... vowel1.1
For a definition of acoustic vowel, see page 79 . 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... general2.1
For the distinction between general and linguistic phonetics, see page [*]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... formant2.2
A formant is a resonance frequency of the vocal tract. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... mathematics;2.3
Chiba & Kajiyama state that ``It was our original goal to write this book without making use of mathematical equations'' (1941:i), though they do not. For accessibility's sake, this goal remains a worthy one. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... particles2.4
Air particles are to be thought of as infinitesimal volumes of air rather than as atoms per se. Air volumes have a pressure, but atoms do not. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... waveform2.5
By velocity waveform, I mean the waveform which describes the velocity of air along the length of the tube. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... 2500Hz.2.6
This assumes a vocal tract 17.1 cm in length, and a speed of sound, c, of 343 meters/second -- that is, in a nearly average male vocal tract in under normal atmospheric conditions 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... rarely2.7
Though see Ohala (1985) for its use in explaining the feature ``flat.'' 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... antinode.2.8
This relation is expressed mathematically in terms of an area function of the vocal tract (which specifies how wide the tube is at each point relative to its average width) and a weighting function which weights constrictions positively at nodes and negatively at antinodes. The frequency change may be found by multiplying the values of the area function by the corresponding values of the weighting function at each point along the length of the tube, and adding (integrating) the results. The sum of positively weighted constrictions is balanced against the sum of negatively weighted constrictions; the larger one determines the direction of change. If the constrictions are positively weighted, and the widenings (negative constrictions!) are negatively weighted, the effect is an even more positive one, and the frequency rises a great deal. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... spectrograms2.9
For example, V. Zue makes this point in spectrogram-reading instructions. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... (1985).2.10
Thanks to Dave Graff for referring me to this paper. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... \invv/.2.11
Dental and retroflex sounds in these contexts are not part of my native phonological inventory; I am a native Californian English speaker raised in Thailand (1-4), Okinawa (4-10), and California (10-23), and my parents are also raised on the West Coast. However, two years of intensive Hindi study in the U.S. and in New Delhi provided some training and experience in producing them for the purpose of communication. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... frequencies.2.12
The formant tracker used is part of the waves+ signal-processing package. Default parameters accurately tracked my voice: The signal is preemphasized using a preemphasis coefficient of 0.7. Then it is divided into overlapping frames, 100 per second, with each frame 49ms long, and the frames are windowed with a cosine7#7 windowing function. Then an autocorrelation algorithm is used to fit 12 LPC coefficients, from which spectral peaks and bandwidths are calculated. Finally, a dynamic-programming algorithm is used to find the globally optimum mapping between the LPC peaks and 4 formants, where optimality is measured by narrowness of bandwidths and continuity of formants over time. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... line2.13
If it is indeed a straight line, then slope = (max(onsetF2)-min(onsetF2))/(max(nucleusF2)-min(nucleusF2)). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... phonology,2.14
See the phonological discussion on page [*]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... antinodes.2.15
For example, see Figure 3, Pettersson & Wood (1987), which displays the nodes and antinodes of F1 through F4 in the vocal tract, adapted from Chiba & Kajiyama (1941). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... acoustics.''2.16
Thanks to David Graff for this formulation. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... sounds.2.17
Ladefoged (1990:401) also suggests that the vowel features (High, Low, and Back, for him) are acoustic features. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... vowels?3.1
The issues of temporal and static structure obviously are interrelated, since the features available at any point depend on the location within the syllable. Consonant features and vowel features are not identical; the set of features distinguishing among nuclei may well be different from those distinguishing among glides; certain features (e.g., place of articulation features) may be available only in certain parts of the syllable onset and not in others. And so on. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...caught?3.2
Apparently the answer is both yes and no. Peterson & Barney's (1952) study of General American deserves mention in this context since the data from that study is continually being recirculated (e.g., Watrous 1991). They collapse together speakers of several dialects differing, for example, in whether they have or don't have the phonological distinction between /ɑ/ and /ɔ/ as in cot and caught. The rather anomalous overlap between measurements of these classes (Figure 8) even in highly-monitored speech is therefore not very surprising. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... dialect.3.3
I lack distinctive categories for /ɔ, a:, 'ærV/ as in THOUGHT, PALM, & marry. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... homophonous.3.4
The layout in this listing is designed so that sets that are not distinguished in one dialect or another are relatively close to one another, where possible. The sequence of columns may be described as: front-r-glides, front-glides, other front vowels, other vowels, back-glides, other-r-glides. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... near-mergers3.5
I discuss ``near merger'' below on page [*]; also cf. Labov, Yaeger, & Steiner (1972, Chapter 6 and Appendix A). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... sets:3.6
Of course, stating a merger using lexical sets does not show what the resulting phonetic forms are. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...sear3.7
My monitored pronunciations of these are [lei, le; mo, mo, sii, si]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... one.3.8
Bailey (1985:162) makes the same point. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... deleted.3.9
