The Surface Phonology of Vowels

This section proposes, discusses, and modifies a structural analysis of the Alabama vowel system. It is not concerned with morphophonological alternations or lexical phonology, but with the structure of the sound system which is the output of the lexical phonology and with the ways in which those structural representations are implemented in actual speech. Thus the object of study is surface phonology, or post-lexical phonology in the sense of Kiparsky (1982), as well as phonetic implementation.

Previous published works on the phonology of this and related Southern dialects include Feagin (1986, 1987, 1990, in press), Sledd (1966), C.-J. Bailey (1969ff, 1980, 1985 and forthcoming, among others), Labov, Yaeger & Steiner (1972), Foley (1972), McDavid (1948), and Kurath & McDavid's PEAS (1961), as well as several studies cited in Feagin (1979:11). McMillan & Montgomery (1989) is an annotated bibliography of studies of Southern States English.

Foley (1972) impressionistically studied the phonetics of vowels in Tuscaloosa County, Alabama, using dialect atlas interview techniques. Tuscaloosa is at about the same latitude as Anniston and Birmingham, but lies on the western side of the state. Anniston is in the east of the state, somewhat north of the South/South Midland isogloss^8.2 of Wood (1961:12) and Foscue (1971:41) that runs east-west through the middle of the state, south of Birmingham. Tuscaloosa is about the same distance north of the Southern/South Midland isogloss as Anniston is, so it could be expected to share most phonological features with Anniston.

The following tabular structure summarizes the phonological vowel classes which I have derived from Foley's and other work on Southern States English.^8.3

Table: Alabama Vowel Structure

	V		Vy		Vw		Vr
high			iy	21#21	21#21	uw	ir	ur
mid			ey	21#21	21#21	ow	ær	or
low	æ		ay	y	aw	w	r	r

As in the discussion in Phonological Preliminaries, I exclude prevocalic glides as in cute, quit, as well as vowel sequences that may be analysed as bisyllabic, such as ruin, fluid, Ayre, mayor, lion, Brian. The symbols used in this table, as in other chapters, are not especially phonetic and are not to be taken as IPA transcriptions of these classes. In particular, the V (phonologically short) vowels /, , , etc./ typically have inglides in stressed monosyllables; /ay/ is commonly a monophthong [a:], and the /-y, -w/ symbols in the Vy, Vw classes do not represent phonetic glides whose endpoints are higher than IPA [i, u], respectively, etc. The symbols used here are intended to make reference to locations within the structure simple and unambiguous.

The following discussion describes the relationships of the structure in Table  to that of Reference American, justifying the different choices made in the structural analysis of this dialect, as compared with that reference dialect. I do not mean to imply that Southern dialects are to be analysed as derived from Northern varieties. In fact, there are more phonological distinctions made in some cases in the South than in the North, so a historical derivation might well go in the opposite direction. However, rather than argue for each detail of an analysis of this dialect from first principles, I have found it simpler to begin with an established system, and modify it as necessary in order to accomodate the partly different and partly similar facts of this dialect. There are a number of notable differences between this (surface phonological) structure and that of Reference American (RA), presented above in Phonological Preliminaries.^8.4 To summarize: First, among the Reference American long vowels (the V: subsystem), the PALM class (RA /a:/) appears to be merged in this dialect with the LOT class (RA //); the THOUGHT class (RA /:/) is a back-raising glide. The Reference American V: subsystem of /i:, e:, a:, :, o:, u:/ has no low counterparts in this system, since /a:, :/ are lost by merger and gliding. The result is a complementary distribution between long vowels and Vy, Vw vowels: none of the former are low, while all of the latter are low. I therefore shift the mid and high long vowels /i:, e:, o:, u:/ into Vy and Vw subsystems, thereby eliminating the entire class of surface-phonological long monophthongs and increasing the number of phonological glides in Southern speech relative to (Northern) Reference American.

Second, the Vy vowels are found to be slightly different positions within the Vy subsystem. Third, the Vr subsystem is both phonologically and phonetically different. By way of justifying the choices made in proposing this particular structure for Alabama English, I will discuss each of these differences below, first the long vowels and the Vy, Vw gliding vowels, and then the vowels before /r/.

