Journal of Voice Vok 3, No. 4, pp. 294-305 © 1989 Raven Press, Ltd., New York Glottal Airflow and Transglottal Air Pressure Measurements for Male and Female Speakers in Low, Normal, and High Pitch * t ~ : E v a B. H o l m b e r g , * t R o b e r t E. H i l l m a n , a n d t J o s e p h S. P e r k e l l *Department of Communication Disorders, Boston University, Boston and ?Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, Massachusetts, U.S.A.; ¢Department of Phonetics, Institute of Linguistics, University of Stockholm, Stockholm, Sweden Summary: Measurements on the inverse filtered airflow waveform and of estimated average transglottal pressure and glottal airflow were made from syllable sequences in low, normal, and high pitch for 25 male and 20 female speakers. Correlation analyses indicated that several of the airflow measurements were more directly related to voice intensity than to fundamental frequency (Fo). Results suggested that pressure may have different influences in low and high pitch in this speech task. It is suggested that unexpected results of increased pressure in low pitch were related to maintaining voice quality, that is, avoiding vocal fry. In high pitch, the increased pressure may serve to maintain vocal fold vibration. The findings suggested different underlying laryngeal mechanisms and vocal adjustments for increasing and decreasing Fo from normal pitch. Key Words: Inverse filter--Transglottal air pressure-Glottal airflow--Pitch--Male and female speakers. male and female voices. The results are to be used to understand more about normal vocal function and, also, as norms for the evaluation of pathological voices (2). The main purpose of this study was to establish quantitative group data of a number of measures of vocal function for normal male and female speakers in normal, low, and high pitch. This is the second part of a larger study, whose first part dealt with intensity variation (1). In addition to measurements of fundamental frequency (Fo) and intensity, we used a noninvasive technique to derive measurements of transglottal air pressure and glottal airflow from oral measurements of pressure and flow. Statistical analysis were performed to determine: (a) relationships among measured parameters, (b) differences in measured parameters when Fo was changed between conditions of normal and low and normal and high pitch, and (c) differences between BACKGROUND Control of F o has been found to be a result of both laryngeal adjustments and subglottal pressure (314). In high pitch, the longitudinal tension of the vocal folds is increased (15), primarily by contraction of the cricothyroid muscle (8,11,12,16). Vocalfold thickness (17) and glottal area (18) decrease for increased F0. The stiffness of the vocal-fold body is increased by contraction of the vocalis muscle (16). Raising of the entire larynx may also contribute to the longitudinal tension, in addition to increasing the tension of the vocal-fold surface in the vertical direction (16). In low pitch, the vocal folds are relatively short Address correspondence and reprint requests to E. B. Holmberg at Research Laboratory of Electronics, Massachusetts Institute of Technology, Bldg. 36, Rm. 521, Cambridge, MA 02139, U.S.A. Presented in part at The 17th Symposium: Care of the Professional Voice, New York, June, 1988. 294 GLOTTAL A I R F L O W A N D T R A N S G L O T T A L A IR P R E S S U R E VS. P ITCH and slack (12,16) and the vocal-fold mass per unit of length is increased (17). Lowering F 0 involves relaxation of the cricothyroid muscle (12). Activity in the extrinsic sternohyoid muscle has been observed to lower the larynx, thereby helping to slacken the vocal folds (J. J. Ohala cited in ref. 12). It has been suggested that when the vocal folds are relatively thick and slack, Bernoulli forces and subglottal pressure have greater influence on the vibration pattern (J. W. van den Berg cited in ref. 12). F o has been found to be more sensitive to changes in subglottal pressure in the low F o region than at higher frequencies (6,12). Due to the inaccessible location of the larynx and vocal folds, most techniques used to study vocal function--such as tracheal puncture for subglottal pressure measurements, electromyography, laminagraphy and high-speed motion pictures for studies of vocal fold movements--are more or less invasive; therefore, most studies have reported data on a relatively small number of subjects. Since there is wide variation of measures of glottal function across individuals (1,19), it should be useful to study large numbers of subjects in order to capture normal variability. The intent of the present study was to examine a relatively large number of speakers using a noninvasive technique that allows us to compare glottal airflow waveform measures with an indirect measure of transglottal air pressure and acoustic measures, thus providing new insight into functional differences across pitch conditions and between male and female voices. METHODS The procedures were designed to make possible noninvasive recordings of laryngeal function for large groups of subjects. A detailed description of the recording, calibration, processing, and analyses procedures, including underlying assumptions and rationale for the measurements, can be found in Holmberg et al. (1). A brief summary of the procedures is provided below. Subjects, speech tasks, and recording Forty-five adult native speakers of American English, 25 men and 20 women, with no history of speech, voice, or hearing problems served as subjects. The average age for the men was 22 years and 5 months with a range of 17-30 years. The women, who ranged from 18 to 36 years, had an average age of 24 years and 3 months. The subjects were nonsmokers with no professional speaking or singing training. 