Lexical Incidence of Shared Phonemes

A. Northern Features

1. root predominantly rimes with foot [Kurath-McDavid 1961: map 113]

2. broom has the vowel of full in New England, less frequently in New York State [ibid.: map 107]

3. gums often has the vowel of fool or of full in folk speech, less commonly among the middle class [ibid.: map 85]

4. goal rimes with fool, especially in New England folk speech [ibid.: map 122]

5. on (adv.) has the vowel of hot [ibid.: map 138]

B. Southern Features

6. poor rimes with pour [Kurath-McDavid: map 42]

7. bulge has the vowel of bull [ibid.: map 84]

8. creek rimes with week [ibid.: map 97]

9. coop has the vowel of foot [ibid.: map 108]

10. the vowel in Tuesday, due, new is preceded by / j / [ibid.: map 163–5]

2.33 Lexical Incidence of Shared Phonemes

Not infrequently the incidence of shared phonemes in the lexicon varies regionally (and especially socially). Though of low rank, these incidental heterophones must be taken into account along with the phonemic and the phonic, when the total area is subdivided in terms of phonological features.

Two separate bundles of incidental heterophones run westward from the Atlantic coast: one through northern Pennsylvania, the other south of Pennsylvania. See Figure 7. Both of these bundles are in substantial agreement with rather prominent bundles of heterolexes.

Other incidental heterophones set off subareas that are fairly congruent with lexical subareas. Thus the / a / in calf, can’t [Kurath McDavid 1961: Maps 68–69], the /æ / in care, stairs [ibid.: Map 39], and the /e / in Mary [ibid.: Map 50] characterize eastern New England; the lack of contrastive vowels in such pairs as morning, mourning, horse, hoarse, and merry, Mary [ibid.: Maps 44 and 49–50] the North Midland; checked / ɑ / in long [ibid.: Map 137] is confined to the Upper South, the midfront vowel /e / in here, beard [ibid.: Maps 34–36] to the Lower South.

All in all, in the Eastern United States the phonological heteroglosses of whatever kind do not diverge markedly from the behavior of the lexical heteroglosses. They exhibit a very similar regional structure of that area, so that the grouping of the speech areas based upon the regional vocabulary is supported by the phonological data to a considerable extent.

However, phonemic heteroglosses make some of the coastal subareas—Eastern New England, Metropolitan New York, the Virginia Piedmont, and South Carolina—more conspicuous; and the loose bundles of phonic and incidental heterophones running through northern Maryland and West Virginia [Figures 6 and 7 above] set the North Midland (Pennsylvania) off from the South Midland more decidedly than does the bundle of heterolexes.

2.4 Morphological Heteroglosses

Structural differences in morphology do not exist in cultivated American English. Even incidental heteromorphs, as of the verb, are largely confined to folk speech and rather sharply recessive. Those that have been revealed by the Linguistic Atlas survey of the Eastern States are rather fully discussed by Atwood [1953].

As the accompanying illustrations show, regional verb forms exhibit the same dissemination patterns as regional words and sounds. Some past tense forms are peculiar to Northern folk speech, though / εt / and / dov / are also current among the middle class and not always avoided by the cultured; others are confined to Southern folk speech. See Figure 8.

There is one aspect of the morphology of the verb that seems to involve a structural difference between folk speech and cultivated speech in American English.

In folk speech (1) the past tense and the past participle are nearly always identical, as in broke, drunk, drove or driv, took or taken; and (2) the past tense and the past participle are leveled to the form of the present in come, fit, give and some other cases. This deviation from cultivated usage is clearly not structural. It only increases the incidence of the patterns represented in cultivated speech (1) by live-lived, find-found, hold-held, etc., and (2) by let, set, cut. Even the not infrequent omission of the auxilliary have, as in I just heard about it, I never seen the like of it, I been looking for you, has its parallel in the informal speech of the upper classes. It does not eliminate the distinction between the simple past tense and the phrasal past perfect in folk speech.

FIGURE 8: Northern Verb Forms

1. /klim / ‘climbed,’ rimes with dim [Atwood 1953: fig. 5]

2. /si / ‘saw,’ rimes with sea [ibid.: fig. 17]

3. /εt / ‘ate,’ rimes with let [ibid.: fig. 9]

4. /dov / ‘dived/ rimes with grove [ibid.: fig. 6]

Southern Verb Forms

5. /hirn / or /hj3n / ‘heard’ [Atwood 1953: fig. 12]

6. /holp / ‘helped’ [ibid.: fig. 13]

7. /maut / ‘might’ (rimes with out) [ibid.: fig. 14]

8. /riz / ‘rose’ (rimes with fizz) [ibid.: fig. 16]

2.5 Concluding Remarks

At this point in our discussion we may ask the crucial question: Is the substantial agreement between lexical, phonological, and morphological heteroglosses in American English typical?

If so, either set of heteroglosses could be taken as fairly representative of the dialect boundaries existing in any given area, and hence of its dialectal structure. Moreover, a rather small number of heterolexes, of heterophones, or of heteromorphs running a similar course could serve as indicators of areal patterning. Even a single heterogloss, unless shown to have a unique course, could be taken to be a significant indicator, a “straw in the wind.”

This would give substance to the practice of many European scholars who intuitively select a small number of phonological heteroglosses, or even a single one, to represent major and minor dialect boundaries within the area investigated. Such choices are usually made by men who have more or less extensive information of one kind or another about the behavior of other heteroglosses. Hence, though intuitive and not statistical, they are apt to be significant. But they are surely not definitive.

If, on the other hand, the findings of the Linguistic Atlas of the Eastern United States should turn out to be atypical, if in other language areas lexical, phonological, and morphological heteroglosses should be shown to diverge from each other more or less sharply, intuitively chosen indicators would lose much of their significance. Since no analysis comparable in scope with that of American English has as yet been made anywhere in Europe or in other parts of the world, the validity of selective indicators in displaying the dialectal structure of an area is still in doubt.