Thus a secretary recently heard answering the telephone as ``Peter Science'' was reinterpreted correctly as saying ``Computer Science'' after normalizing for utterance-initial syllable deletion, /uw/-fronting, and the physical location of the speech event. Or for example in some speech recorded in Southampton, England, an utterance-initial syllable was constituted by a single pitch pulse, 40ms before the onset of the audible part of the utterance. Such observations may be typical of naturally spoken conversation. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... mono\-phthong\-al3.10
As mentioned above, these sound classes often have phonetic inglides in the Southern U.S. -- cf. the Alabama chapter -- but not in the basically Northern dialect of Reference American. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...put.3.11
Phonetically, these may be written [phIt], [pht], [pht], [phUt], respectively. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...bone.3.12
[biy], [biyn]; [beI], [beIn]; [buw], [buwn]; [boU], [boUn], respectively. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... counterparts.3.13
See also Peterson and Lehiste (1960) on ``intrinsic vowel length''. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... dialects.3.14
Los Angeles Chicano, Jamaica Creole, and Wisconsin/Minnesota, to cite three dialects, often have monophthongs for the ``long'' vowels. Cf. also the duration measurements in Appendix 3. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...bough.3.15
[b], [boI], [bo], respectively. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...high).3.16
These may be written phonetically as [hiyd], [hd], [hiy], [h], respectively. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...pp:l.3.17
These may be transcribed as [le], [lei], [mo], [mo], [fi], [fii], [tu], [du]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... identical.3.18
For me, these all end in []. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... identical.3.19
flour and flower may be a poor example, since they are historically derived from the same word, as in ``the flower of the wheat.'' 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...Sue3.20
[siy], [se], [soU], [suw], respectively. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...boor3.21
[bier, be, bo, b]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... off-glides.3.22
For example, as pointed out in Chomsky & Halle (1968), and Janda (1988), German on-glides /w,y/ become fricatives, but off-glides do not. Thus /au/, /aI/ don't become [av], [a]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... (1951).3.23
But see also, inter alia, Bloomfield (1934), Whorf (1943), Swadesh (1947), Pike (1947). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... correspond.3.24
The lexical sets with no place in this table are six: CLOTH, which goes with either THOUGHT or LOT, depending on dialect; BATH, which goes with either TRAP or PALM, depending on dialect; NORTH, which goes with FORCE in most dialects; and HAPPY, LETTER, COMMA, which are lexically unstressed and therefore excluded from T&B's consideration. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... uw/3.25
These symbols are constructed from T&B's table. / w/ is often also written as /ow/. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... dialects.3.26
However, /oy/ is raised to mid or higher in many dialects including my own; I therefore analyse it, below, in Reference American, as containing a mid-back nucleus. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... vowels:3.27
For an early statement, Kenyon and Knott's section, Variations, §92, p. xxxviii, in the Pronouncing Dictionary which has served as the data for much of the generative phonological treatment of English. For recent discussion, cf. Halle & Mohanan 1985, Section 2. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... glide.3.28
In many Southern U.S. dialects in various contexts, /y/ has monophthongized, so the high-front glide is absent. In Reference American, this glide is present. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...put,3.29
These may be pronounced as [pUt, pt, pæt, pt, pt, pUt]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...for3.30
cf. the split within the word horror ([hor]). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... monophthong3.31
This merger did not occur in certain Celtic dialects, where /r/ was not a continuant and thus remained outside the Nucleus-Glide positions. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... consonant.3.32
The exception is the Los Angeles Chicano community, where Santa Ana found that /r/ behaves like a consonant in its effects on /-t,-d/ deletion, an unusual and striking fact that may be attributable to the influence of Spanish contact with that English dialect. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...for3.33
In Reference American of these contain []. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... English.3.34
The argument here must be restricted to English alone, since there are languages in which the glide is the only segment which occupies the coda. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...fort,3.35
[fis], [fo t]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...soy3.36
Phonetically, [siy], [seI], [s], [suw], [soU], [soI], respectively. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... heavy.3.37
Bailey (1985) suggests that English is one of these. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... syllables.3.38
Thanks to David Graff and to William Labov for pointing this out to me. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... /l/.3.39
Part of the answer may be that languages cannot distinguish [15#15round] back glides. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... time.3.40
However, Krohn (1969) and Wang have argued that diphthongs are [+high, +low]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... flawed.3.41
Grammars may incidentally overgenerate, without being flawed. For example Russian requires the [±voice] dimension to distinguish /p/ from /b/, and other features for /X/, and /c/. But //, // are not distinctive, though [], [] do occur (Halle 1957, 1959). The feature system can hardly be at fault for this kind of incidental gap in the system. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... Bell's3.42
The speech pathologist, better known as the father of Alexander Graham Bell, the inventor of the telephone. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... (1867).3.43
For a discussion of the history of this model see Wood (1987). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... height.3.44
Stanley (1967) objected that using an unspecified value for a feature amounts to having three values for it. If [height] is taken as a feature, with values [high height], [0 height] and [low height], then his objection would apply. However, height is a tier, not a feature, and it contains features, not values. Only if tiers are considered to be identical to features is Stanley's objection relevant. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... PALM,3.45
Two other lexical sets which go with these are CLOTH, which in the U.S. goes with THOUGHT and BATH, which in the U.S. goes with TRAP. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...pp:struct.3.46
We may locate Wells' lexical sets in the slots of this structure, thus:

V V: Vr Vy Vw
KIT FOOT FLEECE GOOSE NEAR CURE 21#21 21#21 21#21 21#21
DRESS STRUT FACE GOAT SQUARE NORTH 21#21 CHOICE 21#21 21#21
TRAP LOT PALM THOUGHT (marry) START 21#21 PRICE 21#21 MOUTH

Not displayed are BATH, CLOTH, FORCE, NURSE, and the unstressed sets HAPPY, COMMA, LETTER. BATH goes together with TRAP; CLOTH with THOUGHT; FORCE with NORTH; COMMA with STRUT; LETTER with NURSE; HAPPY with FLEECE (though in the South, HAPPY typically goes with KIT). NURSE is analyzed separately from this system, as argued below. 

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...Murray.3.47
See the discussion of near-merger in appendix 2, page [*]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... not.3.48
Southern dialects may have [æ] in SQUARE, but this may be analysed as lowered /er/ rather than as /æ r/, as discussed in the Alabama chapter. If /ær/ were correct for these dialects, then so much the better for this analysis, since /æ r#/ would then fill this gap in the system. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... [front],3.49
For the sake of consistency in features, I represent the front-back dimension with a privative feature, either [front] or [back]. The choice between these two is discussed below where the use of the non-traditional feature, [front], instead of the usual feature, [back], is justified. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... \scripta/3.50
See for example, L.A. Chicano, page [*]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... low.3.51
Another perspective on this issue is the following. Notice that the glide position is used to encode four things: the three glides /y, w, r/, and length. This requires a front/back distinction, and a high/non-high distinction, but it appears that there is no need for a three-height distinction among glides. The multiple distinctions of height among long vowels may presumably be dealt with at a lower, phonetic level, where the height of the glide is phonetically assimilated to that of the nucleus. If the glide position need not distinguish three underlying degrees of height, then the [low] feature in glide position may be re-interpreted simply as non-high. This would make better sense of the three instances where ``non-high'' was used in the preceding two paragraphs. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... dialects.3.52
r-less dialects retain /r/ in syllable onsets, which may be specified as [rhotic] underlyingly. Where r-less dialects have postvocalic ``intrusive'' /r/, as in ``A vodka or two''[vdkthuw], the intrusive /r/ must be analysed as non-underlying, since it cannot be syllable-initial. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... one.3.53
In most English dialects ``glide'' is also a phonetically appropriate term for long vowels, since they do generally have phonetic glides. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... system.3.54
In phonological and phonetic performance, errors are almost non-existent (Labov 1966), and dysfluencies or ``false starts'' themselves are well-formed according to a simple set of rules (Hindle 1983). Limitations of memory pose no difficulty for native speakers in the production or perception of sequences of sounds. Thus the flaws of performance data adduced for syntactic data do not apply with any force to phonological and phonetic performance. On the contrary, in phonetics and post-lexical, surface phonology, a similar concern about errors in the data leads to the opposite conclusions: speech errors are more frequent when speech is self-conscious. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...u]3.55
Here [y,w] are as Chomsky & Halle (1968) define them: higher than [i, u]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... u/,3.56
Central /⋀/ is not only quite non-peripheral, it also doesn't share the roundness feature with its counterpart /o:/, as in Table [*]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... sounds,4.1
This may be due to the fact that sounds are generally louder when voiced. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... speech)4.2
cf. the Project on Linguistic Change and Variation, Labov (1980). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... another.4.3
What is ``easy'' varies across languages. Just as what is physically easy for some is difficult for others, because of the practice to which their muscles -- or neuromuscular control systems -- are accustomed, different languages and dialects may set the default energy-expenditure level of the various speech articulators at different levels. In this way, the physical sluggishness of the articulators may effectively vary across languages. Cf. Sievers (1901), who offered the explanation of phonological symmetry that there is a different rest position of the organs in speakers of different languages. We may add to Sievers that certain muscle movement patterns are highly practiced in one language and not in another, and are therefore easier, both because of the resulting strength and endurance of the muscles involved, and the more redundant, robust, and fully-trained neural control systems behind the movements involved. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... model).4.4
In an all-pole model, the representation of a spectrum in terms of center frequency and bandwidth of spectral peaks, or poles, is interconvertible with several other representations. Thus formants may be considered as good as any other analytical representation of the speech spectrum, under the assumptions of this model. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... made.4.5
The tape recorder was a high-quality (Nakamichi) cassette recorder; the tape used was Maxell XL-II cassette tape; the microphone was a broadcast quality Shure 570S lavalier mike; the setting was an isolated room in a quiet house, with no machine noise. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... purpose.4.6
Since a moving-average amplitude contour may be spiky rather than smooth, some method is necessary for separating ``true'' crossings of the threshold signifying the beginning of the acoustic vowel, from false crossings due to short-duration amplitude spikes due to single glottal pulses, transient bursts, etc. ``Sloppy-crossing'' is an algorithm developed for this purpose which requires the threshold to be crossed for more than a given fraction, 29#29, of a given amount of time, 30#30, before a ``true'' threshold crossing is considered to have occurred. If these parameters are set to 29#29=80% and 30#30=.1 seconds, for example, then if the parameter stays over the threshold for 80% of any .1-second segment, a threshold-crossing is recorded at the beginning of that segment. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... formants.4.7
cf., Labov, Yaeger, and Steiner 1972:29,32, and Vol. II, Figures, p. 9. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... formant-tracked.5.1
The computer software used in this project for digitizing, for formant-tracking, for display of waveforms, spectrograms, and formant-tracks, and for segmentation, are parts of the waves+ package developed by David Talkin at AT&T Bell Laboratories, and available commercially from Entropic Speech, Inc., Washington, D.C. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... technique5.2
This post-processing technique is described in Talkin (1987). Cf. also Secrest and Doddington (1983), and Dupree (1984). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...S.5.3
Documented in Becker, et al, 1988. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... starts.5.4
Interestingly this is not a positive cue, but rather the non-application of a regular process, which is in a sense a negative cue. Another cue to the presence of this ``editing signal'' is phonotactically impermissible syllable endings, e.g., str-. There are also explicit editing-signal morphemes, such as uh, um, etc. I speculate that another editing-signal cue is an apparent stress on closed class items which shouldn't be stressed, where the appearance of stress comes from a rapid change in the pitch contour at the boundary of the hesitation. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... class,5.5
William T. Reynolds. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... carelessness,5.6
Doddington estimates usual human error rate in making categorical classifications of clear cases at around 4%. However when highly motivated, the level improves to under 0.5%. In the current case, the motivation of the coder (me) is undoubtedly greater than that of a poorly paid, unmotivated experimental subject, but not as great as the financially highly motivated subjects in Doddington's study. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... bootstrap.5.7
Equally well-known is a resampling technique known as the jackknife, so-called because it's good for many purposes. The bootstrap is simpler and even more useful (though it will not quite allow you to pull yourself up by your bootstraps). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... statistics.5.8
A function in the S language (described in Becker, et al, 1988) to bootstrap resampled statistics from a given dataset is given here and described below: 


