Implications of the Glide Slot for Dialectal Variation and Historical Change.

The theory of the glide slot has important and clear implications for sound structure and sound change in English history and dialectology. Changes in the inventory or phonetic quality of segments in the glide slot result in a small set of possible readjustments: merger and breaking. Glides or nuclei or both may merge, or syllable-boundaries may be inserted. A number of important phonetic changes are particularly relevant in this context. One is the ancient vocalization of /r/ discussed above, a 17th century English change which is now widespread in English dialects, in which /r/ changes from an obstruent to a continuant. A second is the more recent ``r-loss'' of Southern British, Boston, some Southern U.S. dialects, and New York City, in which /r/ changed from one kind of continuant (retroflex, or ``sulcal'' []) to another (non-rhotic [] or []). Labov (1966) analyzing NYC desulcalization, called it /r/-vocalization, under the assumption that /r/ is a consonant. A third important phonetic change is the present-day vocalization of /l/, which is developing or developed in various American and British dialects, such as Cockney (Sivertson 1960), Philadelphia (Ash 1982), Pittsburgh and other Western U.S. dialects, etc.

These changes are important, since when a phonetic change occurs so that a post-vocalic consonant becomes vocalic, and is phonologically re-analysed as a glide, all the contrasts among vowels that precede the new glide, which depended formerly on the presence of glide features to distinguish them, must either be re-analysed, or lost, depending on the phonetic forms of the relevant sound classes. Thus there are severe consequences for vowel structure when one of these changes occurs.

First consider the change of /r/ to a vocalic segment. Where post-nuclear /r/ is not a glide, as in those Celtic dialects where the reflex of /r/ is a tap or other non-continuant, there should be no merger among the short vowels: bird, herd, word, and curd should not rhyme. I believe this is correct. As discussed above, where post-nuclear /r/ is a glide, the high and mid short vowels merge before /r/ into a single class of rhotic monophthongs, without distinctive height or backness in the nucleus. In these -- most -- English dialects, bird, etc. do rhyme. Thus it was the phonetic change of /r/ from a consonant to a vowel that was the force behind the merger of earlier short /ir, ur, er, or/. Note that the short low vowels do not merge (as in car [k], carry [kæi]). These apparent exceptions are consistent, on the one hand, with the claim that these changes are phonetically rather than phonologically driven, and on the other hand, with the fact that these low vowels, while phonologically short, are phonetically relatively long, due to the larger mouth-opening gesture required for their production.

The result of this phonetic change of /r/ for the ``true diphthongs'', /y, w, oy/ was breaking. All the segments were retained, but in order to allow the two glides to occur in a row, Nucleus+Glide+r sequences were spread over two syllables rather than one. This is again consistent with the glide-slot theory, since at most one glide can occur in each syllable.

In dialects where the glide is lost from /y, w, oy/, it is predicted that tautosyllabic /r/ should again be possible, so that flour, dire etc., may contain just one syllable. This prediction is successful in monophthongizing Southern U.S. dialects, where these can indeed be monosyllables.

In short, the theory of the glide slot neatly explains one of the most important mergers in the history of English, as well as with phonetic facts in dialects which do not have this merger, as well as syllable-counting facts in dialects which monophthongize /y/ or /w/.

The second change, the de-rhoticization of /r/ in dialects like New York City, was explored in detail by Labov (1966). When /r/ becomes an inglide, for example, merger occurs between vowels containing similar nuclei and either inglides or /r/. Thus the vowel in source [sos] merges with that in sauce [sos], and bared merges with bad (as [bed]), etc. Again phonetic change in the glide position has major phonological consequences.

The third example of such a change is the ongoing vocalization of syllable-final /l/ in various English dialects. This change goes to an extreme in Pennsylvania, both in Philadelphia (Ash 1982) and Pittsburgh, but also occurs in some British dialects (e.g., Cockney, cf. Sivertson 1960), and to some extent in Western U.S. dialects. The empirical consequences of this change have not been fully described, but the known facts are consistent with the theory of the glide slot. Pittsburgh has extreme /l/-vocalization, where the realization of /l/ is a high-back glide. The result is in many cases a merger of post-vocalic glides with /l/: aisle, owl, Al, and Ow are all pronounced identically as [ao]. At the same time, the long-short distinction between certain pairs of vowels may be weakening: the stereotypical example is the local pronunciation of the name of their professional football team, the Pittsburgh Steelers, as the ``Stillers''.

