Acoustic Measure of Hormone Affect on Female Voice During Menstruation

Purpose: The purpose of this study was to use connected speech patterns in acoustic analysis of voice to determine if variations in vocal patterns can be detected during different phases of the menstrual cycle. It was hypothesized that natural hormonal changes during menstrual cycle would alter the acoustic characteristics of voice. Method: This descriptive research design acoustically analyzed 175 audio recorded female voice samples from 35 subjects. The points of measure included the following phases: menstrual, follicular, ovulatory, luteal and premenstrual. Eight parameters were statistically analyzed for significance: mean fundamental frequency, jitter, relative average perturbation, shimmer, peak-to-peak amplitude variation, noise to harmonic ratio, degree of voice breaks, and number of voice breaks. Results: A statistical analysis of data reliability between the 1 st and 2 nd menstrual periods showed no statistically significant differences for any of the eight acoustic parameters. No statistically significant differences were found for any of the acoustic parameters between subjects using estrogen/progesterone contraceptives and those that were not using them. Conclusion: None of the eight acoustic parameters examined demonstrated statistically significant differences in the four different hormonal phases of the menstrual cycle. Results suggest hormones are not implicated in voice change parameters measured during menstrual cycle. The purpose of this study was to use connected speech patterns in acoustic analysis of vice to determine if variations in vocal patterns can be detected during different phases of the menstrual cycle. Recently, a number of studies have suggested the female voice has an increased perceived attractiveness during specific periods of menstrual cycle. The implication is that during these periods hormones peak causing a change in the female voice. For example, Pipitone and Gallup (Pipitone et al., 2008) rated the perception of 38 female voices across four equally spaced periods during the menstrual cycle. A total of 152 voice samples were evaluated by 30 male and 30 female raters. The results indicated a perceived increase in voice attractiveness as subjects approached a greater probability of conception during the menstrual cycle. The perceived effects of voice change during menstrual cycle are well documented among classically trained singers. Ryan and Kenny (Ryan et al., 2007) identified six female singers who indicated their voices were greatly affected during menstrual cycle. Voice recordings from the six subjects taken at the first and mid cycle periods were randomly presented to both the subjects and their instructors for analysis. The subjects were able to identify when during their cycle the recordings were taken as well as reported greater effort required during the menstrual cycle to perform vocal tasks. The perceived differences in voice change during menstrual cycles are supported in studies of vocally trained groups as well. Brodnitz (Brodnitz, 1971) documented a case of sudden dysphonia of a professional female singer occurring one day prior to menstruation onset. Gynecological examination revealed abnormally low levels of corpus luteum hormone caused the recurring dysphonia prior to menstruation. Perceptual characteristics of vocal fatigue, loss of intensity, and variances in harmonics are indicative of premenstrual vocal syndrome (Abitbol, Abitbol, & Abitbol, 1999). In such cases hormone replacement therapy typically restores the voice to within normal parameters. Additional studies implicate the effects of hormones on voice. Baker (Baker, 1999) studied the cases of four women who sought otolarygological examinations following suspected iatrogenic dysphonia. Withdrawal from hormonal medications coupled with voice therapy eventually reduced some of the adverse vocal symptoms. If hormones can have a physiological effect on the female voice then it would stand to reason that oral hormone medication for contraception use would indicate such effects as well. However, research associated with oral contraception effect on voice presents inconclusive. While two studies present with significant changes in vocal perturbation as a result of oral contraception use (Amir, Biron-Shental, Tzenker, & Barer, 2005; Amir, Kishon-Rabin, & Muchnik, 2002) two other studies identify no changes in glottal airflow (Gorham-Rowan & Fowler, 2008) or voice onset timing (Morris, Gorham-Rowan, & Herring, 2009). © Centre for Promoting Ideas, USA www.ijhssnet.com 6 The limited variation on vocal performance by the use of oral contraceptives would indicate hormonal impact may be pre-conditioned in some manner. To examine the pre-conditional effects of hormones on voice several researchers have examined impact during various stages of menstrual cycling. One group of researchers analyzed the fundamental frequency changes occurring during menstrual cycle without hormone replacement therapy and during menopause with two different hormone replacement agents. Differences in fundamental frequency for the three subject groups