Introduction

The first half of the dissertation is about general theoretical issues in phonetics and phonology. An understanding of the phonetics of English vowels must begin with principles of general phonetics. General phonetics is the study of the physical universals of speech production and perception, while linguistic phonetics is the study of language-particular aspects of speech production and perception. Without a fundamental understanding of what vowel sounds are as sounds and of how they are produced, a linguistic phonetic study of vowel production is uninterpretable. Chapter 2, The Mapping from Articulation to Formant Structure, provides a fundamental understanding of these central issues in phonetics. As such it is a contribution to general phonetics, and may be read independently of the rest of the book, but to properly interpret the acoustic-phonetic studies in the body of the book it is necessary to have an understanding of the content of this chapter.

The phonetic substance of vocalic sounds, not just in English, but in all languages (and even vocalic sounds produced by gorillas and baboons) must be understood from both articulatory and acoustic perspectives. Often one perspective or the other, but not both, is taken as fundamental, as exemplified in the shift from acoustics-based features for Jakobson, Fant, and Halle (1952) to articulation-based features for Chomsky and Halle (1968). Studies of formant frequencies undertaken by those who lean toward acoustics may be criticized or insufficiently understood by those who lean the other way; similarly, EMG studies of the activities of the articulators may be poorly understood by those who lean towards acoustics. This chapter explains the relationship between vocal-tract resonance frequencies and vocal-tract shapes, and thus attempts to bridge this gap. It shows how changes in formant frequencies, F1, F2, etc., are derivable from changes in the articulatory configuration of the vocal-tract. It shows that F1 and F2 are precise measures of mouth-opening, and of tongue-body frontness/lip rounding. It thereby helps to justify the importance of the studies of formant-frequency patterns which are taken up in the dialect chapters.

On the other hand, Chapter 2 may be taken as an independent contribution to a fundamental understanding of general phonetics. Though the ideas are quite old, their application to these issues is in many aspects new. It describes how the pattern of resonances of acoustic tubes like the vocal tract can be modified by changing the shape of the tube. The effects of constriction and widening on nodes and antinodes of standing waves in the tube are described. The theory is presented without use of mathematics (though the theory is mathematically expressible), and gives a method of reasoning simply about the effects of movements of the vocal tract on formant frequencies, and conversely, about how to determine what the vocal tract is doing, given a pattern of formant frequencies. The theory is qualitatively understandable, unlike quantitative theories in which equations describing large numbers of concatenated tubes of varying cross-sectional area are solved in order to predict the resonant structure of the tube. The theory is shown to explain quite simply the fundamental empirical facts of the relationship between articulation and vowel acoustics, including some rather mysterious ones (e.g., the triply constricted articulation of []). Also, an experiment is conducted and described which confirms one detailed prediction made about a previously unstudied (to my knowledge) acoustic pattern (the difference between retroflex and dental ``locus'' frequencies). Finally, it is pointed out that the acoustic dimensions that this theory gives prominence to correspond quite straightforwardly to the traditional phonological dimensions of height and frontness: formant frequencies F1 and F2 are shown to reflect the degree of mouth-opening and of tongue-body frontness, and both acoustic and articulatory dimensions may simultaneously be understood as the phonetic dimensions corresponding to phonological height and frontness. This conclusion resolves an ongoing controversy over the nature of phonetic features for vowels, in favor of the traditional dimensions of height and frontness, slightly reformulated.

The surface phonological structure of English is discussed in Chapter 3, Phonological Preliminaries. By surface phonology, I mean the structure of the sound system which is the output of the lexicon. The goal of this chapter is to characterize the phonological structures which are the input to the system of phonetic implementation that is the central focus of study here. For this reason, lexical- or morpho-phonology was not considered; the focus is on the true generalizations and symmetries that hold on the phonological surface. The evidence used in this chapter is that of complementary distribution and phonetic similarity, rather than morphophonemic alternations. After preliminary discussions of the inventory of comparative sound-classes and of the incoherence of ``General American English'' as an object of phonological study, this chapter develops a phonological structure for a hypothetical dialect, ``Reference American'' (which conceptually replaces General American). A comprehensive structural analysis of the English vowel system is given. The structure proposed has two aspects: static and temporal. The static phonological structure is a 3 X 2 (``base-6'') system of nuclei distinguishing three heights and two degrees of frontness. The temporal, or sequential, structure distinguishes a nucleus and a post-nuclear glide within the syllable, where the glide slot is used to represent vowel length as well as high-front, high-back, and rhotic glides, /y, w, r/. Central to the proposed analysis is the argument that postvocalic /r/ is located in this glide slot: various counterproposals and apparent problems with such a view are examined and overcome. Movements of post-vocalic consonants into the glide slot (among other consonants, /r/ and /l/ have undergone this change in various English dialects at different times in history) thus has considerable consequences for the vowel system and explains a number of historical and ongoing vowel mergers in English dialects.

