VALIDATION AND APPLICATION OF EXPERIMENTAL FRAMEWORK FOR THE STUDY OF VOCAL FATIGUE By Mark Leslie Berardi A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Communicative Sciences and Disorders Doctor of Philosophy 20 20 ABSTRACT VALIDATION AND APPLICATION OF EXPERIMENTAL FRAMEWORK FOR THE STUDY OF VOCAL FATIGUE By Mark Leslie Berardi In recent years, vocal fatigue research has been increasingly studied particularly with application to the reduction of its impact on schoolteachers and other occupational voice users. However, the concept of vocal fatigue is complex and neither well defined or well understood . Vocal fatigue seems to be highly individualized and dependent on several underlying factors or concepts. The purpose of this dissertation is to propose and support through experimentation a framework that can identify the factors contributing to vocal fatigue. T he main hypothesis i s that the change in vocal effort , vocal performance , and/or their interaction through a vocal demand ( load ) will implicate vocal fatigue. To test this hypothesis , three primary research questions and experiments were developed. For all three experiments vocal effort was rated using the Borg CR - 100 scale and vocal performance was evaluated with five speech acoustic parameters (fundamental frequency mean and standard deviation, speech level mean and standard deviation, and smoothe d cepstral peak prominence). The first research question tests whether perceived vocal effort can be measured reliably and if so, how vocal performance in terms of vocal intensity changes with a vocal effort goal . P articipants performed various speech task s at cued effort level s from the Borg CR - 100 scale . Speech acoustic parameters were calculated and compared across the specific vocal effort levels. Additionally, the test - retest r eliability across the effort levels for speech leve l was measured . Building from that experiment , the second research question was t o what degree are vocal performance and vocal effort related given talker exposure to three equivalent vocal load level s. Th is experiment had participants performing speech tasks when presented with three different equivalent vocal load scenarios (communication distance, loudness goal, and background noise) ; for a given load scenario, participants rated their vocal effort associated with these tasks. Vocal effort ratings and measures of vocal performa nce were compared across the vocal load levels . T he last research question built on the previous two and asked to what degree do vocal performance, vocal effort, and/or their interaction change given a vocal load of excess background noise (noise load) over a prolonged speaking task ( temporal load) . To test this, participants described routes on maps for thirty minutes in the presence of loud (75 dBA) background noise . Vocal effort ratings and measures of vocal performance were compared thr oughout the vocal loading task. The results indicate that elicited vocal effort levels from the BORG CR - 100 scale are distinct in vocal performance and reliable across the participants. Additionally, a relationship between changes in vocal effort and vocal performance across the various vocal load levels was quantified . Finally, these findings support the individual nature of the complex relationship between vocal fatigue, vocal effort, and vocal performance due to vocal loads (via cluster and subgroup anal ysis) ; the theoretical framework captures this complexity and provides insights into these relationships . F uture vocal fatigue research should benefit from using the framework as an underlying model of these relationships . Copyright by MARK LESLIE BERARDI 2020 v To Luke and Sam , may you never lose your voice. vi ACKNOWLEDGEMENTS F irst , I acknowledge that this research was in part supported by the N ational Institute on Deafness and Other Communication Disorders of the National Institutes of Health under Award Number R01DC012315 . Thank you to the entire Communicative Sciences and Disorders department at Michigan State University for the ir incredible support throughout my PhD journey. Additi onally, thank you to the many students from the Voice Biomechanics and Acoustics Laboratory for their help in testing and running the experiments. I would like to thank my colleagues Miriam van Mersbergen from the University of Memphis and Susanna Whitling from Lund University for their valued collaborations . I would also like to thank my committee members : Dirk Colbry, Peter Lapine, and Jeff Searl for their excellent guidance and support throughout my entire doctoral training. I would additionally like to express the deepest appreciation to my committee chair and mentor, Eric Hunter. Dr. Hunter is not only brilliant and resourceful, but also kind, compassionate, and patient. Wh en reflecting on his impact on this work , I am reminded of the quote by Sir Issac If I have seen further, it is by standing upon the shoulders of giants I was truly blessed to have the opportunity to be mentored by such an extraordinary person. Finally, I am grateful for the support of all my family, in particular , my in - laws David and Ellen, my brother Andrew, and my dad. I am most grateful for my wife Emily who sacrificed so much for me to endure the PhD journey. She was encouraging and understanding when I needed it most and my accomplishments would be nothing without her. I am additionally grateful for her time and expertise in editing this docum ent. vii TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ......................... ix LIST OF FIGURES ................................ ................................ ................................ ..................... x iii CHAPTER I: INTRODUCTION ................................ ................................ ................................ .... 1 CHAPTER I I: BACKGROUND ................................ ................................ ................................ .... 4 2.1 Historical Background of Vocal Fatigue ................................ ................................ .. 4 2.1.a Review of definitions and use of vocal fatigue ................................ ............................... 4 2.1.b Review of potential mechanisms of vocal fatigue ................................ .......................... 5 2.1.c Review of attempts to measure vocal fatigue ................................ ............................... 11 2.1.d Review of experiments in vocal fatigue ................................ ................................ ....... 19 2.2 Historical Background of Vocal Effort ................................ ................................ ... 20 2.2.b Review of measurements of vocal effort ................................ ................................ ...... 21 2.3 Theoretical Framework ................................ ................................ ........................... 25 2.2.a Review of definitions and use of vocal effort ................................ ............................... 25 2.3.a Theoretical framework for vocal effort and related terms ................................ ............ 31 2.3.b Summary of proposed defintions for voice - related terms ................................ ............ 33 2.4 Hypothesis and Research Questions ................................ ................................ ....... 34 2.4.a Main Hypothesis ................................ ................................ ................................ ........... 35 2.4.b Research Questions ................................ ................................ ................................ ....... 35 CHAPTER III: METHODOLICAL APPROACH ................................ ................................ ....... 38 3.1 Measurement ................................ ................................ ................................ ....................... 38 3.1.a Vocal Effort Measurement ................................ ................................ ............................ 38 3.1.b Acoustic Measurement ................................ ................................ ................................ . 39 3.1.c Analysis Software ................................ ................................ ................................ ......... 43 3.2 Experi ment 1 ................................ ................................ ................................ ....................... 43 3.1.a Participants ................................ ................................ ................................ .................... 43 3.1.b Instrumentation ................................ ................................ ................................ ............. 44 3.1.c Stimuli ................................ ................................ ................................ ........................... 45 3.1.d Procedure ................................ ................................ ................................ ...................... 