presented with no significance (Mendes-Laureano, Ferriani, Reis, Aguiar-Ricz, Valera, Kupper, et al., 2006). The study was limited however to the measurement of one vocal parameter and one point of measure during targeted periods. No significance was found when assessing across two points within the menstrual cycle among adolescent females (Meurer, Garcez, von Eye Corleta, & Capp, 2009). Acoustic measures of fundamental frequency, rate of speech, and intensity were taken during follicular and luteal phases of the cycle. These results are in spite of a previous study conducted by Whiteside (Whiteside, 2004) measuring voice onset times during the same phases of cycle examined by the Meurer study (Meurer et al., 2009). Results indicated a significant effect during menstrual cycle phase on the voice onset times of plosive production. However, this study fell short in a low number of subjects (seven) evaluated and in producing significant standard deviation between means. Replicating the Whiteside study by expanding the number of subjects (15 women and 20 men) and contrasting gender groups, Wadnerkar, Cowell, and Whiteside (Wadnerkar et al., 2006) evaluated the means of voice onset times in naturally occurring whole words from gender and low/high estrogen and progesterone samples. Results indicated significant relationship between cycle phase and voice onset timing of plosives. Comparative gender differences within the study implicated hormones as agents. The limited number of subjects continues to call into question the generalization of results. Using a similar research design, Raj, Gupta, Chowdhury, and Chadha (Raj et al., in press) increased the number of subjects (100), contrasted reproductive and postmenopausal age groups, measured five acoustic parameters (H/N ratio, s/z ratio, perturbation, fundamental frequency, & maximum phonation) and included five points of measure during the cycle (menstrual, follicular, ovulatory, luteal, and premenstrual). While the postmenopausal group presented with significance in relationship between hormonal balance and voice changes, the reproductive group presented with no significance in any of the vocal parameters measured. One potential limitation of the study however, was the limited acoustic input resulting from the stimulus of single phoneme production. Still, the inconsistent findings between studies suggest examination of pre-conditional variables is needed. If a similar research design were implemented using a connected speech pattern for acoustic analysis would the outcomes show significant differences? Digital recordings of connected speech samples are standard when analyzing voice characteristics. Commercially available acoustical analysis programs generally do not correlate with each other secondary to the various types of acoustic signals presented in the voice sample. For this reason researchers often assess samples limited to phoneme or syllable stimuli and not to connected speech. However, perceptual analysis of speech implies an interpretation of emotional variance (i.e. seduction) that is accounted for in connected speech patterns not in phoneme or syllable stimuli. Researchers have argued that acoustic measures take into account the contextual need for a broader connected speech sample (Alpert, Pouget, & Silva, 2001; Johnstone, van Reekum, Kirsner, & Scherer, 2005). Alpert, Pouget, and Silva (Alpert et al., 2001) recognized the need for acoustic analysis of “free speech” patterns in order to affirm clinical impressions of depression in the voice. They hypothesized that free speech (spontaneous connected speech) required cognitive activity associated with emotion patterns that would not be acoustically detected in nonfree speech patterns. Results indicated acoustic analysis of connected speech patterns were directly correlated to clinical impressions of emotional speech. Johnstone (Johnstone et al., 2005) corroborated this finding with a study to determine the degree in which emotionally charged speech could acoustically reflect factors other than arousal. Acoustic analysis of speech samples from 30 subjects measuring four vocal parameters correlated with the arousal perceptually identified during elicitation from computer game tasks. Speech recognition software uses this same theory to program computer voice differentiation of energetic states within voice. Acoustically analyzing connected speech patterns of 17 subjects, Krajewski, Wieland, Sommer and Golz (Krajewski et al., 2008) validated a software program’s ability to recognize psychological changes in cognitive speech planning. Using the same speech sample phrase from all of the subjects, the program was able to analyze as many as 8,500 high level speech features not found in non-connected speech samples. It is hypothesized that inconsistent results in voice analysis during different phases of the menstrual cycle may be the result of limited acoustic stimuli. Broadening the acoustic input from phoneme/syllable to connected speech would provide the computer greater variance from which to measure specific parameters. International Journal of Humanities and Social Science Vol. 1 No. 3; March 2011