University of Southern California Dissertations and Theses

An experimental study of perceived differences in efficient and inefficient voice production in low-pitched male voices by acoustic spectrography

(USC Thesis Other)

An experimental study of perceived differences in efficient and inefficient voice production in low-pitched male voices by acoustic spectrography

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content AN EXPERIMENTAL STUDY OP PERCEIVED DIFFERENCES IN EFFICIENT V » AND INEFFICIENT VOICE PRODUCTION IN LOW-PITCHED MALE VOICES BY ACOUSTIC SPECTROGRAPHY by Granville Monroe Sawyer A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (Speech) June 1955 UMI Number: OP32018 All rights reserved INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted. In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion. Disssftafion F\ibi sNng UMI DP32018 Published by ProQuest LLC (2014). Copyright in the Dissertation held by the Author. Microform Edition © ProQuest LLC. All rights reserved. This work is protected against unauthorized copying under Title 17, United States Code ProQuest LLC. 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, Ml 48106 - 1346 1 nts dissertation, written by ........... under the direction ofï^%3..Guidance Committee, and approved by a ll its members, has been presented to and accepted by the Faculty of the Graduate School, in partial fulfillm ent of requirements fo r the degree of D O C T O R OF P H I L O S O P H Y • Dean Date. Jtme 1955 Guidance Committee //I TABLE OF CONTENTS CHAPTER PAGE I. THE PROBLEM AND DEFINITIONS OP TERMS USED .... 1 The p r o b l e m.................................. 5 Statement of the p r o b l e m ................. 5 Importance of the s t u d y ................... 7 Definitions of t e r m s .............. 11 Organization of the s t u d y .................... 16 II. REVIEW OF THE L I T E R A T U R E ...................... 18 Classification of voice types ................. 19 Acoustical analysis ......................... . 24 III. SUBJECTS, MATERIALS AND P R O C E D U R E S ............ 33 Sub j e c t s ................ 33 Subjects for the pilot study ........ 33 Subjects for the production of voice samples................ 33 Subjects for the experimental t a p e ........ 34 Subjects used as judges of voice production . 35 Materials ..... 36 Mechanical equipment for recording ........ 36 Rating sheets for evaluations ........ 37 Voice sam p l e s...............................38 The sound spectrograph ..................... 41 CHAPTER Reproducing unit . . # ................... Measuring devices ......................... Procedures , ............ .................. Procedure for the pilot study ............ Procedure for the collection of the raw data Procedure for constructing the experimental t a p e ..................................... Procedure for constructing the introductory material for the auditory discrimination test tape ............................... Procedure for constructing the auditory discrimination test tape .............. Procedure for obtaining a qualified jury . Procedure for evaluating the voice production of the experimental tape • . . Procedure for spectrographic analysis Making the sonagram . . . ........ Analyzing the sonagram ........ Locating and measuring the formant . . Measuring the amplii^ude of formants Determining the fundamental frequency Special characteristics of sound patterns Statistical procedures ......... iii PAGE 42 42 43 43 44 46 49 51 52 53 53 56 62 63 63 65 66 66 iv CHAPTER PAGE IV. PRESENTATION AND ANALYSIS OP DATA ....... 72 Fundamental frequencies .... ............. 72 Difference between efficient and inefficient voice s a m p l e s ........................ 72 Correlation among efficient and inefficient voice s a m p l e s ....................... .. 76 Formant 1 ............................... 77 Difference between formant frequencies of efficient and inefficient voice samples • 77 Correlation among formant frequencies of efficient and inefficient voice samples . 80 Difference between relative formant amplitudes of efficient and inefficient voice s a m p l e s ........................ 81 Correlation among relative formant amplitudes of efficient and inefficient voice s a m p l e s .................... . 83 Formant 2 ................................ 85 Difference between formant frequencies of efficient and inefficient voice samples . 85 Correlation among formant frequencies of efficient and inefficient voice samples . 88 Difference between relative formant V ! CHAPTER PAGE I ' I I amplitudes of efficient and inefficient voice s a m p l e s ............. . 89 Correlation among relative formant amplitudes of efficient and inefficient voice s a m p l e s .................... . • . 91 Formant 3 * .................... 92 Difference between formant frequencies of efficient and inefficient voice samples . 94 Difference between relative formant amplitudes of efficient and inefficient voice samples .......... 96 Special Pattern Characteristics ...... 100 Weak harmonic registration ......... 101 Erratic harmonic registration ...... 101 Pitch c h a n g e........................ 106 Occurrence of higher f o r m a n t s ............... 106 Detection of efficiency by acoustic characteristics ........ 114 V. SUMMARY AND C O N C L U S I O N S......................... 119 S u m m a r y ........................................119 Conclusions............ 123 Fundamental f r e q u e n c y . 123 Formant 1 .............. 123 VI ! CHAPTER PAGE I Formant 2 ................ . , 124 Formant 3 ................................ 125 Special pattern characteristics ......... . 126 : VI. LIMITATIONS AND IMPLICATIONS FOR FUTURE STUDY . 128 Limitations ................................ 128 Implications................................ 130 BIBLIOGRAPHY................................. 134 , APPENDIXES.............................................. 150 Appendix A— Material for pilot study .............. 151 Appendix B--Material for selection of judges .... 152 Appendix C--Material for final experiment ........ 160 Appendix D— Comparative sonagrams with and without compression ...... 162 Appendix E— Summary of statistical data .......... 164 Appendix P--Recordings of voice samples .......... 170 LIST OF TABLES i i1 'TABLE PAGE• I. The Numerical and Lettering Pattern Used to Indicate Voice Samples .......... 40- II. The Evaluation of the Efficient or Inefficient Production of the Sixty-four Voice Samples in Tape C by the Jury of Authorities .... 54 III. Comparative Summaries of Mean Fundamental Frequencies for Test Vowels of Efficient and Inefficient Voice Production .......... 73 IV. Comparative Summaries of Mean Frequencies for Formant 1 of Test Vowels of Efficient and Inefficient Voice Production .............. 78 V. Comparative Summaries of Mean Relative Amplitudes of Formant 1 for Test Vowels of Efficient and Inefficient Voice Production .... 82 VI. Comparative Summaries of Mean Frequencies of Formant 2 of Test Vowels of Efficient and Inefficient Voice Production .............. 86 VII. Comparative Summaries of Mean Relative Amplitudes of Formant 2 of Test Vowels of Efficient and Inefficient Voice Production .... 90 VIII. Comparative Summaries of Mean Frequencies of Formant 3 of Test Vowels of Efficient and viii TABLE PAGE Inefficient Voice Production .............. 93 IX. Comparative Summaries of Mean Relative Ampli^ tude of Formant 3 of Test Vowels of Efficient and Inefficient Voice Production .... 97 X. Summary of Data on Special Characteristics of Efficient Sonagrams ................. 102 XI. Summary of Data on Special Characteristics of Inefficient Sonagrams.............. 103 XII. Acoustic Characteristics of Efficient and Inefficient Voice Production Samples for Each Subject ........ .............. 116 LIST OF FIGURES I I FIGURE PAGE 1. Types of Spectra Analyses Provided by the i Sona-Graph • • . • 14 2. Sound Portrayal Using the 300 ops Analyzing F i l t e r .............. 58 3. Sound Portrayal Using the 45 ops Analyzing F i l t e r ....................................... 59 4. Spectrogram Prepared for Analysis .............. 64 5. Sample Spectrogram: Weak Harmonic Structure in Inefficient Voice Sample ........ . 104 6. Sample Spectrogram: Harmonic Structure in Corresponding Efficient Voice Sample ........ 105 7. Sample Spectrogram: Erratic Harmonic Structure in Inefficient Voice Sample ................... 107 8. Sample Spectrogram: Harmonic Structure in Corresponding Efficient Voice Sample ....... 108 9. Sample Spectrogram: Illustration of Intonation Pattern . . . . . . 109 10. Sample Spectrogram; Illustration of Multiformant Voice S a m p l e .............. Ill 11. Representative Spectrogram of Efficient Voice as Assessed by Judging Panel ................. 112 X FIGURE PAGE 12* Representative Spectrogram of Inefficient Voice as Assessed by Judging P a n e l ............ 113 CHAPTER I THE PROBLEM AND DEFINITION OF TERMS The lack of adequate specification of voice production has resulted in a twofold problem of considerable magnitude; first, redundancy of classification, and second, teaching procedures derived from casual subjective observation rather than controlled observation. Because of the multi-dimensional nature of voice quality, attempts at establishing objective criteria for types of voice production have yielded an abundance of classifications almost as diverse as the different researchers who reported them. One researcher reported that over two hundred different terms have been used to describe "good" and "bad" voice qualities,^ and that some of these terms are so arbitrary as to be useless for research purposes. Another writer simply lists a series of descriptive terms after the statement "check adjectives which best describe . . . voice quality: . . Still another observer reported that, "The confusion of principle and terminology . . . is disappointing for those who seek specific inform- ^Virgil Anderson, Improving The Child ^ s Speech, (New York: Oxford University Press, 1953) p. 41. ^0. Van Riper, Speech Correction, Principles and Methods, (New York: Prentice-Hall, Inc., 1947) p. 222. *z ; ation for perfecting a teaching technique." The second part of the general problem is directly I related to the first. Empirical observation supports the ! I ' assertion that classification and specification of voice I . : I production significantly determines pedagogical and clini- ■ I cal procedures in voice improvement. Rushes classification ; of voice qualities, for example, influenced teaching and ^ , training procedures in speech for more than a hundred i 4 years. The point is further exemplified by examining clinical techniques in voice training. Such techniques are so : highly subjective that they vary from clinic to clinic, and , often among clinicians within the same clinic. They vary ; I 5 from the "vigorously popping corn" technique of Moser to ' i I I the concept of "anchorage" which is a part of the present ! study. Blanton assessed the problem in 1916 as follows; <2 I Ralph D. Appleman, "A Study by Means of Radiograph,I I Flanigraph, and Spectrograph of the Physical Changes Which i Occur During the Transfer from the Middle to the Upper Reg-' I ister in Vocal Tones," Unpublished doctor’s dissertation, I Indiana University, Bloomington, 1953, p. 27. | ^Don E. Plugge, " ’Voice Qualities’ in Oral Inter- I I pretation," Quarterly Journal of Speech. 28:4 (December, : I 1942), p. 443. c I ^Henry Moser, "A Voice Case," Speech Therapy; a | , Book of Readings by Charles H. Van Riper, TNq w York: Prent-' I rceTq-Iall, Inc. ,1953 p. 73. Some scientific work upon the problem of voice training will accomplish two much-needed results: it will give a body of assured and tested facts I which the teacher can use in the classroom, and it ■ ! will enable him to tell the good books on voice I training from the poor ones. G * The need for "a body of assured and tested facts . is no less today than at the time of Blanton’s assessment; in fact, the need is greater, for the problem has be- . come compounded with the increase in the number of workers , who have contributed to the mass of subjective data. I The present study sought to correct directly the * problem of redundant classification by being delimited to | ! I I efficient voice production, a type of voice that lends it-? ' self to both subjective and objective specification. This ! I i j i concept was singled out for study because it has definite ' I criteria by which it can be evaluated, has specific audi- | ! tory and kinesthetic sensations to characterize it, and ' provides a single goal to strive for in therapy* The efficient voice may be evaluated by the follow- ' : ing criteria. First, the voice must be produced with a i , conservation of energy. It must involve no sensation of t ; vocal strain. Second, the voice must yield a maximum ef- I ! feet for the energy expended. That is, it must be clear and ^Smiley Blanton, "Research Problems in Voice and Speech," Quarterly Journal of Public Speaking, 2 (January, 1916), p. 10. , 4 1 audible and gain its power without strain. Finally, the ! j voioe m u s t be as pleasant to hear as it can be made to be. Not all voices can achieve what the average person might consider as "good quality," but most voices when produced I efficiently can be pleasant to hear and will approximate ’ "good quality" as closely as their potential allows. The distinguishing auditory and kinesthetic charact-; ! I I eristics of the efficient voice are epitomized in the con- j : cept of "anchorage." "Anchorage" here is meant to describe, ! the vocal tone when it subjectively seems to receive a I i - great degree of reinforcement from the region of the stern-j I i ' urn, pass through the throat with no sensation of constrio- i i tion, and yield a firm, vibrant, free tone. Admittedly | ' I the term "efficient" is a subjective one, but the sensa- | , tions associated with it, although they are difficult to ! I describe, nevertheless, can be perceived rather readily by | Î I anyone familiar with this voice type. | I j ! Finally, efficient voice production was chosen for ; study because it meets the primary concern of the clinic- ' I I : ian, a positive goal to work toward. Obviously, the voice ' I ' can be produced incorrectly an infinite number of ways, | but this concept proposes that it can be produced correct- | ' ly only one way. Therefore, rather than expending effort, ' time and money in the analysis and quantification of end 5 less deviant voice qualities, this investigation was foe- ;ussed on the specification of the acoustic properties of - ; the voice as it should be produced. Accordingly the con- 'cept of inefficient voice production was introduced only as ,a reference to assist in the specification of the efficient i voice. Inefficiency was interpreted as any form of voice ■ iproduction that does not have the specific characteristics of efficiency. I The foregoing choice was not intended to suggest that jail existing terminology should be discarded in favor of I ! ! that concerning efficiency. The implication was that the ; I i jconcept of efficient voice production might yield more j I ! ■readily to testing, since it can be recognized as a voice | jtype, and thus might provide the "body of assured and test- | ©d facts" which will give more objective meaning to des- ! 'criptive terminology utilized by voice therapists. , i I I I I I. THE PROBLEM ' ; j statement of the problem. The general purpose of ; the study was to determine certain acoustical bases for per-i oeived differences between efficient and inefficient voice production in low-pitched male voices as analyzed by the sound spectrograph. The hypothesis tested was that the ! perceived differences between efficient and Inefficient j ■ 6 ' voice production can be specified by quantifying (1) fundamental and formant frequencies of selected vowels, and (2) amplitudes of the formants of these vowels, I More specifically the study was broken down into i : the following questions, 1, To what extent does the fundamental frequency i ! of vowel sounds differ between voices of low-pitched male ; subjects perceived as efficient and those perceived as in- | I efficient and what is the nature of such differences? I ' 2, To what extent do the frequencies of formants | ■1, 2, and 3 of vowel sounds differ between voices of low- | j ■ ; pitched male subjects perceived as efficient and those per-j Iceived as inefficient and what is the nature of such dif- i ’ I 'ferences? i i 3, To what extent do the amplitudes of formants 1, I I : 2, and 3 of vowel sounds differ between voices of low- i j •pitched male subjects perceived as efficient and those per-. Iceived as Inefficient and what is the nature of such difi 1 Iferencesf J ; 4, What correlation exists between the fundamental j ! frequency of vowel sounds which were perceived as efficient; % and those perceived as inefficient as spoken by the same i subjects? ! 5. What correlation exists between the frequencies ’ of formants 1 and 2 of vowel sounds which were perceived | : as efficient and those perceived as inefficient as spoken , I by the same subjects? i i . ^ ; 6. What correlation exists between the amplitudes of formants 1 and 2 of vowel sounds which were perceived | ! as efficient and those perceived as inefficient as spoken , : by the same subjects? j I 'I I 7, To what extent are spectra of samples of effic- | I lent voice production descriptively different from those I of inefficient voice production on the basis of special pattern characteristics? Importance of the study. Three important points re-| i commended the present study. First, a consideration of ; I ' vocal efficiency has implications for a practicable voice i I , therapy goal; second, a study of voice types by spectrum | ! analysis has been suggested by researchers as a needed contribution toward the specification of speech; and third, findings of scientific studies of voice types would provide objective bases for the development of voice training techniques, 0 Increasing vocal efficiency is a realistic goal in most cases, It is realistic because it does not necessarI ily imply a change in basic voice quality; thus it is I practical as a clinical approach, Lindsley concluded that 8 since the size and shape of the resonating cavities are i more or less determined by nature {these were thought to i ; be the chief determinants of voice quality}, therapists , and teachers are limited in what they can do toward chang7 : ing basic voice types. Drake agreed with Lindsley, but added . . . it is difficult to Imagine the creator endowing some people with the horrible structures they must possess if we are to judge by their voices. Not ' all the trouble lies in structure. Some of it must ; lie in the use and operation of the structure. Imi provement may be found in this functional a s p e c t . B Such functional improvement must of necessity pro- ' ceed in hit or miss fashion until the physical invariants . peculiar to each voice type have been isolated. The pos- , sibility of observing the physical characteristics which are present from sound to sound in repeated samples of I voice types is inherent in spectrum analysis. I It was felt that the study would contribute to the I research program toward the specification of speech by ; spectrum analysis. Potter, Kopp, and Green suggested apI I plication of acoustic spectrography to the study of voice : Charles P. Lindsley, "Psycho-Physical Determinants : ! of Individual Differences in Voice Quality," Speech Mono- j , graphs. I (1954), pp. 85ff. ; i ^Ormond J. Drake, "Toward An Improved Vocal Quality,"; ; Quarterly Journal of Speech. 23:4 (December, 1947), p. 620.' 9 quality in an early report of the work at Bell Telephone ! Q I Laboratories. Later, Dunn suggested, "Another large field i of study by acoustical analysis is that of factors respons-! I ible for differences in voice quality, between individuals."^^ I . . j Cooper lists three significant reasons for the sup- ! ' 11 I eriority of spectrum analysis. Although his recommend- | ' ations were intended for general application in speech re- ' search, one of the reasons has special importance for the j I I present study. He states that spectrum analysis is the "natural" way to think of auditor^r phenomena in terms of its basic dimensions, frequency, intensity, and time. To i discover the relations between the basic dimensions of , sound produced by efficient voices is the sine qua non of ! . this study. He asserts further that . . the development • of the sound spectrograph ♦ . • has resulted in spectrum I j analysis becoming almost a standard procedure for speech studies."12 ! ^R. K. Potter, G. A. Kopp, and H. G. Green, Visible ; Speech (New York; D. Van Nostrand, 1948), pp. 373ff. ' K. Dunn, "The Calculation of Vowel Resonances, , and An Electrical Vocal Tract," Journal of the Acoustical I Society of America. 22:6 (November, 195077 P# 753. ^ n I "^Franklin S. Cooper, "Spectrum Analysis," Journal * of the Acoustical Society of America. 22:6 (November, 1950), I p. 761. I 12 Ibid.. p. 762. i 10 I Findings from studies such as this present one would: contribute directly to more objective bases for the devel- > 'opment of techniques of voice training. This is a need ; that has been felt strongly in the field. As Tiffin states, . . . training in speech . . . is effective in pro- i portion as it is specific and to the point, rather than haphazard and random. But training cannot be made specific until it is known what physical, physio- ' logical and psychological factors characterize a voice ; which is uniformly judged to be pleasant, effective, i and, to put it plainly,,’easy to listen to,’13 | The most succinct justification for the study on . this ground was presented by Bartholomew. Since every aspect of the sound we hear has its counterpart in the sound wave, careful registration of this sound wave by sufficiently sensitive apparatus should give us a record of the whole story. If we can find significant differences in such records ; between good voices and poor ones and can avoid the j I pitfalls of drawing wrong conclusions concerning | something so transitory and subtle as is the voice, we can perhaps work backwards and deduce the physioi logical structure responsible for various qualities. I With that basis, we would know how to proceed most intelligently in the psychological process of bring- ' ing the voice mechanism under conscious control.14 This study was undertaken with the specific search for physical bases for improved techniques of voice train- ' ing as one of the major objectives. Two companion studies Joseph Tiffin, "Applications of Pitch and Intensity Measurements of Connected Speech," Journal of the Acoustical Society of America, 5:4 (April, 1954), p. 225. ^^Wilmer T. Bartholomew, "A Physical Definition of ’Good Voice Quality’ in the Male Voice," Journal of the Acoustical Society of America. 7 (J u l y l ^ M ] 7 p p ^ Z T 5 ? ^ ._ 11 were devoted to the search for physiological bases and psychological implications of efficient voice production. : The probability of developing insight into the basic problems of classification of voice types, of contributing to the specification of speech, and of pointing the way to improved methodology in voice improvement lends importance to this investigation as well as to the companion studies. II. DEFINITION OF TERMS Efficient Voice Production. Efficient voice pro- ^ I I jduction was interpreted as that voice type which is char- : I ' i j acterized by maximum pleasurable auditory effect with min- ! imum effort. On the basis of clinical observation it was found to possess special auditory and kinesthetic characteristics. For the specific design of this study, it was operationally defined as the voice type perceived as efficient by a panel of expert listeners. ^^Robert Miller, "An Experimental Study of the Relationships Between Efficient Voice Production and Good Vowel Quality as Perceived by an Untrained Audience," A doctoral dissertation in progress at the University of Southern California, Los Angeles, 1954. 16Peggy Harrison, "An Experimental Analysis of Resonator Adjustments in Efficient and Inefficient Voice Production in Low-Pitched Male Voices by X-ray Photography," A doctoral dissertation in progress at the University of Southern California, Los Angeles, 1954. 12 Inefficient Voice Production* Inefficient voice I production was interpreted as any deviation from efficient : voice production as perceived by a panel of expert listenI j ers* No special criterion was thought to be basic to in- , efficient voice production. That is to say, voice may be ! inefficient due to any one or several of many reasons. ! Sonagram. "The spectrogram can be regarded as a : I ■ j I display in time of power density distribution among different frequencies. Or better, it is a display of energy den-; 1 ri '■ I sity distribution in the frequency-time plane." Time is ; ! I ; represented by the ordinate axis, frequency by the abscis- ! i sa, and intensity by the degree of blackness of the stylus j I ' ! trace. Sonagram and spectrogram were used interchangeably ! I in the study. I I Formant. Formant and resonating frequency were in- ; I I terpreted synonymously in accordance with the definition of i Dunn. I . . . different vowels have associated with them different frequency regions in which the sound is ; ^ S. E. Chang, £t al., "The Intervalgram as a Visual, Representation of Speech Bounds," Journal of the Acoustical Society of America, 23:6 (November, 195TJ, p. 674. : 13 I more intense than elsewhere in the spectrum. The name ; "formant" has been applied to these regions. , . ,18 I i i I I Spectrum. Spectrum is the characteristic combin- j I ! ^ ation of the physical dimensions of an audio signal: fre- , : quency, intensity, and time. Two such spectra were used in the study. See Figure 1. i I I Line Spectrum. Line spectrum is the "discrete com- ; • 19 I ;ponant" of audio signal as a function of finite time. It 'is discontinuous in that it is not related by any of its j I ! physical dimensions to any other segment of the signal, | I I I thus permitting absolute measurement. See Figure IB. j ! i ! i Continuous Spectra. Continuous spectra are the com-j ; I ponents of an audio signal; frequency, intensity, and ex- ! ' < I 'tended time. They permit only relative measurement. See i I I Figure lA. ; i I I ! I Resonance. Resonance phenomenon "... occurs when-' .ever there is impressed upon a body the frequency at which ^^Dunn, o£. cit., p. 740. E. Heindsmann, "Acoustic Spectrum Terminology,"; Journal of the Acoustical Society of America, 25:6 (Novem- ; her, 19531, p. 1201. Ik mmfBmfBam f . ÏIOTJHE 1 TOUS or spi c m aha.ltsbs proto)» bi t m sohh-giaph A - Line spectra analysis which plots frequency versus amplitude. rrequency is represented hy the abscissa and amplitude by the ordinate. B - Continuous spectra analysis which displays intensity In a timefrequency plane. Jrequency is represented by the abscissa end time b§ the ordinate. Intensity is indicated by the lightness or darkness of the traces 15 ' it would vibrate if set in motion and then left to itself It is the selective reinforcing and damping of frequencies I'' generated bj the vocal cord* Resonance may be said to be directly related to vocal efficiency. If the resonator is ' in tune with the generator, maximum reinforcement of the cord tone results. On the other hand when the resonator is not in tune with the generator forced vibration results and reinforcement is reduced. "Therefore, the effect of resIonance is to increase the initial tone or to change its quality or both."^^ ! ‘ The type of resonance described in the preceding ■ paragraph is due to maintained vibration. It results when - .................................. I a cavity is caused to vibrate at the same frequency as that. : of the sounding body. Vibrations of this type will contin-^ j I Iue for a period after the sounding body is removed. That ' I type of resonance which is due to forced vibration results ! when a continuous outside force activates a sounding body. ; Vibration stops when the force is removed. I Fundamental Frequency. The fundamental frequency was interpreted for this study as the rate at which the S. S. Stevens and Hallowell Davis, Hearing, Its Psychology and Physiology, (New York; John Wiley and Sons, Inc., 1953), p. 8. 