46 3.1.e Statistical Analysis ................................ ................................ ................................ ........ 48 3.3 Experiment 2 ................................ ................................ ................................ ....................... 50 3.2.a Participants ................................ ................................ ................................ .................... 51 3.2.b Instrumentation ................................ ................................ ................................ ............. 51 3.2.c Stimul i ................................ ................................ ................................ ........................... 52 3.2.d Procedure ................................ ................................ ................................ ...................... 57 3.2.e Statistical Analysis ................................ ................................ ................................ ........ 58 3.4 Experiment 3 ................................ ................................ ................................ ....................... 59 viii 3.3.a Participants ................................ ................................ ................................ .................... 60 3.3.b Instrume ntation ................................ ................................ ................................ ............. 60 3.3.c Stimuli ................................ ................................ ................................ ........................... 61 3.3.d Procedure ................................ ................................ ................................ ...................... 62 3.3.e Statistical Analysis ................................ ................................ ................................ ........ 63 3.3.f Distribution for Collaborative Work ................................ ................................ ............. 67 CHAPTER IV: RESULTS ................................ ................................ ................................ ............ 68 4.1 Experiment 1 ................................ ................................ ................................ ....................... 68 4.1.a Demographics ................................ ................................ ................................ ............... 68 4.1.b Results ................................ ................................ ................................ .......................... 69 4.1 Experiment 2 ................................ ................................ ................................ ....................... 80 4.2.a Demographics ................................ ................................ ................................ ............... 80 4.2.b Results ................................ ................................ ................................ .......................... 81 4.3 Experiment 3 ................................ ................................ ................................ ..................... 116 4.3.a Demographics ................................ ................................ ................................ ............. 116 4.3.b Results ................................ ................................ ................................ ........................ 117 CHAPTER V: DISCUSSION ................................ ................................ ................................ ..... 146 5.1 Experiment 1 ................................ ................................ ................................ ..................... 146 5.1.a Interpreta tions and Implications ................................ ................................ ................. 147 5.1.b Limitations ................................ ................................ ................................ .................. 148 5.1.c Future Recommendations ................................ ................................ ........................... 149 5.2 Experiment 2 ................................ ................................ ................................ ..................... 149 5.2.a Interpreta tions and Implications ................................ ................................ ................. 14 9 5.2.b Limitations ................................ ................................ ................................ .................. 154 5.2.c Future Recommendations ................................ ................................ ........................... 154 5.3 Experiment 3 ................................ ................................ ................................ ..................... 155 5.3.a Interpreta tions and Implications ................................ ................................ ................. 155 5.3.b Limitations ................................ ................................ ................................ .................. 157 5.3.c Future Recommendations ................................ ................................ ........................... 158 5.4 Validation and Application of the Theoretical Framework ................................ .............. 158 CHAPTER VI: CONCLUSION ................................ ................................ ................................ . 163 APPENDICES ................................ ................................ ................................ ............................ 165 APPENDIX A: E xperiment 1 Stimuli ................................ ................................ ..................... 166 APPENDIX B: Experiment 2 Stimuli ................................ ................................ ..................... 167 APPENDIX C: Experiment 3 Stimuli ................................ ................................ ..................... 169 APPENDIX D: Informed Con sent Forms ................................ ................................ ............... 174 BIBL I OGRAPHY ................................ ................................ ................................ ....................... 182 ix LIST OF TABLES Table 3.1 Praat specifications for fundamental frequency calculation ................................ ......... 41 Table 3.2 Features used for VER clustering ................................ ................................ ................. 66 Table 4.1 Abbreviations for results of experiment 1 ................................ ................................ .... 6 8 Table 4.2 Descriptive statistics for F0 and VEL in experiment 1 ................................ ................. 70 Table 4. 3 Multiple comparison statistics for F0 and VEL for experiment 1 ................................ 7 1 Table 4. 4 Descriptive statistics for F0 sd and VEL in experiment 1 ................................ ............. 7 2 Table 4. 5 Multiple comparison statistics for F0 sd and VEL for experiment 1 ............................. 7 3 Table 4. 6 Descriptive statistics for spee ch level and VEL in experiment 1 ................................ . 7 4 Table 4. 7 Multiple comparison statistics for SL and VEL for experiment 1 ................................ 7 5 Table 4. 8 Descriptive statistics for SLsd and VEL in experiment 1 ................................ ............. 7 6 Table 4. 9 Multiple comparison statistics for SLsd and VEL for experiment 1 ............................ 7 7 Table 4. 10 Descriptive statistics for CPPS and VEL in experiment 1 ................................ .......... 7 8 Table 4. 11 Multiple comparison statistics for CPPS and VEL for experiment 1 ......................... 7 9 Table 4. 1 2 Descriptive statistics for Pearson's R for linear regression of SL and VEL for experiment 1 ................................ ................................ ................................ ................................ .. 7 9 Table 4. 1 3 Abbreviations for results of e xperiment 2 ................................ ................................ .. 80 Table 4. 14 Descriptive statistics for VER and communication distance vocal load for experiment 2 ................................ ................................ ................................ ................................ ..................... 82 Table 4. 15 Multiple comparison statistics for VER and communication distance vocal load for experiment 2 ................................ ................................ ................................ ................................ .. 83 Table 4. 1 6 Descriptive statistics for F0 and communication distance vocal load for experiment 2 ................................ ................................ ................................ ................................ ....................... 83 Table 4. 1 7 Multiple comparison statistics for F0 and communication distance vocal load for experiment 2 ................................ ................................ ................................ ................................ .. 85 x Table 4. 