A formal treatment of the structure of the vowel system is given. ``Vowel'' is argued to be a constituent within the syllable, which has an optionally (binary-)branching structure, and which licenses the privative features [front], [high], and [low]. The phonological literature on the problematic dimension of vowel height is reviewed, and a novel solution to the problem is proposed, using privative features on an autosegmental height tier. All the possible combinations of features occur in the nucleus; four of the six possibilities also occur in the glide position. The unit of syllable weight, the mora, is defined in terms of branching structure in the syllable rhyme rather than as a constituent of the syllable. Heavy and light vowels in English are distinguished by branching vs. no branching of the Vowel constituent. The feature [round] is argued to be a redundant feature on certain back vowels.

Finally, a plausible grammar of phonetic implementation for q vowel nuclei is presented, by which the phonological structure is mapped into a default acoustic-phonetic representation, using three phonological and four phonetic rules.

Chapter 4 discusses various theoretical background issues. It makes a central conceptual distinction between linguistic and general phonetics. It characterizes ``vernacular speech'' and argues that this is historically and socially the most important form of language; it also is the form which contains the greatest variety of phonetic variation, which is of central interest in this work. Next, definitions are presented for the analytical units, ``acoustic vowel'', ``acoustic consonant'', etc., and the relations of phonological vowels and consonants to acoustic vowels and consonants are clarified. Criteria for including measurable phonetic features in the catalog of real phonetic features are suggested. Finally, the relationship of F1, F2 measurements to this catalog of features is discussed, and an argument is presented that significantly different F1, F2 measurements, which are larger than the difference limen, must reflect real phonetic differences (that are audible, given training) which are to be accounted for at some level by a theory of phonetic performance. Lisker's (1949) thesis is replicated, with different results.

Chapter 5, Methods, describes the analytical techniques used in describing each dialect's vowel system and its phonetic implementation. Some of these methods are relatively new and unknown, such as the technique for estimating the sampling distribution of a statistic, called the bootstrap, while others are traditional linguistic methods (e.g., complementary distribution, contrast, and phonetic similarity). Statistical regularities in phonetic form are also adduced as evidence for structural phonological questions. Methods of impressionistic coding of phrasal stress are described, along with reliability tests.

In this way, the first half of the book establishes a phonological framework for English vowels and a phonetic theory for understanding the articulatory significance of acoustic measurements. It also clarifies some aspects of the theoretical context in which these measurements are to be understood and provides a set of methods which are likely to lead to interesting phonetic and phonological generalizations in the study of spoken vernacular dialects. The descriptive part of the book follows these initial theoretical and methodological chapters.

The following paragraphs provide the rationale for the approach taken to dialect description in chapters 6, 7, 8, and 9.

The study of vowels in the English languages is approached from a number of directions in this work. ``English'' is not taken to be a single object. There are many different dialects of English, which are often extremely different from one another in surface-phonological inventory, in phonetic realizations of corresponding sound-classes, and in the phonetic processes that partly constitute the performance of phonological forms.

In order to investigate differences and similarities in these aspects of speech performance, several dialects are studied: Chicago White English (CWE), Los Angeles Chicano English (LACE), Anniston (Alabama) English (AE), and (Kingston) Jamaican (mesolectal) Creole (JC). The geographical and social boundaries of these dialects are not established here; what was important for this study was simply that they be as different from one another as possible.

A number of reasons determined the choice of these four dialects. A major goal was to test the assumption of the universality of phonetic implementation. To maximize the phonetic differences and minimize the phonological differences, different but related dialects of a single parent language were examined, since as different ``accents'', they may be expected to share much of the same phonological system, but at the same time to be phonetically quite different. In fact, Bailey (1985) titles an appendix, ``The underlying vowels of most English lects'' -- if most English dialects have the same underlying vowel system, then it may be inferred that abstract phonological differences are minimized by choosing English dialects. On the other hand in order to maximize the differences in phonetic characteristics, dialects of English were chosen that are as different from one another as possible.

A cue was taken from Labov's (1991) paper on the ``Three Dialects of English.'' There he identifies Northern, Southern, and Low-Back Merged as the three main (phonologically defined) divisions among English dialects. Thus one dialect from one of each of these areas was chosen (Chicago, rural Alabama, and Los Angeles, respectively). Further, in case these three dialects turn out to be too similar to one another, a fourth dialect was chosen which was so radically different that it might be considered a separate, though closely related, language.

Another important problem in enhancing diversity in this data was to avoid the effects of linguistic conservatism. The more conservative or standardized the speech, the more similar the dialects might be. For this reason upper-class, highly-educated speakers were avoided, in favor of analysing the more vernacular, more historically advanced speech of working-class speakers.

One problem with studying coarticulatory patterns in the laboratory is that many natural phonetic phenomena of this kind often disappear under laboratory conditions. This is the Observer's Paradox: if you look at something very closely, the act of observing changes the thing observed, so that one can't tell what it would have been like if it hadn't been invasively observed. Formal, self-monitored speech has much less coarticulation and reduction, than natural, normal, informal and un-self-conscious conversation (cf. Labov 1986, Keating and Huffman 1984). Because this study attempts to document significant cross-dialectal differences in phonetic processes of reduction and coarticulation, is important to maximize the effects of phonetic processes and to maximize the geographical differences, and thus to use as data the most unmonitored, vernacular style of speech. The vernacular is important for other reasons (discussed on ); it is the natural, undistorted form of the language used in day-to-day communication among native speakers of the dialect. An excellent way to get relatively unmonitored, continuous vernacular speech is by means of the tape-recorded sociolinguistic interview, so the data comes from previous and ongoing sociolinguistic studies of communities that fit the above constraints.