21 Appleman, 0£. cit., p. 38. 16 ! ' vocal cords vibrate measured in cycles per second (ops). Harmonic. An harmonic was interpreted to be a component of a periodic wave having a frequency which is an | I integral multiple of the fundamental frequency. I I III. ORGANIZATION OP THE STUDY | I j ; The study is organized as follows. Chapter II con- I ^ I tains a review of the literature. An attempt was made to I I cover the pertinent studies and reports which treat the | i I ; classification of voice types and the acoustical analyses I I I I of voice production. | i I I Chapter III pertains to subjects, materials, and j ! j I procedures. A detailed description of the subjects used ^ i i in the study is given. Treatment is given to the nature of, materials used in the study. Instruments of analysis and I , : measurement receive detailed discussion. Procedures for I j the general experimental design are given in chronological I : steps. I ! Chapter IV deals with presentation and analysis of , the data. It contains tables, figures and descriptive text' 'on the statistical treatment of the data. Thirteen questions are asked and answers are given on the basis of the ; 1 evidence presented. i i Chapter V contains a summary of the study, and con- i 17 : elusions are made within the limits of the statistical sigjnificance of the experimental results. I Chapter VI treats limitations and implications for future research. Weaknesses which became evident as the study moved from conception to execution are pointed out. Suggestions for future studies of voice types are outlined on the basis of insights gained in the present study. Chapter VI is followed by the bibliography and appendixes. The bibliography contains lists of articles, books, journals, technical papers, and other selected sources used in preparing the work. The appendixes contain copies of printed instructions to listeners, samples of rating sheets, statistical data sheets, and recorded samples of test materials used in the study. I CHAPTER II I REVIEW OF THE LITERATURE ; I The problem of redundant classification of voice ! i I types and of basing teaching procedures in voice improve- : ment on highly subjective empirical evidence provided the first reference point for the present review of the liter- , : ature. The classification of voice types may be observed j I in the various treatises on oratory by the Greek and Roman ' ' I writers. However, it appeared that no work addressed it- ’ self to the problem of classifying voices directly before ’ , the 19th century. This historical point provided a begin- : i ning for examination of the literature related to this as- ; I pact of the study. i ! The second emphasis in the review was given to ex- | j ; I amination of studies and reports of acoustical analyses | ' of speech sounds independent of perceptual assessment. I This was thought to be necessary since this study of isoI lated speech sounds which were aurally evaluated has its ; I ' ; roots in the laboratory study of speech sounds not assoc- ; ! iated^with specified perception. These analyses were ac- i : complished by many different instruments, all of which were; based upon acoustical principles. Finally, most direct emphasis was given to examin- ! ! ' " " " ' ' ' 19 ! ation of reports of acoustical analyses by the sound speot- ' rograph of speech sounds which had been previously classii I fied by qualified listeners. Although the literature at this point was very negligible, several creditable studies j were found and reported in the concluding section of the chapter. I. CLASSIFICATION OF VOICE TYPES o p Moses cites as a primary difficulty in objective description of voice, the virtual non-existence of nomenclature for expressing primary perception of vocal phenomena. More than a hundred years earlier, James Rush, also a physician, was aware of a similar need and sought to cor23 rect it. The appearance of his book has been accorded recognition as the first to bring vocal phenomena into the . realm of science. He specifically announced this as his purpose, and by many standards he succeeded. Plugge states The classification of "voice qualities" was one of his [Rush’s! contributions. He gave four principal i forms: normal, orotund, whispering, and "falsetto", , although later he mentioned the gutteral.24 .j ^^Paul Moses, The Voice of Neurosis, (New York; Grune and Stratton, 1954), p. 8ff. 23James Rush, The Philosophy of the Human Voice, (Philadelphia: Gregg and Elliott, 1833)", p. 432. 24piugge, op. cit., p. 443. 20 The importance of Rush’s findings is indicated by I ' the fact that they Influenced teachers of speech for more 25 ; than a century. Gray agrees that . . the first attempt to classi-i ; fy and to give specific nomenclature to the "qualities" was; made by Rush . . Gray lists them as whispering, natur- ' I al, falsetto, and pure tone. He states further that in ^ 1844 Goldbury and Russel discussed eight qualities: harsh, ' smooth, aspirated, pure tone, pectoral, gutteral, oral and iorotund; Hamill (1882) listed seven qualities; and Fulton 1 i . . i land Trueblood (1893) named eight qualities. They added two; ! qualities not mentioned by Rush--the pectoral and the | 26 I ’ oral. "It is probable that this text contains the first | i * * 2 7 i classification of the eight voice qualities;" and it 1 I ■ I I marks the end of an unbroken line of classifications dir- | I ; ^ectly traceable to Rush. Woolbert (1920) listed seven qualities: normal, orotund, pectoral, gutteral, aspirated,' - 28 nasal, and falsetto. Gray asserts further that 25&0C. cit. 2®6iles Gray, "The Voice Qualities in the History of Elocution," Quarterly Journal of Speech, 29:10 (Decemb- :er, 1943), p. 475. BTpiugge, op. cit., p. 443. ^^Gray, pp. pit., p. 477ff. 21 Woolbert seems to have bean the first to recognize the traditional nature of the classification of the "qualities", and to raise a question as to the validity of the nomenclature.29 From about 1920 the conventional classification of j I eight voice qualities continued with frequent modifications, ! ' , I each researcher adding or substituting terms which seemed more descriptive than those he inherited. Among the more recent writers Fairbanks^^ listed four deviant qualities; | I nasal, breathy, harsh, and hoarse. More recently Anderson 'reported eight deviant qualities and reports research which i discovered over 200 terms in the literature "... used by | ! * I various writers at various times to describe good and bad ' qualities of the voice."^^ j Perhaps the most recent and most objective attempt ' at classifying voice qualities was that of Thurman.He I recorded 129 samples of deviant voice qualities which were I j classified and described on the basis of terminology found 29lbld.. p. 480. ®®Grant Fairbanks, Voice and Articulation Drillbook. I (New York; Harper & Brothers, 1940) p. 201. I ^^Anderson, pp. cit.. p. 41. ! ^%ayne L. Thurman, "The Construction and Acoustic ; Analyses of Recorded Scales of Severity for Six Voice QualI ity Disorders," Unpublished Doctoral Dissertation, Purdue University, 1955. 22 'In the literature, and by clinical observation. These I voice samples were played to a jury of staff and speech ' students at nine universities in the Midwest. They were : agreed upon by 67 judges as being (1) breathy, (2) nasal, (3) hoarse, (4) harsh, (5) thin, and (6) strident. In general the researchers reported here succeeded in establishing certain criteria, but such criteria did not' I ; i submit to standardization. The specific designation of j I voice as efficient or inefficient seems to have appeared in ; I the literature for the first time in 1931. Holmes attempt-1 I ! led to specify the efficient voice by equating it with I I I"frontal placement." He wrote | ’ ; I Efficient voice production cannot be achieved unless "frontal placement" is established. It means I conservation of energy for the producer of sounds, ' . . . It means better normal quality hence better speech.SS I Further, he asked "... does it [frontal placement! Î I not mean . . . the sensation that comes to the listener I when the voice of the speaker seems to come from the region directly behind the front teeth?"®^ He concluded that I ". . . speech sounds so produced by frontal placement ^^Holmes, pp. cit., p. 236. S^Loc. cit. 23, I have more e n e r g y as shown by the amplitude of the sound wave than has the same sound produced at a ’placement’ other than ’frontal’. Wye was more descriptive but less objective in his 'attempt to distinguish between efficient and Inefficient ' voice production. He described the efficient voice as strong, accurate, and persuasive, as "one that does the work allotted to it . . , with a maximum of effect, a wise ' jconservation of energy and a large degree of esthetic grat- ,ification." He specifies the inefficient voice as high, | light, and displaying "lack of vigor and emphasis, lack of : ! volume and clearness."36 j ; The classification of voice quality seemed to be ' characterized more by what is not good than what is good. ' The two authors who introduced efficiency and inefficiency as reference points seemed to have unconsciously equated I the most efficient voice production with the best voice ! quality. While it is reasonable to suspect a close relatI ion between vocal efficiency and voice quality, experimental evidence was found lacking on the subject. One of the • companion studies to this one was addressed to determining ! ®®Loo. cit. 36,1,.,.^ o£. cit.. pp. 644ff. 24 ! %rf ' if a significant correlation exists between the two. , ; I ^ Upon findings from such studies as these more accurate classifications of voice qualities must depend. ' I Close examination of the many classifications of ; voice types as found in the literature failed to help in I I deciding which is the best choice for intensive analysis. ; ^ This was due to the fact that the myriad classifications I j lack sufficient criteria upon which objective assessments ; I can be made. The efficient voice was chosen because it was found to possess specific criteria to direct the selection ! of that voice type. It provided a realistic positive goal ' I to work for and thus, was thought to be more deserving of , I attention than the many deviant qualities. Further, the I i i ! efficient voice provided.distinguishing auditory and kin- ; esthetic characteristics which the ear could perceive. ' I I Consequently, this voice type was considered to be amenable| I I to exacting acoustical analysis. i I II. ACOUSTICAL ANALYSES i 1 "Instruments and apparatus have long been employed I , sporadically for the study of speech, but the systematic applications of accurate experimental methods may be said ^^Miller, Loc. cit. 26: : to have begun with , . . Abbe HousselotIn 1889, :Rousselot opened the first course in experimental phonet- : ics.^^ He succeeded in building a crude membranous device | i I ' which when activated by voice registered movement on soot I I covered paper. Thirty-three years earlier (1856), however,- Leon Scott had developed the phonautograph. This was followed by Rousselot’s device which has been called the fore-: , runner of present instruments that make use of diaphrag- | I matic principles.40 i : Professor Henrici devised an harmonic analyzer in , 41 ■ I '1894. The Henrici analyzer made possible an instrumental; I Fourier analysis of complex wave audio signals. It remainseven today as a very adequate instrument for certain har- ;monic analyses. In 1920 Glenn N. Merry developed apparatus! I for "analysis of pitch, or voice inflection of s p e e c h . " 4 % | i This was followed by Seashores’ tonoscope which permitted B. W. Scripture, "The Study of English Speech By New Methods of Phonetic Investigation," Proceedings of the 'British Academy. 13 (London: Oxford University Press, 1923),p. 3. Hazel M. Roth, "Vowel Tonality," Humanistic Stud- : ies. (Iowa City: Iowa University Press, 1931) pp. 5-67. i p. 6. ! 41 Dayton C. Miller, The Science of Musical Sounds. ' (New York: The Macmillan Company, 192277 PP* 98ff. 42Roth, 0 £. cit.. p. 6. 2 ^ , direct reading of voice patterns, Blondell and Duddell’s , ; oscillograph, and Miller’s phonodeik.^^ In 1934 Steinberg I hand-constructed the first sonagram. From this beginning I came a series of investigations which led to the develop- j ment of the sound spectrograph in 1946. For the first ; time researchers in speech had an instrument which could register and reproduce complex sound waves in their basic 44 dimensions: time, frequency, and intensity. Such instruments as enumerated above have made ' acoustical analyses possible, and the nature of these anal-' I • j yses as found in the literature is an area for immediate ! i ! : consideration. | I Acoustical analysis of vowel sounds was made by | ' 45 46 47 ' I Crandall in 1926,^ Steinberg in 1934, Lewis in 1936, : I and Black in 1937.4® These researchers were concerned with! ^^Loc. cit. 44Cooper, 0 £. cit.. p. 761. 45 I. B. Crandall, "The Sounds of Speech," Bell System Technical Journal. IV (1925), pp. 586-626. 43J, c. Steinberg, "Application of Sound Measuring Instruments to the Study of Phonetic Problems," Journal of ; the Acoustical Society of America. 6 (1934), pp. 16-24. 47Don Lewis, "Vocal Resonance." Journal of the Acoustical Society of America. VIII (1936), pp. 91-99. ! I 4®John Black, "The Quality of a Spoken Vowel," Archives of Speech. II (1937), pp. 7-27. ' j ...______________ J 27, the locations of dominant resonance areas in the acoustic ' : spectrum. Other researchers^^ studied changes in the : spectra as related to changes in pitch, intensity, and ! I other aspects of the acoustical signal. i A close examination of the literature revealed, how-|., I ' I C A ; ever, that a study by Lewis and Tiffin was the first to analyze distinct voice types by acoustic methods. Six male: I voices were rated by 300 students for pitch, intensity ' range, general effectiveness and pleasantness of quality. ' The Henrici analyzer and an oscillograph were employed in : ! '' I the analyses. The authors concluded that superior speakers; ! j I had marked flexibility in pitch and intensity and those j i ! : rated as inferior were comparatively inflexible. The j study revealed further that for the superior voices there ' i was a noticeable concentration of energy in one partial or i I I in two adjacent partials. On the other hand, the less I pleasant voices were characterized by widely distributed | I ! energy in the spectrum. I c n i Bartholomew ^ analyzed forty-six wave forms of male ^^See G. H. Talley, "Comparison of Converstional ! and Audience Speech," Archives of Speech. 11(1957), pp.28-40. Lewis and J. Tiffin, "A Psychophysical Study ofi Differences in Speaking Ability," Archives of Speech. I (1954), pp. 43-60. 61Bartholomew, 0 £. cit.. pp). 25-33 28 I ' ' singing voices by Henrici methods. The voices had been ; I I ranked "by the judgment of teachers with experienced musi- : ' I cal taste and extended acquaintance with the voices themI selves." Among other things his analysis revealed that in ; I the case of "good" voices there was considerably more in- ; tensity throughout the spectra, and a "strong low formant," ■ ' I , whereas voices of poor quality were characterized by less ; ' intensity in the spectra, weak low formant, and higher I "high" formant. j Analyses of specific voice types by the sound spec- j ' trograph are not generally available probably due to the I I comparative recency of its development* Early work at Belli : ■ I ! Telephone Laboratories and more recent work at Haskins Lab-i I . ' oratories have been centered around specification of sona- ^ ' graphic methodology. However, in the process much normat- i ' ive data have become available on formant location and | : structure. i 52 ' ! Potter and Peterson studied the locations and 'movements of formants in adult voices. Peterson and ! ! ' 53 Barney reported positions of first and second formants ; R. K. Potter and G. E. Peterson, "Representation of Vowels and Their Movements," Journal of the Acoustical ; Society of America. XX (1948), pp. 628-535. i G. E. Peterson and H. L. Barney, "Control Methods ; Used in a Study of the Vowels." Journal of the Acoustical i Society of America. XXIV (1952). _ J 29 for men, women, and children. Both studies show comparât- ^ i ; ively consistent formant locations from speaker to speaker ; i I within a group. | 54 ' Joos presented a lengthy report of his work with I the sound spectrograph relative to formant location. He ! devised an acoustical chart on which the formant values ' (in cycles per second) could he plotted. Variations on ' the chart were found to be consistent with perceptual dif- i I ferences between vowels of different speakers. | 55 I I Delattre sought to find the positions for the two I I main formants of cardinal vowels. He hand-constructed j j I I spectrograms and played them on a pattern playback to stu- j I dents in phonetics. The study revealed that on the basis j i i of 2 formants, 250 and 2900 cps, fi3 received 100 per cent ! : I agreement on identification; at 750 and 1325 cps, [a") re- I I ceived 94 per cent agreement; and 250 and 900 cps, [u] rei I ceived 77 per cent agreement. These findings are consonant with Joos* formant chart which shows perceptual differ56 ences to be more difficult in the back vowel region. ^^Martin Joos, "Acoustic Phonetics," Journal of the ,Linguistic Society of America. XXIV (April, June, 1948). ; I Delattre, nt al, "Observations on One- and ' I Two-Formant Vowels Synthesized from Spectrographic Patterns^" Word. 8 (1952), pp. 195-210. ! ' 5 g I Joos, 0 £. cit.. pp. 83ff. 30 57 I I Potter and Peterson presented mean values for for-; I ‘ mants of vowels used in this study as follows; [13 250/ ^ ; 2150 cps; [a3 675/975 ops; and [u] 265/725 cps for formants 1 and 2 respectively. ' It will be noted that only two formants have been -| : specified in the research reported here. No account of the, 'real significance of formants above 2 were found in the ; iliterature. Dunn stated that "... two formants are es- • I ' sential for good quality in most vowels, while the third adds somewhat to this quality, and may be essential in 1 some sounds . . In his work with synthetic vowels, ! 59 ’ Delattre assumed that the third formant of spoken vowels makes a significant contribution to vowel color in the case I of front vowels, and that the ear averages the second and ! ^ I I third formants when they are relatively close together. ; I i With further regard to the relative importance of formants, I Potter and Peterson said ". . • bar 2 formant 2j is the I most importance as suggested by its activity in speech I patterns. . . i ; I It may be concluded at this point that the bulk of ; ^'Potter and Peterson, 0 £. cit., p. 528. 5®Dunn, 2 £. cit., p. 740. ^^Delattre, op. cit., p. 208. ^^Potter and Peterson, op. cit.. p. 528. 31 : normative data is centered around specification of formant ' location and structure, and that formants 1 and 2 are the essential components of vowel quality and intelligibility. The implication at this point is that on the basis of present findings acoustical analysis should focus upon eviIdence gathered from observation of formants 1 and 2, and I possibly 3. These have consistently yielded the best results. The contributions of higher formants to vowel qualIity appear to be too conjectural at the present time to be 1 : worthy of more consideration in the present study than frequency of occurrence. The most recent studies involving spectrographic 61 analysis of voice types are those of Appleman and Thur- ^ 62 gman. Appleman used six male subjects chosen from voice I students who could make the transition from the middle to : the upper register while singing three vowels. His analy- ! I sis was predominantly by radiology, but cursory spectro- ! graphic analysis indicated "that the manner in which the I 'cavity is coupled is directly responsible for the vowel .formant." Thurman compared the amplitude peaks, formant i \ amplitudes of six voice qualities which were judged to be 61 Appleman, _0 £. cit. 62Thurman, 0 £. cit. 32, deviant. He concluded that spectrographic analysis is ! "impracticable" in determining differences and severity of ‘ voice quality types considered in the study. i I ' The present study differed from those cited in a I very basic manner. In previous studies where spectrograph- : ic analysis was applied to voice quality types, it was applied in a very cursory fashion. That is to say, the ! i i spectrograph was subordinate to other analytic instruments i t , ’ in those studies, and as a result the analyses were not sufficiently detailed with reference to acoustical measurements. This study sought to make use of a major portion of the data provided by the sound spectrograph. The basic t ■ purpose of the investigation was to determine (1) the mean -, I ! frequency locations of formants 1, 2, and 3 of efficient } . and inefficient voice samples; (2) the mean relative ampli-: tudes of formants 1, 2, and 3 of efficient and inefficient voice samples; and (3) the mean fundamental frequencies of ' selected vowels perceptually judged to be efficient or in- 'efficient which were produced by the same subject. This ' threefold purpose was instrumental in determining the experimental design discussed in the following chapter. CHAPTER III I SUBJECTS, MATERIALS, AND PROCEDURES I. SUBJECTS Sub .lectg for the pilot study. Before any attempt was made to select subjects for the final experiment of , this study, the feasibility of distinguishing the auditory I characteristics of efficient voice production had to be I established. To accomplish this phase of the investigation a pilot study was initiated. Six male subjects whose voices were considered by the experimenters as low-pitched were selected. All subjects were engaged in instruction or study at the graduate level in the Speech and Hearing Clinic at the University of Southern California. The mat- ,erial used in this pilot study was recorded on Tape A. I i Subjects for the production of voice samples. Tra1 'ditionally, researchers in psychophysical studies of voice i -have been beset with the singular problem of multi-variSables operating in the test situation. Thus, the selection! I of subjects-for the final experiment was strongly influenc- ,ed by the factor of limiting variables. As a result, sub- : ! ! jects were carefully screened on the basis of their posses- ^sion of low-pitched voices which were capable of being pro- I 34 :duced efficiently. These bases gave maximum assurance that; I j i the selection of subjects would not adversely affect the ' validity of the study. I The selection of the voice type labeled as efficient I was an arbitrary decision based on clinical experience. I This type had certain auditory characteristics that seemed : i to best meet the criteria for efficiency. These character-^ ' I I istics appeared to be more easily discernible in the voice ! I samples of low-pitched males than in any other group. For ' this reason, only such voices were used. 1 A minimum of forty subjects was desired with one , efficient and one inefficient sample from each subject. However, due to the large number of factors that might eli-j minate voice samples, e.g., x-ray difficulties, errors in i 1 I I evaluating voice production, and recording problems, forty-1 I : I seven subjects were used. The original voice samples pro- ; ! duced by these forty-seven subjects became Tape B. j I : i I Sub jects for the experimental tape. Of the forty- i i I seven individuals who produced low-pitched male voice sam- i pies for the experiment, thirty-two were used in the exper-, imental tape. The remaining fifteen samples were rejected : 'because of technical errors in recording and failure to : meat basic criteria of the experiment. Samples for the , ; I 'final experiment were extracted from Tape B and became : 35 Tape G. Sub jeots used as judges of voice production. A jury 'of judges or authorities in voice production was selected on the basis of specified criteria* First, the prospective I voice authority must have had graduate status or above in i I the Speech or Music Department at the University of Southlern California. The search for voice authorities was limi-| I ted to this group for two reasons* It was felt that a jud-; I ge must be skilled in the general area of voice production. ! By selecting only graduate students in the areas of speech or music, this background would be relatively and consistently good. It was also deemed wise to limit the search , for judges to the Speech and Music Departments of the Uni- i I I jVersity of Southern California because individuals who } qualified as judges would be available for consultation rejgarding developments in the experiment. ! The second criterion for a judge was that he demonIstrate superior skill in discriminating between efficient I land inefficient voice production. To qualify as an expert, leach subject had to be able to accurately decide whether the voice sample had the auditory characteristics that deI |termined efficiency or those that determined inefficiency, lit was critical that only judges who were quite capable of subjectively perceiving empathieally the auditory and kin- 36: i ; esthetic differences in voice production be selected. ’ Theoretically, one judge was all that was necessary, but to insure the validity of the discrimination between i efficient and inefficient voices, a panel of seven author- ' ities was used. To obtain these seven judges, forty-five | I ; ' individuals were given an auditory discrimination test de- : * : vised for this experiment. I The discrimination test tape, labeled Tape E, conI tained twenty examples of efficient and inefficient voice ! production. To qualify as a judge, each subject had to : assess nineteen of the twenty samples correctly. Seven of . ; forty-five subjects tested were able to attain this score I I of ninety-five per cent. i ! i i II. MATERIALS I i ! Mechanical equipment for recording. The equipment I . i j used in this experiment was selected to reproduce the voice ! : samples with as great fidelity as possible. The recording i ' was done using a Pentron Multi-Speed Tape Recorder, Model 9T-3G. This instrument records audio signals of 100-8000 ' cps with optimal fidelity. Since researchers agree that components of vowel spectra are negligible above 4000, the Pentron was considered to be adequate for the experiment. Recordings were made at seven inches per second. j 57 I The Pentron was used with Model D22 Dynamic Omnii (Directional Microphone, Manufactured by the American Micro- ; phone Company of Pasadena, California. The frequency resI Iponse was 100-8000 cps, plus or minus 5.0 db. Both the I microphone and recorder were matched for high impedance. I Scotch Brand Sound Recording Tape was used for all j recording because it provided a relatively high fidelity | and noise free response. In addition, it could be readily j I modified for the specific requirements of this study as described in Part III of this chapter. Rating sheets for evaluations. Evaluations of voice | samples were done on rating sheets. Two different rating j sheets were used. The first was prepared for procuring the jury of authorities in voice production. This sheet was used in conjunction with the auditory discrimination test consisting of Tapes D and E. It required that prospective judges rate the twenty voice samples on Tape E as efficient,' inefficient, or undecided. The second rating sheet was used by the judging panel to evaluate the voice samples selected for the final experiment as contained on Tape G. It required that the judges rate the sixty-four voice samples as efficient or inefficient. An example of the second rating sheet is found in Appendix 0. 