1 8 Descriptive statistics for F0sd and communication distance vocal load for experiment 2 ................................ ................................ ................................ ................................ ..................... 85 Table 4. 19 Descriptive statistics for SL and communication distance vocal load for experiment 2 ................................ ................................ ................................ ................................ ....................... 87 Table 4. 2 0 Multiple comparison statistics for SL and communication distance vocal load for experiment 2 ................................ ................................ ................................ ................................ .. 88 Table 4. 2 1 Descriptive statistics for SLsd and communication distance vocal load for experiment 2 ................................ ................................ ................................ ................................ ..................... 88 Table 4. 22 Descriptive statistics for CPPS and communication distance vocal load for experiment 2 ................................ ................................ ................................ ................................ .. 9 0 Table 4. 23 Multiple comparison statistics for CPPS and communication distance vocal load for experiment 2 ................................ ................................ ................................ ................................ .. 9 1 Table 4. 2 4 Descriptive statistics for VER and loudness goal vocal load for experiment 2 .......... 9 1 Table 4. 2 5 Multiple comparison statistics for VER and loudness goal vocal load for experiment 2 ................................ ................................ ................................ ................................ ....................... 93 Table 4. 26 Descriptive statistics for F0 and loudness goal vocal load for experiment 2 ............. 94 Table 4. 2 7 Multiple comparison statistics for F0 and loudness goal vocal load for experiment 2 ................................ ................................ ................................ ................................ ....................... 95 Table 4. 2 8 Descriptive statistics for F0sd and loudness goal vocal load for experiment 2 .......... 96 Table 4. 2 9 Descriptive statistics for SL and loudness goal vocal load for experiment 2 ............. 97 Table 4. 3 0 Multiple comparison statistics for SL and loudness goal vocal load for experiment 2 ................................ ................................ ................................ ................................ ....................... 99 Table 4. 3 1 Descriptive statistics for SLsd and loudness goal vocal load for experiment 2 ....... 100 Table 4. 3 2 Descriptive statistics for CPPS and loudness goal vocal load for experiment 2 ...... 101 Table 4. 3 3 Multiple comparison statistics for CPPS and loudness goal vocal load for experiment 2 ................................ ................................ ................................ ................................ ................... 103 Table 4. 3 4 Descriptive statistics for VER and background noise vocal load for experiment 2 . 104 Table 4. 3 5 Multiple comparison statistics for VER and background noise vocal load for e xperiment 2 ................................ ................................ ................................ ................................ 105 xi Table 4. 3 6 Descriptive statistics for F0 and background noise vocal load for experiment 2 ..... 1 06 Table 4. 3 7 Multiple comparison statistics for F0 and background noise vocal load for experiment 2 ................................ ................................ ................................ ................................ ................... 107 Table 4. 3 8 Descriptive statistics for F0sd and background noise vocal load for experiment 2 . 108 Table 4. 3 9 Descriptive statistics for SL and background noise vocal load for experiment 2 ..... 109 Table 4. 4 0 Multiple comparison statistics for SL and background noise vocal load for experiment 2 ................................ ................................ ................................ ................................ ................... 111 Table 4. 4 1 Descriptive statistics for SLsd and background noise vocal load for experiment 2 . 1 1 2 Table 4. 4 2 Descriptive statistics for CPPS and background noise vocal load for experiment 2 1 13 Table 4. 4 3 Multiple comparison statistics for CPPS and background noise vocal load for experiment 2 ................................ ................................ ................................ ................................ 115 Table 4. 44 Abbreviations for results of e xperiment 3 ................................ ................................ 116 Table 4. 45 Descriptive statistics for VER over time for experiment 3 ................................ ....... 1 18 Table 4. 46 Multiple comparison statistics for VER over time for experiment 3 ........................ 120 Table 4. 47 Descriptive statistics for F0 over time for experiment 3 ................................ .......... 121 Table 4. 48 Multiple comparison statistics for F0 over time for experiment 3 ........................... 123 Table 4. 49 Descriptive statistics for F0sd over time for experiment 3 ................................ ....... 124 Table 4. 50 Multiple comparison statistics for F0 sd over time for experiment 3 ........................ 126 Table 4. 5 1 Descriptive statistics for SL over time for experiment 3 ................................ .......... 127 Table 4. 5 2 Multiple comparison statistics for SL over time for experiment 3 ........................... 129 Table 4. 5 3 Descriptive statistics for SLsd over time for experiment 3 ................................ ...... 130 Table 4. 5 4 Multiple comparison statistics for SL over time for experiment 3 ........................... 132 Table 4. 5 5 Descriptive statistics for CPPS over time for experiment 3 ................................ ..... 133 Table 4. 5 6 F tests for feature significance in k - means clustering of VER for experiment 3 ...... 135 Table 4. 5 7 Independent samples t - test for comparison of means between LVLR and HVLR clusters in experiment 3 ................................ ................................ ................................ .............. 135 xii Table 4. 5 8 Bivariate correlation between features NLR and TLR for experiment 3 ................. 136 Table 4. 5 9 Cluster information including number of cases and centers for NLR and TLR for experiment 3 ................................ ................................ ................................ ................................ 136 Table 4. 6 0 Descriptive statistics for VER across time for HVLR and LVLR clusters for experiment 3 ................................ ................................ ................................ ................................ 138 Table 4. 6 1 Multiple comparison statistics for VER across time for HVLR for experiment 3 ... 140 Table 4. 6 2 Independent sample comparison of VER across time between cluster groups HVLR and LVLR for experiment 3 ................................ ................................ ................................ ........ 141 Table 4. 63 Descriptive statistics for Pearson's R for general linear model for PRE - POST acoustic changes in voice change group (YVC) for experiment 1 ................................ ........................... 142 Table 4. 6 4 Descriptive statistics for all five acoustic voice measures across PRE - POST for each vocal load response - voice change groups for experiment 3 ................................ ....................... 143 Table 4. 6 5 Independent samples t - test comparing each of the five acoustic measures across PRE - POST for each vocal load response - voice change group for experiment 3 ................................ 145 xiii L IST OF FIGURES Figure 3.1 Borg CR - 100 scale used to measure the self - perception of vocal effort . .................... 39 Figure 3. 2 Schematic for the instrumentation used in experiment 1 ................................ ............ 45 Figure 3. 3 Subway map used for the map description speech task ................................ .............. 46 Figure 3. 4 Example of the presentation of the vocal effort scale (left), target vocal effort level (upperright) and speech stimulus (right) during an experimental trial. ................................ ........ 49 Figure 3. 5 Example of the presentation of the vocal effort scale (left) and speech stimulus (right) for a map description task during an experimental trial ................................ ............................... 4 9 Figure 3. 6 Schematic for the instrumentation used in experiment 2 ................................ ............ 52 Figure 3. 7 Example map for map description task in Experiment 2. This map is instructing the participant to describe the route from Clackamas Town Center to Beaverton via Gateway. ....... 