The first part of each dialect study characterizes the surface phonological structure of the vowel system of that dialect. Without getting the phonology right, there is little hope of making sense of the phonetics. Previous literature on the phonological structure of these dialects is used as the primary resource for these sections, in which each dialect's surface phonological (not phonetic) inventory is explored and represented using the approach presented in Chapter 3, Phonological Preliminaries. The structural analyses of Jamaican Creole and Alabama English turn out to be quite different from that proposed for Reference American.

The core of each dialect chapter is a set of acoustic-phonetic studies of vowel quality as found in the natural conversational speech of one, two, or three working-class native speakers of the dialect. Vowels are characterized by taking measurements of the first two formant frequencies, F1 and F2, at a representative point within the syllable. The second chapter justified the articulatory importance of F1, F2 measurements. Formant trajectories throughout the acoustic vowel^1.1 as well as acoustic-vowel duration, were also measured, but are not discussed here.

Thousands of vowels are measured for each speaker, in order to bring to statistical significance the effects on most vowel classes of stress and of many of the consonants that occur in their environment. The easiest way to collect enough data to bear on a large number of these effects is to simply measure all the vowels that occur in a fairly long stretch of continuous speech. Some phonological classes occur very frequently, while others (like /oy/) are under-represented, but the overall result is that many, many effects have sufficient data to make them statistically significant.

The effects of phonological class and of phrasal stress are explored for each dialect. The sound changes that have resulted in the phonetic norms of the various speakers are characterized. A simple phonetic grammar is given for Jamaican Creole vowel nuclei (that is, a set of sequentially applied rules which map the phonological structure to the acoustic-phonetic space of F1 vs F2, and which shift the locations of the nuclei in the phonetic space in order to describe the observed relationships accurately). For the other dialects, only some aspects of the corresponding dialect-specific phonetic implementation system are characterized.

Then the effects of impressionistically coded phrasal stress are explored, showing in general that on average, most vowels shift when unstressed towards a common reduction target, indistinguishable from the realization of schwa. Observed exceptions in various dialects are also explained.

Thus these chapters of phonetic description of vernacular speech production in widely different dialects form an important set of contributions to our knowledge of vowel reduction, of sound-change, and of the partly linguistic-phonetic and partly general-phonetic system of phonetic implementation. The knowledge about reduction and coarticulation gained over the past thirty years in the laboratory is thus extended to natural speech data in different dialects. The knowledge gained by sociolinguistic research on sound change over twenty years is extended by investigating interactions of dialect diversity with stress and consonant effects, factors which have important roles in sound change.

Finally, each of the dialect chapters may also be seen as a contribution to the literature on the respective dialects; thus, to creole studies (I believe this is the first acoustic study of creole phonetics), to the literature on Southern States English, to the study of Northern Cities dialects, and to the study of linguistic diversity in California.

The concluding 10th chapter, focusses on the theoretical existence of phonetic grammar. The effects of two particular consonant environments on vowels are compared across the four dialects studied. First, the different effects of following laterals are explored; these can only be attributed to differences in the system of phonetic implementation across the various dialects. In particular, Jamaican Creole is shown to lack a phonetic implementation rule demonstrated to exist in American English.

Second, the striking lowering effects of following // on vowels for the Alabama speaker, as opposed to the more coarticulatorily natural fronting and raising effects of the same consonant on //, for example, found in each of the other dialects, are not plausibly attributed to differences in the phonological form of // in different dialects, or to the phonological content of the vowels that are affected. Thus there appears to be a dialect-specific coarticulation rule in the phonetic implementation system, which accounts for the lowering of vowels before // in this Alabama speaker's natural speech.

It is a fact that ``physically motivated'' processes are usually not physically necessary, and can vary stylistically, so that the most natural, physically easiest and most simplified phonetic forms are restricted to certain styles; similarly, there are differences in these details across dialects. Thus ``natural'' processes, while physically sensible, are not physically necessary. ``Hard'' coarticulation does exist, since the tongue cannot move infinitely fast. But there is a large realm of ``soft'' coarticulation, which is not due to absolute physical constraints. Dialects can differ in patterns of soft coarticulation that may be physically easy but are not physically necessary. Speakers must learn these patterns because they are part of what differentiates one dialect or language from another. These ``details of implementation'' are a part of the knowledge of native speakers of particular dialects, a part of what they know as speakers of that dialect and not some other, which may be different. It therefore appears that the control of these rather intricate details is a part of the human language faculty,

Linguistic-phonetic grammar includes the learned rules of soft coarticulation and of prosodically governed (stress-related) vowel reduction. Through comparisons of some of these details in the dialects studied, the final chapter confirms the existence of these linguistic phonetic effects, and opens the way to further research on this intricate and interesting aspect of the linguistic system.