38 Voice samples in Tape A. Seven tape recordings were prepared in the process of this study. The first. Tape A, ’ was produced for purposes of a pilot study. It consisted I : ( ; of samples of voice production that the experimenters be- j 1 I lieved to represent various degrees of efficiency and inefficiency. This tape contained isolated vowels and diphthongs, words, and sentences. ; 1 I Voice samples in Tape B. Tape B contained all of ; the raw material obtained from the forty-seven low-pitched I male subjects used to produce voice samples for the experi- : I ment. Each subject was numbered and each voice sample was i I lettered. Retakes of voice samples were given a suffix i ! number. X-ray plates for another study done in conjunction; 66 i with this one were marked the same way. A voice sample , was not numbered if it did not have an x-ray with it as it ; could not be used in this combined experimental program. ■Subjects were numbered one through forty-seven and retakes 'Were numbered one through four. Table I indicates the I I pattern that this numbering and lettering procedure followI ed. Voice samples in Tape £. Tape C was the master tape, for the experiment. It consisted of sixty-four voice sam- , ^%arrison, pp. cit. ;39 , i pies. These represented what the experimenters evaluated ' I , to be one efficiently produced voice sample and one inI ’ ! ; efficiently produced sample from each of thirty-two sub- , jects. All samples of Tape C were spliced sections taken I ' from Tape B. | i Voice samples in Tape D. Tape D consisted of the j I I introductory material for the auditory discrimination test I I tape that was used in selecting the jury of authorities. I ■ This tape included paired samples of efficient and ineffi- - I cient voice production by the same voice. Isolated vowels ; I and diphthongs, words, and sentences were used. Many of ' these examples were taken from Tape A. Additional explan- I I I atory material was inserted to aid the listeners in differ-j entiating the auditory characteristics between the two j ; voice types. ! i I Voice samples in Tape E. The twenty voice samples | ! i I used as the auditory discrimination test comprised Tape E. ' I iTapes D and E were always presented together. Tape E was ; ; composed of voice samples from Tape A as well as samples I : from Tape B that were not used in the construction of Tape ■ ! \0, the final experimental tape. | ! I All recorded voice samples were dubbed on discs and ; ; included as part of the study. The discs comprised Append-; ' ix p.. ________ _ 40 TABLE I THE NUMERICAL AND LETTERING PATTERN USED TO INDICATE VOICE SAMPLES Symbol Identification Symbol Identification ! A ' B 15 1: ' G efficient fi] efficient [a] efficient Cu] inefficient [1] inefficient [a] inefficient [u] extremely inefficient [i] H extremely inefficient fa] K extremely inefficient [uj 1 through 47 as a prefix indicates subject's test number 1 through 4 as a suffix indicates subject's sample retake number NOTE: This table would be read as follows: 6D3 'would indicate the third retake of the sixth subject's sam- ! pie of inefficient • I' 67 The sound spectrograph. The sound spectrograph Is ^a complex signal analyzer which was developed by the Bell ; 'Telephone Company and made commercially available by the I I I 'Kay Electric Company of Fine Brook, New Jersey and the Wes-| ; tern Electric Company of New York. The Sona-graph of the 'Kay Electric Company was used in this study. The Sona-graph analyzes and visually records the I analysis of sound signals in their three basic dimensions--! ifrequency, time, and intensity. The visual representation I ,of the audio signal is recorded on specially treated paper j i 'which bears the trade name Teledeltos. Frequency is repre-! ^ I ;sented by the abscissa, 0-8000 cps; time by the ordinate, ; '0-2.4 seconds; and intensity by the shades of blackness of | the stylus tracing with a maximum range of 36 decibels. In a word, the sona-graph provides a permanent record of the energy in a given signal distributed over a time-frequency field. This is referred to as Display 1. ; I The Sona-graph is capable of a second display. In I ' I addition to the time-frequency-intensity display, the inistrument can be adjusted to yield a two-dimension display ' of amplitude versus frequency. The first display is an ex- ' ! 67 For a technical description of the sound spectrograph, see W. Koenig, "The Sound Spectrograph,” : Journal of the Acoustical Society of America, 18: (July, 194éj, pp. 19-49. ' ' ... 42, ; i ample of continuous spectrum analysis; the second an exam- i I pie of line spectrum analysis. The second display analyz- ' i ; : 0 8 a finite segment of the audio signal in its intensity- i , frequency relationships.See Figure i. I I Reproducing unit. The Sona-graph used in this experiment was in series with a Magnecorder recorder-reprod- j ' I user. Model PT6J. It is characterized by high fidelity j I reproduction. : I ! Measuring devices. The frequency range--0 to 8000 j I ' 1 I cps— of the Sona-graph spans a distance of four inches on j I i i the Teledeltos paper. A plexiglas scale covering the fre- j quency range up to 4000 was prepared, graduated in units of| I 100 cps. This was found to be consonant with good research! methods in acoustical analysis at two leading research cen-i ters. The amplitude portrayal of the Sona-graph is made on uniform decibel progression. An amplitude scale was made accordingly on the basis of 4.5 millimeters per 5 decibels. By applying both the frequency and amplitude scales' to appropriate parts of the portrayals, numerical values ! CO See Kersta, 0 £. cit., p. 798 G®Ibld.. p. 798. 45 were obtained and recorded for statistical analysis. III. PROCEDURES I Procedures for the pilot study. Before endeavoring ‘ ito gather the data"for the major portion of this experiment; ! a pilot study was done to determine the feasibility of dis- ! itinguishing the auditory characteristics of efficiency in . ; ' ! voice production. The purpose of the pilot study was to I I determine if the arbitrary decision of the investigators, based on observation and clinical experience, could be subI stantiated by other experimenters in the field of voice ■ science. Were other speech therapists able to recognize ; efficient voice production subjectively in terms of empath-; lie feelings of chest resonance, of vibrant tone anchored | I I near the sternum rather than originating in the neck, of | I openness in the throat with no sensation of tension, and of tones that "coast" without effort or strain? Could these I be experienced by others? If others with a background of : speech science had been unable to make this differentiation, this study would than have been impracticable. | I The samples for the pilot study included a sentence, ; I three words, isolated vowels, and diphthongs: 1. The ball has fallen over the wall; 2. live, lag, love; I 44: 3. u l , B e ] , ta] , 1^2 ; and 4. (el] , faij , Dx ] . I Each sample was produced both efficiently and inefficiently: I i by several voices, I The results of the pilot study indicated that in | general, individuals with a voice science background could i distinguish between the two voice types. Some appeared ! t - ' much more aware of the differences than others, but there j was sufficient agreement to indicate the appropriateness ! I of further research in the area. The materials used in the, ‘ ; ! ( 'pilot study composed Tape A. See Appendix A for printed instructions regarding the pilot study. Procedure for the collection of the raw data. The : ! , collection of a large number of samples was required to | produce a tape of efficient and inefficient voice produc- j tion to fit the requirements of this study. Forty-seven ! individuals were used. Only low-pitched male voices who : were capable of using their voice mechanisms efficiently : were selected. Again efficiency was determined by the ex- : perimenters on the basis of auditory characteristics cited ' at the outset of this section. Each subject was then given ! fifteen to thirty minutes of instruction in producing the | desired voice type. This short training period also gave ' i the experimenters an opportunity to analyze the voice pro- I 45 'auction just before recording. I Each subject was instructed to sustain the vowel I ; j sounds [Ï], [a], and [u] after a carrier phrase, first i ' efficiently, then inefficiently. The carrier phrase was j ■ . i : essential only In that generally, it helped the subject to i attain the desired voice type. It also served as an ex- ; ! cellent time cue for the x-ray picture. The phrase usual- i I ^ I , ly consisted of the diphthongs , and Ipx] . In i I j some cases the carrier phrase proved unsatisfactory and , was modified or omitted altogether. Generally, however, it; iwas an integral part of the experiment, even though these i I Î I diphthongs were not analyzed spectrographically. j i ! After the subject had been instructed, he was seated I before the x-ray equipment* A dentist's chair with a headrest was used to maintain the individual in a stationary jposition with the saggital plane perpendicular to the x-ray, I I 'machine and parallel to the x-ray plate. The x-ray was ! I ‘ made of the left side of the upper thorax, neck, and head I at the time each vowel was produced. The directional microphone was placed two inches in ! front of and at the level of the subject's mouth at a forty4 i I five degree angle to the saggital plane of the head. This ; I I Iwas done to eliminate as much as possible the x-ray noise and intensity artifacts caused by the subject's speaking .............. 46 directly into the microphone at close range. When the voice ! production of the subject appeared to have the desired : characteristics, the x-ray was taken. This spot was then j I marked on the tape as indicated in Table I. In addition, ithe exact point at which the x-ray occurred was audible on | I the tape. This noise, when heard as a playback on the tape,' was similar to that of a book being dropped on the floor in; ! I an ad jacent room. It was characteristically a noise of | I low-frequency. Altogether 388 samples with x-rays were ; i ; made. The recorded voice samples for these x-rays were j ■ 1 I contained in Tape B. | ; I Procedure for constructing the experimental tape. ; I ; Prom the raw material collected on all subjects in Tape B, | I I the final experimental tape was constructed. This was de- i i signated Tape G. Tape G consisted of sixty-four voice ! i ■ ■ I samples. Sixty-two were composed of the three vowels [i], j [a] , and [u] • The remaining two samples had but two sounds' I each. One consisted of [i] and [a] , and the other fa] and! ! £u]. One subject was not available for later retakes. Due to this unavailability of the subject, the experimenters I were forced either to use his two samples with only two i sounds each or to discard his samples. The former course was taken due to the four relatively desirable sounds he had produced. 47 j Any sounds might have been selected for the experiI ment. However, the sounds fi] , faj, and [u] were chosen because they represented critical points on the traditional acoustic vowel chart which, in turn, represents those areas of production in the mouth.[i] is a high front vowel j produced in the superior anterior section of the mouth. [a] is a back vowel produced in the inferior posterior area of the mouth. [u] is a high vowel produced in the superior posterior portion of the mouth. Vowels were chosen instead of any other voiced sound because the subjective qualities of efficient and/or inefficient voice production appeared to be more easily discernible in sustained production of pure vowels than in consonants. Moreover, the pure , vowels provided better control of the variables in the spectrographic and physiological aspects of the study than would have diphthongs. The two samples that contained only two sounds each were retained in the experiment despite the shortage of a high back vowel sound in one and a high front vowel sound in the other because it was felt they were still adequate examples of the voice production they represented. J. S. Kenyon and T. A. Knott, A Pronouncing Dictionary, (Springfield, Mass.: G. & G . Merriam Company, 1944), pp. xiif. Also Joos, ££. cit., p. 60, and Delattre, 0£, cit., pp. 866ff. : ' 48 I This was later verified by the judges' assessments. ■ I Tape G was made by arranging the voice samples sel- i : ected from Tape B in random order. Since the carrier jphrases were designed only to a id the subject in acquiring ! I the desired voice production of the vowels and as time cues, they were not used in this tape. Only sounds that were produced at the time a subject was being x-rayed were | ! used. ' For each of the individual sounds, six-sevenths of : I j a second of voice production was selected. The section I : I chosen invariably included the point at which the x-ray | ! ■ ; iphoto was made. The use of a voice sample produced at the i ; i i time of the x-ray was essential to the interrelationships I I i I of the three studies. It was imperative that all exper- ' imenters evaluate the same samples. This was true because I it was conceivable that a voice sample might be produced , efficiently for part of its length and inefficiently for i ianother part. Hence, an evaluation of the first part of a | I ■ voice production might reveal a different set of auditory i characteristics from the second part. The longer the sound; was produced, the greater the possibility of difference between the two samples. | I Each of the three sounds to be used in one sample | I was limited in duration to six-sevenths of a second. This i ...... 49 time-segment was used because of the physical characteristics of the sound spectrograph. ! The three sounds of efficient voice production were ' then treated as a unit, and likewise the three sounds of i I inefficient voice productions of the same subject were j I treated as a unit. Each unit appeared on the final tape I ' as six-sevenths of a second of ]1], a blank of six-sevenths. , of a second, then six-sevenths of a second of [a] , a blank ! of six-sevenths of a second, then six-sevenths of a second I i of tu] . Since the tape ran through the recorder seven in- j ' ! ches per second, each sound represented a six inch section ; i of tape. Each unit of voice production was separated by a : six second interval to allow judges time to analyze and i j , record their decision. The finished tape of sixty-four i i I voice samples had a running time of approximately eleven I I jminutes, | The preliminary decision as to which voice sounds | ! I were efficient and which were inefficient was that of the experimenters. This decision was used to choose the voice , i samples from Tape B for Tape G. The final decision as to | : which samples in Tape C were efficient and which were in- ; I ; efficient was to be made by the panel of judges. Procedure for constructing the introductory material for the auditory discrimination test tape. The preparation 60 of the introductory material for the auditory discrimination test tape was the first of three distinct operations I ! ! in obtaining a qualified jury of authorities in voice pro- j j - ' I duction for this experiment. The first step was the preparation of an orientation : portion. This was designated Tape D. It consisted of sam-; , pies of efficient and inefficient production of the same i I voice. In order to serve its purpose of obtaining quali- i I i fied judges of voice production, it was important that each ' sample be a true example of the voice type desired. Thus, I voice production sampIps were selected by the experimenters! on the basis of clinical impression to represent clear-cut j examples of efficiency and inefficiency. i ’ Gross inefficient samples were avoided in this in- : ! I stance for two reasons. First, they would not require finej I enough discrimination between samples to provide a basis ! for selecting only the most qualified judges; and second, i I it was imperative that the sample should not be so ineffi- ! cient that its poor production would be classified under ' any system of evaluation as undesirable. The samples were interspaced with explanations to aid the listener in developing his auditory senses of acute and accurate discrimination. Instructions concerning the auditory discrimination test tape are included in Appendix 51 I B. The sentences, words, and some of the vowel sounds used. ' in Tape Â, the pilot study, were employed in constructing ’ the first part of this tape. These samples were: The ball has fallen over the wall; live, lag, love; and £%] , Cae] , ! and iol , They were presented first efficiently, then in- j ! efficiently, using the same voice. In the latter part of | ; the tape, three different low-pitched male voices produced ' I ■ i ' additional samples of the same sentences, words, and iso- | jlated sounds efficiently, then inefficiently to supplement the material taken from Tape A. At the beginning of the orientation section,a description of the distinguishing ‘ characteristics of efficient voice productionvwas presenti ed. A total of eighteen samples of efficient and ineffic- ; ' lent voice comprised Tape D. A copy of this tape is found j I I in Appendix P. Procedures for constructing the auditory discrimi- ; nation test tape. The second step was the preparation of ! the Auditory Discrimination Tape portion. This was desig- ! I I jnated Tape E. The samples are given in Appendix C. These ' ' samples were vowels or diphthongs only and they appeared in* .random order. Each sample consisted of a group of two, , ' I ; three, or four sounds and each sample was separated by a ! six second interval. The sounds for Tape E were chosen ! ; I ^from the raw material gathered either from the pilot study 52 i and presented in Tape A, or from voice samples in Tape B that were not used in the experiment and thus required for Tape C . As was the case in the preparation of the orientation section of the test tape, it was imperative that each sample be definitely efficient or inefficient in its production. The efficient production had to have no shades of ! inefficiency and the inefficient production had to be sufI j ificiently inefficient that it could be evaluated correctly j I by the more capable judges. Extremely inefficient voice | ! production samples were not used because their exaggerated I inefficiencies would be evaluated as poor production using I almost any criteria, and a fine discrimination between I I efficient and inefficient samples was essential to select the most qualified judges. Procedure for obtaining a qualified jury. The third step in obtaining a qualified jury of authorities in voice : production was to present Tapes D and E to as large a group of individuals with a graduate status or above in the Speech I 'or Music Department at the University of Southern Californ-' ia as possible. Because of the desirability of having the ; : individuals who qualified as judges available for consult- ; i ation regarding developments in the experiment, it was de- I cided to limit the search for judges to the Speech and I • I I Music Departments of the University of Southern California.I 53 Of the forty-five individuals who took the test, seven met i the standards established by the experimenters. To meet i these standards, not more than one of the twenty samples i could be evaluated wrong, so that each member of the jury : had a score of ninety-five of one hundred per cent. Procedures for evaluating the voice production of the experimental tape. After the completion of the experimental tape which was labeled Tape C, it was evaluated by the jury of seven authorities in voice production. They ; evaluated each voice as either efficient or inefficient. I The results of their evaluation are shown in Table II. ! This table groups the samples showing the number that all judges found efficient or inefficient. It also shows the number of voice samples on which the judges' decisions were split and the strength of that split as to how many judges rated the samples efficient and how many rated them inefficient. A more complete breakdown of judges' decis- ,ions is given on each specific voice sample in Appendix D, in which the samples are also listed from the most efficient to the most inefficient. ! Procedure for spectrographic analysis » The original signals were recorded on a Pentron recorder whereas the re- ,producer used in the spectrographic analysis was a Magne- 54 TABLE II THE EVALUATION OP THE EFFICIENT OB INEFFICIENT PRODUCTION OF THE SIXTY-FOUR VOICE SAMPLES IN TAPE C BY THE JURY OF AUTHORITIES No. of Voice Samples No. Judges Who Rated Samples Efficient No. Judges Who Rated Samples Inefficient 12 0 6 5 1 2 4 3 3 4 8 6 2 1 5 6 18 0 Total 64 56 icorder. A question arose at once. Would the alignment | characteristics peculiar to each instrument significantly ' i ■ ' I affect the acoustic properties of the signal? To answer this question accurately the six-seventh blank between each sound was removed and the Master Tape was dubbed on a Mag- ! neoorder unit with the hope that any alignment differences would be cancelled. Samples from both tapes, the dubbed copy and the Master Tape, were analyzed on the Sona-graph I 'and it was discovered that no appreciable acoustical differences could be observed. Therefore, the original Master ;Tape was used in all spectrographic analyses in this study. Although extraordinary recognition was given to moni-; jtoring the gain aurally in the final recordings for Tape G, ! there were noticeable differences in the intensity levels I at which the signals were recorded. This, added to the ini 71 herent difference in phonetic power of the test vowels, 'posed a serious question in spectrographic analysis. It I I meant that frequent changes in the indicator settings might be necessary to get satisfactory registration of the isignal components. The alternative w&s to find a mid-point, for instrument readings and make minimum readjustments to take care of extremely weak or unusually strong signals. 71Harvey Fletcher, Speech and Hearing in Communication. (New York: D. Van Nostrand Co., Inc., 195^, p. 84. 56/ The mid-point had to be determined experimentally. Since the bulk of normative data concerning spectrum; analysis treats formant structure and location, the major ! question to be answered here was, what adjustments will yield the best formant portrayal? This question could be ; I approached directly. I The Sona-graph contains a compression circuit. The I ; purpose of this circuit is to emphasize the harmonics in j I the high frequency regions of sound spectra. With full compression, maximum emphasis is given the normally weaker I higher harmonics, and often the formants do not stand out I with sufficient contrast. The major question, then, was I partially answered; minimum compression should be used to ■avoid over-distortion of the harmonic structure, and posI :sible occlusion of formants. Experimentally, it was disI covered that with the compression setting at 5, sufficient I ! detail was present for analysis, and further distortion of ; the harmonic structure was avoided. Appendix E shows comjparative sonagrams with and without compression. Samples I ! with compression were made on a Western Electric analyzer ! at the University of Michigan. All other sonagrams used ! in this study ware made on the Kay instrument at the I !University of Texas, Austin, Texas. Making the Sonagram. After intensive expérimenta- r ....... 5 7. ' 'tion with widely selected samples from the Master Tape, the following settings ware adopted as standard procedure. 1. Gain on Magnecorder Reproducer -- 9 2. RECORD on Sona-graph — 6 3. REPRODUCE on Sona-graph — 7 to 9 j ; 4. HIGH SHAPING circuit — ON : 5. Filter Selection — NARROW BAND I ; S, MARK LEVEL -- 6 | I 7. SECTIONER -- OFF (For Display #1) 8. SECTIONER — ON (For Display #2) 9. COMPRESSION — 5 The narrow hand filter was selected because it provided more accurate details for analysis. The alternative i was the wide band filter--500 cycles per second. Although ; i the wide band filter makes formant representation stronger | and generally more intelligible, the representation is so gross that large errors in calculation of resonance frequen- ! icies are inescapable. Figures 2 and 3 show two sonagrams iof the same voice sample. Figure 2 was made using the 300 'cps analyzing filter; Figure 3 used the 45 cps filter. It i may be readily seen in Figure 2 that when the formants approach each other it is extremely difficult, if not impossible, to distinguish them. Spoh is not the case in ' Figure 3. 58 2000 1000 Formant 1 Formant 2 Formant 1 2000 Formant 2 Indistinct Formant Structure 5000 Indistinct Formant ,,,Structure Formant 3 M l 2 l’ hiU - 4000— ,1, 3000 PIGTISB 2 SOURD POBTSàYAD USING THE 300 GPS AKàLYZING FILTER When formants lie close together es ia fa] it is very difficult to determine the f***equency of the center of each separate formant if the 300 cps analyzing filter is used. Compare with Figure 3. 59 Formant 1 Formant 2 Formant 1 Formant 1 Formant 2 Formant 2 Formant 3 Formant 3 Formant 4 Formant 3 Formant 4 Formant 4 4000 _ 4000 Formant 4 Formant 4 3000 Formant Formant 3 Formant 3 2000 J Formant 2 Formant 2 Formant 1 1000 armant 2 Formant JL FIGURE 3 SOURD PORTBAY&L USIBG TBE 43 OPS k M L Z Z i m FILTER The 45 cps analyzing filter provides distinct formant structure and permits more accurate measurement of formant frequency. 60 I j The Magnecorder reproducer was tied in directly to the recording circuit of the Sona-graph. The signal was monitored through head phones. Once the signal was trapped; iln conformance with the time limitations of the Sona-graph,: I the switch was instantly thrown to REPRODUCE. Gain on the ' I ; Sona-graph was then increased until the sound was audible. If the signal was not sufficiently audible, or if part of , it were not recorded, the signal was erased and the process| i repeated. If the signal was trapped in its entirety and with ! sufficient audibility, the drum upon which the Teledeltos | paper was placed was caused to rotate at reproducing speed.* I ' ' The VU indicator showed the relative intensity level at * which the signal was recorded. It was determined experi- ; mentally that if the needle fluctuated consistently between! ! i I-7 and 0 readings, roughly equivalent to from 45 to 100 , ' i decibels, the portrayal would be very satisfactory. HowI lever, it was not possible to get this reading with all the i signals because of the differences in intensity levels at * which the original recordings were made. It was necessary I to reproduce parts of some signals at an increase in gain, and a decrease in others. [i] for example, in Sample #58 I had to be selectively recorded at 2 steps higher than other | ! I components of the sample. Conversely, Sample #26 had to be; 61 , recorded at 1 step lower than other samples to avoid over- : distortion of the signal. It was not feasible to increase i the gain linearly in the first instance. To do so would ! have introduced undesired distortion in the stronger com- ' ponents of the audio signal. With such necessary adjustment completed the Teledeltos paper was placed on the drum and as the drum rotated^ the stylus was brought into contact with the paper. The I preparation of display 1, time-frequency-intensity analyi sis, from this point was purely mechanical. Only half of I i the total range of 8000 cps was used. Previous researchers i I have established that the important components of vowel ! sounds are present below 4000 cps. The Western Electric | Î i Acoustic Analyzer as used at the University of Michigan ' ' outs off at 3600 cps. j i ; Because only half the range was used,only the bottom! j ' I half of the paper was involved in display #1. This made it| I ^ I : possible to put display #2, frequency versus amplitude, on , the upper half of the paper. Further, the frequency scale ! is inverted when the sectioner is used in 4000 cps analy- I sis. Thus, at the end of display #1 it was necessary only I ! to stop the drum, switch in the SECTIONER circuit; select points for frequency-intensity analysis, and start the drum, rotating again. In making an amplitude section analysis, j 62 display #2, the stylus was brought into contact with t he drum at 4000 cps rather than 0 cps as in display #1. j i From this point, making the section analysis was | ■ ' I mechanical. When the stylus had reached the end of its ex-| i Gursion, the drum was stopped, the stylus returned to 0 cps . and the completed sonagram removed. Each step in spectrographic procedure was checked repeatedly with the instruc- ! 