Similarly, in the Western U.S., two studies have found mergers of long and short vowels before /l/. Labov, Yaeger and Steiner (1972: 236ff) studied the merger of fool and full in Albuquerque, and DePaolo (1988) documented mergers of certain pairs of long and short vowels before /l/ in Salt Lake City. Unfortunately these studies do not state what the quality of the /l/ is in these dialects. My impression of a couple of speakers (namely, myself and another speaker) I have analysed informally, from Southern California, a closely related Western U.S. dialect, is that some vocalization of post-vocalic /l/ does occur. Also, I count two syllables in oil and in Owl. These are consistent with a role for /l/-vocalization in the mergers found by LYS and by DePaolo.

The other result of /l/-vocalization which is consistent with the theory is the breaking of sequences of long or gliding vowels plus vocalized /l/ into two syllables. Thus owl could be pronounced [aU$]; also, ``feel'', ``fail'', and possibly even ``fool'', ``foal'' could become bisyllabic. The phonetic back-round glide on the long vowels in ``fool'', ``foal'', ``fowl'' may be indistinguishable from the back-unround glide which results from /l/ vocalization, so these may be more likely to undergo glide coalescence rather than breaking into two syllables.^3.38

These facts are tantalizing, and provide support for the theory that a glide slot exists in English syllables. The logic of the argument is as follows. There are four conceptually unrelated changes that seem to occur: (1) the phonetic realization of tautosyllabic /l/ changes into a back vocalic segment; (2) the short-long distinction between certain pairs of vowels is lost through merger; (3) various post-vocalic glides undergo coalescence, or merger with post-vocalic /l/; (4) some syllables with post-vocalic /l/ undergo breaking into two syllables. These changes would seem to be quite unrelated, and the fact that they might be found together might be thought of as a coincidence. However, when considered in terms of the theory of the glide slot, all these changes can be seen as consequences of a single change. The formal description of the change, according to this logic, is that /l/ shifts into the glide slot from its former coda position. The glide slot does not license the [lateral] feature (cf. page  below), so it is lost; this accounts for (1), the phonetic vocalization (by which the realization of /l/ is no longer phonetically [lateral]). At the same time, the Glide slot into which the /l/ has moved had formerly represented the long-short distinction; if /l/ now occupies the post-vocalic Glide slot next to formerly long as well as short nuclei, the formal distinction between the long and short vowels disappears; this accounts for the long/short mergers, (2). The same logic accounts for the coalescence of glides with the realization of /l/, (3). Both /-w/ and /l/ cannot at the same time occupy the same slot and also be temporally distinct from one another. Also, as back, [15#15round], vocalic glides, they are also phonetically quite similar. The result of coalescence of /l/ with /-w/ is explained if /l/ moves into the slot occupied by /-w/, and is phonetically indistinguishable from /-w/. In cases in which the features of the glide which precedes /l/ remain temporally and phonetically distinct from those of /l/, a sequence of two glides results, and the restriction in the theory of the glide slot, that states that only one glide may occur in a syllable, predicts that a sequence of two syllables will result, that is, that observation (4), syllable-breaking, will occur.

Thus these quite different changes may be seen as one. The fact that they seem to co-occur in known cases is support for any theory that explains why they should do so, as the theory of the glide slot does. These very different changes are explained as deriving from a single actual change, which in combination with the theory of the glide slot makes sense of all four otherwise unrelated outcomes of /l/-vocalization that were noted above.

Much research remains to be done to establish whether or not the merger of long and short vowels before /l/ is precisely correlated with the phonetic vocalization of /l/, to chart the course of the mergers as they occur (for example, is there an intermediate stage, where the /l/ is entirely vocalic, but has not been analysed phonologically as a glide?), to explain why some pairs may merge and others may not. An important question is, Why should it be that when /r/ vocalized, the short vowels lost their nuclear qualities, becoming //, but as /l/ vocalizes, the short vowels retain their nuclear qualities and their (back) glides merge into the /l/.^3.39 A descriptive study of the social, geographic, phonetic, and phonological factors influencing these changes is called for.

In this section I have defined the temporal structure of English vowels; in fact I gave a technical definition for ``Vowel'', a certain temporal sequence that is a constituent within the syllable. Let us next consider what are the static components within the sequence, that is, the features which may fill the nucleus and glide positions.