295 The speech material was chosen to allow for estimates of glottal function from oral measurements (20-22) and consisted of strings of five repetitions of the syllable/p~e/. The subjects produced the strings in three pitch conditions, namely normal pitch, lower than normal pitch, and higher than normal pitch. In the interest of keeping the productions as similar as possible to the subjects' natural speech, the subjects were not asked to attain prescribed levels of pitch (or loudness), but instead were free to use "comfortable" levels for each condition. The only instruction given to the subjects v i s a vis pitch was to not increase or decrease F o so much that there was a possibility of using falsetto or fry registers. The subjects produced the syllable strings five times per speech condition at a rate of -1.5 syUables/s. The recordings were made in a sound-isolated booth. Intraoral air pressure for the occlusion of the stop consonant/p/was transduced using a translabial catheter connected to a differential pressure transducer (Glottal Enterprises). Oral volume velocity (airflow) for the vowel/ae/was transduced by means of a so called "Rothenberg mask" (a hightime resolution pneumotachograph, attached to a circumferentially vented facemask; Glottal Enterprises) (20). The acoustic signal was recorded with an external microphone (Model ECM 50; Sony) at a fixed distance of 15 cm from the subjects' lips. The signals were recorded on FM tape along with appropriate calibration signals, and monitored on an oscilloscope. Data analyses The pressure, flow, and intensity signals were low-pass filtered at different frequencies, digitized simultaneously at two different sampling rates, and processed further in software to produce "slowly varying" and "glottal airflow" signals. From these signals, data were extracted interactively, and measures were calculated (from the extracted data). Figure 1 shows an example of the slowly varying signals and measurement procedure. Average transglottal air pressure (cm H20) for the vowel was estimated by interpolation between peak intraoral air pressures for the stop consonants in the syllable string (1,20-22). Average airflow (L/s) through the glottis, estimated from the smoothed oral airflow, and intensity were measured for the vowel. Pressure, flow, and intensity values were also used to define two general measures of glottal function: "vocal efficiency," the ratio of intensity to the Journal of Voice, Vol. 3, No. 4, 1989 296 E. B. H O L M B E R G E T AL. MV MV MY I t t IIo ~-mnlll _I AIRFLOW < 0 ----~ Ltl .9. > PRESSURE T r T o - ~ ' ' ' L b ae ; I p ae:- ; ' p _ cl . 'ae :}P ae :p ae u I ~ec FIG. 1. Example of slowly varying signals for one sequence of five syllables and the mid-vowel (MV) location for data extraction. The traces from top to bottom correspond to intensity, average airflow, and intraoral air pressure versus time, product of air pressure times average flow (I/cm H20 x L/s) (19,23) and "glottal resistance," to airflow, the ratio of air pressure to average flow (cm H20/L/s) (22). These measurements were made at a vowel midpoint for the middle three syllables in each string (making 15 tokens per pitch condition) and averaged. The Fo (Hz) was measured (PM Voice Analyzer; Voice Identification, Inc.) for each of the 15 tokens per pitch condition and averaged to obtain a F0 for each pitch condition. The one token out of the 15 for each pitch condition that was closest to the mean for that condition in terms of F0 was located and used for inverse filtering of the first formant (Fl). (The effects of formants above F~ had been cancelled previously by low-pass filtering in hardware.) The glottal airflow measures were made and averaged from the four middle cycles of the inverse filtered token. Figure 2 shows a stylized cycle and the glottal airflow measurements. Time-based measures included: the fundamental period (7) (msec); open quotient, the ratio of open time to the period [(h + t2)/T)]; speed quotient, the ratio of opening to closing times (t~/t2); and closing quotient, the ratio of closing time to the period (t2/T) (cf. ref. 27). Amplitude-based parameters (measured in L/s) were peak flow, measured from the zero flow baseline to maximum flow; minimum flow (or "DC flow offset"), measured from the zero flow baseline to the minimum during the closed phase (cf. refs. 23 and 28); AC flow, calculated as peak flow minus minimum Journal of Voice, Vol. 3, No. 4, 1989 FIG. 2. Airflow versus time for one glottal cycle. The glottal airflow parameters are: the fundamental period (T), open quotient [(t] + t2)/T], speed quotient (t]/t2), closing quotient (tz/T) , peak flow (l/s), minimum flow (L/s), AC flow (L/s), AC/DC (RMS of AC flow around its mean/mean of AC). [In addition, maximum airflow declination rate (L/s 2) was extracted as the maximum amplitude of the negative peak from the first derivative.] flow (cf. ref. 29); and the AC/DC ratio, the root mean square (RMS) of the AC portion around its mean divided by the mean of the AC portion (30). In addition, the maximum flow declination rate (L/s 2) was extracted as the maximum amplitude of the negative peak of the first derivative of the airflow signal. Statistical analyses Calculated summary statistics include means, SDs, and ranges for the parameters in normal-, low-, and high-pitch condition for men and women separately. In analyses within low and high pitch and for change between normal- to low- and normal- to high-pitch conditions, F o is given as a percentage of the Fo value in normal pitch. The