72 I tion book provided by the Kay Company, and selected re- I ' search reports as cited previously in this study. i Analyzing the sonagram. Immediately upon removal , ! from the drum the sonagram was inspected for general appear- , ance. If formant structure was too distorted, or formant ; , location too doubtful, the sonagram was discarded and an- | ’ other made before erasing the s ignal. If the sonagram | I I evidenced special promise in any way, a second one was made; I for possible reproduction. It was thought that a subseI quent decision to reproduce a sonagram would be subject to ^ question due to changes in power voltage, and similar fact-; I i ors. A reproduction made at the time of analysis insured validity of this aspect of the experiment. i ^^"The Sona-graph,” Instruction Book, published by ; the Kay Electric Company, 14 Maple Avenue, Pine Brook, Jersey, 1955. 63 Locating and measuring the formant. Formants were 73 located in accordance with Joos, and described on the ; basis of the frequency of their centers. When the center i I was found to be between harmonics, a thin line was drawn : parallel to the harmonics, and this served as the measur- I I ; I ing point for determining frequency locations. If the ; center coincided with an harmonic, measurement was taken I 1 from line 0 to the center of the harmonic. See Figure 4. ! I Formants 1 and 2 were located before the signal was erased I from the drumhead of the sona-graph. Higher formants, not j I i I peculiar to all the signals, were located later. All form- | : ants were measured after construction of the scale as dis- I i I cussed under MATERIALS above. i "■ I ! ! I I Measuring the amplitude of formants. It was desired i that the amplitude sectioner of the Sona-graph be callI brated in conformance with the nature of the signals to be ,analyzed in this study. Such was not feasible and as a I result, to measure the amplitude of formants, amplitude ,peaks were measured in millimeters and converted to decibels on the basis of 4.5 millimeters equals 5 decibels. I It was felt that the validity of the study would not be in , I I jeopardy for want of instrument calibration at this point. 73 Joos, 0 £. cit.. p. 48 64 Formant 1 Formant 2 Formant 1 41 db (4,5 mm 5 db ) Formant 1 Formant 2 Formant 3 Formant 2 42 db (4,5 mm 5 db) Formant 3 3000 Formant 3 38 db (4,5 mm 5 db) - 4000 Formant 4 4000 ormant 3 Formant 4 Formant 3 _ Formant 2 Formant 3 2000 Formant 2 Formant 2 Formant 1 _ 1000 Formante 1 F I G U m 4 SFECTROGB&M PEEE&EBD FOR ABKLTSIS Formants are located and labeled in both analyses. In display 2, line analysis, formants are measured in millimeters for amplitude quantification as indicated in [i]» i for such calibration would only yield a constant of the order of 45 decibels which would be added to the decibel i rating obtained by using the method described here. This | would make the obtained decibel rating more in keeping with| more familiar decibel ratings for normal speech, but it ! would not affect the amplitude ratio as was found between the efficient and inefficient groups, for the same scale was .used throughout the study. The scale was described in the I preceding section. I I I I I Determining the fundamental frequency. The scale | I as used for locating the frequency regions of formants was i I also used in determining the fundamental frequency. The | jtenth harmonic was located in display #2. The scale was i .placed on the amplitude portrayal, and the frequency at ^ •which the tenth harmonic was found was recorded. This value I •divided by ten yielded the fundamental frequency. This pro-, ' I icedure has been employed successfully by other researchers. ' |Often the harmonic structure was distorted by x-ray noise or for some unknown reasons. l%ien this successfully oc- ' I eluded determining the tenth harmonic, ten harmonics were j I counted in other regions of the signal, say from the 9th -through the 18th. In each case the value arrived at as the fundamental frequency was checked against the distance be- - tween harmonics registered in display #1 for verification. 66 Special characteristics of sound patterns. After obtaining the quantitative values for each signal, certain | : aspects of the sound pattern became noticeable. Some sam- | j I pies had more than two formants, others had abnormal pitch | I patterns, still others, erratic harmonic registration. | These special characteristics were singled out and placed in this category on a summary data sheet. They were so recorded in event the frequency of occurrence should become significant. All descriptive assessments together with numerical data were checked a minimum of three times to , ! insure maximum accuracy and objectivity. | j Some researchers have attributed differences in voice I I ; quality to the number and relative strength of the over- | tones. Miile this experimenter did not ignore this fact in the present study, he was not able to discover any normative data treating this area of investigation among reports of spectrum analyses by the sound spectrograph. Thus, this study was held to the analysis of (1) fundamental frequenicies, (2) frequencies of formants 1, 2, and 3, (3) amplitudes of formants 1, 2, and 3, and (4) general description of special pattern characteristics. I Statistical procedures. The final experimental tape consisted of sixty-four voice samples representing what the experimenters had judged to be 32 examples each of efficient 67 and inefficient voice production. However, the panel of judges perceived little difference between eighteen samples ; of nine subjects. Since the present study was based upon difference "as perceived by a panel of authorities," the ; eighteen samples in which little perceptual difference was noted were omitted from the analyses. The remaining 46 samples of efficient and inefficient voice production of i ,23 subjects were treated on the basis of three statistical , instruments. The data were in the form of measures for mean fundamental frequency; mean frequencies of formants 1, 2, and 3; and mean decibel rating for amplitudes of formants 1, 2, | i i and 3. Each sample had a minimum of five measures: (1) | I i fundamental frequency in cycles per second; (2) frequency ! of formant 1 in cycles per second; (3) frequency of formant ! I ^ 12 in cycles per second; (4) mean decibel rating for ampli- 'tude of formant 1; and (5) mean decibel rating for ampliI I tude of formant 2. Most of the samples had two additional ' measures, frequency and amplitude for formant 3. I I Since the data were in the form of mean values and | the primary purpose of the study was to determine differences, the first statistical instrument used sought to | determine the significance of difference between means. i The method used for testing the significance of a differ- 68 . ©nee between the means of two correlated samples was the ' Jb ratio. The formula used was 74 Mx - My t a ------ dmxy , where Mx was the mean for the efficient sample, and My the , i : I mean for the inefficient sample. The obtained t was inter75 ■ preted according to Guilford. In computing the standard ' : I error of the difference, the correlation coefficient des- j I cribed below was employed. i ' When the _t ratio was obtained, inspection of the ; standard deviations often suggested a significant difference between the variability among the two sets of means. ' The statistical instrument used to evaluate such difference: was the F test of difference between variances. The form- , ula used was ! O J ^2. < where Ox represents the larger variance,Oy the smaller | ; variance, and the degrees of freedom for both variances 1 equal 23. Interpretation of the obtained F ratio was based' P. Guilford, Fundamental Statistics in Psycho- ,logy and Education. (New York: McGraw-Hill Book Company, |TnoT, 1950) pp. 214ff. TSibid.. pp. 609ff. j TGlbid.. p. 232. : ' ■ 69 rtrj on the Snedecar Table F, The P ratio equals the larger obtained standard deviation value squared, divided by the smaller standard deviation value squared. If, then, upon .inspection the standard deviation values did not indicate I a difference significantly greater than 2, no F test was j 'applied. In all doubtful cases the P test was used. It was further desired to know if significant re- | lations existed between the two sets of means. Three in- ! struments were considered: the scattergram, the Rank- ; I Difference correlation technique, and the Pearson Product- | ! ! : Moment Method of correlation* The Pearson formula was used! ' I ! for it permitted computation from the original measurements,' I ! hence was the most sensitive of the three instruments. The I I formula used was i V8 I where X * measurements of efficient voice samples, Y « meas-j urements of inefficient voice samples, and N * 23. Interpretation of the Pearson r was based upon the 7Q ' description of coefficients by Guilford. The Pearson r ! 77Ibid.. pp. 612-613. 78lbld.. p. 159. p. 164f. I 70: was not applied to measures for formant 3 of the voice sam-' I ! Iples. This was true because formant 3 was not present in I ' I I all vowels of all samples to permit a valid test of correl-i I i Iation. Only tests of significance of difference were used , in this connection. In summary, forty-seven male subjects were chosen ^ ! who ware judged by the experimenters to have low-pitched | voices that tended under normal speaking conditions to be | efficient. All subjects were instructed to produce three vowel sounds , \a], and ]u] first efficiently then in- i 'efficiently. The basic purpose of the research was to es- ; :tablish parameters of the efficient voice. Inefficiency ! • ■ ■ j I was introduced only as a reference point to facilitate dis-' 'tinguishing the stated characteristics of efficiency. It j 1 I was desired that the physical, physiological, and psychol- , ogical aspects of the efficient voice be determined in re- | lated studies, therefore, x-ray pictures were made simul- i taneously with voice recordings. A total of 388 x-ray i ■ i photos was made with accompanying voice recordings. I ' : The final tape consisted of sixty-four voice samples which equally represented the two voice types as assies sed by the experimenters. However, only forty-six of these samples were perceived as adequately different by a | panel of expert judges who were selected on the basis of ...................... .................... ... ........... ” 7Ï ' above average performance on a test of auditory discrimin- ■ atlon. These samples were used in the present study for | ; analysis by the sound spectrograph. : I Data were in the form of mean values for (1) funda- | | : mental frequency, (2) frequency and decibel rating of am- ' I I ' I ' plitude of formants 1, 2, and 3. Three statistical instru- I i , ^ ! ments were used: t test of significance of difference be- 1 i : tween means, F test of significance of difference between | j standard deviations, and the Pearson Product-Moment coI efficient of correlation. I CHAPTER IV PRESENTATION AND ANALYSIS OF DATA : In Chapter I the problem was broken down into seven i I . I I questions. Five of the inquiries yielded more than one j ; ! I area for discussion. The resultant thirteen questions are to be treated in the present chapter. It was found more convenient and meaningful to the organization of the study to present analyses in reference to subject areas rather than in reference to each of the questions in turn. The 1 I subject areas represented by the questions were (1) funda- | I ! •mental frequencies; (2) frequencies and amplitudes of form-| jant 1; (3) frequencies and amplitudes of formant 2; (4) frequencies and amplitudes of formant 3; and (5) special ! pattern characteristics. I. FUNDAMENTAL FREQUENCIES ! Two questions were asked in regard to the fundament- | al frequencies of test vowels used in the study. The first ; question was. To what extent does the mean fundamental frequency of vowel sounds differ between voices of low-pitched male subjects perceived as efficient and those perceived as inefficient and what is the nature of such differences? Table III presents comparative summaries of the mean 73 TABLE III COMPARATIVE SUMMARIES OF FUNDAMENTAL FREQUENCIES FOR TEST VOll/ELS OF EFFICIENT AND INEFFICIENT VOICE PRODUCTION. (Cycles per second) s : EFFICIENT : INEFFICIENT S t EFFICIENT : INEFFICIENT 1 92 92 13 113 150 2 107 96 14 110 112 3 101 150 15 107 135 4 125 152 16 121 127 5 97 107 17 115 122 6 106 117 18 107 137 7 109 130 19 95 166 8 120 120 20 113 130 9 97 131 21 97 116 10 93 103 22 98 102 11 108 160 23 100 113 12 122 245 W iMeans: 2,453 3,013 EFFICIENT 106.65 INEFFICIENT 131 CT 2.03 <r 6.67 jfc 4 . 04 r .456 ■ ' 74 fundamental frequencies of the test vowels used in the i study. In twenty of the twenty-three subjects the inefficI ient voice samples had a higher mean fundamental frequency I j than did the efficient samples. This finding was of particular interest since considerable effort was expended attempting to get inefficient voice productions that matched the pitch of the efficient productions. Subjects 1 and 8 had the same mean frequency for both samples. Subject 2 had a higher mean frequency for his efficient sample than for his ! inefficient sample. Clearly, the fundamental frequencies of voices produced efficiently seemed to be lower than for the same voices produced inefficiently. However, some voices produced efficiently had higher pitches than others produced inefficiently, so the comparison made was between ' the same voices produced efficiently and inefficiently. It ; may be noted in Table III that the mean fundamental frequency of the test vowels is 106.65 cycles per second (ops) ,when the efficient voice was used, and 131.00 cycles per second (ops) when the voice was used inefficiently. On the ; basis of general inspection a difference of 24.35 cps would appear to be of importance. i t The importance of the difference between the mean I fundamentals of the two voice types was demonstrated when a I test of significance of difference was applied. A t ratio 75 ' of 4*04 was obtained. For the degrees of freedom permitted I by the number of samples used in the study, a jb of 2.819 : ! : would be significant at the .01 level of confidence. The ; obtained value of the jb test of difference between funda- j ! mental frequencies was significant beyond the .01 level, | ' and indicated that the mean fundamental frequencies for efficient and inefficient voice samples were distinct , enough to come from different populations. It will be noted also in Table III that the standard : deviation from the means was 2.03 for the efficient, and I - i 6.67 for the inefficient voice samples. The difference be-! I I j tween the standard deviation scores would appear to be sig-j j ! Inificant upon inspection. But in order to get a more ac- j • i : curate assessment of the difference between the two meas- : ures, an F test of the difference between standard deviaItion scores was made. The result was an P ratio of 10.75 ! which was also significant beyond the .01 level of confid- , I I ©nee « The significant F ratio indicated that there was | more homogeneity among the efficient samples than among the I inefficient samples. This would be expected, for the rel- ;ative stability of the pitch of efficient samples was pro- ' I bably influenced by the specific criteria governing efficient voice production. The efficient voice could be ach- i 76 ieved only one way, while the inefficient was categorized I as any deviation from efficiency* The second question concerning fundamental frequencies was. What correlation exists between the mean fundamental frequencies of vowel sounds which were perceived as efficient and those perceived as inefficient as spoken by the same subjects? The Pearson product-moment formula of correlation was I applied to the frequency measures as obtained from the : spectrograms of spoken vowels used in the study* Table III shows the correlation coefficient to be .456. Since this j obtained coefficient was significantly greater than zero j I at the five per cent level of probability, it was concluded! that a positive, moderate degree of association existed be- ; tween the two variables. They tended to vary interdependently. This may be interpreted another way. Examination of the fundamental frequency of vowel sounds characterized as efficient on the basis of criteria used in this study would provide a moderate degree of predictability in regard • to probable fundamental frequency of the same vowel produced inefficiently. In summary, it may be stated that the difference be- j I tween fundamental frequencies of vowel sounds perceived as efficient and inefficient voice productions was found to be 77 significant. That is, efficient voices had consistently higher pitches than the same voices produced inefficiently. i Analysis of mean fundamental frequencies for each group of I I samples yielded a critical _t ratio of 4.04, and an P ratio j I I I of 10.75. Both values were significant beyond the .01 lev-I el of confidence. A Pearson product-moment correlation coefficient of .456 indicated that a moderate positive correlation existed between the two sets of measures. Variations among the fundamental frequencies of efficient and ineffic-• lent samples were moderately interdependent. A change in fundamental frequency of the efficient voice sample was moderately relatéd to a similar change in fundamental frequency of the corresponding inefficient voice sample. II. FORMANT 1 Four questions were asked concerning the nature of formant 1 in the samples of efficient and inefficient voice used in the study. The first was. To what extent do the : frequencies of formant 1 of vowel sounds differ between .voices of low-pitched male subjects perceived as efficient and those perceived as inefficient and what is the nature of such differences? Table IV presents comparative summaries of mean frequencies for formant 1 of the test vowels used in the study. 78 TABLE IV COMPARATIVE SUMMARIES OF FREQUENCIES FOR FORMANT 1 OF TEST VOWELS OF EFFICIENT AND INEFFICIENT VOICE PRODUCTION. (Cycles per second) s : EFFICIENT : INEFFICIENT S : EFFICIENT : INEFFICIENT 1 440 533 13 448 523 2 492 503 14 437 542 3 453 497 15 460 442 4 487 520 16 460 533 5 513 492 17 453 505 6 440 347 18 460 475 7 490 470 19 430 333 8 438 492 20 467 500 9 508 527 21 477 475 10 447 500 22 475 483 11 483 467 23 453 442 12 555 508 Ts 10,766 11,109 Means: EFFICIENT 468.09 cT 6.21 r t INEFFICIENT 483.00 CT 10.96 .28 1.36 79 It will be noted that the frequency of formant 1 was generally higher when the voice was used inefficiently than whenj it was used efficiently. This followed the same pattern ! as reported in the treatment of fundamental frequencies, ' though not to the same degree. Eight of the twenty-three subjects, slightly more than one-third, produced inefficient samples at a lower formant 1 frequency than that used in efficient voice production. It was necessary to state, nonetheless, that generally, inefficient voice samples were I characterized by higher frequencies of formant 1 than the I i efficient samples. ' I j i I The mean frequency of formant 1 for the efficient I : : samples may be observed in Table IV to be 468.09 cps, and that of the inefficient samples to be 483, The difference of 15 cps would hardly be perceptible as a group characterIistic. Formant 1 has been observed consistently in previous research as having very negligible movement in the I : audio spectra, and does not vary too noticeably within a 'rather restricted frequency band. The negligible differ- ;ence in frequencies of formant l a s found in this study was I in keeping with such findings. The t test for significance of difference was applied to the mean values for formant frequencies. The result was ^included as part of Table IV. The analysis yielded a critical ratio of 1.36. This was not large enough to be signi- 80 ficant at the required .05 or .01 level of confidence. On the basis of the obtained ratio, it may well be that such differences as were observed in the frequencies of formant 1 may be due in part to chance. I Study of the standard deviation scores presented an ^ interesting analysis. Table V shows the standard deviation I of efficient samples to be 6.20, and that of the inefficI ient samples 10.96. The homogeneity of measures for the ! efficient samples and the comparative heterogeneity of ; measures for the inefficient samples remained consistent with previous findings in the present study. The second question asked regarding the nature of formant 1 was. What correlation exists between the frequencies of formant 1 of vowel sounds whioh were perceived as ; efficient and those perceived as inefficient as spoken by ; ' i the same sub jects? The Pearson product-moment formula of correlation I was applied to the frequency measures of formant 1 as obtained from the spectrograms of efficient and inefficient | voice samples. The correlation coefficient was found to be *28. See Table V. This obtained coefficient was not I ‘ sufficiently large to reject the null hypothesis. The low i coefficient indicated that there was little association be- ; tween the frequency of formant 1 when the voice was used | i '____________________ I 81 efficiently and when the same voice was used inefficiently. The frequencies of formant 1 tended to vary independently ! when the two voice types were produced. < The third question pertaining to formant 1 was. To i ; what extent do the mean relative amplitudes of formant 1 of ; vowel sounds differ between voices of low-pitched male subjects perceived as efficient and those perceived as inefficient and what is the nature of such differences? Comparative summaries of mean relative amplitudes of formant 1 in decibels are contained in Table V. Upon inspection it appeared that the decibel ratings for the | ; I amplitudes of formant 1 for both efficient and inefficient ; voice samples ware quite similar. This might have been due to the limited range of the Teledeltos paper on which the spectrographic analysis was made. From the lightest stylus trace to the darkest stylus trace represents a range of ; I 135 decibels. The range of speech is potentially much great-- 1er. This would tend to suggest that a minor difference in I decibel rating for the sets of measures might be signifijcant. This was later found to be the case. In thirteen of the twenty-three cases the mean relative amplitude of formant 1 of the efficient samples was greater than that of the inefficient samples; in two cases it was the same; and in the remaining eight cases the mean 82 TABLE V COMPARATIVE SUMMARIES OF RELATIVE AMPLITUDES OF FORMANT 1 FOR TEST VOWELS OF EFFICIENT AND INEFFICIENT VOICE PRODUCTION (Decibels) s : EFFICIENT ; INEFFICIENT S : EFFICIENT : INEFFICIENT 1 42.5 39.2 13 41.8 41.5 2 42.6 40.8 14 41 41.8 3 41 41 15 42.1 35.8 4 41.3 42.3 16 41.3 42.3 5 42.1 42 17 41.5 41.7 6 42.3 42.1 18 40.8 42.0 7 41.8 42.5 19 40.8 40.7 8 42.6 42.6 20 43.0 42.0 9 42.3 42.2 21 42.6 41.5 10 42.1 40.5 22 42.1 41.5 11 42 42.1 23 42.2 42.5 12 42.2 42 Ts 963.8 950.6 Means; EFFICIENT 41.904 ^ .13 INEFFICIENT 41,33 CT .307 r .00786 t 3.54 85 relative amplitude was greater in the inefficient samples than in the efficient samples. Since more than half the | 1 total number of efficient samples had greater decibel rat- ; I ' j I ings for the mean relative amplitude of formant 1, a gener-l I alization that the efficient samples tended to have a greatf I : ; er amplitude than the inefficient samples seemed in order. : The mean relative amplitude for formant 1 of the efficient samples was 41.904, and for the inefficient sam- , pies, 41.33. A t test of significance of difference be- ; tween means was applied. The results indicated a critical I 1 i I ratio of 3.54 which is significant at the .05 level of con-| ! i j fidence. ; I The standard deviation scores were submitted to the | I F test for significance of difference. It was found that a ! ratio of 5.58 existed between them. This obtained value I was significant at the .01 level of confidence. It indi- 'cated that there was more variability among the decibel jratings for amplitudes of formant 1 in samples of ineffici Iient voice production than among the decibel ratings for the I I jsamples of efficient voice production. The fourth question asked in relation to formant 1 was. What correlation exists between the mean relative am- 'plitudes of formant 1 of vowel sounds which were perceived : as efficient and those perceived as inefficient as spoken 1 'by the same sub jects? 84 The Pearson product-moment formula of correlation ; was applied to the decibel ratings for mean amplitude of formant 1 as taken from the spectrograms of test vowels : used in the study. The result as included in Table V yielded a coefficient of .00786. Since this obtained coefficient was not significantly greater than zero, it was assumed that negligible association existed between the two variables. Decibel ratings of formant 1 of efficient and inefficient voice samples tended to vary independently. ! ! : The nature of formant 1 may be summarized as fol- |lows. No significant differences were found between the i frequencies for formant 1 when the test vowels used in the I study were produced efficiently and inefficiently. Analysis revealed a t ratio of 1.36. The lack of significant differences between mean frequencies for formant 1 of the two voice types represented in the study was consonant with findings of other researchers who have reported the relat- 'ively restricted movement of formant 1 in the spectra from signal to signal. A significant difference was found between the mean decibel ratings for relative amplitudes cf formant 1. Efficient samples had significantly more energy in the formant than corresponding inefficient samples. This was supported by a t ratio of 3.54. The difference between mean decibel 85: ratings of the formant was further supported by an F ratio of 5.58 which resulted from analysis of standard deviation : scores. ^ ; I The correlation coefficient was not significantly I I I I different from zero. The obtained coefficient was .00786. ! I ! I I III. FORMANT 2 i Four questions were asked concerning the nature of formant 2 in samples of efficient and inefficient voice production of test vowels used in the study. The first ^ i 'question was. To what extent do the frequencies of formant | ; 2 £f vowel sounds differ between voices of low-pitched male j subjects perceived as efficient and those perceived as in- : : efficient and what is the nature of such differences? : Table VI presents comparative summaries of mean frequencies for formant 2 of the test vowels used in the study. In fifteen of the twenty-three cases formant 2 of the inI i I efficient samples had a higher frequency than formant 2 of : I I the corresponding efficient samples. In the remaining 'eight cases the frequency of formant 2 of efficient samples was higher than that of the corresponding inefficient samiples. The generalization that a higher formant 2 tended to 'distinguish the Inefficient samples from the efficient sam- ,ples was consistent with findings in relation to fundamentI al frequency and frequency of formant 1 already reported in 86 TABLE VI COMPARATIVE SUMMARIES OP FREQUENCIES FOR FORMANT 2 OF TEST VOWELS OF EFFICIENT AND INEFFICIENT VOICE PRODUCTION (Cycles per second) s : EFFICIENT : INEFFICIENT S : EFFICIENT : INEFFICIENT 1 1407 1580 13 1280 1597 2 1356 1460 14 1405 1700 3 1323 1390 15 1525 1523 4 1362 1353 16 1430 1830 5 1322 1283 17 1493 1567 6 1293 1508 18 1322 1325 7 1377 1536 19 1327 1190 8 1407 1612 20 1610 1470 9 1547 1578 21 1389 1437 10 1293 1300 22 1563 1338 11 1520 1413 23 1343 1293 12 1630 1667 Means EFFICIENT 1414.09 cT 22.15 r t INEFFICIENT 1476.09 cT 35.77 .29 1.71 87 ' this study. I The mean frequency of formant 2 for the efficient samples was 1414.09, and for the inefficient samples, ; ; 1476.09. The difference between the means did not appear i : j j significant from inspection. In order to get a more accur-i ate and reliable assessment of the difference between the means, the t test of significance of difference between means was applied. The result was also included as part of Table VI, The difference between the mean frequencies of formant 2 for the samples of the two voice types was char- , ; acterized by a ^ ratio of 1.71. On the basis of the degrees of freedom permitted by the number of subjects used I . in the study, the obtained value of t did not indicate a . significance at the required levels of confidence. The significance of difference between the mean fre- 'quencies of formant 2 for efficient and inefficient voice samples represented an interesting point in the present i ianalyses. This was due to the general character of formant ‘2 as reported in the literature, and as observed experimentally. Dunn^^ and Potter and Peterson^^ were cited in i Chapter II for their contributions to establishing the im1 portance of formant 2. Its importance was pointed out on 65 Dunn, 0 £. cit., pp. 740-753. 