bootstrap 34#34 function(data,stat,nResamplings) { 


N 34#34 length(data) 


result 34#34 vector(mode="numeric",length=nResamplings)  


for (i in 1:nResamplings) {   


result[i] 34#34 stat(data[round(runif(N,min=-0.5,max=N-0.5))])

} 


return(result) 

}

This function works thus: Inputs are the sample itself, data; the function, stat, which generates the statistic (e.g., mean, sum, variance, etc.); and the desired number of times to resample the data, nResamplings. The length of the data vector is the number of observations in the sample, N. The output of the function, result, is a vector of numbers as long as the number of resamplings desired (this should be on the order of 200 to 1000). The substance of the function is a single line inside a loop, which repeatedly does the resampling and the calculation of the statistic. The key line works from the inside out, thus:

runif(n,min,max) generates a list of n random floating-point numbers from a uniform probability distribution between min and max, thus generating random floating-point numbers between -0.5 and N-0.5. round(vector) rounds off each element of a list (or vector) of numbers to integers; in this context it rounds off all the random numbers to integers between 0 and N-1, inclusive. data[vector] uses the vector of integers as a list of indexes into the sample data, and extracts the indexed elements. Thus the random numbers are interpreted as indexes into the data array. Note that multiple references to the same index can occur freely, so that if data was the vector (6, 7, 8, 9), then the expression, data[(1,2,1,3,3,1)], would return the vector, (6, 7, 6, 8, 8, 6). stat(vector) interprets vector as a data sample, and computes the given statistic using the numbers in that sample. In this context, then stat() takes the indexed data-points picked by the random-number generator, and calculates the statistic using them. Finally, result[i] 34#34 stat(..) sets the i'th element in the result[] vector to be the value that is returned from the calculation, stat(..).

This code is not especially fast, and it is elaborated when applied to multi-dimensional data, but it is effective both for doing the task itself in the one-dimensional case and for showing how to do (and how easy it is to do) bootstrap resampling. 

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... study.6.1
As was Jim from Chicago, for example. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... Roasta,6.2
Unstressed /ɝ/ becomes /a/ in Jamaican, thus I might call him ``Roaster'', a nickname that comes from his avocation, ``roasting'', or moonlighting with his employer's equipment. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... census.6.3
This information is from the Statistical Yearbook of Jamaica, 1986. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... Wells.6.4
LePage (1960), Cassidy & LePage (1967), Wells (1973), Wells (1982). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... (1982):6.5
As discussed in Phonological Preliminaries, page [*], Wells' comparative categories are not the same as the historical word classes whose mergers, splits, and phonetic changes resulted in the modern form of the language. These categories are not primarily intended to show the changes by which Middle English developed into Early Modern English and into Modern Jamaican, but simply to show the lexical correspondences that now hold between this dialect and others.

35#35 The uncertain, possibly unmerged status of the low back phonemes /, :/, which extensionally correspond to the lexical sets LOT, THOUGHT, NORTH, is discussed below and in Appendix 2. 

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... Cassidy's.6.6
The corresponding lexical sets are shown in a footnote on page [*]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... /bwai/.6.7
Although avoid is /avaid/ rather than /avwaid/ in Roasta's speech, so merger has indeed occurred in some cases. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... offglide.6.8
The offglide could be [] or [], [o]. Distinguishing these possibilities is a difficult matter, since [i, uo], and [i, u] are all ingliding vowels. It may have been equally appropriate if Cassidy had written these vowels as /i, u/. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... tier6.9
In Phonological Preliminaries, phonological height is analysed as an autosegmental tier, which may contain a single privative height feature or none, where [high] and [low] are the two features available. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... Shift,6.10
If we may reason from synchronic rules to diachronic rules, it would appear that the Great Vowel Shift did not proceed symmetrically in the front and the back. It is indeed a fact that the two vowels that underwent diphthongization and lowering from earlier /iy/i:, u:/uw/ to modern /ay, aw/ in many dialects of English did not both fall to low in Jamaican Creole. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... back.6.11
The set of rules stated here could possibly be simplified if the natural rules of nucleus-glide differentiation (in height features) and nucleus-glide assimilation (in backness features) were attributed to general principles, and if they did not need to be stated explicitly in a grammar of Jamaican Creole. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...jc:struct.6.12
The lexical sets corresponding to the slots in Table [*] are as follows:

V V: V+{i,u,r}
KIT FOOT FLEECE GOOSE NEAR/SQUARE CURE/FORCE
DRESS STRUT FACE GOAT PRICE/CHOICE MOUTH
TRAP/LOT PALM/THOUGHT/CLOTH START/NORTH 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... speakers,6.13
The numbers of measurements broken down by class and speaker are given in Appendix 3, along with mean vowel durations. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... vowels.6.14
A display of single means would show the same structure, but would not show how significant the relationships between the means are. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... phonemes.6.15
Liljencrantz and Lindblom's theory of maximal dispersion may be criticized in two ways. Labov 1982 has pointed out that their results are based on phonemic data, not the kind of phonetic data which would be necessary to support their claims. If i, a, and u are used in the phonological transcription of most of the languages of the world, it cannot be inferred that most of the languages of the world have vowel phonemes with the phonetic qualities of IPA [i, a, u]. A language might well have, for example, [i, , u] as the main allophones for /i, a, u/. Any three-vowel system is likely to be transcribed with these symbols, no matter what the phonetic targets are. Bessell's (1991 and forthcoming) studies of the phonetics of vowels in Interior Salish languages make this quite evident. The fundamental distinction between phonetics and phonology is ignored when this inference is made from phonology to phonetics.

L&L's functional theory suggests that sounds should be maximally far apart, but a principle of minimum effort is also necessary, because sounds are not always maximally separated. For example, languages with just two tones often use the minimum degree of phonetic difference to carry the contrast. Thus there is a balance between maximal distinction and minimal effort. But this amounts to saying simply that sounds are more or less distinct, which is vacuous.

Something must be salvaged from L&L's theory in order to use it in deriving (2) and (3). 

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... phonetics6.16
For the distinction between general and linguistic phonetics, see page [*]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...I]6.17
That is, the Reference American vowels /ay, aw, oy/, or Jamaican /ai, ou, ai/, corresponding to the lexical sets PRICE, MOUTH, and CHOICE. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... stress.6.18
See Harris (1969, 1983). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... vernacular.7.1
The vernacular of most African American speakers is substantially different, in Northern U.S. cities, from the vernacular of white speakers in those cities. African American English in the northern cities has its origins in Southern U.S. speech and in the Caribbean creoles. It has been imported via migration to the North, and is relatively homogeneous throughout the U.S. Thus pen/pin are likely to be homophonous, /y/ is variably a monophthong before voiced consonants and in free position, and final consonants are frequently deleted (so that told, toll, and toe may be homophonous). The de facto segregation of the black and white communities has resulted in continuing linguistic divergence. See Labov and Harris (1986) for a discussion of this situation in Philadelphia. Thus it is necessary to analyse the white and black dialects separately. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... tape.7.2
Nagra produced the highest quality portable tape-recorders available before digital audio tape. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... 1990,7.3
The interview was conducted by Sharon Ash as part of the fieldwork done for the project on Cross-Dialectal Comprehension with William Labov. I would like to express again my thanks to her and to the members of CDC project. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... speakers,7.4
Speakers from other Northern Cities, such as Detroit and Buffalo, are equally advanced, sometimes more advanced in some of the changes. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... speech.7.5
Casual speech is the least-monitored style of speech, in which speakers pay less attention to how they are speaking than to the content of what they are saying. Casual speech is formally defined as speech that is directed to people other than the interviewer and narratives, in which the telling of important events takes precedence over the constrained self-consciousness of the interview situation. Careful speech is all other extemporaneous speech. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... advanced7.6
That is, phonetically advanced along the paths of the sound changes in progress in her dialect. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... speakers7.7
Including 5 from Detroit, 4 from Buffalo, 1 from Rochester, and 4 from Chicago 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... distribution7.8
FLEECE-type words have lexically stressed /iy/, while HAPPY-type words have a lexically unstressed vowel. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...cwe:struct.7.9
The lexical sets corresponding to these phonological classes are:

V V: Vr
KIT FOOT FLEECE GOOSE NEAR CURE
DRESS STRUT FACE GOAT SQUARE FORCE/NORTH
TRAP/BATH LOT/PALM   THOUGHT/CLOTH marry START