following statistical analyses were performed. (a) Pearson product moment correlations were calculated to examine pair-wise linear relationships between all parameter pairs. This was done within each of the pitch conditions and for change (difference values) between normal to low and normal to high pitch. In correlation analyses of parameters with SPL, logt0 of each parameter was used. Two GLOTTAL A I R F L O W A N D T R A N S G L O T T A L A IR P R E S S U R E VS. P ITCH variables were considered to be highly correlated if r was i>0.70, which means that at least 49% of the variance is accounted for (i.e., r 2 = 0.49). (b) Multivariate (Hotelling's T2) and univariate t tests for paired comparisons were performed to determine which parameters were significantly altered when pitch was manipulated (normal- to low-, and normal- to high-pitch condition). Initial results from the correlation analyses had shown that no parameters were strongly related to F 0, whereas there were high correlations between SPL and some of the parameters. Therefore, in addition to the above t tests, analyses of covariance with SPL as the covariate were performed to examine parameter differences across the pitch conditions. For men and women separately, values were " a d j u s t e d " for the SPL difference that occurred across the pitch conditions (28,30). Thus, the effect of intersubject S P L variation was reduced. To offset the problem with artificially increased significant differences that can be associated with such repeated univariate analyses, the tests for significant differences across pitch conditions were performed at the conservative level of p = 0.025. (In the following text, differences with p values between 0.050 and 0.025 are referred to as "nearly significant.") (c) Analyses of covariance (covariate SPL) were performed within each pitch condition to test for significant differences between men and women. low 0;5 1 o ~ pitch /~ 297 Values were " a d j u s t e d " for the cross-subject SPL variation within each pitch condition (28,30). This analysis is a type o f post hoc correction for the fact that in all pitch conditions, men were 2-3 dB louder than women. The tests for significant differences between men and women were performed at the conservative level of p = 0.025. RESULTS General observations on the glottal airflow waveforms Figures 3 and 4 show glottal airflow waveforms for low, normal, and high pitch for six male and six female speakers, respectively (A-F). The waveforms were chosen to show typical men's and women's waveforms as well as to illustrate qualitatively the large intersubject variation that is reflected in the quantitative data (cf. refs. 1 and 19). As seen in both figures, the inverse filtering result was not always completely successful, and residuals of the F 1 are superimposed on some of the waveforms. The residuals were seen mostly during the opening and closed phases of the waveforms, and should not affect our measurements to any great extent (cf. ref. 1). As seen in Figs. 3 and 4, in many cases for both men and women, the flow waveform did not decrease c o m p l e t e l y to 0; there was a " D C flow ncrmal pitch high pitch ~ j ~ x . ~ J ~ tll c 0 .=J 11 a: FIG. 3. Glottal airflow versus time waveforms in low, normal, and high pitch for six male speakers. .C ,.J A_.A_A_./ E .A_/k_/ 0 [~ .02 )~ TiME ($EC) Journal of Voice, Vol. 3, No. 4, 1989 298 E. B. H O L M B E R G ET A L . low pitch normal pitch high pitch 0.5- B A IXI O ,,d t, FIG. 4. Glottal airflow versus time waveforms in low, normal, and high pitch for six female speakers. ¢ Q -d t~ o.s ]_~ jXJ%/~AJX.A 0 L.r .02 , ,;~ TIME (SEC) offset" during the portion of the waveform when the glottis is assumed to be closed. This DC flow offset was often larger than the 0.02-0.03 L/s that can be due to vertical movements of the folds (K. N. Stevens, 1987, unpublished data). Such a DC flow offset has been found and discussed previously (cf. refs. 1,16,20,30,31). If there is a DC flow offset in combination with a flat and welldefined closed portion (cf. Fig. 3, subject E, low and normal pitch; Fig. 4, subject D, low pitch), the membranous part of the glottis is likely to be closed during the flat (closed) portion of the glottal waveform. The most plausible explanation for the DC flow is that air is shunted through a posterior glottal opening--a chinkmbetween the arytenoid cartilages (cf. ref. 32). In waveforms having a DC flow offset in combination with a more rounded closed portion (cf. Fig. 3, subject B in high pitch; Fig. 4, subject A in high pitch), the closure and opening are likely to be gradual with the glottal opening extending into the membranous part of the glottis. However, there are also waveforms with small or no DC flow offset (cf. Fig. 3, subject D in normal and low pitch; Fig. 4, subjects B in low pitch and E in normal pitch). The DC flow offset is represented by the measurement "minimum flow." In the following, we will refer to the portion of the glottal waveform having the lowest flow as the "closed portion," even in cases of a DC flow. Journal of Voice, Vol. 3, No. 4, 1989 The " f l a t n e s s " of the closed portion varied across subjects and pitch conditions. In all pitch conditions, some waveforms for both men and women showed well-defined closed portions (cf~ Fig. 3, subjects A and E in low and normal pitch, subject C in normal pitch, and subject F in low and high pitch; Fig. 4, subject C in low and normal pitch, subject D in low pitch, and subject E in normal and high pitch). However, there were also waveforms with more gradual beginnings and ends of the closed portions (cf. Fig. 3, subject A in high pitch, subject B