66Potter and Peterson, 0 £. cit., pp. 528. 88, the basis of its wide range of movement in vowel spectra, and the fact that the best one-formant approximations of | ! : ■ i I spoken vowels in experimental work with synthetic vowels i i ! i was often observed to be a single formant near the normal | ! i ; position for formant .2. Apparently, a change in voice | quality was not a significant function of the frequency of formant 2 which has been found in other studies to be an ; important determinant in the perception of vowels. Table VI also shows that the standard deviation scores for frequency of formant 2 were 22.15 and 35.77 for ! the efficient samples and inefficient samples respectively. I I An F test was made of these scores. The analysis yielded i an F ratio of 2.87 which was significant at the .05 level .of confidence. Again, more variability was found among the inefficient samples than among the efficient samples although no significant difference was observed between jmeans. j The second question asked regarding formant 2 was 'What correlation exists between the frequencies of formant :_2 of vowel sounds which were preceived as efficient and I I those perceived as inefficient as spoken by the same sub- 'jects? The Pearson product-moment correlation formula was ! applied to the frequency measures of formant 2 as obtained 89 from spectrograms of efficient and inefficient voice samples nsed in the study. The correlation coefficient was ^ found to be .29. The obtained coefficient was not signifi- ; cantly greater than zero, which indicated that little re- , lation existed between the two variables. The frequencies i of formant 2 for efficient and inefficient voice samples in the study tended to vary independently. The third question asked relative to formant 2 was. To what extent do mean relative amplitudes of formant 2 of ; vowel sounds differ between voices of low-pitched male sub- ; jocts perceived as efficient and those perceived as ineffi- : ; ! ‘d e n t and what is the nature of such differences? j ’ Table VII presents comparative summaries of mean 1 relative amplitudes for formant.2 of the test vowels used in the study. Fifteen of the twenty-three subjects had higher decibel ratings for peak amplitude of inefficient samples than for peak amplitude of their corresponding effijCient samples. The increase in amplitude of formant 2 app- ; eared to vary concomitantly with the increase in frequency, ; which tended to follow findings of previous studies involving frequency and intensity variables. The mean decibel rating for amplitude of formant 2 ; of efficient samples was 39.28, and for inefficient samples,, 39.21. A t test of significance of difference between means TABLE VII 90 COMPARATIVE SUMMARIES OF RELATIVE AMPLITUDES FOR FORMANT 2 OF TEST VOWELS OF EFFICIENT AND INEFFICIENT VOICE PRODUCTION (Decibel) s 2 EFFICIENT : INEFFICIENT S 2 EFFICIENT 2 INEFFICIENT 1 38.8 38.3 13 39.6 41.5 2 41.5 36.7 14 37.8 41.8 3 37.0 37.2 15 39.3 35.5 4 37.6 39.2 16 39.1 41.1 5 41.0 38.3 17 37.0 38.3 6 37.5 41.6 18 41.3 42.0 7 41.3 40.0 19 39.3 39.7 8 42.1 41.6 20 40.3 34.4 9 38.0 40.2 21 41.0 41.3 10 35.5 38.8 22 36.0 37.5 11 40.0 40.6 23 41.3 36 12 41.3 40.3 Ts Means : EFFICIENT cr 39.28 .405 903.5 INEFFICIENT <f 901.9 39.21 .463 1 r .114 .0083 91 , resulted in a critical ratio of .114. An F test of significance of difference between standard deviation scores > obtained an P ratio of 1.011. In both tests the ratios I were not significant at the required levels of confidence, i ! I I The fourth question asked concerning formant 2 was, i What correlation exists between the mean relative ampli- > i tudes of formant 2 of vowel sounds which were perceived as efficient and those perceived as inefficient as spoken by the same sub jeots? ! ' The Pearson product-mement correlation formula was | i ' ! applied to the mean decibel ratings of relative amplitudes ; I ■ I of formant 2 for efficient and inefficient voice samples | ; used in the study. The analysis yielded a correla tion co- i efficient of .0083. Since the obtained coefficient was not significantly different from zero, it was assumed that no relation existed between the mean decibel ratings of formant 2 for efficient and inefficient voice samples. The Imean decibel ratings tended to vary independently as the 'two voice types were produced. In summary, no significant difference was found between mean frequencies for formant 2 of the samples of efficient and inefficient voice production. Likewise, no significant difference was observed between the mean decibel ratings of peak amplitudes for formant 2 of the same sam- 92 pies. Analyses of the standard deviations of the means resuited in an P ratio of 2.87 for frequency which was significant at the .05 level of confidence, and an F ratio of I 1.011 for decibel ratings which was not significant at the I required levels of confidence. I Application of the Pearson product-moment correlation formula to mean decibel ratings of amplitude of formant 2 revealed a coefficient of .0083. The low obtained coefficient indicated that the mean decibel ratings between the efficient and inefficient voice samples were unrelated. The correlation coefficient of .29 for mean fre- ! ■; quencies of formant 2 was also too low to indicate an interdependence between these measures. IV. FORMANT 3 Formant 3 appeared in the spectra of most of the voice samples, but not in all vowel sounds of the samples I with consistency. It was present in at least one of the ' three vowels in all of the efficient samples, and in at I ileast one of the test vowels in twenty-one of the inefficient samples. Variation in occurence of formant 3 among the I vowels of both voice samples was parallel in only seven of i the twenty-three cases, and in only five cases did the formant appear in all vowels of all samples. This inoonsist- 93 TABLE VIII COMPARATIVE SUMMARIES OP FREQUENCIES FOR FORMANT 3 OF TEST VOWELS OF EFFICIENT AND INEFFICIENT VOICE PRODUCTION. (Cycles per sedond) s : EFFICIENT : INEFFICIENT S : EFFICIENT t INEFFICIENT 1 2340 2547 13 2600 2833 2 2433 2065 14 2237 2732 i ^ 2000 2033 15 2425 3000 i 4 2853 1150 16 3050 2526 i s 2562 2700 17 2787 2313 ! 6 2195 2895 18 1998 2550 7 2610 19 2125 1775 8 1948 2525 20 3060 1800 ■ff 9 2092 3083 ^21 2430 2413 i&O 1893 2707 22 3125 1675 11 2823 23 1383 1900 12 3660 2600 Ts 55,929 49,821 Means ; EFFICIENT 2431.70 INEFFICIENT 2372.43 t ,371 {("Formant 3 appeared in all vowels of efficient and inefficient samples. 94 ency made correlation invalid as an instrument of analysis for treatment of formant 3. Thus, only differences were considered. Two questions were asked relative to formant 3. The first inquiry was, To what extent do the frequencies of formant 3 of vowel sounds differ between voices of lowpitched male sub.jects perceived as efficient and those perI QQivQd as inefficient and what is the nature of such dif- ^ Terences? Table VIII presents comparative summaries of ■ mean frequencies for formant 3 of the test vowels used in I the study. Formant 3 appeared in all three vowels of both I efficient and inefficient samples of only five of the I twenty-three subjects. These were indicated in Table VIII by asterisks. Wo valid generalization can be made about ; the comparative frequencies of the remaining sixteen subjects. For the five cases where formant 3 was consistently present, the mean formant frequency for the efficient samples was 2097.6 cps, and for the inefficient samples, 2563 cps. This represented the greatest difference in formant ■frequencies for the two voice types of the three formants treated in this study. It followed the general pattern of higher fundamental and formant frequencies for the ineffiIcient voice, but to a much greater extent. 95 Because the voice samples were assessed as a single perceptual entity, and not on the basis of three individual i vowel sounds, it was thought that presenting the mean fre- : I I I quency for formant 3 in those vowels in which it occurred | I would be interesting even if its occurrence was so erratic | as to make meaningful statistical treatment very limited. | Table VIII shows the mean frequencies for formant 3 was 2431.70 cps for the efficient voice samples, and 2372.43 for the inefficient voice samples. However, the negligible, ! difference may be misleading. I j Whenever a single formant appeared only once in the | I i : spectrum in a high position as in |^i] , or in a low posi- | tion as in Ja], the resultant mean for the entire sample was expected to be biased; for the mean difference between their normal position was estimated to be about 1000 cps. So, if the only formant 3 in the sample was present at 3500 cps in [i] it appeared in the table as such. If the I only formant 3 in the sample was present at 2500 cps in [a] it appeared the same in the table. Thus, the mean frequency of formant 3 as included in Table VIII was considered to be a gross approximation at best. The difference between an estimate of the true mean frequencies of formant 3 of efficient and inefficient samples was thought to exist somewhere between the mean frequency represented by the small 96 group of five subjects and the larger group of eighteen. 'It would be consistent with caution to suspect it would lie : nearer the smaller group. : ' I A t test of significance of difference was applied | i I i to the means for the formant in all cases where it occured I I I I at all. This represented the twenty-three cases of effic- ■ lent voice production, and twenty-one of the inefficient ' voice samples. A critical ratio of .371 was obtained. This low t ratio was not significant at the required level 1 of confidence. Such finding was to be expected in light j I I ! of the preceding discussion. ' j i The second question asked in relation to formant 3 I I was. To what extent does the mean relative amplitudes of i ! formant 3 vowel sounds differ between voices of lowpitched male sub jects perceived a s efficient and those perceived as inefficient and what is the nature of such dif- ; : I Iferenoes? ' ' Table IX presents comparative summaries of mean rel- ' 1 : lative amplitude ratings in decibels for formant 3 of the I ■test vowels used in the study. The five subjects whose ! samples contained formant 3 in all three vowel sounds of both voice types are noted by asterisks as in Table VIII. The same generalization relative to the inconsistency of ithe presence of formant 3 in the spectra held in this por- 97 TABLE IX COMPARATIVE SUMMARIES OP RELATIVE AMPLITUDES FOR FORMANT 3 OF TEST VOWELS OF EFFICIENT AND INEFFICIENT VOICE PRODUCTION. (Decibel) s ; EFFICIENT : INEFFICIENT S : EFFICIENT ; INEFFICIENT 1 32.3 37 13 35.7 34.5 ^ 2 34.6 33.5 14 33.5 35.5 3 36 32.6 15 25.6 28.2 4 32.8 42.5 16 34.5 41 5 29.7 32.5 17 26 37.5 6 37.5 36.5 18 38.6 39 7 41.5 19 32.4 36.4 8 38.1 33.5 20 37.5 38 9 51.3 29 *21 35.5 30.5 *10 28.8 27.8 22 29.5 31.3 11 30 23 35.2 36 12 32 30 Ts Means ; EFFICIENT 33.42 6 .566 768.6 722.8 INEFFICIENT 34.42 6 .893 .847 ^Formant 3 appeared in all vowels of efficient and inefficient samples. 98 tion of the analysis. However, the means for the small group were not as distinctly different from those of the i eighteen subjects who did not produce matched samples in ) ; regard to presence of formant 3. This was thought due in | ' i I part to limitations of the Sona-graph as related to ampli- | I , tude section analysis, ! The mean decibel rating of relative amplitude of formant 3 of efficient voice samples wherever it appeared was 33.42, and for the inefficient samples, 34.42; a difference of 1,0 decibels. In the smaller group where samples were matched relative to occurrence of formant 3, the mean . I ' ' I decibel rating for the efficient samples was 34.14, and for; I the inefficient samples, 33.56; a difference of .58 decibels. In the latter case the difference was not thought important. Generally, formant 3 of inefficient voice had a higher decibel rating for relative amplitude than that of efficient voice. A t test of significance of difference between the mean decibel ratings for relative amplitudes of formant 3 of the test vowels used in the study was made. A critical ratio of .847 resulted. This obtained t was not significant at the required level of confidence. An F test of the significance of difference between standard deviation scores was also made. It was found that 99 this difference was represented by an P ratio of 2.26 whichi is significant at the .05 per cent level of confidence, ' Again, it was indicated that more variability existed among i the inefficient samples than among the efficient samples. = j ; I In summary, it was discovered that formant 3 was i i I : not present in all vowels of all samples of the twenty- : three subjects used in the study. It was present in at least one vowel of all efficient samples, and in twenty-one of the inefficient samples. Formant 5 was present in all vowels of only five of the twenty-three subjects. , A negligible difference was found between the mean I frequencies for the efficient and inefficient samples when : formant 3 was considered even though it existed in only a portion of the voice sample. When the sanpies were matched for occurrence of formant 3, the difference between formant , frequencies was found greater than between frequencies for formants 1 or 2 of the two voice types. ; The mean decibel rating of relative amplitude of formant 3 was not found to be significantly different between the efficient and inefficient voice samples in cases where it was present in only a portion of the voice samples, nor where it was present in all of the samples alike. The difference of 1.0 decibel between the decibel rating of formant 3 of the two voice types tended to suggest the pre- 100 senoe of more energy among the higher formants of ineffic- ; ient voice samples than among corresponding higher formants : of efficient voice samples. A significant P ratio of 2.26 indicated that there j j I ' was more variability among the inefficient samples than i ■ ! among the efficient samples. This finding was consistent with previous findings already reported in this study. V. SPECIAL PATTERN CHARACTERISTICS ; One question was asked in relation to special char- | i I ^ acteristics of acoustic patterns. It was. To what extent 1 i ^ are spectra of samples of efficient voice production descriptively different from those of inefficient voice production on the basis of special acoustic pattern characteristics? After recording quantitative data for statistical treatment, it was discovered that certain visual clues were present in many sonagrams which did not fit into any of the i quantitative categories of the study. These characteristics were not always directly related to the obtained measures, and were thought to warrant separate treatment in the study. That is to say, they were not always a function of frequency or amplitude. The special pattern characteristics were conveniently divided into four categories; (1) 101, I weak harmonic registration; (2) erratic harmonic registration; (3) abnormal pitch change; and (4) frequency of occurrence of higher formants. These characteristics were sum- : ' ! imarized in Tables X and XI. ! Weak harmonic registration. Many of the spectro- ; grams had noticeably weak harmonic structure, especially in ' the high frequency regions. While it is true that vowel sounds such as [a] and (u] normally have comparatively weakIer intensities in the high frequency range of the spectra j jthan does [i], for example, inefficient voice samples were | i ■ I jconsistently associated with weaker harmonic registration of Ihigher harmonics than were corresponding samples of effic- j ient voice. This factor was observed as indicated in Figure 5. The corresponding efficient sample is contained in Figure 6. Erratic harmonic registration. It was further ob- ! i served that frequently in spectra of both efficient and in- I [efficient voice samples, gaps appeared in the vertical dimlension of the amplitude section analysis. This could not be accounted for in relation to either of the two dimensions analyzed, frequency and amplitude. It was thought that stylus pressure might be a causative factor, but readjustment of stylus contact on the Teledeltos paper did not cor- TABLE X 102 SUMMARY OP DATA ON SPECIAL CHARACTERISTICS OF EFFICIENT SONAGRAMS s 1 O 03 •H ^ a -p c o ra o g ^ »H Jh ejû-p / CO ® / W « n & •P CO / «H 0H / Ai 44 ra O -P c ?4 CO J * S 1 O CO •H JL| c-p n o CO O e ^ *H U Î30-P y CO © c O -P -p © / 44 ra O -P C ® a / I I / ^ 1 4 13 E 4 2 E 4 14 WE 5 3 3 16 E 3 4 WE 4 16 E p 4 5 E P 4 17 E 4 6 E P 4 18 4 7 P 5 19 E 4 8 WE P 5 20 3 9 4 21 4 10 WE 6 22 WE 4 11 E 3 23 5 12 WE P 3 'W - Weak Harmonie Registration E - Erratic Harmonic Registration ip - Pitch Pattern TABLE XI SUMMARY OP DATA ON SPECIAL CHARACTERISTICS OP INEFFICIENT SONAGRAMS 103 s 1 o © •H Pi C P C O 03 O g *H vH U bOP , CS <D / w cu c Pi0 xi p Ü p p © / «HAi L (p 03 O P C Pi © © g . . . . . S 1 o © •H Pi C P c o © O g »H «H Pi W P y © © c c © xi p O P p © , 44 Ai _ 4-1 W O P C Pi © y . . . . . . . . . . . . . . . . 1 E 3 13 p 4 2 E 4 14 WE 4 3 P 7 15 W 3 4 W P 5 16 p 3 5 W 3 17 p 5 6 E 4 18 WE 3 7 WE P 2 19 E p 3 8 E P 3 20 m 3 9 E 4 21 WE 3 10 E 3 22 E 3 11 W P 2 23 WE 3 12 P 3 W - Weak Harmonic Registration E - Erratic Harmonic Registration P - Pitch Pattern 104 4000- Weak 3000 Harmonic Structure 4000 UL Weak Earmcnic Struc iVeak Harmonic Structure 3000- 2000. PIGURE 5 SAMPLE SPECTROGRAM: WEAK HASNOaiO STRUCTURE IB INEFFICIENT VOICE SAMPLE 105 4000 _ 3000 — 2000 ___ Strong li armor lo Structure LL Strong Harmonic Structure (Cf. Corresponding sample in preceding Figure) tn.- 1000 _ FIGURE 6 SAMPLE SPECTROGRAM: HARMONIC STRUCTURE IB CORBESPOBDIBG EFFICIENT VOICE S 106 rect the condition. The gaps in harmonic structure were tantamount to harmonics being missing without apparent cause. Although this phenomenon occurred in samples of both efficient and inefficient voice production, it was more frequently observed in inefficient samples. It proved to be a consistent clue to identifying inefficient samples by inspection, %here erratic harmonic registration occurred in efficient voice samples it was not as frequent nor as severe. A comparative view of this condition is found in Figures 7 and 8. Pitch change. One of the most readily observed characteristics of certain sonagrams was abnormal change in pitch. In one instance the subject changed from a high of 245 cps to about 120 cps in one vowel sound, a full octave. The immediate question at this point was. How did the judges tend to rate voice samples characterized by such abnormal pitch change? It was discovered that in the two most severe cases all seven judges agreed that the voice samples were inefficient, and in moderate cases there was general agreement on inefficiency. See Figure 9 for illustrative sonagram of this category. Occurrence of higher formants. It was noticed that many vowel samples contained more than the two essential formants. Formant 5 appeared with regularity, though it was 107 4L UL 1000 __ 2000 _ 4000 _ 4 0 0 0 __ 3000 Erratic Harmonic Registration 2000 __ 1000 Erratic Harmonic Registration f FIGURE 7 SAMPLE SPECTROGRAM: ERRATIC BARMOBIO STRUCTURE IB INEFFICIENT VOICE SAMPLE 108 OL u. 4000 __ 3000 2000 fr 1000 _ UL^BgKHepeaSBI 2000 3000 4000 Regular Harmonic Registration Regular Harmonic Registration (Of. Corresponding sample in preceding Figure) FIGURE 8 SAMPLE SPECTROGRAM: EARMOBIC STRUCTURE IE CORRESPOEDIEG EFFICIENT VOICE SAMPLE. SEE FIGURE 7* 109 4000 IPOO 2000 _S Sharp Change of Pitch IQPO _ è % FIGURE 9 SAMPLE SPECTROGRAM: ILLUSTRATION OF INTONATION PATTERN 110 not present in all vowels consistently. In thirteen of the twenty-three samples of efficient voice, formant 4 was present. In only five of the inefficient samples did formant 4 appear. Of the twelve efficient samples receiving 100 per cent agreement from the judges, eleven had formant 4, and one had only three formants. Of the eighteen inefficient samples receiving 100 per cent agreement from the judges, only six had formant 4, one had only two formants and the remainder had three formants. Formant 5 was present in five of the twenty-three cases of efficient voice; while it appeared in only three cases among the same number of inefficient samples. The highest number of formants was observed in a sample of inefficient voice. Sample number 44 had six formants in [u] and seven formants in \[a]. This was the voice of a trained singer. Judges were 71 per cent in agreement that the sample was inefficient. Figure 10 illustrates a multi-formant voice sample. This spectrogram represents the voice of a trained singer as in the previous case, but it was judged to be efficient. Representative sonagrams of both voice types on the basis of judges* assessments and spectrographic findings are found in Figures 11 and 12. The acoustical discoveries tended to support the perceptual evaluations vary closely. The spectrogram of sample number Ill 100 2000, 3000. 4000 — 4000 3000 _* 2000 _ 1000 _ 0 Formant 1 Formant 2 formant 3 Formant 4 Formant 4 Formant 3 Formant 2 Formant 1 LL Formant 1 Formant 2 Formant 3 formant 4 Formant 5 Formant 1 Formant 2 Formant 3 Formant 3 — /" Formant 4 Formant 3 Formant 2 Formant 2 Formant 1 FIGURE 10 SAMPLE SPECTROGRAM: ILLUSTRATION OF MULTI-FORMANT VOICE SAMPLE 112 1000 2000 3000 4000 _ 4000 Formant 1 Formant 2 Formant 3 Formant 4 Formant 1 Formant 2 Formant 3 Formant 4 Formant 4 Formant 1 Formant 2 Formant 3 Formant 4 Formant 4 3000 Formamt^y^l^ Formant 3 Formant 3 Formant 2 2000 Formant 2 Formant 2 1000 Formant 1 ormant i jHMüiaM nq. t ormaïll FIGURE 11 REPRESENTATIVE SPECTROGRAM OF EFFICIENT VOICE AS ASSESSED 3T JUDGING PANEL 113 1000 2000 3000 4000 4000 Formant 1 Formant 2 (L Formant 1 Formant 2 Formant 3 Formant 1 Formant 2 3000 — 2000 __ logo _ FIGURE 12 REPRESENTATIVE SPECTROGRAM OF INEFFICIENT VOICE jASI JLS8ŒSI5BD ]3T JtSDGIIB IRAJMEI, 114 ! 55, for example, is representative of both significant per- , ceptual difference and acoustical difference. ; In summary, four descriptive categories were struct-i lured for consideration of special pattern characteristics: I I(1) weak harmonic registration; (2) erratic harmonic registration;' (3) abnormal pitch change, or intonation pattern; and (4) frequency of occurrence of higher formants. It was discovered that acoustic patterns characterized by weak and erratic harmonic registration and abnormal pitch change I were consistently related to voice samples perceptually ! I I jjudged to be inefficient. Though erratic and weak regis- ; * j itration were not infrequently observed in patterns of effi- | i i cient samples, they were much less severe. Higher formants ■ were present in efficient samples to a much higher degree than in inefficient samples. However, the greatest number of formants appeared in the spectra of an inefficient voice I sample. I Generalizations in this section held also for the jfour cases of efficient voice production and the five cases lof inefficient voice production on which the judges perceived little difference. I I VI. DETECTION OP EFFICIENCY BY ACOUSTIC CHARACTERISTICS Since this study developed out of a clinical need for 115 better understanding of voice production, the concern must eventually return to the individual. Therefore, Table XII , : was designed to summarize the data on each subject^s effi- I ' , 1 I cient and inefficient voice production to determine by ob- ; I ! servation if any acoustic characteristic or combination of i characteristics could consistently reveal individual vocal efficiency. Specifically, the question was whether or not the dominant group characteristics of efficiency could be utilized to recognize vocal efficiency in the individual by : spectrum analysis. That is, were lower fundamental fre- ; i I Î quency, greater relative amplitude of formant 1, lesser ! i ! relative amplitude of formant 3, stronger regular distinct | : harmonic registration, less extreme pitch changes, or more of formants 4 or 5 distinguishing clues in the differentiation of efficiency from inefficiency in any one subject? Inspection of Table XII shows that with but one exception a lower fundamental frequency distinguished each isubject*s efficient from his inefficient voice. Although I two subjects had the same pitch for both types of voice, ■ still a relatively lower fundamental for efficiency was the most reliable predictor of efficiency in the study. This is not a surprising finding since inefficient production invariably required more vocal tension by the subjects than did efficient production. lABLE SU ACOUSTIC CmmCTEEISTICS OF SPFICIEUT AI® IIEFFICIIHT VOICE PEODUCTIOU SAMPLES FOR EACH SUBJECT 116 4^ ÿ-i •OS 0 0) I Î I I / g A / I t & •s , l f / I t m k, I I / m 0 sS -H j ^ < ! 1 I / i / J O S 0 A *H 0 'H Si ri fi a 0 a) s4 -t> 0 ai PI "3 ,lsl E ■ 92 hhQj 4 2 .3 1407 *234 o' * 32.3 X 1 I 92 ,'j3 3 . _ 39n.2 . 1580 38.3 *2547 3 ^ 0 X E 107 492 4 ^ 6 13.56. 4 1 .5 *2433 *3 4 .6 X X 2 I 96 503 4 0 .8 1460 3 6 .7 *2065 • 33.5 X X S 101 453 4 1 .0 1323. 3 7 .0 2000 36.0 3 I _ 150 497 4 1 .0 1390 .......3.7.2 2033 3 2.6 . X X X X X E 125 487 4 1 .3 1362 3 7 .6 2853 32.8 X X X A I _ 152 520 4 2 .3 ■ 1353 3 9.2 1150 42.5 X X % X E 97 513 4 2 .1 1322 41.0 2562 29.7 X X X 5 I 107 492 4 2 .0 1283 3 8 .3 2700 32.5 X 1 106 440 4 2 .3 1293 37.5. . 2195 3 7 .5 X X X ° I . 117 347 .4 2 .1 1508 41.6 2895 3 6.5 X X E 109 490 4 1 .8 1377 4 1 .3 2610 4 1 .5 X X X 7 I 130 470 _ 4 & ^ l # 6 4 0 .0 X X X E 120 438 4 2 .6 1407. 4 2.1 1948 38.1 X X X X X I 120 492 4 2 .6 1612 4 1 .6 2525 33.5 X X E 97 508 4 2 .3 1547 38.0 *2092 *31.3 X 9 I 131 527 4 2 .2 1578 4 0 .2 *3083 *29.0 X X 10 E 93 447 4 2.1 1293 3.5..5_ *1893 *28.8 X X X X X I _ .103. 500 4 0 .5 1300. 38.8 *2707 *2 7 .8 X 11 E 108 483 4 2 .0 1520 4 0.0 2823 3 0.0 .. X I 160 467 42.1 1413 4 0 .6 X X 12 E 122 555 4 2 .2 163.0. 41.3 -3660 32.0 X X X I 2k5 .__.5 0 8 ._ 4 ^ 0 1667 4 0.3 2600 3 0.0 X 13 E 113 448 4 1 .8 1280 39.5 2600 3 5.7 . X X I 150 ._ .,:5 2 i. 41.5 1597 41.5 2833 3 4 .5 ........ X X lA B 110 .... .43.7. _ 4 l *0 1405 37.8 2237 3 3 .5 X X X I 112 542 4 1 .8 1700 41.8 2732 3 5.5 X X X 13 E 107 450 4 2.1 1525 39.3 2425 25.6 X I 135 442 3 5 .8 1523 35.5 3000 28.2 X 16 E 121 460 41.-3 1430 39.1 3.050 3 4.5 X X X I 127 ■533 4 2 .3 1830 41.1 2525 41.0 X 17 E 115 453 4 1 .5 149.3... 37.0 2787 2 6.0 X X I 122 .. 505 _ 4 1 .7 38.3 2313 37 .i X A X 18 E 107 460 40.8 1322 4 1 .3 1998 38.6 X I 137 475 4 2 .0 1325 42.0 2550 3^^ . X X 19 E _ . _95. ... 430 4 0 .8 1327 _39.3 2125 3 2.4 X X I l 66 333 4 0 .7 1190 3 9.7 1775 36 X X 20 1 113 467 43:0 1610 40.3 3060 3 7 .5 I 130 500 4 2 .0 1470 34.4 1800 3 8.0 X X 21 E 97 477 4 2 .6 _ 1339.. 4 1 .0 *2430 *35.5__ _ X I 116 475 4 1 .5 1437 4 1 .3 *2413 *30.5 X X 22 E 98 . 475. _ 42.1 1 # 3 36.0 3125 29.5 X X X I ■ 102 483 4 1 .5 1338 3^5 1675 31.3 X 23 E 100 453 4 2 .2 .1.343. 41.3 1383 3 5.2 X X I 113 442 4 ^ ^ .. . 1 2 2 1 .,,. 36,0_ 1900 36.0 , I X ^Form ant 3 appeared i n a l l vow els o f th e sample. X - The a c o u s tic c h e r s .c te r is tic was p re s e n t.