Vy Vw Unstressed  
21#21 21#21 21#21 21#21 HAPPY  
21#21 CHOICE 21#21 21#21 COMMA  
21#21 PRICE 21#21 MOUTH LETTER   
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... Belt.7.10
See LYS for the first discussion of this shift that I am aware of. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... above.7.11
Some ingliding does occur with this vowel (4 of the first 30 stressed /æ/ tokens), but none of the first ten sounded ingliding to my ear, perhaps because these were mostly non-final, thus not especially lengthened tokens. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... [\invv].7.12
In some productions // moves downward towards [æ]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... 7.13
Bloomfield (1934), when he phonemicized the TRAP, BATH classes (here, /æ/) as /æ/ , and the STRUT class (here /⋀/) as /o/, would seem to have been prescient (unless he was influenced by the Northern Cities Chain Shift already in progress, which is unlikely), since /æ/ rises towards [e~ ] and /⋀/ backs towards /ɔ/. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... 3.7.14
But cf. recent revisions in Labov, forthcoming. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... fronted.7.15
Fronting is relative to the values in Reference American, where /u:, / are realized as [w, ]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... ow,7.16
The vowel /ow/ is sometimes monophthongal, as is more typical of Minnesota or Wisconsin, but more often has an upglide. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... measured.7.17
In an important study of the effects of situational style-shifting on a speaker's vowel system, Hindle (1980) measured even more tokens (33#33 10,000) in the speech of a single Philadelphia speaker, recorded during work, at dinner with her family, and during a bridge game with friends. The situation-related vowel shifts are closely related to the historical changes ongoing in Philadelphian sound system, in that sounds in more vernacular, informal-style speech is more historically advanced. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... corrected,7.18
See page [*]ff for discussion of the post hoc examination of outliers. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... efforts.7.19
Failure to decode the sounds of speech does occur; language does not always function successfully, as Labov's project on Cross-Dialectal Comprehension shows in detail. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... shift,7.20
However, according to the discussion on page [*], /æ/-raising is the second step, which follows the loss of the Reference American /a:/ phoneme, broad A, by merger with /ɑ/. It was shown above that /a:/ is not a distinct class in Chicago. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... diphthongs7.21
Chapter 3 argues that postvocalic /r/ is a glide, and forms diphthongs with preceding nuclei. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... words7.22
``Clitics'' are defined in the Methods chapter. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... minimized.8.1
Feagin (1991) summarizes the general social picture regarding the disappearance of Southern r-lessness and gliding:

Anecdotal evidence supports this hypothesis. A young upper-class boy from Anniston -- the son of a banker -- who was attending Amherst College in Massachusetts told me recently that people often are surprised that he is from the South. That is probably because of the stereotype of Southern speech which is that Southern States English is R-less and full of diphthongs on vowels which in Northern States English have none. Actually, the stereotype is not altogether incorrect, for a few older working class people, but it is out of date. The majority of young Southerners -- of whatever social class -- have R's now. What the stereotype is depicting for the glides turns out to be working class speech, at least for younger people -- or the speech of older women, of whatever class. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... isogloss8.2
Based on distributions of lexical items rather than phonological features. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... English.8.3
To ease the usually difficult task of understanding what sounds are being referred to, I present here the lexical sets of Wells (1982) that correspond to the sound classes in the given structure. (For further explanation, see the section on English Lexical Sets in Phonological Preliminaries.)