53 Figure 3. 8 Researcher's template for Experiment 2. This template includes a colorless map to be highlighted as the route is described by the partic ipant. ................................ ............................... 54 Figure 3. 9 Visual description of the communication distance vocal load. The graphic in the bottom left corner with the headset represents the participant, the ears represent the potential locations of the listener during the experiment. The jagged walls are reminders that the experiment was contained in an anechoic chamber ................................ ................................ ...... 55 Figure 3. 10 Example of map stimulus for the participant in Experiment 2. This particular example would be for the loudness goal condition of 66 dB. The participant is shown their current level and if it is lower than the target, a red arrow (shown in the upper - right) reminds them to talk louder. ................................ ................................ ................................ ....................... 56 Figure 3. 11 Visual description of the background noise vocal load. The graphic in the bottom left corner with the headset represents the participant, the ear represents the location of the listener (1 m). The loudspeakers are shown to be in a 2 m arc from the participant 30 degrees off axis. The jagged walls are reminders that the experiment was contained in an anechoic chamber. ............ 57 Figure 3. 12 Schematic for instrumentation used in ex periment 3 ................................ ................ 61 Figure 3. 1 3 Example of a map for Experiment 2. Each map had a compass in one of the corners showing north. Additionally, each map had a starting point denoted by a red circle, an ending point denoted by a red "X", and a dotted line between the two denotin g the route to be described. ................................ ................................ ................................ ................................ ....................... 62 xiv Figure 4.1 Boxplot of F0 and VEL for experiment 1 ................................ ................................ ... 70 Figure 4.2 Boxplot of F0sd and VEL for experiment 1 ................................ ................................ 72 Figure 4.3 Boxplot for SL and VEL for experiment 1 ................................ ................................ .. 74 Figure 4.4 Boxplot for SLsd and VEL for experiment 1 ................................ .............................. 76 Figure 4.5 Boxplot for CPPS and VEL for experiment 1 ................................ ............................. 78 Figure 4.6 Boxplot for VER and communication distance load for experiment 2 ....................... 82 Figure 4.7 Boxplot for F0 and communication distance load for experiment 2 ........................... 84 Figure 4.8 Boxplot for F0sd and communication distance load for experiment 2 ....................... 86 Figure 4.9 Boxplot for SL and communication distance load for experiment 2 ........................... 87 Figure 4.1 0 Boxplot for SLsd and communication distance load for experiment 2 ..................... 89 Figure 4.11 Boxplot for CPPS and communication distance load for experiment 2 .................... 90 Figure 4.12 Boxplot of VER and loudness goal vocal load for experiment 2 .............................. 92 Figure 4.13 Boxplot of F0 and loudness goal vocal load for experiment 2 ................................ .. 94 Figure 4.14 Boxplot of F0st and loudness goal vocal load for experiment 2 ............................... 96 Figure 4.15 Boxplot of SL and loudness goal vocal load for experiment 2 ................................ . 98 Figure 4.16 Boxplot of SLsd and loudness goal vocal load for experiment 2 ............................ 100 Figure 4.17 Boxplot of CPPS and loudness goal vocal load for experiment .............................. 102 Figure 4.18 Boxplot for VER and background noise vocal load for experiment 2 .................... 104 Figure 4.19 Boxplot for F0 and background noise vocal load for experiment 2 ........................ 106 Figure 4.20 Boxplot for F0sd and background noise vocal load for experiment 2 .................... 108 Figure 4.21 Boxplot for SL and background noise vocal load for experiment 2 ........................ 110 Figure 4.22 Boxplot for SLsd and background noise vocal load for experiment 2 .................... 112 Figure 4.23 Boxplot for CPPS and background noise vocal load for experiment 2 ................... 114 Figure 4.24 Graph of VER over time for experiment 3 ................................ .............................. 119 xv Figure 4.25 Graph of F0 over time for experiment 3 ................................ ................................ .. 122 Figure 4.26 Graph of F0sd over time for experiment 3 ................................ .............................. 125 Figure 4.27 Graph of SL over time for experiment 3 ................................ ................................ . 128 Figure 4.28 Graph of SLsd over time for experiment 3 ................................ .............................. 131 Figure 4.29 Graph of CPPS over time for experiment 3 ................................ ............................. 134 Figure 4.30 Scatter plot of clustered data with cluster centers. Square markers are the LVLR group, triangle markers are the HVLR group, and the circle markers are the cluster centers as labeled. ................................ ................................ ................................ ................................ ........ 136 Figure 4.31 Line graph of VER over time separate d clusters HVLR and LVLR for experiment 3 ................................ ................................ ................................ ................................ ..................... 139 Figure 4.32 Number of participants in each of the four vocal load response - voice chang e groups ................................ ................................ ................................ ................................ ..................... 142 Figure 5.1 Distribution of males and female participants across the vocal load response - voice change groups ................................ ................................ ................................ ............................. 161 Figure B.1 Map used as example during tutorial ................................ ................................ ........ 167 Figure B.2 Map used as practice during the tutorial ................................ ................................ ... 167 Figure B.3 Map used for D01, L60, and N71 ................................ ................................ ............. 167 Figure B.4 Map used for D04, L60, and N53 ................................ ................................ ............. 167 Figure B.5 Map used for D02, L66, and N53 ................................ ................................ ............. 167 Figure B.6 Map used for D04, L54, and N62 ................................ ................................ ............. 167 Figure B.7 Map used for D01, L60, and N71 ................................ ................................ ............. 168 Figure B.8 Map used for D01, L66, and N62 ................................ ................................ ............. 168 Figure B.9 Map used for D02, L54, N71 ................................ ................................ .................... 168 Figure B.10 Map used for D04, L54, N62 ................................ ................................ .................. 168 Figure B.11 Map used for D02, L66, and N53 ................................ ................................ ........... 168 Figure B.12 Map used by communication partner ................................ ................................ ...... 168 xvi Figure C.1 Image used in the map descri ption task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ..................... 169 Figure C.2 Image used in the map description task during the vocal loa ding task for experiment 3 ................................ ................................ ................................ ................................ ..................... 169 Figure C.3 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ..................... 169 Figure C.4 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ..................... 169 Figure C.5 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ..................... 170 Figure C.6 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ..................... 170 Figure C.7 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ..................... 170 Fi gure C.8 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ..................... 170 Figure C.9 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ..................... 170 Figure C.10 Image used in the map description task during th e vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ................... 