* 117 Table XII also reveals that aside from fundamental frequency no other acoustical characteristic could be used i singly as a determinant of vocal efficiency. However, it ; ! i j was found that efficiency could be frequently differentia- | ted from inefficiency in a subject's voice by examining the relationships among relative amplitude of formant 1 and the presence of formants 3 and 4. The relative amplitude of formant 3 and the special pattern characteristics of harmonic registration, extreme pitch change, and the occurrence of formants 5, 6, and 7 were such inconsistent char- ■ Iacteristics as to be of little help in differentiating one ; I ' I voice type from another. | One pattern of efficiency was found when the relative amplitude of formant 1 for efficient voice excelled that ,for inefficient voice and formant 4 was present only in the efficient sample. This pattern was found in only four instances but variations on it were numerous. When formant 1 I relative amplitudes were approximately equal the presence 'of formants 3 or 4 only in efficient samples indicated effioieney in five instances; when formant 1 efficient ampli- !tudes were less and formants 3 or 4 were present only in ! : : efficient samples three efficient voices were distinguished. Another variation around the same pattern was found when formant 1 relative amplitudes were greater for efficiency 118 and formants 3 and 4 were present or missing in both voice types. This variation accounted for efficiency in three cases. The remaining eight subjects did not show the above; patterns, but they could be differentiated since the funda-l I i I mental frequency of efficient production was lower in all | cases than it was for inefficient production. However, three of the subjects did not show a great difference. Apparently, then, at least three variables operated to distinguish the two types of voice production. Whenever for the efficient voice the fundamental frequency was rel- | : atively low, the relative amplitude of formant 1 was rel- | i atively high, and formants 3 or 4 were present only for the; : efficient sample, a clear distinction existed between the two voice types. Furthermore, the distinction did not seem to depend on all three characteristics occurring together. If the formant amplitude pattern was as stated above, the I fundamental frequency pattern could apparently be reversed ! without altering the distinction; if the fundamental frequency and formants 3 and 4 pattern remained, the amplitude pattern could be reversed; and if the fundamental frequency pattern was present, then both the amplitude and formants 3 and 4 pattern could be reversed and efficiency and inefficiency could still be discriminated. CHAPTER V ' SUMMARY AND CONCLUSIONS I ■ I I ' I I. SUMMARY I I I I1 I The general purpose of the study was to determine ifi I ' j certain acoustical bases existed for the perceived differ- ' ences between efficient and inefficient voice production as; I I analyzed by the sound spectrograph* The hypothesis tested ^ I was that the perceived differences between efficient and ! I : I I i inefficient voice production can be specified by quantify- j ing (1) fundamental and formant frequencies of selected vowel sounds, and (2) relative amplitudes of the formants of these vowels* j Three important points recommended the study. Firstj ■ the concept of vocal efficiency has implications for a ' ; practicable voice therapy goal; second, a study of voice types by spectrum analysis has been suggested by research- ' > I Iers as a needed contribution toward the specification of ' I I speech; and third, findings of scientific studies of voice I types would provide objective bases for the development of | I better techniques of voice training. ' Thirty-two male subjects whose voices were judged by ; the experimenters to be low-pitched were instructed to produce the vowels [i], [a] and (u] efficiently and ineffic- ; 120 iently. The voice samples were recorded and placed in random order on a magnetic recording tape. The final tape of , sixty-four voice samples was played to a jury of seven ' ' "expert” listeners who had demonstrated superior skill in distinguishing between the two voice types on an auditory : discrimination test. Significant agreement on efficiency or inefficiency was reached by the judges on forty-six of | : the sixty-four samples. I 1 Spectrograms were made of the forty-six samples agreed upon by the judges. Instrument settings were experimentally determined by examining acoustic patterns which i yielded the best information concerning formant structure and location. ' Seven questions were asked in the study. 1. To what extent does the fundamental frequency : of vowel sounds differ between voices of low-pitched male : j subjects perceived as efficient and those perceived as in- ' I ' _ ! efficient, and what is the nature of such differences? : 2. To what extent do the frequencies of formants i 1, 2, and 3 of vowel sounds differ between voices of low- ! ; pitched male subjects perceived as efficient and those per-' ceived as inefficient and what is the nature of such differences? 3. To what extent do the relative amplitudes of ; 121 i I formants 1, 2, and 3 of vowel sounds differ between voices ■ ! of low-pitched male' subjects perceived as efficient and i those perceived as inefficient and what is the nature of : such differences? 4. What correlation exists between the fundamental ! ! : frequencies of vowel sounds which were perceived as effic- ; I ient and those perceived as inefficient as spoken by the i I ; 'same subjects? j j 5. What correlations exist between the frequencies I I I of formants 1 and 2 of vowel sounds which were perceived as| i I I efficient and those perceived as inefficient as spoken by j I I I the same subjects? : ‘ I 6, What correlations exist between the relative j I amplitudes of formants 1 and 2 of vowel sounds which were | I perceived as efficient and those perceived as inefficient j I i I as spoken by the same subjects? | I I 7. What descriptive differences exist in the spect-j 1 I j ra of samples of efficient voice production and inefficient I voice production on the basis of special pattern character-, istics? ; : I I The results of the study indicated the following. | 1. The mean fundamental frequency for efficient ' voice production was lower than for inefficient voice proI duction. The difference was found significant beyond the ! 122 *01 level of confidence. 2. The frequencies of formants 1, 2, and 3 of vowel ! sounds produced efficiently tended to be lower than for the | same formants of vowel sounds produced inefficiently. The difference was not significant at required levels of con- , I ; fidence. 3# The relative amplitude of formant 1 was greater i for efficient voice production than for inefficient voice ''production. The difference was found significant at the ,.05 level of confidence* The relative amplitude of formant; : I ; 2 tended to be slightly greater for efficient voice pro- | auction, whereas the relative amplitude of formant 3 was I : greater for inefficient voice production. The difference ! in formant 3 was found significant at the .05 level of con-* ; I fidence. ! i 4. A moderately strong relation existed between the! I fundamental frequency of efficient voice production and in-1 I I ; efficient voice production. A correlation coefficient of I.456 was obtained. I 5. Correlations of slight relationships existed between frequencies of formants 1 and 2 of efficient voices and inefficient voice-s. ; 6. Correlations of slight relationships existed b e - : tween amplitudes of formants 1 and 2 of efficient voices ' ]U23 I and inefficient voices, 7. Spectra of efficient voice production differed 'descriptively from those of inefficient voice production in that (a) harmonic registration was stronger, more regular and distinct; (b) they had no abnormal pitch change; I I(c) they had consistently more of the higher formants; and (d) energy was concentrated in the lower frequency region. ' i i I 1 II. CONCLUSIONS ! Fundamental frequency. The fundamental frequency of I I ! Îefficient voice production differs from the fundamental ' I i I i frequency of inefficient voice production for the same voice I I I in two distinct ways. First, the efficient voice tends to i have a lower fundamental frequency, and second, it does not; vary as widely. The extent of these differences was indi- | ! 'cated by a ^ ratio of 4.04, and the nature of the differences by an F ratio of 10.75. Both ratios are significant | : beyond the .01 level of confidence. It is consistent with caution to say that one of the parameters of the efficient ; voice is a relatively low pitch for an individual. Con- I versely, a parameter of vocal inefficiency may be high I pitch relative to the individual's pitch for vocal efficiency. Formant 1. Formant 1 of efficient voice production | differs from that of inefficient voice production in one I basic manner. While no appreciable difference appears beI i ^ tween the frequency of formant 1 for the two voice types, ; I i I the difference between the relative amplitude is signifi- I : cant. The efficient voice has a strong low formant 1 while^ ,the inefficient voice has a correspondingly weaker and I ; slightly higher formant 1. The extent of the difference in I decibel ratings for the amplitudes of formant 1 was indi- !cated by a t ratio of 5.58 which is significant at the .01 i .level of confidence. These findings support those of I Bartholomew^^ who found a "strong low formant" among good I male singing voices. I I ; Formant 2. Formant 2 of efficient voice production differs from formant 2 of inefficient voice production in I three ways. First, there is a slight tendency for the jefficient voice to have a lower formant 2 than for the in- ,efficient voice. Although a t ratio of 1.71 is not signiI jficant at the required levels of confidence, it does apI iproach significance and Indicates a rather definite tend- ! ency. Second, formant 2 of the efficient voice is more stable, the frequency location in the spectra is not as 'spasmodic as is that of the inefficient voice. The nature 79Bartholomew, 0 £. cit.. pp. 25-33. ' 125 of the difference is indicated by an P ratio of 2.87. Third, the nature of the difference between the two vari- , ables is further illustrated by a correlation coefficient Î : I of .29. On the basis of the obtained coefficient we may j I expect some relation to exist between the two variables. : The degree of relationship is low, however. Frequency of j formant 2 for the two voice types tends to vary with little interdependence. No significant difference nor correlation exist between the mean decibel rating of formant 2 for efficient and inefficient voice production. There is a slight tend- , ency for formant 2 of the efficient voice to be more stable : ; with reference to amplitude. This is indicated by an P ratio of 1.011. Formant 3. Although formant 3 does not appear in all samples of efficient nor in all samples of inefficient voice production, whenever it appears it may be expected to I : differ in two ways. First, formant 3 is generally lower I in frequency for the efficient voice than for the inefficient voice. The difference in frequency is not significant, but it tends to be significant when formant 3 is considered in matched voice samples. Under these conditions formant 3 of the efficient voice is lower than that of the inefficient by a greater mean difference than for either I 126 . I : I formant 1 or 2. Second, formant 3 is stronger for ineffic1 ient voice production than for efficient voice production. The difference in decibel rating of formant amplitude was ; not found to be significant at the required levels of con- ! I fidence, but it was definitely in favor of the inefficient j voice. This finding supports the observed tendency of ' energy to be widely distributed and skewed to the higher | frequencies in acoustic patterns of inefficient voices. It| j Q Q I also supports the findings of Lewis and Tiffin who discovered that less pleasant voices are characterized by widely distributed energy in the spectrum, while superior ! i ' 1 voices are characterized by a concentration of energy in ; one partial or two adjacent partials. Third, the relative ‘ ; amplitude of formant 3 of the efficient voice is more stab-; I I jle than that of the inefficient voice. The nature of this | I difference is indicated by an F ratio of 2.26 which is I i significant at the .06 level of confidence. Because of j I ! , the inconsistent appearance of formant 3 in both the samI . pies of efficient and inefficient voice, correlation was i 'not thought to be valid instrument of analysis at this i point. I i . , - Special pattern characteristics. The acoustic pat- i ®®Lewis and Tiffin, op.'^ cit., pp. 43-60. 127 i 4 terns of efficient voice production seem to be distinguish- | ^ f ed from those of inefficient voices by (1) stronger, more if ? : I ^ regular, and more distinct harmonic registration; (2) less j| extreme pitch change; (3) more of the higher formants 4 and I 5; and (4) stronger low components in the spectrum# CHAPTER VI ' LIMITATIONS AND IMPLICATIONS I FOB FUTURE STUDY I . LIMITATIONS ! I Limitations of the study were considered on the basis of the three major entities of the investigation: subjects, materials, and procedures. The three points are I interrelated in the weaknesses to be pointed out in this i I ' i 'section. ! I i I One of the chief limitations of the study was the | ! I I number of subjects used. It was thought that the judging i ' Ipanel would agree consistently with the experimenters* pre- ' 'evaluation of the voice samples. This was not the case. ;Although the number of subjects was not insufficient for I statistical analyses, additional subjects would have inIcreased the reliability. The procedure for selection of I I Isubjects should have made use of the jury of listeners bejfore, or at the time of recording. Thus, only samples already agreed upon by the judges would have been recorded, and all would have been suitable for analysis. The second weakness of the study was concerned with the Sona-graph. Because intensity was considered to be a critical factor in the study, it was highly desired that the 129 amplitude seotioner of the Sona-graph be calibrated to conform to the nature of the signals to be analyzed. This was! not feasible. It was pointed out that the validity of the | : study was not threatened, nevertheless, for calibrating thei Instrument would have only yielded a constant which would have been added to each obtained value from amplitude analysis. Still, calibration was desired as an integral part of research procedure involving instruments such as were used in the study. I With reference to procedure, a more sensitive moni- | I : storing system could have been employed at the initial re- ; I 1 joordings. The visible light beam of the Pentron was too \ I gross for adequate monitoring, hence, all components of each voice sample were not recorded at intensity levels optimal for spectrum analysis. A general limitation may be stated as the remoteness i of personnel qualified to give technical assistance during ispectrographic analysis. One Sona-graph in the area remains unusable for want of technical adjustments and repair. It would have been desirable to make the spectrograms at one of the established speech research centers using such accessory devices as the new amplitude unit which provides amplitude analysis of continuous spectra. In an effort to account in part for this limitation, the author visited one 130 suoh research center for the express purpose of receiving needed technical advice. II. IMPLICATIONS FOB FUTURE STUDY I i I Formant 2 has been specified as the most important I formant in vowel spectra* On the basis of its importance it was thought that any real difference between voice types should be supported by a significant difference in the nature and frequency location of the second formant. It may be recalled that the differences between formant 2 of I the two voice types treated in this study was not signifileant, but was thought to indicate a tendency toward significance. The major recommendation, then, is that a larger number of subjects be used in a study to determine more accurately the nature of the difference in frequency location of formant 2 as suggested in this study. Such a study made with regard for limitations as outlined in the preIceding section should contribute materially to the informI ation already at hand. Further, it was indicated that the higher formants I ,4 and 5 appeared with more consistency among the efficient voices. Yet it was discovered that the largest number of I formants appeared in a sample of inefficient voice. The implication may well be that if more than four or five for- 131 I mants are present in the vowel spectrum, the sound will ; tend to be judged inefficient. The question might be ■ I ! raised, how many vowel formants characterize the efficient ; i voice and how many must be added to make the vowel be per- ; I ceived as inefficient? A selection of vowels that charact-; i : . eristically have multi-formant structure would seem to facilitate such a study. Although an increase in frequency was generally concomitant with change from efficient to inefficient voice I production, in three cases this was not so. In one of i ' these cases inefficiency was produced at the same frequency! I i I and in the remaining two cases, it was produced at lower i ! ; I frequencies than was efficient voice. The implication may i ; be that the judges were at times assessing something other ' than increase of frequency. A study in which voice samples ; of efficient and inefficient voice are matched for fundaI mental frequency should prove of interest, since one of the basic assumptions of the present study is that the voice can be inefficient at the same pitch of the efficient voice. ' Would the conclusions found in this study obtain in cases where efficient and inefficient voice production characterize two separate groups of subjects? The impli- ; cations from the present study would tend to suggest an 132 , affirmative answer, however, it appears that a study in which two separate populations are represented would be war-* ^ ranted. Study of such voice samples by acoustical methods ! i should yield reliable estimates of basic parameters of each; j voice type. ' ! Several instruments were reported in the literature ; : which would seam to have special bearing on future acoustical analysis of voice types. Among them are the Purdue Pitch M a t e r , a device which plots visually the pitch of 1 ; I voice on a logarithmic scale, and an automatic formant ex- | ! go I I tractor which would eliminate probable error when locat- i i i I ing formants manually. The amplitude display unit of the I Sona-graph was mentioned in the preceding section. This unit is now commercially available. It produces amplitude variations on recording paper in conjunction with the conventional display of time-frequeney-intensity. Voice quality by its very nature is subjective, and I objective investigations must inevitably take into account I ! perceptual assessment of sound production. In a word. 81m . E. Dempsey, "The Purdue Pitch Meter - A Direct Reading Fundamental Frequency Analysis,” Journal of the Acoustical Society of America. 22:6 (November, 195077PP*Ôdé^6. 8^James L. Flanagan, "Automatic Extraction of Formant Frequencies from Continuous Speech,” A paper read at the convention of Acoustical Society of America, Austin, Texas, November, 1954. 133 I acoustical analysis of voice types must come back to the ; I question, what did the listener hear and how did he evalI uate it? Thus, adequate attention must be given to speci- i ; ! I fying criteria for listening authorities. When such crit- ; : I I eria are well defined and the physical invariants of a | voice type have been isolated, comparison with "what was heard" should yield the best probable information. Although it is hoped by people in various fields that eventually speech sound units can be accurately and invariably specified in terms of physical dimensions, the canonical form of such dimensions has yet i to be found.83 ; 8^Chang, ^ ^ . , 0 £. cit., pp. 675-679. BIBLIOGRAPHY BIBLlb&RAPHY A. BOOKS iAustin, Gilbert, Ghironomia. or A Treatise on Rhetorical j Delivery, London; W. Bulmer and Go., 1806, 585 pp. | I Avery, Dorsey and Sickles, First Principles of Speech Training. New York: D. Appleton Publishing Company, 1928, 518 pp. Bell, Alexander, Visible Speech. London: Simpkin, Marshall and Company, 1867, 126 pp. , The Mechanism of Speech. New York: Funk and Wagnalls, 1916, 8th Edition; n. p ! I jBrodnitz, Friedrich S., Keep Your Voice Healthy. New York; | ! Harper & Brothers Publishers, 1953, 229 pp. | Burke, Joanna Beljora (Arrington), The Fundamentals of Tone I Production. New York: Schroeder and Gunter, 1928. pp. Colby, M. Y., Sound Waves and Acoustics. New York; Henry Holt and Company, 1938, 356 pp. Culver, Charles A., Musical Acoustics. London: Blakiston, 1951, 194 pp. Dickens, Milton, Speech; Dynamic Communication. New York: Harcourt. Brace and Company, 1954, 440 pp. Edwards, A. L., Experimental Design In Psychological Research. New York: Rinehart Company, 1950, 446 pp. Eisenson, Jon, Basic Speech. New York; Macmillan Company, ■ 1950, 280 pp. Fairbanks, Grant, Voice and Articulation Drillbook. New York: Harper and Brothers Publishers, 1940, 234 pp. Fletcher, Harvey, Speech and Hearing. New York: D. Van Nostrand Company, Incorporated, 1948, 331 pp. , Speech and Hearing In Communication. D. Van Nostrand Company, Incorporated, 1953, 461 pp 136 ProescheIs, Emil, Twentieth Century Speech and Voice Correction. New York; Philosophical Library, 1948, 321 pp. ' Pulton, Robert I., and Trueblood, Thomas C., Practical Ele-i I ments of Elocution. Boston: Ginn and Company, 1893, j 610 pp. I i I I Goldbury, John, and Russel, William, The American Common- | School Reader and Speaker. Boston; Tappanand Whittenj more, 1844, pp. ; Grim, Harriett E., Practical Voice Training. New York: ' Guilford, J. P., Fundamental Statistics in Education. New i York: McGraw-Hill Book Company, Inc., 1950, 633 pp. | ! Hamill, S. S., Science of Elocution. New York; Phillips j I Hunt, 1882, pp. I I I jHeffner, R-M. S., General Phonetics. Madison; University j ; of Wisconsin Press, 1949, pp. | I Helmholtz,. H, L. F., Sensations of Tone. London: Longman,' Green, and Company^ 1895, [Translated by A. J. Ellis) pp. XIX - 574. Holmes, P. Lincoln D., Voice and Diction. New York: F. S. Crofts and Company, 1948, 279 pp. Judson, L. S. and Weaver, A. T., Voice Science. New York: F. S. Crofts and Company, 1942, 377 pp. Meyer, Ervin, Electro-Acoustics. London: G. Bell and Sons; Ed., 1939, 117 pp. Miller, Dayton C ., The Science of Musical Sounds. New York: The Macmillan Company, 1922, 286 pp. __, Sound Waves - Their Shape and Speed. New York: The Macmillan Company, 1937, 164 pp. Miller, George A., Language and Communication. New York; McGraw-Hill, 19 , 275 pp. Moses, Paul J., The Voice of Neurosis. New York: Grune and Stratton, 1954, 131 pp. 137 i : ' ; Negus, V. E,, Comparative Anatomy and Physiology of the Larynx. London; Grune and Stratton, 1949, 528 pp. .Ogilvie, Mardel, Terminology and Definitions of Speech De- ; fects. New York; Columbia University Press, 1942. I Paget, R. H. S., Vowel Resonances. Published by the Inter- ; national Phonetic Association, 1922, 12 pp. Pike, K. L., Tone Languages. Ann Arbor; University of Michigan, 1948, 187 pp. ■Potter, Ralph, £t al., Visible Speech. New York: D. Van ^ Nostrand Company, Incorporated, 1947, 430 pp. I I 'Raubicheck L., Davis, E. H., and Caroll, L. A. Voice and I Speech Problems. New York; Prentice-Hall, Inc., 1946, 569pp. I ! ! iRichardson, E. G. Ed, Technical Aspects of Sound. Houston;| Elseview Press, 1953, 544 pp. I Robb, Margaret, Oral Interpretation of Literature. New I i York: H. W. Wilson Company, 1941, 242 pp. j Robbins, S., Hawk, S. and Russell, G. 0., Standard Diction- ' i ary of Terms Dealing with Disorders of Speech, rev. and abr .^America Speech Correction Association ) : 1951, 112 pp. . I Rush, James, The Philosophy of the Human Voice. Philadel- i I phia; Grigg and Llliott, 1833, 432 pp. I iRussell, G. Oscar, Speech and Voice. New York: The Mac- i I millan Company, 1951, 250 pp. I ; The Vowel. Columbus : Ohio State University Press,! 1928, 354 pp. Scripture, E* W., The Elements of Experimental Phonetics. New York: Charles Scribners' Sons, 1902, 696 pp. :Stevens, S. S., and Davis, H,, Hearing. Its Psychology and I Physicology. New York; John Wiley and Sons, Incorpor- . ated, 1938, 489 pp. Travis, L. W., Speech Pathology. New York: D. Appletonj Century, 1931, 331 pp. 138 Van Dusen, C. Raymond, Training the Voice for Speech. New York: McGraw-Hill Book Company, 1943, 232 pp. Van Riper, Charles.Speech Correction; Principles and Methods. New York: Prentice-Hall, Incorporated, 1963, 470 pp. _______ , Speech Therapy, A Book of Readings. New York; Pr en tic e-Hall, Inc., 1953, 320 pp. ^ Vennard, William, Singing the Mechanism and the Technic. Ann Arbor; Edwards Brothers, 1949, 171 pp. Watson, Floyd R., Sound. New York : John Wiley and Sons, Incorporated, 1935, 219 pp. Woolbert, Charles H., Fundamentals of Speech. New York; Harper and Brothers, 1920, Rev. 1927, 383 pp. B. PERIODICAL ARTICLES jBartholomew, Wilmer T., "A Physical Definition of 'Good ; ' Voice Quality' in the Male Voice," Journal of the Acoustical Society of America. VII (July, 19341 pp. 25-33. Biddulph, R.,"Short-Term Autocorrelation Analysis and Correlatograms of Spoken Digits," Journal of the Acoustical Society of America. 26:4 (July, 19647# pp. 539-541. Bennett, W. R., "The Correlatograph— A Machine for GontinI uous Display of Short Term Correlation," Bell System I Technical Journal. 32:5 (September, 1953), pp. 1173-1186. 'Black, John W . , "Natural Frequency, Duration, and Intensity of Vowels in Reading," Journal of Speech and Hearing Disorders. 14 (May, 1949), pp. 216-221. ,"The Quality of a Spoken Vowel," Archives of Speech. TT1937), pp. 7-27. Blanton, Smiley, "Research Problems in Voice and Speech,"- Quarterly Journal of Public Speaking. 2 (January, 1916), 139 Bogert, B. P*, "On The Band Width of Vowel Formants," Journal of the Acoustical Society of America, 25:4 (July, , 1953T7 pp. 791-792. iCarhart, Raymond, "The Spectra of Model Larynx Tones," i I Speech.Monographs, 8 (1941), pp. 78-84. i jChang, S. H., Pihl, 0, E., and Wiren, J., "The Intervalgram| I as a Visual Representation of Speech Sounds," Journal I of the Acoustical Society of America, 23:6 (November, I 195Î77 pp. 675-679. Cooper, Franklin S., "Spectrum Analysis, " Journal of the Acoustical Society of America. 22:6 (November, 1950), pp. 761-762. __ , ^ , "Some Experiments on the Perception of Synthetic Speech Sounds," Journal of the Acoustical Society of America, 24:6 (November, 1952), pp. 592-606,, I I . I Crandall, I. B., " The Sounds of Speech," Bell System Techni4 I cal Journal, IV (1925), pp. 586-626. I i , "Dynamical Study of the Vowel Sounds II," Bell ‘ System Technical Journal, II (March, 1927) pp. 79-83. Crandall, I. B., and Mackenzie, D., "Analysis of the Energy Distribution in Speech," Physical Review, 19:3 (1922), pp. 221-232. Curry, R., "Mechanism of Pitch Change in the Voice," Journ- ' al of Physiology, 9:3 (1937), p. 254. ; Curtis, James P., "The Case for Dynamic Analysis in Acoustic Phonetics," Journal of Speech and Hearing Disorders, 19:2 (June, 1954T, pp. 147-157. Davenport, Wilbur Jr., "An Experimental Study of SpeechWave Probability Distributions," Journal of the Acoustical Society of America, 24:4 (July, 19527, pp. 390-399. Delattre, Pierre, "Physiological Interpretation of Sound Spectrograms," PMLA, LXVI (September, 1951), pp. 864- 875. ^______ , 2È q 1*."An Experimental Study of the Acoustic Determinants of Vowel Color; Observations on one-and twoformant Vowels Synthesized from Spectrographic Patterns," Word. 8 (1952), pp. 195-210. 140 Dempsey, M. S., "The Purdue Pitch Meter - A Direct Reading Fundamental Frequency Analyses," Journal of Speech and Hearing Disorders, 15:2 (June, 1950), pp. 135-141. , Drake, Ormond J., "Toward An Improved Vocal Quality," The Quarterly Journal of Speech, 23:4 (December, 1937), ! pp. 620-626. i I Dreyfus-Graf, Jean, "Sonograph and Sound Mechanisms," JourI nal of the Acoustical Society of America, 22:6 (NovemI ber, 1950),pp. 731-738. Dunn, H. K., "The Calculation of Vowel Resonances, and an Electrical Vocal Tract," Journal of the Acoustical Society of America, 22:6 (November, 1952), pp. 740-753. Edmondson, Harold and Horwitz, Elinor M., "Cues for Vowel Discrimination," Journal of Speech and Hearing Disorders, 15:3 (September, 1950), pp. 202-206. ' Fairbanks, Gr^nt, "A Physiological Correlative of Vowel Intensity," Speech Monographs, 17:4 (December, 1950) I pp. 390-395. ! Fairbanks, Grant, £t al., "Experimental Study of Vowel Intensities," Journal of the Acoustical Society of Amer- ; ica, 22:4 (July, 1950T, pp. 457-459. ,Fischer-Jorgenson, Eli, "Phonetic Basis for Identification of Phonemic Elements," Journal of the Acoustical Society of America, 24:6 (November, 1952), pp. 611-617. Fletcher, Harvey, "Loudness, Pitch and the Timbre of Musical Tones and their Relation to the Intensity, the Fre- ; quency and the Overtone Structure," Journal of the ; Acoustical Society of America, 6: (April, 1935), pp. 59- ; 69. IGemelli, Agostino, "La Strutturazione psycologica del luigu- ■ aggia studiata mediante I'analisi elettrocustica." (The ! psychological structuring of the language studied through electroacoustic analysis.) Pontificia Academic ; Scientific Scripts Varia, 8 (1950), pp. 53. _______ , "La Strutturagione del luiguaggia studiata mediante I'analisi elettroacustics," (The structuration of language studied with the help of electro-acoustical analysis. ) Dialectics, 4 (1950), pp. 255-267. 