English vowels may be named and enumerated in many ways, depending on what one wishes to do with or to claim about the resulting list. One may enumerate ``surface phonemes,'' ``systematic phonemes,'' ``word classes,'' ``sound classes,'' etc. Among the central purposes of this thesis is to compare dialects, and for this purpose, the relevant conceptual entity is the ``lexical set.'' John Wells (1982, vol. I) defined a number of lexical sets, which I will refer to frequently in this thesis. These consist of sets of words that are cognate in the different dialects; they provide an economical way of specifying the set of words which contain any given phoneme in any given dialect. The fact that these classes are listed and thoroughly characterized in a published work makes them especially useful, because this allows most words to be categorized unambiguously. Even more useful would be a dictionary which listed all words and the lexical set(s) that each belongs to.
Lexical sets are something like phonemes, and even more like the historical ``word class'' category used by Labov (e.g., LYS 1972:24ff). The terminology is initially disconcerting, since it implies that single sounds are identical with lists of words. All of the concepts, ``phoneme'', ``lexical set'', ``word class'', etc., may be understood as having a similar intensional and extensional meaning. The intension is ``a linguistic class of sounds'', and the extension is ``the words containing that sound class''. The names suggest one or the other kind of meaning: ``Lexical set'' and ``word class'' suggest the extension, while ``phoneme'' or ``sound class'' suggest the intension. Nonetheless they all have, to this extent, the same meaning at both semantic levels.
The very important differences among them are the relevant domains of analysis: the set of words in the extension is claimed to behave as a group or unit in the domain of the concept: phonemes are a unit that (synchronic) phonological rules apply to; word classes are a unit that sound changes operate on; lexical sets are a unit that comparative statements refer to. Thus different approaches to linguistic sound classes make use of different terms.
Why make lists and sounds synonymous? A list of words is not a single sound! Even so, it is useful to refer to single sounds by the list of words which contain it, so as to avoid having to say exactly what the sound itself is. Suppose for example that a sound becomes something else across dialects or through history. Then a name (e.g., the IPA symbol) for the sound itself cannot be used to refer to this changing unit, since the name may be inappropriate once the sound has changed. In such a situation, as in the comparison of dialects or the study of historical change, the word-list approach to defining sound categories is more concrete and well-defined.
A historical word class is list of words which changes as a unit in the sound shifts, mergers, and splits operating on a dialect at a particular stage of history. If the dialect is traced farther back into history, to the time before certain mergers, say, then more word classes must be specified, since the classes that merged into one must now be distinguished with their own separate list. Thus the historical word classes necessary for analysing sound change in any particular dialect or language depends on the time period being studied and the classes which need to be distinguished in that period.
Similarly a ``lexical set'' may be defined as a list of words which constitutes the comparative categories of speech sounds that are necessary in the comparison of a particular set of dialects. More precisely, a lexical set is a list of cognate words for which, in all the dialects being described, all the words in the set contain a particular phoneme. Thus if the words containing phoneme A in one dialect have one of two phonemes B or C in another dialect, then there are two lexical sets, which might be labelled AB and AC. The first set contains the cognate words which have A and B in the two respective dialects, and the second contains the cognates with A and C, respectively.
The purpose of dialect comparison is better served by comparative classes like Wells' lexical sets than historical classes such as the Middle English inventory of vowel phonemes, etc., even though the totality of historical word classes which are relevant to the description of sound changes in any English dialect are never entirely identical to Wells' enumeration of lexical sets. Without detailed study of the history of English phonology, references to historical word classes are largely incomprehensible. I do not assume that the reader is an expert in English historical phonology. The task in dialect comparison is to find out what lists of words correspond to the phonological classes that are distinctive in a particular dialect, and this is the task that lexical sets do.