in low and high pitch; and less clearly in Fig. 4, most subjects in high pitch). Statistical results In the Appendix, Tables A1-A6 show the summary statistics (unadjusted values) for normal, high, and low pitch for male and female speakers. For men, mean F 0 for normal pitch was 116 Hz (SD = 12.0 Hz), with a range of 93-135 Hz. For the high-pitch condition, the men increased Fo an average of 33% (of normal F0), with a range of 5-74% for the individual speakers. For the low-pitch condition, the men decreased F0 an average of 13%, with a range of 4-26%. For women, mean Fo for normal pitch was 205 Hz (SD = 24.0 Hz), with a range of 162-238 Hz for the individual speakers. For the high-pitch condition, the women increased F0 an average of 27% (of nor- GLOTTAL A I R F L O W A N D TRANSGLOTTAL AIR P R E S S U R E VS. PITCH mal F0), with a range of 9-43%. For the low-pitch condition, the women decreased F0 an average of 11%, with a range of 4-21%. Differences across pitch conditions As described above, analysis of covariance (covariate SPL) was used to test for significant differences among pitch conditions. Tables 1 and 2 summarize mean values in low, normal, and high pitch for men and women speakers, respectively, after adjustment for SPL variation a c r o s s the p i t c h c o n d i t i o n s , i . e . , the "adjusted" mean values that resulted from the analysis of covariance. Significant (as well as nearly significant) differences across pitch conditions are indicated with p values. Differences between normal and high pitch For both men and women, SPL increased significantly (p < 0.001) between the normal- and highpitch conditions. Men displayed significant differences between normal and high pitch in the unadjusted mean values for all parameters but glottal resistance, speed quotient, and AC/DC ratio. For women, the unadjusted values showed significant 299 differences for all parameters except glottal resistance, AC flow, minimum flow, speed quotient, and AC/DC ratio. However, as seen in Tables 2 and 3, most of these changes were no longer significant after including SPL as a covariate. Therefore, these changes were considered to be more strongly related to SPL than to F 0. Aside from the obvious significant increase in Fo for the higher pitched voices of both men and women, the adjusted mean values for men showed significant differences between normal and high pitch for the following parameters: increased average flow (p = 0.007), increased open quotient (p < 0.001), and decreased AC/DC (p = 0.007). Nearly significant differences were shown for the following parameters: decreased glottal resistance (p = 0.037), increased minimum flow (p = 0.030), and decreased speed quotient (p = 0.027). For women, pressure increased significantly (p = 0.025) and there was a nearly significant decrease in AC/DC (p = 0.039). Differences between normal and low pitch There was no significant difference in SPL between normal and low pitch for men or women. TABLE 1. Adjusted mean values and significant differences between normal and low, and normal and high pitch in men a Normal Low PER (ms) MFDR (L/s 2) ACFL (L/s) PKFL (L/s) MNFL (L/s) OPQU [(t I + t2)/T] SPQU (tl/t2) CLQU (tz/T) AC/DC F o (Hz) PRES (cm H20 ) AVFL (L/s) GLRE (cm H20/L/s ) VOEF (I/cm H20 x L/s)b 9.8 246 0.24 0.35 0.10 0.58 1.86 0.20 0.60 101 7,1 0.17 44 11 p < 0.001 p < 0.001 p < 0.001 p = 0.001 Low High 8.7 255 0.25 0.36 0.10 0.60 1.80 0.22 0.56 115 6.0 0.18 34 11 8.7 309 0.27 0.38 0.09 0.59 1.84 0.21 0.62 119 6.6 0.18 38 16 High 6.6 309 0.26 0.43 0.14 0.68 1.55 0.25 0.45 151 6.8 0.23 32 15 p < 0.001 (p = 0.030) p < 0.00t (p = 0.027) p = 0.007 p < 0.001 p = 0.007 (p = 0.037) " Significant (p < 0.025) and close to significant (p < 0.050) differences between normal and low and normal and high pitch for adjusted values (covariate SPL) of the following parameters: period (PER), maximum flow declination rate (MFDR), AC flow (ACFL), peak flow (PKFL), minimum flow (MNFL), open quotient (OPQU), speed quotient (SPQU), closing quotient (CLQU), AC/DC, Fo, pressure (PRES), average flow (AVFL), glottal resistance (GLRE), and vocal efficiency (VOEF). Values for normal pitch can differ for analyses of change between normal and low, and normal and high pitch because the number of subjects in the two comparisons are different. This is because three subjects did not lower Fo for the low-pitch condition. Therefore, two values for normal pitch are listed in the tables: one value from analysis between normal versus low pitch, the other from normal versus high pitch. Nearly significant p values are shown in parentheses. b Values for VOEF have been multiplied by 105. Journal of Voice, Vol. 3, No. 4, 1989 300 E. B. H O L M B E R G E T A L . TABLE 2. Adjusted mean values and significant differences between normal and low, and normal and high pitch in w o m e n a Normal Low PER (ms) MFDR (L/s2) ACFL (L/s) PKFL (L/s) MNFL (L/s) OPQU [(tl + t2)/T] SPQU (tl/t2) CLQU (tz/T) AC/DC F 0 (Hz) PRES (cm H20) AVFL (L/s) GLRE (cm HzO/L/s) VOEF (I/cm H20 × L/s) b 5.5 170 0.14 0.24 0.08 0.68 1.57 0.26 0.41 182 6.0 0.13 46 9 p < 0.001 p = 0.024 p < 0.001 (p = 0.027) Low High 4.9 160 0.13 0.22 0.08 0.74 1.62 0.28 0.37 203 5.8 0.14 41 9 4.8 175 0.14 0.23 0.08 0.74 1.62 0.28 0.39 207 6.0 0.14 42 11 High 3.9 168 0.13 0.23 0.09 0.87 1.48 0.35 0.32 252 6.3 0.16 41 11 p < 0.001 (p = 0.039) p < 0.001 p = 0.025 a See Table 1, footnote a for details regarding data and definitions of abbreviations. b Values for VOEF have been multiplied by 105. Therefore, there were only small differences between the unadjusted and adjusted mean values, and few parameters changed significantly. Aside from the obvious significant