V Vy Vw Vr
KIT FOOT FLEECE 21#21 21#21 GOOSE NEAR CURE
DRESS STRUT FACE 21#21 21#21 GOAT SQUARE FORCE
TRAP LOT/PALM PRICE CHOICE MOUTH THOUGHT START NORTH 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... Preliminaries.8.4
It has been claimed (for example, by C.-J. Bailey 1985:205) that as dialects of English, Southern and Northern (and by extension, Reference American) share an underlying vowel system. The systems under discussion here are surface-phonological, or post-lexical in the theory of Lexical Phonology, and they are not the same across these dialects. Surface phonological structures are the output of the lexical phonology, and their details of implementation are assumed not to be subject to the direct influence of morphological or grammatical features in specifying their phonetic realization. The topic here is not lexical phonology, but post-lexical phonology. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... dialect.8.5
In fact, 56% of non-native-born Tuscaloosans in 1850 were from the Carolinas or Virginia (Foley 1972:3) 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... 1982)8.6
This is, or was, a Southern shibboleth: monophthongs in the phrases ``nice white rice'', or ``bright light tonight'' are stigmatized (Feagin, p.c., C.-J. Bailey, p.c.). Thus, speakers that monophthongize /ay/ before voiceless obstruents are marked as relatively lower-class speakers, though all speakers monophthongize in some other environments (C.-J. Bailey 1980:171). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...{,}].8.7
Here as in SPE, () marks an optional element, and {} marks a choice among elements. It may be pointed out that this use of () notation is incompatible with C.-J. Bailey's use of () within phonetic brackets, [..(X)..], which signifies a particular kind of systematic ``optionality'', namely that X is present in monitored styles, and absent in unmonitored styles. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... (1990).8.8
See also McDavid (1948) for a very early sociolinguistic study of /r/ in another Southern variety. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... Island,8.9
Steven Peloquin, an undergraduate linguistics student at the University of Pennsylvania, has discovered a merger of chair and cheer, bear and beer, hair and here, etc., in his Rhode Island dialect. I have found two older Rhode Islanders that do not share the merger, according to minimal pair tests. But another one, a younger man, seems to share the merger, according to minimal pair tests, and furthermore merges the corresponding back vowels, the sets CURE and FORCE (also NORTH), so that bore and boor, tore and tour, etc., sound the same. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... dialect,8.10
cf. Foley (1972:47, #13) and Kenyon, American Pronunciation, 10th ed., pp366-372, among many references. For this reason the Northern dialects which neutralize this distinction cannot be used as a starting point for the description of Southern dialects: it would be impossible to derive the membership of the FORCE and NORTH classes in a dialect that distinguishes them from a dialect which merges them. Wells' lexical sets are crucial in this context because they can be used to define most of the relevant sound classes in all English dialects, including this contrast in Alabama speech. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... whites.8.11
The merger of the Mary, merry classes occurs in many other dialects, including Los Angeles Chicano English, and Chicago White English, and Jamaican Creole. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... pronunciations).8.12
C.-J. Bailey (1985:162) is the source of the Ayre/heir/air triplet. Another given there is they're:their:there. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... /ay\$r/?8.13
Here, $ represents a syllable boundary, and /y/ represents not phonetic [y] but an abstract feature that is sometimes realized by up-gliding and sometimes by length. Surface phonetic [ay] sequences do occur, from phonological /a$y/ sequences. But while onset /y/ may surface as [y], coda /y/ does not. This is illustrated by the (r-less) Southern minimal pair, Maya [may], vs. Myer [ma:], pointed out to me by C.-J. Bailey (see, for example, 1985:118). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... 1972:36),8.14
It should also be noted that /aw/ before /r/ remains disyllabic (as in ``flowers'', Foley 1972:36), as does /y+r/, just as in Reference American. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... monophthongs.8.15
This is surprising, since monophthongal /ay/ may be categorical in some environments. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... vowel.8.16
In r-less Tidewater Southern speech (as in Boston), the /r/ class is pronounced with a low-fronted vowel [a:]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... speech'',8.17
According to Foley (1972:50), [] is more restricted than [], occurring in /r, or/ words for blacks, but only in /or/ words for whites. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... pronunciation.8.18
C.-J. Bailey points out to me that Southern boy as a call to a dog, or a threat to a youth, can undergo the identical shift as French moi: [bwa:], and similarly that goin' may be rendered as gwine. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... immigration9.1
The resident Mexicans, or ``Californios'', who remained in California after the Mexican war in which California was taken away from Mexico, are a tiny population, mostly assimilated to the Anglo community. The entire recorded population of Los Angeles in 1820 was 615, and of California at the time of the 1848 war, about 15,000. The populations discussed here, both Chicanos and Anglos, is entirely composed of later immigrants. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... etc.9.2
As in FLEECE, KIT, GOOSE, FOOT, respectively. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... /\er/9.3
As in TRAP, NURSE. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... American.9.4
The issue here is whether or not there was exposure to natively-spoken English in the household during childhood. The first person in the family with native competence is the first-generation speaker, in this view. All those in the family with a native-English speaking parent are exposed during childhood to natively spoken English, whether or not there is also non-native English in the household, spoken by family members that are immigrants. Thus if even a single grandparent is a native English speaker, the grandchild of such a speaker is third-generation. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... opportunities,9.5
The population in 1930 of East L.A. was 70% of 1970's population, whereas in West L.A. it was 3%. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... community.9.6
We have collaborated on the acoustic analysis of this and other speakers, working together to some extent since our interests are complementary. His work focuses on social variation within the Los Angeles Chicano English speech community, considering consonant deletion, and variation in the phonetic quality of stressed vowels, while my work focuses on the internal sources of variation in vowel quality, according to consonant environment and stress. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... sets''.9.7
Using the term ``merger'' is misleading because actual phonological mergers have not necessarily taken place in each case in this particular speech community. Nevertheless, this remains a convenient term when referring to lexical sets that are not distinguished. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... U.S.9.8
Thus, my parents, both born and raised on the West Coast, maintain the distinction, while I and my sisters do not. Previous studies of this merger are discussed below. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... KIT).9.9
cf. Bailey, 1985:187. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...lace:struct.9.10
I view ambiguity as a greater sin than lack of correspondence between graphemic and phonological representation. The typography used, with /i:, I, u:, U/ as the high front and back vowels, is intended to avoid the ambiguity of /i/ and /u/ which arises when these 4 vowels are alternately symbolized with the sets /i, I, u, U/ or /i:, i, u:, u/. Similarly, ``e'', ``o'' are avoided due to this ambiguity. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...9.109.11
The lexical sets (of Wells, 1982) that correspond to these phonological classes are given below in order to clarify the correspondences of these classes with those of other dialects, including the reader's. See page [*]ff for discussion of what lexical sets signify and how they are useful.

V V: Vr Vy Vw
KIT FOOT FLEECE GOOSE NEAR CURE 21#21 21#21 21#21 21#21
DRESS STRUT FACE GOAT SQUARE NORTH 21#21 CHOICE 21#21 21#21
TRAP 21#21 21#21 LOT 21#21 START 21#21 PRICE 21#21 MOUTH
                 

BATH is included with TRAP; FORCE is included with NORTH; PALM, CLOTH, and THOUGHT are included with LOT, Unstressed vowel classes HAPPY, COMMA, LETTER, and the /ɝ/ vowel, NURSE, are not shown here. 