170 Figure C.11 Image used in the map description task during the vocal loading task for exper iment 3 ................................ ................................ ................................ ................................ ................... 171 Figure C.12 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ................... 171 Figure C.13 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ................... 171 Figure C.14 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ................... 171 Figure C.15 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ................... 171 Figure C.16 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ................... 171 xvii Figure C.17 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ................... 172 Figure C.18 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ................... 172 Figure C.19 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ................... 172 Figure C.20 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ................... 172 Figure C.21 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ................... 172 Figure C.22 Image used in the map description task during th e vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ................... 172 Figure C.23 Image used in the map description task during the vocal loading task for exper iment 3 ................................ ................................ ................................ ................................ ................... 173 Figure C.24 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ................... 173 Figure C.25 Image used in the map description task during the vocal loading task for experiment 3 ................................ ................................ ................................ ................................ ................... 173 1 CHAPTER I : INTRODUCTION The prevalence of occupational voice problems results in a financial burden on society . Within the United States, 25% to 35% of the working population are occupational voice users (Titze, Lemke, & Montequin, 1997 ). Occupational voice users are (Hunter & Banks, 2017) . Teachers represent 4.2% of th e working population, however, teachers are about 20% of the cases in voice clinics (Titze et al., 1997) . A 2001 study (Verdolini & Ramig, 2001) estimates the societal cost of voice problems in teachers in the U.S. to be on the order of $2.5 billion annually . A more recent 2015 study in Colombia found the average indirect cost of a teacher with voice symptom s to be around US$492 a month (Cantor Cutiva & Burdorf, 2015) . Among the various voice problems reported in teachers, vocal fatigue is the most common. Calas and colleagues (1989) state that out of 100 teachers referred for dysphonia (difficulty in speaking), 96 reported experienc ing vocal fatigue. Similarly, Gotaas and Starr (1993) report that 80% of teachers reported symptoms of vocal fatigue as compared to 5% of the general population. Hunter and Banks (2017) reported that teachers exhibited higher levels of vocal fatigue as compared to vocally - healthy adults. Other occupations that report elevated levels of vocal fatigue include but are not limited to: call - center workers (Ben - David & Icht, 2016; Lehto, Laaksonen, Vilkman, & Alku, 2008) , aerobics instructors (V. I. Wolfe, Long, Youngblood, Williford, & Olson, 2002) , radio broadcasters (Cantor - Cutiva, Bottalico, & Hunter, 2018; Guzmán, Malebrán, Zavala, Saldívar, & Muñoz, 2013; Timmermans, De Bodt, Wuyts, & Van de Heyning, 2003) , and singers (Carroll et al., 2006; Kitch & Oates, 1994; Tepe et al., 2002; Yiu & Chan, 2003) . 2 Despite the prevalence of v ocal fatigue within occupational voice user s , this condition is not well defined or well understood . Previous attempts to quantify vocal fatigue are generally inconclusive and in s everal cases contradictory (Laukkanen & Kankare, 2006; Welham & Maclagan, 2003) . One reason for this is that vocal fatigue is a complex condition that has multiple possible underlying mechanisms . For example, in some cases vocal fatigue is used as a symptom (clinical history; Nanjundeswaran, Jacobson, Gartner - Schmidt, & Verdolini Abbott, 2015) and in other cases it describes the physiological alteration f ro m vocal over use (i.e. fatigued tissue; Boucher & Ayad, 2010) . The lack of a consensus definition may be a major contributor to the inconclusiveness in vocal fatigue results , and the lack of coherent framework prevents the interpretation of this misunderstanding . Another possible factor relating to th is inconsistency and misunder standing is the person to person variation in vocal fatigue. Several studies support the notion of vocal fatigue being characteristically different for each individual (Kitch, Oates, & Greenwood, 1996 ; Ternström, Bohman, & Södersten, 2006) while others showed inconsistencies within an individual (Remacle, Garnier, Gerber, David, & Petillon, 2018) . While these studies are methodologically sound, they are missing a framework for vocal fatigue that allows for a complex interaction of mechanisms (e.g. physiological and perceptual) and individual variation. In the present work, a framework is developed through the connection of vocal fatigue and related concepts of vocal load, vocal loading , vocal performance , and vocal effort since these concepts are better defined . The purpose of this dissertation is to propose and support through experimentation a fra mework for vocal fatigue that can identify the factors contributing to vocal fatigue. The remainder of the chapters are outlined as follows. Chapter 2 provide s the relevant background information o n vocal fatigue and related concepts including vocal load, vocal 3 loading, vocal performance, and vocal effort. From this background, a theoretical framework of the relationships between these concepts to guide the study vocal fatigue is presented . T he main hypotheses and research questions are presented based on the gaps in the literature and the framework . Chapter 3 provides the methodological approach to the experiments. Chapter 4 is a presentation of the results of the experiments. Chapter 5 provides a discussion of the results. Finally, Chapter 6 contains a su mmary of the conclusions from the discussion chapter. Additionally, the references and appendices are included at the end of this document. 4 CHAPTER II: BACKGROUND 2.1 Historical Background of Vocal Fatigue In recent years , vocal fatigue research ha s been increasingly studied (Cantor - Cutiva, Banks, et al., 2018) . This research demonstrates vocal fatigue as more than a singular mechanism but a dynamic system of multiple possible underlyin g and contributing mechanisms. To better delineate these uses and mechanisms, this section reviews th e definitions and use of vocal fatigue, the previously proposed mechanisms, and attempts at measur ing these mechanisms and their associated experiments . Ad ditionally, this section highlights the complexity of vocal fatigue that has resulted in inconclusive attempts to quantify vocal fatigue. 2.1. a Review of d efinitions and u se of v ocal f atigue Vocal fatigue is a common term that carries intrinsic meaning. However, within the literature there is not a common definition of vocal fatigue (Hunter & Titze, 2009) . In a review of vocal fatigue, Welham and Maclagan (2003) ptual, acoustic, or physiologic concept, indicating undesirable or unexpected changes in - perceptual characteristic of fatigue, Vilkman (2004) Emphasizing the physiological nature of vocal fatigue, McCabe and Titze (2002) central and/or peripheral fatigue of the respiratory subsystem, the phonatory subsystem, and the 5 Clinically, vocal fatigue is defined by its symptoms (Sapienza, Crandell, & Curtis, 1999; Solomon , 2008) . Various symptoms of vocal fatigue have been reported. Kostyk and Rochet (1998) summarize 18 primary symptoms of vocal fatigue: hoarse vocal quality, breathy vocal quality, loss of voice, pitch breaks, inability to maintain typical pitch, reduced pitch range, lack of vocal carrying power, reduced loudnes s range, increased vocal effort, running out of breath while talking, unsteady voice, tension in neck or shoulder, throat/neck pain, throat fatigue, throat tightness or constriction, pain on swallowing, increased need to cough or throat clear, and discomfo rt in chest, ears or back of neck (Kostyk & Rochet, 1998) . Other important symptoms include an increase in fatigue throughout the day and improvement following rest (Colton, Casper, & Leonard, 2011; Gotaas & Starr, 1993; Kitch & Oates, 1994; Solomon, 2008) . These definitions and symptoms provide a foundation to understand the underlying mechanisms, quantifications, and assessments of vocal fatigue. 