decrease in F0, the adjusted mean values for men showed significant increases in pressure (p < 0.001) and glottal resistance (p = 0.001), whereas for females open quotient decreased significantly (p = 0.024). The pressure increase for women was nearly significant (p = 0.027). TABLE 3. Adjusted mean values f o r men and women and significant differences in normal pitch between men and women a Differences between men and women Tables 3-5 compare men's and women's adjusted mean values within the normal-, high-, and lowpitch conditions. Significant (as well as nearly significant) differences between men and women are indicated with p values. As expected, men had significantly lower Fo in all pitch conditions (p < 0.001). In the normal pitch condition (Table 3), in comparison with women, men had significantly higher maximum flow decliT A B L E 4. Adjusted mean values Jbr men and women and significant differences in high pitch between men and women a Adjusted mean F o (Hz) PRES (cm H20 ) AVFL (L/s) GLRE (cm H20/L/s) VOEF (I/cm HzO x L/s) b MFDR (L/s2) ACFL (L/s) PKFL (L/s) MNFL (L/s) OPQU [(q + t2)/T] SPQU (tl/t2) CLQU (tJT) AC/DC Adjusted mean Men Women p 119 5.9 0.19 33 9 239 0.23 0.35 0.10 0.61 1.79 0.22 0.52 207 6.1 0.14 44 12 117 0.14 0.23 0.08 0.74 1.67 0.28 0.40 <0.001 -0.005 0.006 (0.033) <0.001 <0.001 <0.001 -<0.001 -<0.001 0.009 Adjusted mean values for men and women in normal pitch. Significant differences are indicated with p values. Nearly significant p values are shown in parentheses. b Values for VOEF have been multiplied by 105. See Table 1, footnote a for definitions of abbreviations. Journal of Voice, Vol. 3, No. 4, 1989 F o (Hz) PRES (cm H2O) AVFL (L/s) GLRE (cm H20/L/s) VOEF (l/cm H20 × L/s) b MFDR (L/s z) ACFL (L/s) PKFL (L/s) MNFL (L/s) OPQU [(t~ + tz)/T] SPQU (tl/t2) CLQU (tz/T) AC/DC Men Women p 151 6.9 0.23 31 15 304 0.26 0.42 0.14 0.66 1.56 0.24 0.46 263 7.1 0.16 45 20 220 0.16 0.26 0.09 0.85 1.50 1.34 0.37 <0.001 -0.002 0.002 0.015 <0.001 <0.001 <0.001 (0.026) <0.001 -<0.001 (0.051) a Adjusted mean values for men and women in high pitch. Significant differences are indicated with p values. Nearly significant p values are shown in parentheses. b Values for VOEF have been multiplied by 105. See Table 1, footnote a for definitions of abbreviations. GLOTTAL A I R F L O W A N D T R A N S G L O T T A L A IR P R E S S U R E VS. P ITCH T A B L E 5. A d j u s t e d m e a n values for men and w o m e n and significant differences in low pitch between m e n and w o m e n a Adjusted m e a n Fo (Hz) P R E S (cm H20 ) A V F L (L/s) G L R E (cm H20/L/s) V O E F (I/cm H 2 0 x L/s) b M F D R (L/s 2) A C F L (L/s) P K F L (L/s) M N F L (L/s) O P Q U [(t~ + t2)/T] S P Q U (q/t 2) C L Q U (tJT) AC/DC Men Women p 100 6.6 0.16 42 8 214 0.23 0.33 0.10 0.59 1.79 0.21 0.56 200 6.2 0.14 46 11 182 0.15 0.24 0.08 0.67 1.59 0.26 0.42 <0.001 ----(0.028) <0.001 0.002 -0.011 -0.005 0.006 a Adjusted m e a n values for m e n and w o m e n in low pitch. Significant differences are indicated with p values. Nearly significant p values are s h o w n in parentheses. b Values for V O E F h a v e been multiplied by 105. See Table 1, footnote a for definitions of abbreviations. nation rate (p < 0.001), higher peak flow (p < 0.001), higher AC flow (p < 0.001), higher average flow (p = 0.005), smaller open quotient (p < 0.001), smaller closing quotient (p < 0.001), higher AC/DC ratio (p = 0.009), and lower glottal resistance (p = 0.006). A nearly significant difference was found for vocal efficiency (p = 0.033), which was lower for the men. No significant differences between men and women were found for pressure, minimum flow, or speed quotient. In the high-pitch condition (Table 4), in addition to the significant differences between men and women found for normal pitch, men had significantly lower vocal efficiency (p = 0.015). Minimum flow (p = 0.026) and AC/DC (p = 0.051) were higher for the men by a nearly significant difference. No significant differences between men and women were found for pressure or speed quotient. In the low pitch condition (Table 5), there were fewer differences between men and women in parameters than for the other conditions. Men had significantly higher AC flow (p < 0.001), higher peak flow (p = 0.002), smaller open quotient (p = 0.011), smaller closing quotient (p = 0.005), and higher AC/DC (p = 0.006). The higher maximum flow declination rate for men was close to significant (p = 0.028). No significant differences between men and women were found for pressure, average flow, glottal resistance, vocal efficiency, minimum flow, or speed quotient. 301 DISCUSSION Large intersubject variation was found in all pitch conditions. This finding agrees with reports in previous studies of air pressure and airflow (1,19). Some of the variation in our pitch data could be due to the possibility that some subjects may have used falsetto register in the high-pitch condition in spite of instructions designed to elicit the chest register. To avoid this problem in future studies, it may be preferable to control pitch to a greater extent when the recordings are made (33). No parameters correlated strongly with Fo, whereas there were high correlations between some parameters and SPL. Some of the variation in parameters across subjects and conditions was therefore likely to have been due to variation in SPL, which was the primary reason that SPL was treated as a covariate in testing for significant differences. It should be noted that the results of such analyses cannot be interpreted as directly reflecting the physical significance of the measures that we obtained. One somewhat more direct way of studying which