Restricting the discussion now to phonological temporal structure, I will examine evidence for three degrees of vowel length and complexity, as opposed to the traditional two degrees of long vs. short. In the end, I will reject the claim that there are three degrees of phonological vowel length in English, but not before the evidence relevant to this question leads to a fundamental re-thinking of the structure of the English vowel system.
The first degree of phonological vowel length or complexity is exemplified by the monophthongal3.10 ``short'' vowels, as in pit, pet, putt, put.3.11 These vowels do not occur in open stressed syllables: when lexically stressed, they must be followed by a tautosyllabic consonant.
The second more complex (or in this case, longer) set of vowels are the traditional mid and high ``tense'' vowels, which occur in both free and checked syllables, as in be, bean; bay, bane; Boo, boon; Bo, bone.3.12 Giving acceptable names to this class of vowels and to the difference between it and the first class of vowels above is a problem that has occupied phonologists and phoneticians for decades. They were traditionally analysed as ``long'' vowels, as opposed to the ``short'' vowels above. This discussion will use these terms, though each of the other terms has some value. They have been called ``tense'', as opposed to ``lax'', a term which is applied to the short vowels. The tense/lax distinction is notoriously ill-defined. Bloomfield (1934:101), for example, says the ``simple [i.e., short] vowels are pronounced with loose muscles...[while] [ij, ej, uw, ow] seem to be somewhat tense throughout'', but overall vocal-tract muscle tension does not necessarily differentiate these vowels. Labov abandoned the term, ``tense'', and proposed the acoustical term, ``peripheral'' (as in ``peripheral in acoustic vowel space'') to replace some of its senses. Since there are also non-peripheral, long vowels, such as stressed /ʌ/, raised /ay/, etc., which are pronounced centrally when articulated with the greatest force, and which count as long vowels, I will not use this as an underlying phonological dimension. Bailey (1985:205) calls long vowels and diphthongs ``heavy'', and short vowels ``light''. Mora-counting also differentiates them as one versus more than one morae. Wells uses the distributional terms ``checked'' (never syllable-final and stressed) vs. ``free''(sometimes syllable-final and stressed) to differentiate them. This is the main distributional fact that holds most English dialects together as similar to one another.
The vowels I call ``long'' vowels may be considered ``longer'' than the ``short'' vowels according to the following metaphor. Suppose that, in order to be stressable, syllables require a certain quantity or length of segmental material. Then, vowels that occur only in closed syllables when stressed may be considered ``not long enough'' for this purpose, since additional following segments (which close the syllable) are required to form a stressable syllable. Vowels that can occur in open syllables on the other hand are ``long enough'' in this sense, since they provide the necessary quantity of material to form a stressable syllable on their own. This explicates the traditional distinction between ``short'' and ``long'' vowels, a distinction which is based on this distributional difference. The phonological picture is paralleled by the phonetic facts: these ``tense'' and peripheral vowels are indeed longer in duration, other things being equal, than their ``lax'', non-peripheral counterparts.3.13 Further justification for the term ``long'' is that these vowels are indeed phonetically long monophthongs in various dialects.3.14
A third set of vowels, however, is even ``longer'' than the other classes, according to a similar metaphor. This class consists of the diphthongs in buy, boy, bough.3.15 This class is not distinguished from the ``long'' vowels by the checked vs. free distribution, since both occur in checked as well as free syllables (e.g., heed, hide, he, high).3.16 They are different from the mid- and high- long vowels in that they occur in a smaller set of (checked) environments. In particular, postvocalic /r/ cannot follow them within the syllable. Consider the distribution of post-vocalic /r/ with the merely ``long'' vowels in Table .3.17
The contrast between the two columns, as mentioned earlier, may be attributed to syllable count: one versus two. But when I try to find such pairs for the third class of vowels, I fail. For all the speakers whom I have asked (of those whose dialects are r-ful and have a front-raising glide in buy), the pairs hire and higher, mire and (Oscar) Mayer are identical.3.18 A root-beer company promoting its connection with the Philadelphia ice-hockey team prints on its cans the phrase, ``Hires Flyers'', which rhyme suggests the same identity.
While phonetic evidence must be interpreted cautiously, the following spectrogram (Figure ) of the phrase fires 'n' wars (taken from some of the vernacular speech analysed in the Chicago chapter) shows two relevant examples. These are in fully stressed, list intonation, in the longer context, It's lasted through, fires, and wars, and God knows what. (where ``,'' signifies final lengthening and a fall in pitch), so both /yr/ and /or/ occur in (final) lengthening environments. An impressionistic transcription of this sequence is [f:znwo:z]. The first word, fires contains an extremely long, steady-state  segment (~ 215ms, measuring from the midpoints of the transitions from and to adjacent segments), while the second contains an /r/ realization whose formant structure changes continuously and which is only ~ 140ms in duration. These phonetic differences are consistent with phonologically syllabic and non-syllabic /r/ for fires and wars, respectively.
Similarly, hour rhymes with power, and flour and flower are identical.3.19 Even our, a closed-class word which may be expected to reduce as much as possible, either rhymes with flower, or else it is identical with are, where the high-back glide is lost along with the syllabicity of the /r/. I find no examples of monosyllabic /oyr/ sequences which might correspond to foyer and (in my dialect) lawyer, etc.
Since /r/ cannot follow these diphthongs within a single syllable, we may infer that the position in the syllable that post-vocalic /r/ occupies is filled up. Thus these diphthongs fill up more of the syllable than the mid- and high- long vowels in see, say, sew, Sue3.20 which do allow following /r/ in the same syllable (as in beer, bear, bore, boor3.21). This class is thus different from, and longer than the ``long'' vowels. Our metaphor may now therefore be re-examined: The three classes, ``short'', ``long'', and ``longer'' are, respectively: not long enough to close a stressed syllable, long enough, and so long that some segments can't fit in the syllable with them. Table summarizes the constraints.
These three degrees of apparent vowel length will be explicated in a theory of the temporal structure of vowels within the syllable in the next sections.
Yet another degree of length might be added by considering vocalic on-glides such as /y, w, r/. However, this leads to uninteresting questions like ``Is the vocalic part of yet (which could be written /ye/) longer or shorter than the vocalic part of ate (which could be written /ey/)?'' Also on-glides are more consonantal than off-glides.3.22 We will not discuss on-glides further. These facts would appear to support the idea that there are three phonological degrees of vowel length. The basic linguistic observation is that there is a complementary distribution, within the syllable, of the glides in buy, bough, boy and postvocalic /r/. This observation may be interpreted in a different way.