2.1. b Review of potential mechanisms of vocal fatigue In exercise science, fatigue is defined in terms of central and peripheral; central fatigue themselves (Davis, 1995) . As stated above, th is approach has been reflected in discussions of vocal fatigue where central vocal could manifest as an increase in effort or feeling of increased muscular tension, and the peripheral vocal fatigu e consists of the neuromuscular and biomechanical factors within the respiratory, phonatory, and articulatory subsystems (Mccabe & Titze, 2002) . Additionally, the etiology of vocal fatigue is proposed to be either organic or functional. An organic factor of 6 phonation, and ultimately, the (Mccabe & Titze, 2002) . One of the subs ystems of voice production that potentially influences vocal fatigue i s the respiratory system. Previous work has shown a relation between pulmonary function and reports of vocal fatigue (Hunter, Maxfield, & Graetzer, 2019) . The respiratory system can be broken down into the airway (bronchi, trachea, larynx, pharynx, mouth, and nose), the lungs, and the muscles of respiration. The major muscles in respiration are the diaphragm and the intercostal muscles. There are also accessory muscles used (e.g. sternocleidomastoid and scalene muscles). The respiratory system is one of the main subsystems of the voice; it provides the air supply and pressure needed for phonation. Therefore, in theory , fatigue in the respiratory system may result in vocal fatigue. However, this effect has not been well observed (Welham & Maclagan, 2003) . n general, the capacity of the pulmonary system far exceeds the demands required for ventilation and gas exchange during exercise respiratory system is only limiting in elite athletes (McKenzie, 2012) . Leanderson and Sundberg (1988) note that only 50% of vital capacity is used in initiating speech as opposed to up to 100% of vital capacity used in ini tiating singing. Another area of potential respiratory fatigue is in individuals with pulmonary disorders. Dysphonia was noted for patients with obstructive (e.g. asthma (Abdul Latif Hamdan et al., 2017) ) and restrictive (e.g. cyst ic fibrosis ; Lourenço, Costa, & Da Silva Filho, 2014; Willis, Michael, Boyer, & Misono, 2015) pulmonary disorders. While speech within hea lthy adults would not fatigue the respiratory system, task - specific events and disordered conditions may contribute to respiratory fatigue. The phonatory system could potentially have muscle or tissue fatigue. Muscle fatigue could exist within the intrins ic or extrinsic muscles. The intrinsic muscles are a set of paired adductor 7 muscles (closing of the glottis; e.g. cricothyroid, lateral cricoarytenoid, transverse arytenoid, oblique arytenoid, and thyroarytenoid muscles). The only abductor muscles (opening of the glottis) are the posterior cricoarytenoid muscles (Rosen & Simpson, 2008) . As air flows through the larynx, adducted thyroarytenoid m uscles (vocal folds or cords) vibrate, providing a harmonic acoustic source for voiced speech sounds. Peripheral fatigue in this system can be categorized as either neuromuscular or biomechanical, in other words, laryngeal muscle fatigue or laryngeal tiss ue fatigue respectively (Mcc abe & Titze, 2002; I. R. Titze, 1999) . exercise - (Boyas & Guével, 2011) . Prolonged or repeated muscle contractions lead to changes in the chemical state o f the muscle as the body tries to maintain the level of force being produced and resist the fatigue (Boyas & Guével, 2011) . The chemical changes include depletion of energy compounds (e.g. glycogen and adenotriphosphate [ATP]) and the accumulation of lactic acid (Mccabe & Titze, 2002; Welham & Maclagan, 2003) . The glycogen depletion is related to long - term submaximal muscle contractions, while lactic acid accumulation is related to short - term maximal muscle contractions (Katch, 2009; Welham & Maclagan, 2003) . It is suggested that in occupational voice users, the laryngeal muscles activate more than 1,800 times per hour (Titze, Hunter, & . This implies that the laryngeal muscles are contracting in a prolonged and repeated manner. The sustainability of muscle contraction and its inherent fatig ue properties can be predicted by the motor unit of the muscle (Potvin & Fuglevand, 2017) . Boucher and Ayad (2010) summarize the motor unit muscle fibers from Pette and Staron (1990) the slowest, most fatigue - resistant, and generate less force compared wit h Type IIa fibers, which 8 Studies using artificial stimulation of feline and canine thyroarytenoid muscles describe high levels of fatigue resistance and that human thyroarytenoid muscles contained more than twice the proportion of Type I to Type II fibers (implicated greater fatigue resistance) than in canine and four times the proportion than in feline specimens (D. S. Cooper & Rice, 1990; Edstrom, Lindquist, & Martensson, 1974; Mascarello & Veggetti, 1979; Zealer, 1983) . One histochemical study of human laryngeal muscles by Claassen and Werner (1992) indicates that on average, the thyroarytenoid muscles consist of 53% of Type I, 36% of Type IIa, and 5% of Type IIb. The lateral cricoarytenoid muscles consist of 51% of Type I, 35% of Type IIa, and 14% of Type IIb. This fiber distribution suggests an overall slow, fatigue - resistant property of the intrinsic laryngeal muscles (Solomon, 2008; Welham & Maclagan, 2003) . However, the adductor muscles in humans are actually quite fast (Hast, 1969; Sahgal & Ha st, 1974) . Boucher and Ayad (2010) fibers in facial and laryngeal muscles. Hoh (2005) states that previous histochemical analyses are invalid because they ignore the presence of these hybrid fibers and th at single - fiber protocols must be used. Wu, Crumley, Armstrong, and Caiozzo (2000) applied a single - fiber protocol of the thyroarytenoid muscle and report 30% Type I fibers, 49% Type II fibers, and 21% Type IIx fibers. Shiotani, Westra, and Flint (1999) report that the thyroarytenoid muscle has an average of 13.5% Type I fibers, 49.2% Type IIa fibers, and 37.3% Type IIx fibers, while the lateral cricoaryt enoid muscle contained 18.8% Type I fibers, but 57.1% of Type IIa and 24.1% of Type IIx fiber s. Hoh (2005) explains that this variance is likely a result of plasticity effects. These studies showing a higher proportion of Type II fibers support the reports of the high contraction speeds of the laryngeal adductor 9 muscles . The implications are that the intrinsic laryngeal muscles may be affected by neuromuscular fatigue and that there may be a significant amount of individual variation in the fatigue of laryng eal muscles. In addition to the intrinsic laryngeal muscles, Solomon (2008) suggests that neuromuscular adjustments in the extrinsic laryngeal muscles (e.g. suprahyoid and infrahyoid [or strap muscles] groups) can contribute to the overall stiffness of the larynx a nd tension in the laryngeal muscles. This excessive tension can result in co - contraction of agonistic muscles and other inefficient muscle contractions that contribute to vocal fatigue (Solomon, 2008) . Prior to a discussion on biomechanical fatigue , consider the composition of the vocal folds. A vocal fold consists of five main layers. The superficial layer (epithelium) consists of squamous epithelium (which is continuous throughout the trachea, pharynx, and mouth). The next three layers have differing consis tencies but are often referred to together as the lamina propria. The superficial layer of the lamina propria can be compared to the consistency of soft gelatin and is ng of this layer due to fluid; often associated with smokers). The next two layers are sometimes referred to together as the vocal ligament. The intermediate lamina propria is characterized by elastic fibers (a rubber band - like consistency). The deep lamin a propria primarily contains collagenous fibers (thread - like consistency). The final layer is the vocalis muscle of the thyroarytenoid which is the main body of the structure and is very stiff (Rosen & Simpson, 2008) . In the case of biomechanical fatigue, parallels are often drawn to the use of fatigue in the from stress (force (Solomon, 2008) . During regular phonation, the lamina propria is subject to continuous stress 10 and strain. Stress from the numerous collisions of the two folds and strain from anterior - posterior posturing (elongation or shortening) as well as other deformations from the various laryngeal muscles. Titze (1994) discusses mechanical stresses on the vocal folds and concludes that to vibrate on the order of 1.4 million times during their work day (I. R. Titze et al., 2007; E Vilkman, 2004) . The magnitude of the number of collisions for a population with high prevalence of vocal fatigue is an intriguing correlation. However, two reviews of vocal fatigu e concluded that while the mechanical stresses exist, their contributions to lamina propria fatigue (and vocal fatigue in general) are not well understood (Solomon, 2008; Welham & Maclagan, 2003) . Other biomechanical factors of vocal fatigue include increased viscosity, stiffness, an d lesions (I. R. Titze, 1994) . absorpt ion ( Solomon, 2008) . McCabe and Titze (2002) suggest that a change in tissue viscosity circulation throughout