Eduardo Castro-Sierra,1,2 (eduardcs@yahoo.com.mx),
Miguel Zenker,2 Adrián Poblano3 & Alberto Ramírez-Treviño4
1Hospital Infantil de México Federico Gómez.
Mexico City.
2National School of Music, National Autonomous University of Mexico,
Mexico City.
3Institute of Human Communication, National Center for Rehabilitation,
Mexico City.
4Center for Research and Advanced Studies, Mexico City
Popular version of paper 3pMUb6
Presented Wednesday afternoon, November 17, 2004
148th ASA Meeting, San Diego, CA
Formants of phonemes (vowels or consonants) in different languages are produced
when air from the lungs passes through the vocal cords and is filtered by constrictions
at various points in the vocal tract. The fundamental frequency (F0), or pitch,
of vocal emissions is the frequency of the cycle of opening and closing of the
cords, whereas the degree of opening and the position of the tongue in that
cavity will determine two peaks of resonance, or formant frequencies R1 and
R2, respectively, that will characterize each phoneme. Further anatomical strictures
from the vocal cavity to the larynx will determine higher peaks of resonance,
R3, R4, R5, etc., and provide a more detailed acoustical characterization of
phonemes.
R1 and R2 of every phoneme have distinct values in each person. Since the dimensions of the vocal cavity will not change in the process, these frequencies will not be substantially modified when the phoneme is spoken. However, the singer must adapt his/her vocal tract to varying conditions when singing at different pitches, and then the formants of phonemes may change. At National School of Music, of the National Autonomous University of Mexico in Mexico City, an investigation has begun to analyze the formant frequencies of the vowels produced by 11 sopranos with at least 5 years of training in classical operatic singing when speaking or singing at high, mid or low pitch speech sequences from diverse melodies in Spanish. The Spanish language has a five-vowel system, /a/, /e/, /i/, /o/ and /u/, similar to that of Italian.
This investigation has been undertaken employing Computerized Speech Research Environment 45 (CSRE 45Ò), which is a digital program for spectrographic voice analysis. Each of 11 sopranos was asked to pronounce under 4 different conditions the following sequence belonging to the first verse of a popular Mexican children's song:
1) "Naranja dulce" ("Sweet orange")
2) "limón partido" ("split lemon")
3) "deme un abrazo" ("give me a hug")
4) "que yo le pido..." ("I am asking you for..")
The first condition was while speaking (F0 avg. = 250 Hz), the second while singing at mid pitch (F0 avg. = 250 Hz), the third while singing at highest pitch without forcing the singer's voice (F0 avg. = 1050 Hz), and the fourth while singing at lowest pitch without forcing the singer's voice (F0 avg. = 190 Hz). Not forcing the voice of each of the participants will prevent the vocal productions of singers from going to higher or lower vocal registers and, thus, not allow interval modifications in their vocal productions making comparisons between different persons' voices impracticable.
In the first segment, "naranja dulce," there are two stressed (tonic) vowels, second /a/ of "naranja" and /u/ of "dulce"; in the second segment, "limón partido," there are two stressed vowels, /o/ of "limón" and /i/ of "partido"; in the third segment, "deme un abrazo," there are two stressed vowels, first /e/ of "deme" and, again, second /a/ of "abrazo". After obtaining the spectrograms of each segment, an analysis of each tonic vowel was undertaken.
The analyses of the voices of the sopranos participating in this investigation provided evidence that the values of R1 and R2 of the vowels while either speaking, singing at lowest pitch or, in some cases, singing at mid pitch varied with each singer's vocal production, and did not follow any special pattern. However, the values of R1 and R2 of the vowels of each soprano while singing at highest pitch or, in some cases, at mid pitch were separated by a constant interval distance, thus forming a straight line in a graph. This may be seen especially for the back vowel /o/ (here, the interval R2/R1 lay between 1.98:1 and 2.01:1)::
Joliveau and her associates1, at the University of New South Wales in Australia, have found that when sopranos sing in English they raise the frequency of R1 to match that of F0. Since singers trained in Western operatic tradition need to make themselves heard in large auditoria, and F0 lies within the range of greatest hearing sensitivity, tuning R1 to match F0 will produce a sound whose loudness and timbre will vary less with pitch and which is louder for a given constant effort. Thus, singers are taught to lower their jaw or to "smile," when they sing at higher frequencies, since both procedures will increase mouth opening and, thereby, R1. In the study from Australia, 8 professional or student sopranos sang the vowels in the words "hard," "hoard," "who'd" and "heard" for 4 seconds, without vibrato and in an ascending scale. When F0 was less than the value of R1 for normal speech, the resonance of the vowel was roughly constant. However, when F0 exceeded that value, R1 was tuned to follow F0. When singing beyond 1000 Hertz, the trend continued for front vowels, i.e., those in "hard" and "heard," but for back vowels with lip rounding, i.e., those in "hoard" and "who'd," the data fell below the tuning line (R1 = F0). Apparently, with lips rounded it becomes uncomfortable to raise R1 to 1000 Hz.
The results from our own study of sopranos singing in Spanish add some interesting information to these data. It appears that when singing at high pitch (ca. 1000 Hz) in this language, sopranos will be able to maintain a constant interval distance between R2 and R1, allowing a clear distinction of each of the 5 vowels of Spanish. This is possible since the Spanish vowel system is simpler than that of English, and R1 or R2 of each vowel may vary within wider limits. For instance, the minimum and maximum values of R1 and R2 of the central low vowel /a/ sung at high pitch by the 11 sopranos of our study were 372 Hz and 602 Hz (R1), and 727 Hz and 1024 Hz (R2), respectively, whereas those of the back mid vowel with lip rounding /o/ were 461 Hz and 803 Hz (R1), and 923 Hz and 1595 Hz (R2), respectively. Thus, the frequency limits for R1 and R2 of each of these vowels were clearly differentiable from those of the other vowel. It will be rewarding to study sopranos singing at highest pitch in languages with different vowel systems in order to determine if they follow either of the language patterns discussed here, or if there are other, unexpected findings.
1Joliveau E, Smith J, Wolfe J. Tuning of vocal tract resonance by sopranos.
Nature 2004; 427:116