parameters are related to pitch change would be to ask the subjects to change pitch while keeping intensity constant. However, since a change in number of air pulses per unit of time in itself would contribute to a change in SPL, the subjects might have to manipulate their phonation in some way to keep SPL constant, and the result would not reflect normal vocal function. Therefore, in this study, we considered the statistical post hoc adjustments for SPL variation to be potentially less confounding than attempting to directly control F o and SPL. Since SPL is inseparable from F 0 in natural speech, the only way to study pitch and loudness separately may be with modeling. Within pitch conditions, air pressure was somewhat higher for men than for women, but when the influence of the higher male SPL was factored out, there was no significant difference between men and women in pressure. Therefore, glottal waveform differences between men and women (post SPL adjustment) were not primarily due to differences in respiratory force. With increasing F0 from the normal- to high-pitch condition, SPL increased significantly for both men and women, and a majority of the speakers increased pressure. However, for men after adjusting for the SPL increase, the pressure increase was not significant, suggesting that the pressure increase Journal of Voice, Vol. 3, No. 4, 1989 302 E. B. H O L M B E R G ET AL. was more related to the increased SPL than to F 0. In agreement with previous work (cf. ref. 12), these results suggest that for men increased F 0 between the normal- and high-pitch conditions was achieved primarily by laryngeal adjustments rather than by increased respiratory forces. In contrast, for women the pressure increase between normal and high pitch was significant after adjusting for SPL. The women's results suggest, at least statistically, that pressure was related not only to the SPL increase but also to the increased F 0. Further research is needed to explore the physical significance of these differences b e t w e e n men and women. Previous research has generally shown a positive relationship between subglottal pressure and F0 during sustained vowel phonation (4,10), and during sudden pushes on subjects' chest (13). Subglottal pressure has also been found to drop with Fo at sentence endings (12). Therefore, the present findings of increased pressure (in adjusted as well as unadjusted values) for low pitch were unexpected. We speculate that the discrepancy between the previous results and our data may be due to the use of different speech tasks. It is conceivable that Fo during sustained phonation of vowels may covary with subglottal pressure in a more systematic way, whereas syllable repetitions in low- and high-pitch conditions may demand different vocal adjustments and represent different speech modes (cf. refs. 6 and 12). Our findings of increased pressure in low pitch are somewhat supported by results reported by Kunze (4). He found for 10 male speakers that subglottal pressure increased with increased F 0 with the exception of the lowest F0, where pressure was somewhat higher than for the second lowest F0 level. Increased pressure in high pitch was accompanied by increased AC flow (cf. refs. 1 and 26), whereas in low pitch the increased pressure instead was accompanied by decreased AC flow. We speculate that for the speech task used in this study (i.e., syllable repetition) there may be different mechanisms for increasing and decreasing pitch, and that pressure may have different functions in the high and low pitch conditions. In high pitch, when the vocal folds are tense and stiff, increased pressure may be necessary to sustain vocal fold vibration. In low pitch, when the vocal folds are relatively slack, increased pressure may serve as a stabilizing factor to maintain voice quality, i.e., to prevent vocal fry. Journal of Voice, Vol. 3, No. 4, 1989 Further work is needed to explore the mechanism underlying these findings for pressure. Holmberg et al. (1) found that many of the glottal airflow parameters measured in this study changed significantly with increased or decreased intensity across conditions of soft, normal, and loud voice. As described above, in this study several of these parameters (maximum flow declination rate, AC flow, open quotient, speed quotient, and closing quotient) changed significantly across the pitch conditions, but, after covariation with SPL, these changes were no longer significant. These results, in combination with low correlations between F 0 and all of the parameters, and high correlations between SPL and some of the parameters (pressure, maximum flow declination rate, AC flow, and vocal efficiency), suggest that the measures used in this study are not primarily indicative of parameters determining Fo, such as tension and stiffness of the vocal folds. Instead they may be more directly related to characteristics of the sound source that determine intensity (cf. ref. 34)[ CONCLUSIONS The results of this study indicate that several of the glottal airflow measurements that were obtained are more directly related to voice source characteristics that determine intensity than to factors that determine F 0, such as vocal fold tension and stiffness. Some results suggest that for the speech utterance used (syllable repetition), the conditions of low, normal, and high pitch may represent different speech modes rather than reflecting conditions in which there is a continuous change of Fo. Results further suggest that the subglottal pressure may have a different function in the production of this utterance at low versus high pitch. Even though pitch variation may be less important