The durations of the single phases of an EGG period are analyzed regarding the impact of gender, stress, tenseness and vowel type factors.
The closing phase is divided into two parts (start and end of closing) according to the description given in section 12 (compare Fig. 22). Preliminary analysis showed that the duration of the start of closing does not depend on the factors mentioned above, thus it was eliminated from further studies.
The mean duration of the end of closing (called closing duration in the subsequent figures and tables) differs significantly for both gender groups (repeated measures ANOVA with factor sex, F(1,5242)=2819.8, p < 0.0005). This phase lasts about twice as long (relatively to the pitch period duration) for females than it does for males. The effects of other factors were tested on standardized data.
The repeated measures MANOVA analysis within both gender groups exhibits significant dependences across all examined factors. Unfortunately, the factors interact. In the female group the values also depend on measurement repetition. Thus, it is necessary to analyze the data within smaller groups. From the careful analysis of the data summarized in App.I we can draw the conclusion that the duration of the end of the closing phase depends on stress in an unpredictable manner - it varies between vowel groups and tense/lax vowels. However, the relation to tenseness shows a tendency towards a longer closing phases (Fig.29). The dependency is significant for the /a/,/i/,/u/ vowel groups in unstressed tokens, whereas in stressed tokens the effect is strong in /o/ and /u/ groups. The tendency towards a longer closing duration is similar for both genders.
The tense vowels are pronounced with an increased contact phase (i.e. closing phase) that is 7% longer than that of lax vowels (Fig.29). This is a highly significant result, considering the fact that in our experiment every single pitch period was analysed. Comparing the outcome of the experiment with the experiment of Claßen et al. (1996), where the same phonetic material was analyzed, we found a high correlation of this EGG parameter to the acoustic parameter "rate of closure".
It may be hypothesized that the effect was caused by the higher tongue position in tense vowels, which results in narrowed constriction in the vocal tract, which in turn has consequences for the interaction between vocal tract and larynx. It has been shown, that a narrow constriction in the vocal tract has an effect on the vocal source (Bickley & Stevens, 1986). Other explanation of the tenseness effect are based on the mechanical infleunce of tongue root advancement on the status and functioning of the vocal folds (Denning, 1989). A more elaborate analysis of the observed tenseness effect is described in section 18.4.
The tendency of back and close vowels towards an increased duration of the closing phase should be also noted.
The contact duration has been defined as the time at which the EGG amplitude is above 90% of the peak-to-peak amplitude. The statistical analysis again shows that the gender effect is strongest (F(1,5242)=1822.1, p < 0.0005). For females this segment (contact duration) lasts about 15% of the pitch period duration compared to ca. 19% for males. The shorter contact for females suggests a poorer contact between the vocal folds, which is generally in line with the direct, glottographic study of female voices (Holmberg et al. 1988, 1989; acoustic parameters: Hanson, 1995).
| Figure 29. Standardized duration of the end of closing plotted against vowel group for all speakers. The lax tokens are marked with empty circles, the tense tokens with full ones. The standard deviations are plotted as whiskers. |
|
The analysis of full contact duration across other factors yields mixed effects. Regarding data standardized against speakers, statistical analysis does not reveal any definite relations because of the strong interrelations between factors. For tense vowels there seems to be a tendency towards a longer duration of the fully closed phase. The effect is particularly strong in /u,/ vowels. Based on this analysis one can state that the linguistic phenomena of stress and tenseness do influence the "quality of contact" in the sense of full contact duration. The evidence is, however, masked by the interference between factors and by the differences between the analyzed vowels.
The next part of the EGG waveform corresponds to the opening of the glottis, i.e. the decreasing contact of the vocal folds. Due to the nature of the signal it is to be expected that this part is less informative. Particularly the open phase (when the folds are apart) has no representation in the EGG domain (section 9). However, the decline of the contact depends on laryngeal behavior.
In the experiment described further above, the duration of decreasing contact (the time which is needed for the amplitude to fall from 90% of the peak-to-peak value to 10%) is divided, similarly to the closing part, into two parts - start and end of opening. The opening instant, as defined in section 10.2 (Fig. 22), serves as a demarcation between those segments. A preliminary analysis showed that the second part of the opening phase is not influenced by stress and tenseness. It was thus eliminated from further investigations[8].
The duration of the start of opening (subsequently called opening duration) is analyzed similarly to the previously described parameters. Regarding the dependence on gender, it was found to be significant (F(1,5242)=29.9, p< 0.0005), however, the distinction was not as strong as for other parameters. The opening duration is about 1% (of the pitch period duration) longer for males than for females.
The data presented in Table 10 is slightly misleading as it exhibits significant differences between the tense/lax and stressed/unstressed tokens. After the standardization of the data and the statistical investigation using repeated measures ANOVA with vowel, stress and tenseness factors, it was found that significant effects are caused by vowels in both groups and by stress for women. The stress factor interacts with the vowel factor. As is shown in App.I no definitive direction of the changes is observed.
The maximal no-contact phase duration defines the Open Quotient II and has already been described in the preceding section.
In summary, the durations of particular EGG phases depend primarily on gender. The recordings of male speakers (compared to female speakers) are characterized by:
* longer start of closing
* shorter end of closing
* longer full contact phase
* longer start of opening
* shorter end of opening
* shorter open phase
It is important to note that "weaker" (i.e. shorter) contact phases are found for females. This explains the higher skewness of EGG pulses of male speakers (lower SQ).
Durations of EGG phases strongly depend on the pronounced vowel. A general tendency towards longer closing and shorter opening is observed for more closed and more back vowels.
Word stress affects closing and opening duration. Its effects are dependent on the vowel type. Tenseness contributes more regularly: tense vowels have a longer closing phase than lax ones.
8. the second part of the opening does not show dependence on any factor besides sex (F(1,178)=29.651, p < 0.0005).The end of the opening phase is longer for females than males (17% compared to 11.5%)
