The complementary distribution pointed out above suggests that there may be a position of some kind in the syllable for /r/ (and for the glides in the diphthongs of buy, bough, boy). Let us explore this idea, by examining a theory of English vowels which takes this position quite seriously. We will first present the theory, and then point out certain internal problems and some modifications that are necessary in order to make the theory compatible with the above-given observations about vowel length and complexity and with current views of syllable structure.
Since The Sound Pattern of English (henceforth SPE), treatments of English vowels as sequences have been out of fashion: the diphthongs in high, and how, for example, were derived from monophthongs by rules that recapitulate the Great Vowel Shift. But before SPE, English vowels were defined in terms of sequential structures involving slots for particular classes of segments. Characterizations of English vowels in terms of temporal structures (i.e., vowel-vowel sequences) may have incorrectly been rejected because of the emphasis on static phonology. With the return of the syllable to phonological theory, we may take up this work again.
Consider the American structuralist description of (American) English vowel systems. This trend reached a pinnacle in the work of the George Trager with Bernard Bloch (1941), and with Henry Smith, in (1951).3.23 Take, for example, the treatment of the structure of English syllabic phonemes by Trager & Bloch (henceforth T&B). They establish 6 ``short'' vowel phonemes, called /i, e, , o, , u/. These short vowels are concatenated with other phonemes to form temporally complex sequences of the following 8 classes or ``sub-systems'': /V/, /Vj/, /Vw/, /Vh/, /Vr/, /Vjr/, /Vwr/, /Vhr/. V represents one of the short vowels; /j,w/ are up-glides to the front and back, respectively (I will use /y/ for /j/, below); /h/ is a centralizing off-glide, which can also represent length; /r/ is the rhotic semi-vowel or its non-rhotic reflex in ``r-less'' dialects. The term ``sub-system'' will often be used to refer as a unit to sets of vowels with the same glide. As LYS defines it:
``By sub-system, we mean a set of vowel nuclei or peaks which have the same glides or satellites and the same super-segmental features. Vowels within a sub-system differ only in the three dimensions of [height], backing or rounding.''(p. 219)
Labov (1989) shows that a greater degree of confusion or misunderstanding occurs in natural speech comprehension between vowels within sub-systems than is found between vowels across sub-systems.
Many of the possible combinations of the short vowels with the glides, constituting different sub-systems, are exemplified by large classes of words; others are exemplified by small word classes with as few members as one; yet others are entirely empty. The following table is taken from T&B (p. 243), substituting the names of Wells' lexical sets in capital letters for T&B's example words in those cases where the vowel classes correspond.3.24
| A | B | C | D | E | F | G | H | |
| /V/ | /Vj/ | /Vw/ | /Vh/ | /Vr/ | /Vjr/ | /Vwr/ | /Vhr/ | |
| i | KIT | FLEECE | -- | idea | mirror | NEAR | -- | -- |
| e | DRESS | FACE | -- | yeah | merry | eyrie | -- | SQUARE |
| a | TRAP | PRICE | MOUTH | PALM | marry | Irish | cowrie | START |
| o | LOT | CHOICE | -- | THOUGHT | sorry | Moira | -- | FORCE |
| STRUT | -- | GOAT | huh | hurry | -- | -- | NURSE | |
| u | FOOT | -- | GOOSE | -- | jury | -- | CURE | -- |
These combinations may be more economically expressed as in (1), or more abstractly as in (2):
(1) {i,e,a,o,,u}({y,w,h})(r)
(2) Nucleus (Glide) (r)
Labels for the first two positions are somewhat arbitrary. What is essential to T&B's claim is that these positions are distinct and can be filled only by the corresponding list of segments. I have chosen to label them with familiar phonetic terms, ``Nucleus'' and ``Glide''. ``Peak'' and ``satellite'', or other terms might also be used.
The distinctions made in T&B's analysis are insufficient to represent the contrast discussed above between long vowels /iy, ey, w, uw/3.25 and longer vowels /y, w, oy/. For example, there is no formal difference of length between /ey/ and /y/. If /r/ occurs in a position following the glide slot within the syllable (that is, in the syllable coda), why shouldn't it occur there in words like hire, flour? It seems that it cannot, since these words break into two syllables, merging with higher, flower. Further, there are words like heist, mounds, etc., which do have segments in the coda. If /r/ is a coda consonant, and coda consonants are acceptable after the longer vowels /y, w, oy/, then monosyllabic forms like /yr/, /wr/, /oyr/ should be possible. However, they are not.
What, then, is the difference between long vowels /iy, ey, uw, ow/ and longer vowels /y, w, oy/ that accounts for this contrast? One difference is that the latter are low vowels in some dialects.3.26 However, the true answer to this question will become clear only after we clarify the analysis of /r/ within the syllable, as is done in the following section.