the lamina propria to remove inhib itory elements such as lactic acid, or an - pitched phonation, increased tissue viscosity was measured, as well as increased frictional energy loss and an increase in heat dissipation (Donald S. Cooper & Titze, 1985) . Solomon (2008) concludes that an interaction - muscu lar biomechanical properties on vocal function makes the study of vocal fatigue more complex than the study of fatigue involving most other 11 The third subsystem relating to vocal fatigue is the articulatory and resonance system. There are three main resonating cavities in this system: the pharynx (comprising the laryngopharynx, oropharynx, and nasopharynx cavities), oral cavity, and nasal cavity. The major articulators used in speech are the velum (or soft palate), hard palate, teeth, tong ue, lips, and cheeks. The muscles used in the articulator and resonance systems primarily consist of fatigue - resistant Type I and IIa muscle fibers (this is the case for the tongue, orbicularis oris [lips], and buccinator [cheek] muscles (Solomon, 2006) ). Like the respiratory system, normal speech in healthy individuals is not fatiguing. However, individuals wi th varied disorders may experience fatigue in the articulatory and resonance muscles (e.g. lingual fatigue in myasthenia gravis (Wenke, Goozee, Murdoch, & LaPointe, 2006) or lingual and lip fatigue in laryngectomees (S earl & Knollhoff, 2018) . 2.1. c Review of attempts to measure vocal fatigue Many methods are used to assess and quantify vocal fatigue. In general, these methods attempt to measure either the physiological mechanisms of vocal fatigue or the perception of vocal fatigue experienced by the individual. These measurement methods are ca tegorized as physiological, aerodynamic, acoustic, or self - report assessments. Physiological methods include assessments where the muscles (particularly the intrinsic laryngeal muscles) are measured directly. Aerodynamic methods are assessments that measur e flow or pressure of the aerodynamic source of phonation. Acoustic methods include both objective and perceptual measures where objective measures are calculated from acoustic samples and perceptual measures are judgements of the acoustic samples by liste ners. Finally, self - report assessments are include multiple measurement methods. 12 Physiological assessment of vocal fatigue attempts to directly measure the mec hanism of the larynx. This is done through imaging, electroglottography (EGG), electromyography (EMG), or through physical and computational models. The types of laryngeal imaging used to assess vocal fatigue include rigid endoscopy (Eustace, Stemple, & Lee, 1996; A L Hamdan, Sibai, & Rameh, 2006; Kelchner, Toner, & Lee, 2006; Linville, 1995; Niebudek - - Kowalska, 2007; Pearl Solomon & Stemm le DiMattia, 2000) , flexible nasal endoscopy , and high speed endoscopy (Doellinger, Lohscheller, McWhorter, & Kunduk, 2009; Whitling, Lyberg - Åhlander, & Rydell, 2017) . Various measures are observed and calculated from laryngeal imaging. These include vocal fold edema (Scherer et al., 1987) , glottal closure patterns (subjective evaluation of the open and closure of the vocal folds (Eustace et al., 1996; Kelchner et al., 2006; Linville, 1995; Niebudek - Bogusz et al., 2007; Stemple, Stanley, & Lee, 1995) ), laryngeal lengt h - to - width ratio of the glottis (Yiu et al., 2013) , amplitude of vocal fold vibration (Eustace et al., 1996; Niebudek - Bogusz et al., 2007; Solomon, Glaze, Arnold, & van Mersbergen, 2003; Vintturi et al., 2001) , phase symmetry of vibration (Eustace et al., 1996) , and quality of mucosal wave (Eustace et al., 1996; Niebudek - Bogusz et al., 2007) . Other studies use laryngeal imaging as a screening to confirm the presence or absence of la ryngeal pathology . Some of the studies report no laryngeal imaging changes associated with vocal fatigue (Eustace et al., 1996; Kelchner et al., 2006; Niebudek - Bogusz et al., 2007; Whitling et al., 2017) , while others reported contradicting changes (Solomon, Glaze, Arnold, and van Mersbergen , 2003) report a decrease in amplitude of vocal fold vibration associated with vocal fatigue while Vinturri et al. (2001) report an increase). Observations of incomplete glottal closure, such as an anterior glottal chink or a spindle - shaped glottis (Linville, 1995; Pearl Solomon & Stemmle DiMattia, 2000; Solomon et al., 2003; Stemple et al., 1995) , suggest effects 13 of neuromuscular fatigue on the adductor musc les. Observations of vocal fold edema and changes in quality of the mucosal wave support the theories of tissue changes in vocal fatigue. The major drawback to laryngeal imaging is that it is a relatively invasive measurement requiring particular equipment . Additionally, the quantitative measurements of imaging are computationally expensive and partly subjective (Doellinger et al., 2009) , although recent research aims to address those concerns (Naghibolhosseini, Deliyski, Zacharias, de Alarcon, & Orlikoff, 2018; Poburka, Patel, & Bless, 2017; Zacharias, Deliyski, & Gerlach, 2018) . Elec troglottography (EGG) measures the level of contact between the vocal folds during voicing. It does this with a pair of electrodes placed on the surface of the neck such that the larynx is between them. The electrodes measure the electrical impedance of th e larynx which changes with the contact area of the vocal folds. This can be useful in estimating the time the vocal folds spend open and closed (Childers, Hicks, Moore, Eskenazi, & Lalwani, 199 0) . Due to obvious variations in biology and therefore neck impedance, EGG cannot be used to absolutely measure glottal contact area. The most commonly used EGG measure is the closed quotient (the ratio of closed time and combined closed and open time). Electroglottography presents a less invasive physiological measurement of vocal function. However, the usefulness of EGG in studies of vocal fatigue is inconclusive (Buekers, 1998; Laukkanen, Mäki, & Leppänen, 2009; V. I. Wolfe et al., 2002) . Electromyography (E MG) is a more direct measurement of the electrical activity in muscles and has been used to show changes in the laryngeal muscles (the lateral cricoarytenoid in particular) as a result of vocal fatigue (Boucher, Ahmarani, & Ayad, 2006; Boucher & Ayad, 2010; Rubin et al., 2005) . These results strongly support the proposition that the laryngeal muscles are not fatigue resistant. Although EMG is useful in investigating neuromuscular 14 fatigue, it is highly invasive and not practical in most vocal fatigue research. Some recent work has used surface EMG on the neck to estimate laryngeal muscle activations with the intent to apply the methods to vo cal fatigue research (N. R. Smith et al., 2016) . As stated above, the stress and strain of vibration o n vocal fold tissues may lead to tissue fatigue and, thus, vocal fatigue. Directly measuring these stresses and strains in vivo is problematic, therefore modeling becomes a useful tool. Researchers use physical or synthetic (Drechsel & Thomson, 2008; Spencer, Siegmund, & Mongeau, 2008 ; I. R. Titze, 1994) , computational , and ex vivo human and animal (Doellinger & Berry, 2006; Matsushita, 1975; Katherine Verdolini, Chan, Titze, Hess, & Bierhals, 1998) models to study the vibratory properties of the vocal fo lds. Models are continually being used to study biomechanical changes to vocal folds. Aerodynamic measurements serve two primary assessments of the vocal fatigue. First, to investigate biomechanical changes in the vocal folds and how that might impede wit h the aerodynamic output. Second, to assess the pulmonary capacity of the research subjects. The most common and consistent aerodynamic measure used in vocal fatigue research is phonation threshold pressure (PTP). This is the minimal level of lung pressure below the vocal folds required for sustained phonation. True PTP is measured with a tracheal puncture, but it can be well estimated noninvasively by measuring the oral pressure during a bilabial stop consonant (Fisher & Swank, 1997) . Many (but not all) have found significant increases of PTP with vocal fatigue (Chang & Karnell, 2004; Enflo, Sundberg, & McAllister , 2013; Erickson - Levendoski & Sivasankar, 2011; Kagan & Heaton, 2017; Pearl Solomon & Stemmle DiMattia, 2000; Erkki Vilkman, Lauri, Alku, Sala, & Sihvo, 1999; Whitling et al., 2017) . This finding suggests 15 biomechanical changes in the vocal folds resulti ng in increased viscosity and therefore more pressure is required for them to begin sustained oscillation. This is further supported by the relationship between PTP and hydration (Pearl Solomon & Stemmle DiMattia, 2000; Mahalakshmi Sivasankar, Erickson, Schneider, & Hawes, 2008) . Spirometry is used to measure pulmonary capacity in studies of vocal fatigue (Koufman & Blalock, 1988) . Maxfield, Hunter, and Graetzer (2016) used spirometry to show a relationship between vocal fatigue and pulmonary function. Acoustic assessment is the most common type of assessment