The ``broad A'' class, RA /a:/, as in PALM, has apparently merged with the low-back checked vowel //, as in LOT. Thus, the PALM-class words calm, father words ``are assigned to the checked vowel //, though the phones are prolonged and frequently backed, as [:]''(Foley 1972:31).

Foley seems unwilling to assign a set of vowels to a distinct phonological class when it is distinguished primarily by duration rather than by quality. Since a difference of length can be sufficient for a phonemic distinction, this doesn't seem necessary: PALM could be distinct in this dialect. However, McDavid (1940) in a study of the low-back vowels of the Carolina Piedmont, says that the broad A class is not distinct in that closely related dialect.^8.5 Further research seems called for in light of the length difference found by Foley; for now, I will assume his analysis is correct in assigning broad A to the // class. If the classes are distinguished, then broad A would fill one of the empty low-vowel slots in the V: subsystem.

The THOUGHT vowel class, which is represented in RA as a long low-back vowel, /:/, is a back-raising diphthong in this dialect, most commonly [, o], according to Foley as well as in my data, so I have shifted it from its location in the RA V: subsystem to the Vw subsystem, a step which is structurally sensible given that the nucleus of RA /w/ has fronted to [a, æ] for Foley, and even higher and fronter in my data. (For further support of this point, see the impressionistic transcriptions below, as well as Figure  and the related discussion.)

With the long-low-front slot of RA /a:/ considered to be empty (following Foley and McDavid), it might also be reasonable to reanalyse Southern monophthongized /ay/ (PRICE) as having shifted into that slot, since it is often phonetically long, monophthongal, low, and front. However, in Foley's data, the sound corresponding to RA /ay/ is invariably a front-raising diphthong, which suggests that it belongs in the Vy subsystem (where it is located in Table ).

This class of vowels, however, is quite commonly a monophthong, [a:], in Southern varieties. If this form were taken as fundamental instead of the also-occurring form [ae], then it would make sense to consider it to be a low, front, long monophthong and to write it as /a:/ rather than /ay/. This slot in the V: subsystem of Reference American was vacated by the assumed merger of RA /a:/ (broad A, or the PALM lexical set) with // (short O, or the LOT set), so that the phonological prerequisite for a sound shift from low-front Vy to low-front V: is present: the target slot is empty. This may be part of the phonetological motivation for the original monophthongization process; one analysis might put this vowel class simultaneously in both classes, while it shifts gradually from one to the next.

It should be pointed out that monophthongization of /ay/ is, or was, a conscious sociolinguistic variable (as is typical of late stages of long-established sound changes, before they go to completion, see Labov 1982)^8.6and the speakers Foley interviewed were undoubtedly using their most formal and self-conscious style, which is therefore likely to result in more gliding forms for /ay/. Thus it may still turn out that in this dialect the RA /ay/ class shoul dbe considered a long low-front monophthong at the post-lexical phonological level.

Unlike the relationship between RA /y/ and //, where the nuclei are phonetically and phonologically identical, the nucleus of /ay/ is quite distinct from that of // in this dialect. Unlike the realization of Reference American /oy/, which has a mid-back nucleus, Foley transcribes words of this class with phonetically lower-mid vowels. His transcriptions may be summarized as [(:){,}].^8.7These two facts suggest that these forms should be categorized as /ay, y/: low-front and low-back up-gliding diphthongs. These four vowels thus form a symmetrical set of low upgliding diphthongs, with front and back nuclei and glides:

Table: Low Upgliding Vowels in Alabama English.

		Nucleus
		front	back
Glide	front	ay	y
	back	aw	w

With the assumed merger of RA /a:/ (broad A, PALM) with // (short O, LOT), and the shift of RA /:/ (THOUGHT, CLOTH) to the Vw subsystem, there remain no long low vowels in this vowel system. This brings out a complementary distribution of the sound-classes within the structure: none of the V: vowels are low, while all of the Vy, Vw vowels are low. We might improve the elegance of the analysis by collapsing the long vowels into the Vy, Vw classes, as in the base-5 system discussed in Phonological Preliminaries. The difference here is that the system remains base-6, and there are appropriate places in the system for the RA low vowels /æ, , a:, :/ (namely as Alabama /æ, , , and w/, respectively), as in Table  above.