141 , Gray, Giles Wilkeson, "The Voice Qualities in the History of Elocution," Quarterly Journal of Speech, XXIX:X (December, 1943), pp. 476-483. Harris, Cyril M., "Study of the Building Blocks in Speech,": Speech Synthesizer, 25:5 (September, 1953), pp. 962- I 969. Hawley, M. E. and Kettler, A. H., "Apparent Source of Speech in the Mouth," Journal of the Acoustical Society of America. 22:3 (May, 1960T, pp. 365-368. Heindsmann, T. E., "Acoustic Spectrum Terminology," Journal of the Acoustical Society of America, 25:6 (November, 1953)7 p. 1201. Henley, Homer, "Training the Male Voice," Etude, 54 (1936), p. 46. j I Hirsh, Ira, et al., "Intelligibility of Different Speech | ^ MaterialT”" Journal of the Acoustical Society of America] ! 26:4 (July,“ l95477 pp. 530-538. i I I 'Holmes, P. Lincoln D., "The Problem of Voice Placement," Quarterly Journal of S p e e c h , 17.2 (April, 1931), pp. 236-245. Huggins, W. H., "Phase Principle for Complex-Frequency Analysis and its Implications in Auditory Theory," Journal of the Acoustical Society of America, 24:6 (November, 1 9 5 2 ) , pp. 582-589. jJoos, Martin, "Acoustic Phonetics," Journal of the Linguistic Society of America, 24:2 Supplement TÂpril, June, 1948)7pp. 136. Kersta, L. G., "Amplitude Cross-Section Representation with the Sound Spectrograph," Journal of the Acoustical Society of America, 20:6 (November, 1948), pp. 796-801. .Kock, Winston E., "Problem of Selective Voice Control," Journal of the Acoustical Society of America, 24:6 ^November, 19527, pp. 625-628. Kock, W. E. and Harvey, P. K., "A Photographic Method for Disploying Sound Wave and Mierowave Space Patterns," Bell System Technical Journal. 30:3 (July, 1951), pp. 564-587. 142 Kock, W. E.. and Miller, R. L., "Dynamic Spectrograms of Speech, Journal of the Acoustical Society of America, 24:6 (November, 1952), pp, 783-784. Koenig, W,, £t £l., "The Sound Spectrograph," Journal of i the Acoustical Society of America, 18:1 (July, 19467, i i pp. 19-49. . i Koenig, W., and Ruppel, "Quantitative Amplitude Representation in Sound Spectrograms," Journal of the Acoustical : Society of America, 20:6 (November, 1948), pp. 787-795. Kordan, Walter, "Acoustic Method for the Measurement of Vibration Amplitudes," Journal of the Acoustical Society of America, 26:3 (May, 1954), pp. 428-433. Kopp, G. A. and Green, H. G ., "Basic Phonetic Principles of Visible Speech," Journal of the Acoustical Society | I 2Î. America, 18:1 (July, 194677 PP. 79-89. j I : Knower, F. ÏÏ., and Emerson, M., "Indices of Achievement in j 1 Voice Instruction," Journal of Speech and Hearing Pis- | i orders, 11 (1946), pp. 159-163. I ! Laase, L., "Effect of Pitch and Intensity on the Quality of Vowels in Speech," Archives of Speech, II (1937), pp. I 41-60. Lewis, Don, "Vocal Resonances," Journal of the Acoustical Society of America, 8: (1936), pp. 91-99. Lewis, D., and Tiffin, J., "A Psychophysical Study of Individual Differences in Speech," Archives of Speech, 1 (1934), pp. 43-60. I jLicklider, J. G. R., "On the Process of Speech Perception," Journal of Acoustical Society of America, 24:6 (Novem- ' ber, 19527, pp. 590-594. jLindsley, Gharles P., "Psycho-Physical Determination of I Voice Quality," Speech Monographs, 1:1 (1934), pp. 79- 116. Morgan, C. T., "Pitch and Intensity," Journal of the Acous tical Society of America, 23:6 (November, 195l7, pp. 658-664. 143' Moses, Paul J., "Vocal Analysis," Archives of Otolaryngol- i ogy. (August, 1948), pp. I ! Î Munson, W, A., and Wiener, Francis, "Sound Measurements for; I Psychophysical Tests," Journal of the Acoustical Society I of America, 22:3, (July, 1952), pp. 382-386. I Murray, Elwood, and Tiffin, Joseph, "An Analysis of Some ; Basic Aspects of Effective Speech," Archives of Speech,: : I (1934), pp. 61-83. Obata, J., and Kobayashi, R., "An Apparatus for Directrecording the Pitch and Intensity of Sound," Journal of the Acoustical Society of America, 10 (October, 1938), pp. 147-149. "A Direct-Reading Pitch Recorder and Its Applicaj tions to Music and Speech on Pitch Logarithmic Scale," i Journal of the Acoustical Society of America,9:2 (Oct- ; I ober, 1937), pp. 156-161. I I I jOlson, Harry F., "Frequency Range Preference for Speech and | I Music," Journal of the Acoustical Society of America, I ; 19:4 (July, 194777 pp. 549-555. I Peterson, 0. E., "Design of Visible Speech Devices," Journal of the Acoustical Society of America, 26:3 (May, 195477 pp. 406-412. "Information-Bearing Elements of Speech," Journal , of the Acoustical Society of America, 24:6 (November, 195277 pp. 629-636. [Peterson, Gordon, "Parameter Relationships in the Portrayal ! of Signals with Sound Spectrograph Techniques," Journal I ££ Speech and Hearing Disorders, 17:4 (December, 1952), I pp. 427-432. jPeterson, G. E., and Barney, H. L ., "Control Methods Used i in a Study of the Vowels," Journal of the Acoustical ; Society of America, 24:2 (March, 1952), pp. 175-184. i Peterson, G. E., and Rarsbeck, "Measurement of Noise with the Sound Spectrograph," Journal of the Acoustical Society of America, 25:6 (November, 1953), pp. 1157-1162. •Peterson, G. E., "Frequency Detection and Speech Formants," I Journal of the Acoustical Society of America, 23:6 144 (November, 1951), pp. 668-774. Pike, Kenneth, "Operational Phonenics in Reference to Ling-: uistic Relativity," Journal of the Acoustical Society of America. 24:6 (November, 1952), pp. 618-624. : Plugge, Domis E., " 'Voice Qualities' in Oral Interpretation]" ! Quarterly Journal of Speech. 28:4 (December, 1942), I pp. 442-444. i Pollack, I., "On the Identification of Speakers by Voice," ; Journal of the Acoustical Society of America. 26:3 i (May, 1954), pp. 403-405. Potter, Ralph K., "Objectives for Sound Portrayal," Journal of the Acoustical Society of America. 21:1 (January, 1949)7 pp. 1-5. :_______ , "Introduction to Technical Discussion of Sound Portrayal," Journal of the Acoustical Society of America. ! 18 (July, 1946), pp. 1-3. I ■_______ "Visible Patterns of Sound," Science. 102:2654 | (November, 1945), pp. 463-470. j Potter, R. K., and Peterson, G. E., "The Representation of Vowels and Their Movements," Journal of the Acoustical Society of America. 20, (1948) pp. 528-535. Potter, R. K., and Steinberg, J. C ., "Toward the Specification of speech," Journal of the Acoustical Society of America. 22:6 (November, 1950), pp. 807-820, p. 812. Pronovost, W., "Research Contributions to Voice Improvement," Journal of Speech and Hearing Disorders. 7 (1942), pp. 313-318. Robbins, Samuel D., "Principles of Nomenclature and Classification of Speech and Voice Disorders," Journal of Speech and Hearing Disorders. 12 (1947), pp. 17-22. Scripture, E. W., "The Nature of the Vowels," The Physical Society. (June, 1931), p. 47. Sherman, Dorothy, "The Merits of Backward Playing of Connected Speech in the Scaling of Voice Quality Disorders," Journal of Speech and Hearing Disorders. 19:3 (September, 1954), pp. 312-321. 145 Sivian, L. J., "Speech Power and Its Measurement," Bell System Technical Journal* 8: (1929), p. 646, ^ , ! Steer, Max D., and Tiffin, Joseph, "A Photographic Study of! the Use of Intensity by Superior Speakers," Speech | I Monographs* 1 (1934), pp, 72-78, | i \ I Steinberg, John G,, "Application of Sound Measuring Instru-i I ments to the Study of Phonetic Problems," Journal of J The Acoustical Society of America, 6 (1934T, pp. 16-24. Steinberg, J. C ., and French, N. R., "Thr Portrayal of Visible Speech," Journal of the Acoustical Society of America* 18 (July, 1946), pp. 4-18. Stephens, K. N., "Autocorrelation of Speech Sounds," Journal of the Acoustical Society of America, 22:6(November. 195077 pp* 769-771. Stevens, S. S., £t al., "Methods of Measuring Speech 1 Spectra," Journal of the Acoustical Society of America. -, i 19:5 (September, 194777"PP* 771-779. | 'Stout, B., "Harmonic Structure of Vowels in Singing in Re- | lation to Pitch and Intensity," Journal of the Acoustical Society of America. 10 = (1938), pp. 137-146. Straus, Oliver, "Relation of Phonetics and Linguistics to Communication Theory," Journal of the Acoustical Society of America. 22:6 (November, 1950), pp. 709-712. Taylor, H* G., "The Fundamental Pitch of English Vowels," Journal of Experimental Psychology. 16 (1933), pp. 565- I 582. ! Tiffany, Wm. R., "Vowel Recognition as a Function of Dur- ! ation. Frequency Modulation and Phonetic Context," Journal of Speech and Hearing Disorders. 18:3 (October, 1953), pp. 289-301. ,Tiffin, Joseph, "Applications of Pitch and Intensity Meas- ' urements of Connected Speech," Journal of the Acoustical Society of America. 5:4 (April, 19347, pp. 225-234. Van Dusen, G. R., "A Laboratory Study of the Metallic Voice," Journal of Speech and Hearing Disorders, 6 (1941), pp. 137-140. I 146 I Villarreal, Jesse J., "Consistency of Judgements of Voice I Quality," The Southern Speech Journal, XV (September, I 1949), pp. 11-18. I Ward, W. D., "Subjective Musical Pitch," Journal of the j Acoustical Society of America, 26:3 (May, 195TJ, pp. I 369-380. I Wye, B. C. Van, "The Efficient Voice in Speech," Quarterly ; Journal of Speech, 22 (December, 1936), pp. 642-648. C . JOURNALS I Psychological Abstracts, Published by American Psychologi- ' cal Association, Incorporated, Lancaster, Pennsylvania. 1I Speech, Official Organ of British Society of Speech TheraI pists, London. Journal of Speech and Hearing Disorders, Published by the American Speech and Hearing Association, Dansville, ! Illinois. Speech Monographs, Published by the Speech Association of America, University of Missouri, Columbia. Southern Speech Journal, Published by the Southern Speech Association, Convention Press, Jacksonville, Florida. Word, Published by the Linguistic Circle of New York, New York. PMLA, Published by the Modern Language Association of America, Menasha, Wisconsin. Journal of the Acoustical Society of America, Published by the American Institute of Physics, Lancaster, Pennsylvania. I American Standard Acoustical Terminology, Published by the American Standard Association, New York, New York. Archives of Otolaryngology, Published by the American Medical Association, Chicago, Illinois. 147 ! The Bell System Technical Journal, Published by the Ameri- ; can Telephone and Telegraph Company, New York. ; Journal of Experimental Psychology, Published by the Ameri-' can Psychological Association, Incorporated, Lancaster,’ Pennsylvania. ; ! The Physical Review, Published by the American Physical Society, Lancaster,' Pennsylvania. D. ESSAYS Garcia, Manuel, "Observations on the Human Voice," Royal Society of London, Proceedings, VII, 1885, p. 410. E. PARTS OF SERIES I Scripture, E. W., Researches in Experimental Phonetics, j Carnegie Institute of Washington, Pub1icatlon #44, : Press of Gibson Brothers, 1906, pp. 204. "Technical Aspects of Visible Speech," Bell Telephone System Technical Publications, Monograph B-1415. F. UNPUBLISHED MATERIALS Apple man, Ralph D., "A Study by Means of Radiograph, PlaniI graph, and Spectrograph of the Physical Changes which Occur During the Transfer from the Middle to the Upper Register in Vocal Tones," Ph.D. dissertation, Indiana University, Bloomington, Indiana, September, 1953. Beard, Raymond S., "An Experimental Study to Determine the Qualities that Make Up a Good and a Poor Speaking Voice." A Ph. D. dissertation prospectus. Northwestern ! University, Chicago, Illinois. Curtis, James P., "Experimental Study of Wave Composition of Nasal Voice Quality." Ph. D. dissertation. State University of Iowa, 1942. piehl, Charles P., "The Effect of Voice Quality on Commun!- 148 cativeness." Ph. D. dissertation, Pennsylvania State College, 1949. Edmondson, Harold, "A Spectrographic Investigation of the i Acoustic Similarities and Differences of the Five Front’ Vowels as Spoken by 20 Male Speech Students." A ‘ I Masters’ Thesis, University of Michigan, 1949. : i ! I Goodman, Allan Cooper, "A Spectrographic Analysis of the Rep i lationship Between Fundamental Frequency and the Acous-i tic Spectra of the Vowels, 'EE* and * A ’." A Masters* Thesis, University of Michigan, 1949. Leedman, Jean, "Certain Factors Involved in the Discrimination of Vowel and Vowel-like Sounds." Ph. D. dissertation, University of Wisconsin, 1951. Philhour, Charles W., "An Experimental Study of the Re- , : lationship Between Perception of Vocal Pitch in Con- ; I nected Speech and Certain Measures of Vocal Frequency."! I A Ph. D. dissertation. University of Iowa, 1948. | i > •Sullivan, Ernest G., "An Experimental Study of the Relationj I ships Between Measurable Physical Characteristics and | Subjective Evaluation of Male Voice Quality in Singing." ! Ph. D. dissertation in progress, Indiana University, ■ ! Bloomington, Indiana, 1953. ' Thurman, Wayne Laverne, "An Experimental Investigation of Certain Vocal Frequency-Intensity Relationships Concerning Natural Pitch Levels." A Masters' Thesis, State University of Iowa, 1949. I I Thurman, Wayne, "The Construction and Acoustic Analyses of I Recorded Scales of Severity for Six Voice Quality Disorders." A Ph. D. dissertation, Purdue University, ■ 1953. G. TECHNICAL PAPERS Flanagan, James L., "Automatic Extraction of Formant Frequencies from Continuous Speech," Air Force Cambridge Research Center, Cambridge, Massachusetts. Haden, Ernest F., "A Spectrographic Analysis of the Cardinal Vowels," University of Texas, Austin. 149 ' Licksllder, J., Hawley, M., and Walkling, R., "Influences of Variations in Speech Intensity and Other Factors Upon the Speech Spectrum," Massachusetts Institute of | Technology, Cambridge, Massachusetts, and Radio Gorpor-; ation of America, Camden, New Jersey. | : ' I jStevens, K. N. and Bastide, R. P., "Electrical Synthesizer | I of Continuous Speech," Massachusetts Institute of Tech- i I nology, Cambridge, Massachusetts. > • APPENDIXES I APPENDIX A INSTRUCTIONS TO JUDGES OF PILOT STUDY The following voloe samples have been chosen as being representative of efficient or inefficient voice production. The efficient voice has been characterized by vibrancy anchored near the sternum, by openness in the , throat, and by freedom from tension or strain in the speech musculature. The inefficient voice has been characterized by the absence of the qualities that specify the efficient voice. You are requested to listen to each sample care- . fully, and indicate whether you think it is efficiently or inefficiently produced. Record your judgments in consecutive order for subsequent reference. TEST ITEMS OP PILOT STUDY The test items consisted of twenty-four voice samples; twelve efficiently produced, and twelve inefficiently produced. The items were randomized from the following list. 1. The ball has fallen over the wall. 2. Live, lag, love. 3. cm . Lae}, CaJ, 4. tei] , [ g i ] , Lail . APPENDIX B ITEM I MATERIAL IN THE AUDITORY DISCRIMINATION INSTRUCTION TAPEi^ The following recordings are designed to test your ability to discriminate between efficient voice production and various types of inefficient voice production. Before the test begins listen carefully to the preliminary exam- : I j pies of efficient and inefficient voice production in con- | I nected speech, single words, and in isolated speech sounds.! These can be played as many times as is necessary for you ; to recognize clearly the auditory characteristics that distinguish the two voice types. Example one is efficient voice production. The ball has fallen over the wall. Example two is the same voice produced slightly inI efficiently. The ball has fallen over the wall. I Now you will hear these same samples presented together first efficiently, then inefficiently. #The material in the Auditory Discrimination Instruction Tape is a transcript of the tape as used to instruct potential judges. Material is given in Appendix F in record ' form. 153, The ball has fallen over the wall, (efficiently I produced) 1 The ball has fallen over the wall. (inefficiently ; I I ; produced) , i I j Example three efficient voice production. | ' I ! Live, lag, love. : Example four is slightly inefficient voice production. Live, lag, love. ! Now you will hear the same examples presented to- I I I jgether, first efficiently and than inefficiently. | I Live, lag, love (efficiently produced). j I ! I Live, lag, love (Inefficiently produced). Example five is efficient voice production* Example six is slightly inefficient voice production. I Now you will hear the same samples presented toI {gather, first efficiently, then inefficiently, i (efficiently produced) (inefficiently produced) . I I You will now hear samples of efficient and inefficient voice production using sentences, words, and individual sounds. Listen carefully to these samples as they are 154 played first efficiently, then inefficiently. The ball has fallen over the wall. (efficiently ; produced) I The ball has fallen over the wall. (inefficiently I I produced) Live, lag, love (efficiently produced) Live, lag, love (inefficiently produced) (efficiently produced) (inefficiently produced) The ball has fallen over the wall. (efficiently : produced) The ball has fallen over the wall. (inefficiently I produced) Live, lag, love (efficiently produced) Live, lag, love (inefficiently produced) (efficiently produced) (inefficiently produced) The ball has fallen over the wall. (efficiently I produced) The ball has fallen over the wall. (inefficiently ; produced) Live, lag, love (efficiently produced) Live, lag, love (inefficiently produced) (efficiently produced) (inefficiently produced) 155 You are about to listen to some samples of voice production. Some are believed to be efficient and some inefficient. As these samples are played, try to be as active, I i Iin your listening as possible. Attempt to empathize with i I ■ i ithe subject so that you feel as they must have felt while I producing the voice quality you hear. The distinguishing ^ characteristics of efficient voice production are independent of pitch and volume, so disregard variations in loudness and pitch. The efficiently produced samples may be loud or 'almost whispered, high or low, but they will have in common 1 ! J ithe subjective feelings of chest resonance, of vibrancy, of | I I being anchored near the sternum rather than originating in I ‘the neck, of openness in the throat with no sensation of tension, and of being tones that will coast along without effort or strain. Please rate your reaction to them as efficient, inefficient, or undecided, by placing a check mark opposit the sample number in the appropriate column. 'You may replay the samples as many times as is necessary for I you to make a decision. 156 ITEM II MATERIAL IN THE AUDITORY DISCRIMINATION TEST TAPE* TEST TOMBER SOUNDS PRODUCED EFF. OR INEFF. 1 . I 2 . I 3. Cai], E 4. I a. texl, W I 6 . Ce. 1 7 Lax], t>i3, M E [ex], L W . C p x ] , W I 8 . [ci], \A^X PI], E 9. [ex], [?x] E 1 0 . L3X], to.] E 1 1 . lex], fâîl, Dx ] E 1 2 . [ex], Di], M I 13. [ex], ^^I3 E 14. [exx Gar], E -îî-The material in the Auditory Discrimination Test Tape is a transcript of the tape as used to test potential judges. Material is given in Appendix P in record form. The third column gives the evaluation of the experimenters as to efficiency or inefficiency of the sample. 157 ITEM II (Continued) HUMBER MATERIAL IN THE AUDITORY DISCRIMINATION TEST TAPE SOUNDS PRODUCED EPF. OR 15. tex], ipx], t a i Ë 16. [ei], CPXI, W I 17. Leil, c a n , rvL] I 18. urn, IPX], t o E 19. t<5xi, Dx], tw.] I 2 0 . LexX o n , t o E 158 ITEM III JUDGES DECISIONS ON THE TWENTY SAMPLES OF THE AUDITORY DISCRIMINATION TEST TAPE Sample Numbers Experimenters Evaluation i* 1 2 3 Judges 4 5 6 7 1 . I I I I I I I I 2 . I I I I I I I I 3. E E E E E E E E 4. I I I I I I I I 5. I I I I I I I I 6 . E E E E E E E E 7. I I I I I I I I 8 . E E E E E E E E 9. E E E E E E E B The first column gives the voice sample numbers. ;Tbe second column gives the evaluation of the experimenters as to efficiency or inefficiency of the samples. The following seven columns give the decision of the judges who are numbered one through seven. An evaluation of efficient [voice production is marked E, inefficient voice production 'l. An deviation of a judge from the decision of the experimenters is underlined. Judge number four was female, all the remaining judges were male. Jundes one, two, and seven were over thirty years of age. Judges three, four, five and six were between twenty and thirty years of age. 159 ITEM III (Continued) | JUDGES DECISIONS ON THE TWENTY SAMPLES OP THE j AUDITORY DISCRIMINATION TEST TAPE 1 Sample Numbers Experimenters’ Evaluation 1 2 3 Judges 4 5 6 7 1 0 . E E E E E E E E 1 1 . E E E E E E E E 1 2 . I I I I I I I I 15. E E I E I E E E 14. E E E E E E E E 15. E E E E E E E E 16. I I I I I I I I 17. I I I I I I I I 18. E E E E E E E E 19. I I I I I I I I 2 0 . B E E E E I E E APPENDIX G ITEM I JUDGE’S RATING SHEET* ; Directions: I You are about to listen to some samples of voice proj duction. Some are believed to be efficient and some inef- i ficisnt. As these samples are played, try to be as active in your listening as possible. Attempt to empathize with the subject so that you feel as they must have felt while producing the voice quality you hear. The distinguishing characteristics of efficient voice production are independent of pitch and volume, so disregard variations in loud- : I ness and pitch. The efficiently produced samples may be | loud or almost whispered, high or low, but they will have I I in common the subjective feelings of chest resonance, of | [vibrancy, or being anchored near the sternum rather than j I originating in the neck, or openness in the throat with no j I sensation of tension, and of being tones that will coast j [along without effort or strain. Please rate your reaction to them as efficient, inefficient, or undecided, by placing a check mark opposite the sample number in the appropriate column. You may replay the samples as many times as is necessary for you to make a decision Sample Eff. Ineff. Undecided Sample Eff. Ineff. Undecided 1 . 33. 2 . 34. 3. 35. 4. 36. *The first column indicates the number of the voice sample. The three columns marked "Eff.", "Ineffand "Undecided" were for the evaluator’s decision as to efficient, inefficient or undecided. The judge’s rating sheet was used by individuals taking the auditory discrimination test as well as by the judges when evaluating the sixtyfour samples of voice production. Although space was available for a decision of undecided, all individuals were encouraged to make a definite decision of efficient or inefficient. Voice samples could be played as often as desired. 161 ITEM I (Continued) JUDGE'S BATING SHEET Sample Eff. Ineff. Undecided Sample Eff. Ineff. Undecided 5. 37. 6 . 38. 7. 39. 8 . 40. 9. 41. Ï0 . 42. 1 1 . 43. 1 2 . 44. 13. 45. 14. 46. 15. 47. 16. 48. 17. 49. 18. 50. 19. 51. 2 0 . 52. 2 1 . 53. 2 2 . 54. 23. 55. 24. 56# 25. 57. 26. 58. 27. 59. 28. 60. 29. 61. 30. 62. 31. 63. 32. 64. 162 Ut'11% 'pi Narrow Band Corribession wide Ban Ji^.'.i^.iitrt'. ■’ i j t Sk W . * 1 APPENDIX D ITEM I SONAGEAM MADE WITE MAXIMUM GONPBBSSIOE In both narrow and wide band compression formants are noticeably distorted. Note especially points X, Y , and Z of narrow band sonagram. See following page. 163 Formant 1 Formant 2 Formant 3 2000 Formant 4 4000 _ 4000 3000 j; 2000 j C T — Formant 2 1000 Formant 1 FormanV 4 Formant 1 Formant 2 Formant 3 Formant 4 Formant 4 4y_Fo™B.t3- 3 -- Formant 2 Formant Formant 1 Formant 2 Formant 3 Formant 4 Formant 2 Formant i Formèm??''^ ITEM II SONAGEAM MADE V/ITE OPTIMAL COMPPESSION Formant structure of same voice sanqple is more distinct when compression is reduced. APPENDIX E ITEM I JUDGES RATING OP DECISIONS OP EPPICIENCY OR INEPPICIENCY FOR EACH VOICE SAMPLE ON EXPERIMENTAL TAPE SAMPLE SUBJECT DECISION BY TESTER JUDGES * EFF. DECISIONS INEFF. 61 # 1 . Stedraan Eff. 7 , 52 Richard Porter Iff. 7 1 13 Carl Shultz Eff. 7 ! 58 Henry Lewis Eff. 7 r 1 4 Leon Rudeman Eff. 7 1 27 # 1 . Carver Eff. 7 ! 2 0 Milton Shapiro Eff. 7 1 24 B. R. Murphy Eff. 7 ' 54 Krald Ashbough Eff. 7 42 Luther Sartor Eff. 7 : 2 1 Wm, Perkins Eff. 7 15 Edward Singleton Eff. 7 1 1 Ohm Pauli Eff. 6 1 45 Pressley Keys Eff. 6 1 30 # 1. Carver Ineff. 6 1 1 0 Norman Leer Eff. 6 1 : 43 James MCAree Eff. 6 1 1 7 Kenneth Shanks Eff. 6 1 : 56 Victor Heyden Eff. 6 1 1 49 McCoard Eff. 6 1 1 64 Bob Alkorn Eff. 6 1 31 Carl Shultz Ineff. 5 2 60 Arnold Morton Eff. 5 2 38 Bob McCambridge Eff. 5 2 50 Frank Anderson Eff. 5 2 5 Robert Gillen Eff. 5 2 : 36 Richard Porter Ineff. 4 3 2 Hal Bargelt Eff. 4 3 46 B. R* Murphy Ineff. 4 3 19 Glenn Smith Eff. 4 3 57 John Albright Eff. 4 3 36 %i. Vennard Eff. 3 4 1 Chet Millar Eff. 2 5 ITEM I (Continued) 165 SAMPLE SUBJECT DECISION BY TESTER JUDGES * EPF. DECISIONS INEFF. : 51 Edmund Thile Eff. 2 5 39 Edmund Thile Ineff. 2 5 34 Bob Alkorn Ineff. 2 5 40 Norman Leer Ineff. 2 5 1 2 Prank Anderson Ineff. 2 5 44 Henry Lewis Ineff. 2 5 23 #n. Stedman Ineff. 2 5 41 Edward Singleton Ineff* 1 6 2 2 John Albright Ineff. 1 6 37 James MoAree Ineff. 1 6 : 6 Wm. Vennard Ineff. 1 6 I 29 Dan Austin Eff. 1 6 1 48 Ohm Pauli Ineff. 1 6 1 55 Vfei. Perkins Ineff. 7 1 62 Wm. McCoard Ineff. 7 ; 18 Victor Heyden Ineff. 7 1 28 Luther Sartor Ineff. 7 : 4 Kenneth Shanks Ineff. 7 16 Robert Miller Eff. 7 47 Robert Miller Ineff. 7 59 Kraid Ashbough Ineff. 7 63 Robert Gillon Ineff. 7 8 Glenn Smith Ineff. 7 3 Milton Shapiro Ineff. 7 25 Hal Bargelt Ineff. 7 17 Leon Rudeman Ineff. 7 53 Robert McCambridge Ineff. 7 9 Arnold Morton Ineff. 7 33 Dan Austin Ineff. 7 26 Pressley Keys Ineff. 7 52 Chet Millar Ineff. 7 166 ITEM II SUMMARY OF DATA FOR EFFICIENT VOICE SAMPLES : Fundamental: Formant Frequencies Subject: Frequency : Formant Amplitudes (decibels )__ Number : (ops) ; 1 : 2 : 3 : 1 : 2 : 3 1 92 440 1407 2340 42.3 38.8 32.3 2 107 492 1356 2433 42.6 41.5 34.6 3 1 0 1 453 1323 2 0 0 0 41 37 36 4 125 487 1362 2853 41.3 37.6 32.8 5 97 513 1322 2562 41.1 41 29.7 6 106 440 1293 2195 42.3 37.5 37.5 7 109 490 1377 2610 41.8 41.3 41.5 8 1 2 0 438 1407 1948 42.6 42.1 38.1 9 97 508 1547 2092 42.3 38 31.3 1 0 93 447 1293 1893 42.1 35.5 28.8 1 1 108 483 1520 2823 42 40 30 1 2 1 2 2 555 1630 3660 42.2 41.3 32 13 113 448 1280 2600 41.8 39.5 35.7 14 1 1 0 437 1405 2237 41 37.8 33.5 15 107 460 1525 2425 42.1 39.3 25.6 16 1 2 1 460 1430 3050 41.3 39.1 34.5 17 115 453 1493 2787 41.5 37 26 18 107 460 1322 1998 40.8 41.3 38.6 167 SUMMARY OF DATA FOR EFFICIENT VOICE SAMPLES (Continued) :FundamentalîFormant Frequencies :Formant Amplitudes (decibels) Number : (ops) : 1 : 2 ; 3 : 1 : 2 : 3 19 95 430 1327 2125 40.8 39.3 32.4 2 0 113 467 1610 3060 43 40.3 37.5 2 1 97 477 1389 2430 42.6 41 35.5 2 2 98 475 1563 3125 42.1 36.0 29.5 23 1 0 0 453 1343 1383 42.2 41.3 35.2 168 SUMMARY OP DATA FOR INEFFICIENT VOICE SAMPLES i :Fundamental;Formant FrequenciesîFormant Amplitudes ; I Subject: Frequency ; (cps) _______ :____ (decibels) • ; Number : (cps) ; 1 : 2 ; 3 : 1 : 2 : 3 ' 1 92 533 1580 2547 39.2 38.3 37 2 96 503 1460 2065 40.8 36.7 33.5 3 150 497 1390 2033 41 37.2 32.6 4 152 520 1353 1150 42.3 39.2 42.5 5 107 492 1283 2700 42.0 38.3 32.5 6 117 347 1508 2895 42.1 41.6 36.5 7 130 470 1536 42.5 40 8 1 2 0 492 1612 2525 42.6 41.6 33.5 9 131 527 1578 3083 40.2 40.2 29 1 0 103 500 1300 2707 40.5 38.8 27.8 1 1 160 467 1413 42.1 40.6 1 2 246 508 1667 2600 42 40.3 30 13 150 523 1597 2833 41.5 41.5 34.5 14 1 1 2 542 1700 2732 41.8 41.8 35.5 15 135 442 1523 3000 35.8 35.5 28.2 16 127 533 1830 2525 42.3 41.1 41 17 1 2 2 505 1567 2313 41.7 38.3 37.5 18 137 475 1325 2550 42 42 39 19 166 333 1190 1775 40.7 39.7 36.4 169 SUMMARY OF DATA FOR INEFFICIENT VOICE SAMPLES (Continued) :Fundamental:Formant Frequencies :Formant Amplitudes (cps )_________ :____ (decibels ) Number : (cps) : 1 : 2 : 3 ; 1 : 2 : 3 2 0 130 500 1470 1800 42 34.4 38 2 1 116 475 1437 2413 41.5 41.3 30.5 2 2 1 0 2 483 1338 1675 41.5 37.5 31.3 23 113 442 1293 1900 42.5 36 36 University of S o u th e rn C a,|fo rn ,a APPENDIX F