than intensity variation as a way of eliciting useful measures of glottal function with these techniques, this data base should be of importance for research of normal voice function, as well as for clinical research of voice diagnostic and therapy techniques that require patients to manipulate F 0. Acknowledgment: This research was supported by the National Institutes of Health, grant number RO1 NS21183. We wish to thank Kenneth Stevens for many helpful comments. APPENDIX TABLE A1. Summary statistics: normal pitch in men TABLE A3. Summary statistics: low pitch in men (n = 25) (n = 25) Slowly varying parameters" SPL (dB) 79 F o (Hz) 116 PRES (cm H20 ) 6.3 A V F L (L/s) 0.19 G L R E (cm HzO/L/s ) 38 V O E F (I/cm H20 × L/s) b 16 Glottal waveform parameters ¢ SPL (dB) 79 PER (ms) 8.6 M F D R (L/s 2) 280 P K F L (L/s) 0.38 A C F L (L/s) 0.26 M N F L (L/s) 0.12 0.60 OPQU [(q + tz)/T] SPQU (ta/t 2) 1.82 0.22 C L Q U (t2/T) AC/DC 0.60 PRES (cm H20) 6.4 A V F L (L/s) 0.20 G L R E (cm HzO/L/s ) 37 V O E F (I/cm H20 × L/s) b 16 SD Range 3 12 1.4 0.07 17 13 73-86 93-135 4.2-9.6 0.10-0.35 12-77 2-62 3 1.0 90 0.09 0.07 0.07 0.07 0.28 0.04 0.23 1.4 0.07 18 13 73-87 7.3-10.9 140-470 0.21-0.54 0.16-0.42 0.02-0.30 0.46-0.77 1.32-2.58 0.14-0.27 0.23-1.12 4.3-9.7 0.0%0.36 14-85 2-62 a Averaged over 15 tokens. b Values for V O E F have been multiplied by 105. c Averaged over four cycles of the selected token. See Table 1, footnote a for definitions of abbreviations. TABLE A2. Summary statistics: high pitch in men (n = 25) Slowly varying parameters a SPL (dB) 83 Fo (Hz) 155 PRES (cm H20) 7.5 A V F L (L/s) 0.24 G L R E (cm HzO/L/s ) 36 VOEF (I/cm H20 × L/s) b 30 Glottal waveform parameters C SPL (dB) 83 PER (ms) 6.6 M F D R (L/s 2) 365 P K F L (L/s) 0.45 A C F L (L/s) 0.29 M N F L (L/s) 0.16 OPQU [(t 1 + t2)/T] 0.66 1.65 SPQU (tl/t 2) CLQU (t2/T) 0.24 AC/DC 0.53 PRES (cm H20 ) 7.4 A V F L (L/s) 0.25 GLRE (cm HzO/L/s) 35 VOEF (I/cm H20 x L/s) b 28 Slowly varying parameters a SPL (dB) F o (Hz) PRES (cm H20 ) A V F L (L/s) G L R E (cm HzO/L/s) V O E F (I/cm H20 x L/s) b Glottal waveform parameters" SPL (dB) PER (ms) M F D R (L/s 2) P K F L (L/s) A C F L (L/s) M N F L (L/s) OPQU [(t 1 + t2)/T] SPQU (tl/t 2) C L Q U (t2T) AC/DC PRES (cm H20) A V F L (L/s) G L R E (cm HzO/L/s ) V O E F (I/cm H20 x L/s) b Range 4 19 1.7 0.09 17 32 7%92 122-191 5.2-12.4 0.13-0.47 13-93 5-136 4 0.8 148 0.10 0.07 0.08 0.08 0.46 0.06 0.23 1.9 0.10 19 33 77-92 5.2-8.1 208-827 0.27-0.65 0.20-0.43 0.03-0.32 0.4%0.79 0.50-2.66 0.10-0.39 0.26-1.11 4.6-13.0 0.11-0.44 12-94 3-124 a Averaged over 15 tokens. b Values for V O E F have been multiplied by 105. c Averaged over four cycles of the selected token. See Table 1, footnote a for definitions of abbreviations. SD Range 79 101 7.0 0.18 46 12 3 10 1.6 0.08 20 7 74-85 82-124 4.%11.1 0.09-0.40 12-84 2-27 78 9.9 245 0.36 0.25 0.11 0.59 1.88 0.21 0.60 6.9 0.18 44 t0.3 3 1.0 90 0.10 0.07 0.06 0.10 0.50 0.06 0.15 1.7 0.07 17 6.7 73-84 8.0-12.1 124-430 0.20-0.58 0.15-0.39 0.04-0.32 0.45-0.83 0.96-2.88 0.12-0.31 0.24-0.86 4.6-11.0 0.0%0.40 12-71 1.7-27 " Averaged over 15 tokens. b Values for V O E F have been multiplied by 105. c Averaged over four cycles of the selected token. See Table 1, footnote a for definitions of abbreviations. TABLE A4. Summary statistics: normal pitch in women (n = SD X 20) Slowly varying parameters a SPL (dB) 76 F o (Hz) 205 PRES (cm H20 ) 5.8 A V F L (L/s) 0.14 GLRE (cm H20/L/s) 42 V O E F (I/cm H20 × L/s) b 11 Glottal waveform parameters C SPL (dB) 76 PER (ms) 4.9 M F D R (L/s 2) 164 P K F L (L/s) 0.22 A C F L (L/s) 0.14 M N F L (L/s) 0.09 OPQU [(t~ + tz)/T] 0.76 1.65 SPQU (tJt2) 0.29 C L Q U (t2/T) AC/DC 0.38 PRES (cm H20) 5.8 A V F L (L/s) 0.14 G L R E (cm HzO/L/s ) 43 V O E F (I/cm H20 × L/s) b 11 SD Range 4 24 0.9 0.03 8 9 67-81 162-238 4.4-7.6 0.09-0.21 28-93 1-38 4 0.6 58 0.06 0.05 0.04 0.10 0.30 0.04 0.13 1.1 0.03 9 9 67-81 4.1-6.2 91-279 0.12-0.37 0.0%0.25 0.01-0.16 0.56-0.95 1.1%2.33 0.20-0.37 0.25-0~80 3.6-8.1 0.09-0.22 30-62 2-36 Averaged over 15 tokens. b Values for V O E F have been multiplied by 105. c Averaged over four cycles of the selected token. See Table 1, footnote a for definitions of abbreviations. 304 E. B. HOLMBERG TABLE A5. S u m m a r y statistics: high pitch in w o m e n (n = 20) X Slowly varying parameters a SPL (dB) 80 Fo (Hz) 261 PRES (cm HzO) 6.7 AVFL (L/s) 0.17 GLRE (cm H20/L/s) 43 VOEF (I/cm H20 × L/s) b 18 Glottal waveform parameters c SPL (dB) 80 PER (ms) 3.8 MFDR (L/s 2) 201 P K F L (L/s) 0.25 ACFL (L/s) 0.15 MNFL (L/s) 0.10 0.87 OPQU [(q + t2)/T] SPQU (tl/t2) 1.52 0.35 CLQU (tE/T) AC/DC 0.36 PRES (cm H20 ) 6.6 AVFL (L/s) 0.16 GLRE (cm HEO/L/s) 44 VOEF (I/cm H20 x L/s) b 19 SD Range 4 30 0.8 0.04 11 12 72-85 204-309 4.6-8.1 0.09-0.24 30-75 3-51 4 0.5 68 0.07 0.05 0.04 0.12 0.33 0.04 0.10 1.0 0.04 ll 18 70-86 3.1-4.9 98-317 0.14-0.39 0.07-0.26 0.03-0.20 0.61-0.99 0.90-2.08 0.27-0.42 0.18-0.60 4.5-8.6 0.09-0.26 27-70 3-83 a Averaged over 15 tokens. b Values for VOEF have been multiplied by 105. c Averaged over four cycles of the selected token. See Table 1, footnote a for definitions of abbreviations. TABLE A6. S u m m a r y statistics: low pitch in women (n = 20) Slowly varying parameters a SPL (dB) F o (Hz) PRES (cm HzO) AVFL (L/s) GLRE (cm HzO/L/s ) VOEF (I/cm H20 × L/s) b Glottal waveform parameters c SPL (dB) PER (ms) MFDR (L/s 2) P K F L (L/s) ACFL (L/s) MNFL (L/s) OPQU [(tl + t2)/T] SPQU (tl/t z) CLQU (t2/T) AC/DC PRES (cm HzO) AVFL (L/s) GLRE (cm H20/L/s) VOEF (I/cm H20 x L/s) b SD Range 77 182 6.1 0.14 49 11 4 22 0.8 0.05 23 10 6%80 134-214 4.5-7.6 0.05-0.26 2%112 2-41 77 5.5 180 0.25 0.15 0.10 0.68 1.58 0.27 0.44 6.1 0.14 49 12 4 0.8 68 0.08 0.06 0.05 0.09 0.19 0.04 0.21 1.0 0.05 23 13 68-81 4.5-7.4 78-296 0.15-0.42 0.07-0.27 0.01-0.20 0.52-0.89 1.26-1.87 0.1%0.38 0.20-0.08 4.7-8.1 0.05-0.26 31-108 2-57 a Averaged over 15 tokens. b Values for VOEF have been multiplied by 105. c Averaged over four cycles of the selected token. See Table 1, footnote a for definitions of abbreviations. Journal of Voice, Vol. 3, No. 4, 1989 ET-AE. REFERENCES 1. Holmberg EB, Hillman RE, Perkell JS. Glottal airflow and transglottal pressure measurements for male and female speakers in soft, normal and loud voice. J Acoust Soc Am 1988;84:511-29. 