next up previous
Next: Outlier Analysis Up: Methods Previous: Locating Acoustic Nuclei


Exclusions, Deletions, Total Measurements.

Within the segment of speech chosen, ranging from a few minutes to nearly half an hour, an attempt was made to examine every single occurring vowel. A small percentage of the vowels are devoiced or otherwise deleted in speech, while another, usually small subset was excluded due to background noises or overlapping speech. The overall numbers of tokens in these various categories are shown in Table [*]. For each dialect and speaker, it lists the total number of vowels examined, the number of vowels which were found to correspond to no acoustic vowel (that is, those which were phonetically deleted, devoiced, etc.), the number of vowels which were excluded due to overlapping speech, or background noises, and the total number of vowels measured.


Table: Numbers of tokens, deletions, exclusions, measurements.
Dialect Speaker Total = n(VØ) + n(excl.) + n(meas.). %(VØ)
JC Juba B. 2891 82 129 2680 3.0%
JC Roasta M. 1872 75 124 1673 4.3%
CWE Rita S. 4821 211 140 4470 4.5%
CWE Jim C. 2856 121 406 2329 4.9%
CWE Judy H. 1775 105 55 1615 6.1%
AE James H. 1818 137 43 1638 7.7%
LACE Vince M. 2190 104 196 1890 5.2%

%(VØ) = 100*n(VØ)/(n(VØ)+n(meas.)).

The percentage of phonological vowels which were found to be associated with no measurable acoustic vowel segment is calculated as the number of phonetically deleted segments out of the total number of measurable segments.

The deleted/devoiced tokens might be used as data for another study, examining the conditioning of vowel deletion and/or devoicing. My impression, as might have been expected, was that vowels deleted most often between voiceless obstruents. A more detailed characterization of these exclusions will have to await further analysis beyond the scope of this book.


next up previous
Next: Outlier Analysis Up: Methods Previous: Locating Acoustic Nuclei
Thomas Veatch 2005-01-25