Linked assets

An experimental study of perceived differences in efficient and inefficient voice production in low-pitched male voices by acoustic spectrography

Conceptually similar

PDF

An experimental study by X-ray analysis of some resonator adjustments in efficient and inefficient voice production in low-pitched male voices

PDF

An experimental study of evaluations by untrained listeners of efficient and inefficient voice production as to quality

PDF

An exploratory study of the efficacy of health care in Saudi Arabia: Objective and subjective assessment

PDF

A study of the occupational and self concepts of librarians in Southern California

PDF

Changing health-care delivery organizations in Costa Rica: an inter-institutional approach

PDF

An experimental investigation of the intelligibility of Japanese-born American-speaking male college students

PDF

Simultaneous effects on vowel duration in English: a covariance structural modeling approach

PDF

An experimental study of the relationship of parental attitudes to functional disorders of articulation in children in two different cultural environments

PDF

ISSEMYM: A study in institutionalization

PDF

An experimental study of the effects of rate of speaking upon listenability

PDF

An experimental study of the diagnostic value of vocal cues in psychosis

PDF

A clinical investigation of hearing deficiencies and some possible etiological factors in a group of cerebral palsied individuals

PDF

An experimental study of auditory threshold acuity of cerebral palsy children by PGSR and other techniques

PDF

An experimental investigation of the passive-dependency component in adult male stutterers

PDF

Cardiac cell membrane changes induced by aconitine in the production of atrial arrhythmias

PDF

An experimental investigation of the effects of phenobarbital on stage fright

PDF

An experimental study of listeners' evaluations of speech content as compared with speech delivery

PDF

An evaluation of an innovative master's degree program for administrators of ambulatory health care centers

PDF

A political-economic analysis of hospital receptivity to social work

PDF

Stability and disposition studies of cyclophosphamide metabolites

Asset Metadata

Creator Sawyer, Granville Monroe (author)

Core Title An experimental study of perceived differences in efficient and inefficient voice production in low-pitched male voices by acoustic spectrography

School Graduate School

Degree Doctor of Philosophy

Degree Program Speech

Degree Conferral Date 1955-06

Tag OAI-PMH Harvest

Format application/pdf (imt)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c36-0

Unique identifier UC11241453

Identifier DP32018~001.pdf (filename),usctheses-c36-0 (legacy record id)

Legacy Identifier DP32018

Dmrecord 0

Document Type Dissertation

Format application/pdf (imt)

Type texts