For further discussion and justification of the concept of lexical sets, see Wells (1982, Vol I, p117ff). This group of lexical sets is somewhat arbitrarily chosen (it is the sets formed by the comparison of British Received Pronunciation and what Wells takes as ``General American'', namely a conglomeration of different American dialects), but it is nonetheless quite widely useful. It is important to note that Wells does not include many lexical sets which are necessary for the comparative description of particular dialects, such as the sets corresponding to the meet/meat distinction in Hiberno (Irish) English (J. Harris 1985), the can (noun)/can (aux. verb) distinctions in New York City, and Philadelphia, and other differences, both ancient and new, in many dialects. New sets must occasionally be defined for comparison of particular dialects.
In Table 2.1 I list Wells' lexical sets by name, in capital letters. Wells gives a detailed enumeration by environment of examples of each of his lexical sets. Examples of the classes that I find difficult are discussed below and given in Table 2.1, in hopes that this will aid the reader. The names for the sets were presumably chosen to be members whose pronunciation is not identical with the name of any other set in any dialect. Because of the many vocalic mergers in various dialects, the names include different consonants and on-glides, which serve the function of keeping them apart. Thus there should be no dialect in which the names for any two sets are homophonous.3.4
Some lexical sets do not correspond to single phonemes in any dialect. For example, in British Received Pronunciation the vowel in Wells' CLOTH set is pronounced the same way as the vowel in the LOT set, while most Eastern American dialects pronounce the CLOTH set with the vowel used in the THOUGHT set. Wells makes no claims of phonological reality for such sets; this does not reduce their usefulness both in this research and in exposition.
With any or no training, you can immediately identify the sound class in your own dialect which corresponds to the sound class in the dialect under discussion, simply by noting which of your phonological classes occurs in the word that names the lexical set. Of course, you cannot infer that all words with that phoneme in your own dialect are members of the lexical set referred to, since mergers and splits have rearranged your system as well as the one under discussion. When I see the name of the set, LOT, I know it refers to some of the words in my /ɑ/ phonological class, though not all, since CLOTH words and THOUGHT words are also in that class of mine.
At least the first problem of reading comprehension, namely that of identifying the relevant phonological class in the reader's own dialect, is solved better by using a word which contains the sound than a symbol which may refer to any of several possible classes in Middle English or some other dialect, which may be unknown to the reader. In practice, the differences between description in terms of historical classes and in terms of lexical sets are often quite small.
Wells' lexical sets are used here to compare English dialects, the purpose they were designed for. Many of the phonological differences among English dialects may be described by stating which lexical sets are identified as the same in one dialect or another. For example, we may state a number of conditioned mergers and near-mergers3.5 which have occurred in particular American dialects in terms of Wells' lexical sets:3.6
| Dialect | minimal pair | Lexical Set characterization |
| Southern | pen/pin | DRESS and KIT merge before nasal |
| Californian | Mary/marry/merry | SQUARE merges with TRAP, DRESS in /_rV. |
| Rhode Island | beer/bear | NEAR and SQUARE merge. |
| Philadelphia | merry/Murray | DRESS, STRUT near-merge in /_rV. |
The differences among lexical sets which I find most confusing are NORTH vs. FORCE, and CLOTH vs. THOUGHT vs. LOT vs. PALM. NORTH and FORCE constitute a single sound class for me, but many dialects differentiate them (including Alabama and Jamaica, studied here). Similarly, CLOTH, THOUGHT, LOT, and PALM all contain the same vowel in my dialect, but they are distinct for many. There are traditional names for three of these classes: THOUGHT is called ``long open O'', LOT is called ``short O'', and PALM is called ``broad A''. CLOTH is that set of words which is pronounced like the LOT set in British Received Pronunciation, and like the THOUGHT set in Wells' ``General American''. These sets are exemplified in Table 2.2.
Wells' list of lexical sets provides, by stipulation, one answer to the question, ``What are the English vowels?'' This answer is justifiable for the task of comparing dialects, which is usefully approached using categories like these. But we may also ask for a synchronic, phonological explication of the concept, ``English vowel''.