2. Hillman RE, Holmberg EB, Perkell JS, Walsh M, Vaughan C. Objective assessment of vocal hyperfunction: an experimental framework and initial results. J Speech Hear Res (in press). 3. van den Berg JW. Sub-glottal pressure and vibration of the vocal folds. Folia Phoniatr (Basel) 1957;9:65-71. 4. Kunze L. Investigation of changes of subglottal pressure and rate of airflow accompanying change of Fo, intensity, vowels and voice registers in adult male speakers [Dissertation]. Iowa City, Iowa: State University of Iowa, 1962:1-119. 5. Ladefoged P. Some physiological parameters in speech. Lang Speech, 1963;6:109-19. 6. Stevens KN. Theory of vocal-cord vibration and its relation to laryngeal features. J Acoust Soc A m 1973;55:383(A). 7. Stevens KN. Vibration modes in relation to model parameters. In: Stevens KN, Hirano M, eds. Vocal fold physiology. Proceedings of the vocal fold physiology conference, Kurume, Japan. Tokyo: University of Tokyo Press, 1981:291301. 8. Hirano M, Ohala J, Vennard W. The function of laryngeal muscles in regulating fundamentalfrequency and intensity of phonation. J Speech Hear Res 1969;12:616-28. 9. Hixon T, Klatt DH, Mead J. Influence of forced transglottal pressure changes on vocal fundamental frequency. J Acoust Soc A m 1971;49:105(A). 10. Shipp T, McGlone R. Laryngeal dynamics associated with voice frequency change. J Speech Hear Res 1971;14:761-7. 11. Collier R. Physiological correlates of intonation patterns. J Acoust Soc Am 1975;58:249-55. 12. Atkinson JE. Correlation analysis of the physiological factors controlling fundamental voice frequency. J Acoust Soc Am 1978;63:211-22. 13. Baer T. Reflex activation of laryngeal muscles by sudden induced subglottal pressure changes. J Acoust Soc Am 1979;65:1271-5. 14. Titze IR, Durham PL. Passive mechanisms influencing fundamental frequency control. In: Baer T, Sasaki C, Harris K, eds. Laryngeal function in phonation and respiration. Proceedings from vocal fold physiology conference, New Haven, Connecticut, 1985. Boston: Little Brown & Co., Inc., 1987:304-19. 15. HoUien H. In search of vocal fold frequency mechanisms. In: Bless DM, Abbs JH, eds. Vocal fold physiology: contemporary research and clinical issues. Proceedings from vocal fold physiology conference, Madison, Wisconsin. San Diego, California: College-Hill Press, Inc., 1981:361-7. 16. Stevens KN. Physics of laryngeal behavior and larynx modes. Phonetica 1977;34:264-79. 17. HoUienH, Colton RH. Four laminagraphic studies of vocal fold thickness. Folia Phoniatr (Basel) 1969;21:179-98. 18. Hollien H, Coleman RF. Laryngeal correlates of frequency change: a STROL study. J Speech Hear Res 1970;13:271--8. 19. Schutte HK. The efficiency of voice production. Groningen: Kemper, 1981:1-192. 20. Rothenberg MR. A new inverse-filtering technique for deriving the glottal air flow waveform during voicing. J Acoust Soc Am 1973;53:1632-45. 21. Smitheran JR, Hixon TJ. A clinical method for estimating laryngeal airway resistance during vowel production. J Speech Hear Res 1981;46:138-46. 22. Lofqvist A, Carlborg B, Kitzing P. initial validation of an indirect measure of subglottal pressure during vowels. J Acoust Soc A m 1982;72:633-4. G L O T T A L A I R F L O W A N D T R A N S G L O T T A L A I R P R E S S U R E VS. P I T C H 23. Tanaka S, Gould WJ. Relationships between vocal intensity and noninvasively obtained aerodynamic parameters in normal subjects. J Acoust Soc Am 1983;73:1316-21. 24. Hirano M. Clinical examination of voice. New York: Springer-Verlag, 1981. 25. Rothenberg MR. Measurements of airflow in speech. J Speech Hear Res 1977;20:155-76. 26. Sundberg J, Gauffin J. Waveform and spectrum of the glottal voice source. In: Ohman S, Lindblom B, eds. Frontiers of speech communication. L o n d o n : A c a d e m i c P r e s s , 1979:301-22. 27. Isshiki N. Vocal efficiency index. In: Stevens KN, Hirano M, eds. Vocalfold physiology. Proceedings of the vocal foM physiology conference, Kurume, Japan. Tokyo: University of Tokyo Press, 1981 : 193-208. 28. Winer BJ. Statistical principles in experimental design. New York: McGraw-Hill, 1971:1-905. 305 29. Tabachnick BG, Fidell LS. Using multivariate statistics. San Francisco: Harper and Row, 1983:1-507. 30. Rothenberg MR. Source-tract acoustic interaction in breathy voice. In: Titze IR, Scherer RC, eds. Vocal fold physiology: biomechanics, acoustics and phonatory control. Proceedings of the international conference on physiology and biophysics of the voice, Iowa City, Iowa, 1983. Denver: Denver Center for the Performing Arts. 1985:465-81. 31. Karlsson I. Glottal waveforms for normal female speakers. Quart Prog Stat Rept 1985;1:31-6. 32. Koike Y, Hirano M. Glottal-area time function and subglottal pressure variation, J Acoust Soc Am 1973;54:1618-27. 33. Sodersten M, Lindestad PA. Vocal fold closure in young adult normal-speaking females. Phoneatric and Logopedic Progress Report: Huddinge University Hospital, Karolinska Institute, Stockholm 1987;5:12-29. 34. Fant G. Glottal source and excitation analysis. Quart Prog Stat Rept 1979;1:85-107. Journal of Voice, Vol. 3, No. 4, 1989