The Shape of Vowel Space

Three Chicago speakers were analysed, so as to give weight to the claim that the patterns found are characteristic of the dialect, rather than idiosyncrasies. F1-F2 measurements of their vowels are given in Figures , , .

The first speaker analysed, Rita, is the subject of one of the larger acoustic analyses of a speaker's vowel system, with 4821 vowels examined, and 4479 measured.^7.17 542 were not measured either because of extraneous noises (e.g., overlapping speech) that degraded the signal, or due to deletion or devoicing of the vowel. F1-F2 measurements of all tokens for the three speakers are displayed in Figures , , .

Figure: 4470 F1/F2 measurements. Interviewed 1990.

49#49

Figure: 1614 F1/F2 measurements. Interviewed 1988.

50#50

Figure: 2328 F1/F2 measurements. Interviewed 1970.

51#51

The F1-F2 vowel spaces are somewhat different from one another; the most obvious visual difference is the overall density of points on each graph, which is due simply to the different numbers of tokens measured. The three charts share fundamental properties. The distributions are roughly triangular, with somewhat of an asymmetry between the front and the back: the high-back corner is relatively sparsely filled. This reflects the frequencies of the various vowels: high-back vowels are less frequent than other vowels for all the Chicago speakers. These frequencies are shown in Table , where the counts are ordered from most to least frequent for each speaker.

Horizontal lines distinguish frequent, common, and infrequent vowels.

Table: Phoneme counts ordered by frequency for each speaker.

Rita		Jim		Judy
	564		505		266
ay	555	ay	325		210
	543		259	ay	177
	427	ow	170	iy	162
iy	491	iy	167	ow	128
	352	æ	149		113
ow	341		146		102
æ	278		108	æ	95
	220	ey	108		91
	178		101	ey	87
ey	177		94		64
uw	169	aw	62	aw	39
	75		51	uw	33
aw	50		45		26
	40	uw	37		15
oy	5	oy	1	oy	6

Vowels between horizontal lines vary in relative frequency across speakers; sets separated by horizontal lines do not vary in relative frequency in this data. In particular, notice that the high-back vowels /uw, / are among the least frequent vowels for all speakers. This accounts for the lower density of tokens in the high-back corners of these vowel spaces.

Similar factors may be responsible for the front-back asymmetry found also in Chicano, Alabama, and Jamaican Creole. The main difference between these charts and similar charts made for other dialects (pages , , , .) is that these are more perhaps more evenly distributed over the entire space. The Chicano speaker's tokens are more concentrated in the upper half of the space, while the Alabama speaker very sparsely fills both the high and low regions, and the Jamaicans have a mode in the low corner and (for Juba especially) a large gap in the high-central region. These interesting and suggestive patterns of the overall envelope of variation for other dialects can best be interpreted with reference to the phonological system of the dialect.