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... syllable.9.12
This is formalized on page [*] for Reference American: Stressed rhymes branch. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... u:/9.13
These vowels are sometimes written more phonetically, as they occur in other dialects: /iy, ey, ow, uw/. Both transcriptions can be accurate at the same time, depicting different stages of the derivation of the phonetic form. There is no problem of ambiguity among the classes referred to, and the particular transcription chosen is an arbitrary decision. However, in this dialect, these vowels are frequently monophthongal -- probably an ethnic marker that the LA Anglos do not share. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... /mer\$i/.9.14
Note that $ signifies a syllable boundary. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... contrast9.15
For example, a near minimal pair would have been marry vs. sorry. The contrast between them here is not simply front vs. back, but also mid vs. low. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... [\openo].9.16
This may have first been noticed by Labov for Philadelphian white vernacular, and has been shown to occur much more widely, especially throughout the American South, by Thomas, Bailey, & Benson, 1990. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... English.9.17
See discussion of this point on page [*]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... Ana.9.18
For justification of the phonetic importance of F1 and F2, see the Acoustics chapter (e.g., page [*]). For the defininition of acoustic vowel, see page [*]. For the measurement procedure, see the Methods chapter (page [*]ff). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... or/.9.19
Of the remaining back-round vowels, /:/ has merged with //, and /ur, oy/ are too infrequent in this data to display in this form (n=0, n=5, respectively). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... or/.9.20
This supports the point made with respect to Reference American (page [*]) that // is not back but central. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... Anglo9.21
A pre-medical student at UC San Diego, she is an upwardly mobile middle-class speaker. She participates in, though probably is not among the leaders of, ongoing Anglo sound changes. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... nuclei.9.22
Acoustic vowel is defined on page [*]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... words.9.23
For the precise definition of this distinction see page [*]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...l/.9.24
It should also be noted that the word you, which occurs quite frequently, was phonologized as /y/. It never sounds back or rounded, even when stressed. The single token of stressed // that occurs in the clitics-included chart is from the word, ``you''. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... level9.25
Two-tailed t-test, 5% level of confidence, as described on page [*]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... features9.26
For example, // is underlyingly a Glide slot specified with the feature [rhotic]. This feature is then linked, after nucleus-insertion, to the nucleus timing slot. See Phonological Preliminaries, page [*]. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... England'.''9.27
Herold, 1990:8, quoting Trudgill, 1986:147. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... claimed9.28
cf. DeCamp, 1959; Johnson, 1974; limitations of these studies are pointed out in Herold (1990:8-11) and are overcome in systematic instrumental work by Moonwomon, 1991. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... vowels,9.29
See, for example, Labov, Yaeger, & Steiner (1972:94). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... English.9.30
See Labov, 1991. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... kingdom.10.1
The wing movements of hummingbirds and bees, for example, though more rapid, appear to be less complex (as well as smaller in physical scale), since they follow a single, repetitive pattern of motion, while the human tongue moves in leaps and contortions that, while not entirely unpredictable, are not yet well understood. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... end.A.1
Axiomatized: For every time t1, there is a time t2 after t1, and a time t3 before t1. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... followA.2
Let F(e) be a function from elements to sets of elements such that for any pair of elements, e1 and e2, if e1 is in F(e2), then F(e1) is a proper subset of F(e2). If such a function exists, the set is called a linear sequence. Note that F(e1) does not contain e1, since if e1 were in F(e1), then F(e1) would be a proper subset of itself, which cannot be true. We may define P(e) as a function from elements to sets of elements: P(e) =  (e 72#72 F(e)). In words, P(e) is ``everything else except e and F(e)''. The union of P(e), e, and F(e) thus exhausts all the elements in the universe under discussion; further, they do not intersect. We may say that elements in P(e) are 'before' e, and elements in F(e) are 'after' e, though we could as well say that P(e) is after and F(e) is before e, and linearity would still be satisfied. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... StaticityA.3
In discrete time staticity may be defined as the property that an event may occur at more than one time, s.t. the times at which it occurs form a ``contiguous sequence''. A set of times T forms a contiguous sequence iff there is no proper subset T1 of T for which T1 and T 73#73 T174#74 (the complement of T1) have no ``adjacent'' members. Times t1 and t2 are adjacent iff there is no time t3 such that either t119#19t319#19t2 or t219#19t319#19t1, where ``19#19'' may be defined trivially using the functions F() and P(): t119#19t2 iff t1 75#75 P(t2). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... events/states/elementsA.4
As implied above, these terms are synonymous within this discussion. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... forward.A.5
Thanks to Richard Janda for this observation. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... representations.A.6
This paper further opens up questions of the cognitive representations of actual time which are not often considered. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... GVS.B.1
At the same time, other aspects of the GVS did not go as far in Jamaican Creole as elsewhere: the lowering of Middle English /u:/ (as in MOUTH, discussed above) did not go ``to completion'' as it did in other dialects. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... approximated?B.2
Herold (1991) provides a detailed examination of the conditions of merger. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... dialect.B.3
For example, after several years as a graduate student in a relic area where a merger that occurs in my native dialect had not occurred (/  /), spending my time explicitly studying speech sounds, I may have acquired the distinction, at least in some words where spelling gives the correct clue. Thus I was recently unable to understand a low-back-merger speaker claiming someone was a [frd], because I expected [] in the word fraud and somehow couldn't understand what a frodd could possibly be. So insofar as I can be confused by the wrong phonetic form in a word, I may be functioning rather like a relic-area speaker. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... one.B.4
Cf. Graff, Labov, and Harris, 1986, and other papers in that group, which showed that blacks successfully imitating Philadelphia white speakers front /ow/ but not /aw/. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.