of vocal fatigue. This is likely because it is a noninvasive measurement using relatively common equ ipment (i.e. microphone and signal recorder). The acoustic properties of speech are directly related to physiological mechanisms that produce speech (air supply, vocal fold vibration, and pharyngeal, oral, and nasal articulations). The most obvious relatio nship is that the measured speaking fundamental frequency (F0; usually reported as cycles per second or Hertz but sometimes as semitones which are 1/12 th of a doubling of frequency ) relates to the number of vocal fold vibratory cycles per second. If there are neuromuscular or biomechanical changes, then the acoustic signal should change accordingly. Many different objective measures are calculated from the acoustic signal. The most (Cho, Yin, Park, & Park, 2011; Stemple et al., 1995) , mean eseleer et al., 2016; Jonsdottir, Laukkenen, & Siiki, 2003; Laukkanen, Ilomäki, Leppänen, & Vilkman, 2008; Laukkanen & Kankare, 2006; Laukkenen et al., 2004; Lehto, Laaksonen, Vilkman, & Alku, 2006; V. I. Wolfe et al., 2002) , maximum Sivasankar, 2002 ) , and variance (Ben - David & Icht, 2016; Cho et al., 2011; V. I. Wolfe et al., 16 2002) . The voicing intensity or speech level (reported as a sound pre ssure level in decibels) is also commonly measured in terms of its minimum al., 1999) , mean (Ben - David & Icht, 2016; Jonsdottir et al., 2003; Laukkanen et al., 2008; Laukkanen & Kankare, 2006; Laukkenen et al., 2004; Lehto et al., 2006; V. I. Wolfe et al., 2002) , and variance (Ben - David & Icht, 2016; Bottalico, Cantor Cutiva, & Hunter, 2017; Cho et al., 2011) . Sometimes the range of F0 and the range of voicing intensity are combined to calculate a voice range profile (Damsté, 1970; I. R. Titze, 1992) which is used to investigate changes with vocal fatigue (E. Holmberg, Ihre, & Söder sten, 2007; Wingate, Brown, Shrivastav, Davenport, & Sapienza, 2007) . Measures of perturbation, which include jitter (cycle - to - cycle frequency instability), shimmer (cycle - to - cycle amplitude instability), and harmonic - to - noise ratio (HNR; ratio of the harmonic energy to noise energy in the acoustic signal), are also commonl y used in vocal fatigue assessment Andrews, & Schmidt, 1991; Lau kkanen et al., 2008; Laukkanen & Kankare, 2006; Scherer et al., 1987; Verstraete, Forrez, Mertens, & Debruyne, 1993; V. I. Wolfe et al., 2002) . Less common but important acoustic measures include relative fundamental frequency (RFF; measure of the stabi lity of t he o ff set and o n set of vocal fold vibration in speech) and cepstral peak prominence (CPP; a robust spectral - cepstral measure of vibratory periodicity). Vocal fatigue studies have only recently started to use these two measures, but they present mo re promising results than the previously mentioned measures (Fujiki, C hapleau, Sundarrajan, McKenna, & Sivasankar, 2017; Gorham - Rowan, Berndt, Carter, & Morris, 2016; Kagan & Heaton, 2017) . A recent study measures changes in the formants (acoustic resonances of the vocal tract) associated with vocal fatigue (Pellicani, Fontes, Santos, Pellicani, & Aguiar - Ricz, 2018) . Several studies utilized the commercial multidimension al voice program (Deliyski, 1993) to calculate F0, perturbation 17 measures, and a dozen other acoustic parameters (Boucher, 2008; Boucher & Ayad, 2010; . Although speech acoustic measur es are commonly used, their results vary. In almost every common acoustic measure mentioned above, studies report increases, decreases, or no change associated with vocal fatigue. Kitch, Oates, and Greenwood (1996) report the acoustic parameters changing in opposite directions with different participants within the same study. Another vocal fatigue study reports consi derable inter - and intra - subject variability with these acoustic measures (Remacle et al., 2018) . Another way the acoustic signal is used is in perceptual ra tings of voice quality. In these cases, listeners judge the quality of voice based on predetermined scales. The two most common instruments are the Grade, Roughness, Breathiness, Asthenia, Strain (GRBAS) scale (Cho et al., and the Consensus Auditory - Perceptual Evaluation of Voice (CAPE - V) scale (Kagan & Heaton, 2017; Kempster, Gerratt, Abbott, Barkmeier - Kraemer, & Hillman, 2009) . In studies o f vocal fatigue, these instruments are used as either measurement variables or screenings for vocal health. Another instrument, inability to produce soft voice (IPSV) is a self - evaluation from a subject after they try to produce a high - pitch, low - amplitude vocal sound (Bastian, Keidar, & Verdolini - Marston, 1990; Halpern, Spielman, Hunter, & Titze, 2009; Hunter & Titz e, 2009; Kagan & Heaton, 2017) . This is different from other self - ratings (described below) as it is based on the performance of a task that would be more difficult for someone experiencing vocal fatigue, rather than a rating of internal feelings of fat igue or effort. The previous measurements aim to assess peripheral vocal fatigue. Central vocal fatigue is assessed through self - reporting of the subjects, typically in the form of questionnaires. A variety 18 of self - reporting instruments have been used to study vocal fatigue. One of the most common instruments is self - ratings of vocal or phonatory effort during the vocal fatigue experiment (Carroll et al., 2006; Chang & Karnell, 2004; A L Hamdan et al., 2006; Mccabe & Titze, 2002) . Other commonly asked self - rating questions include laryngeal pain, laryngeal fatigue, voice quality, or prior vocal complaints (Buekers, 1998; Fabron et al., 2015; Kelchner et al., 2006; Laukkanen et al., 2008) . Most of these instruments ask subjects to provide a rating from 1 to 10 (or something similar). These instruments are continually developed using mechanisms fro m the survey literature including visual analog scales (VAS; Lehto et al., 2008; Pellicani et al., 2015) , Likert scales (Whitling et al., 2017) , and the Borg CR - 10 scale (Ford Baldner, Doll, & van Mersbergen, 2015; van Leer & van Mersber gen, 2017) . Within the general field of voice assessment, indexes have been developed and a common one is the Voice Handicap Index (VHI or VHI - 1 ; Jacob son et al., 1997) . This has been applied to vocal fatigue research (Cho et al., 2011; Niebudek - Bogusz et al., 2007; Wingate et al., 2007) but also led to the creation of an index specifically for vocal fatigue. Nanjundeswaran and colleagues developed the Vocal Fatigue Index (VFI) to be u sed in self - reporting of trait vocal fatigue (Nanjundeswaran et al., 2015; Nanjundeswaran, van Mersbergen, & Morgan, 2017) . Although it is a recently developed instrument, it has already been used to show differences in vocal fatigue in occupational voice users (dos Santos, Silverio, Dassie - Leite, Costa, & Siqueira, 2018; Hunter & Banks, 2017) . Self - ratings are important because they aim to measure the perceived experience of vocal fatigue . This contributes to understanding a different facet of fatigue not detected using other assessments. However, the measurement of peripheral vocal fatigue may only be quantifying the trait or systemic factor of fatigue. As a result, these instruments cannot be used to meas ure changes in vocal fatigue throughout a communication event thereby limiting its application . 19 2.1.d Review of experiments in vocal fatigue Quantification of vocal fatigue requires a subject who is vocally fatigued. There are different methods of experim entation that attempt to induce vocal fatigue . The most common type of vocal fatigue experiment is a vocal loading task (VLT ; the term vocal loading is described in more detail in the following section ). Here the subjects undergo some variation of a prolonged phonation task and it is assumed that due to the VLT the subject will experience some amount of vocal fatigue. Vocal loading tasks have taken many different forms. In a review of VLTs by Fujiki and Sivasankar (2017) , it was found that th e most common duration of VLT was two hours with some being as short as 15 minutes and as long as 3.75 hours. Some of the VLT involved multiple experimental sessions which ranged from two to seven sessions. A few of the studies allowed for the subjects to terminate the experiment when they felt fatigued. The most common type of loading task was prolonged, loud reading. When a loud voice was elicited, s everal used ambient noise in the VLT (De Bodt, Wuyts, Van De Heyning, Lambrechts, & Abeele, 1998; Erickson - Levendoski & Sivasankar, 2011; Whitling et al., 2017) , while others required a vocal loudness target (Gelfer et al., 1991; Linville, 1995; Neils & Yairi, 1987) . Others have investigated the relationship between acoustic environment (in particular noise and reverberation) and vocal fatigue (Bottalico, Cantor Cutiva, et al., 2017; Bottalico, Gra etzer, & Hunter, 2016; Kristiansen et al., 2014) . One limitation to many of these studies is that measurements are only taken at the beginning and the end of the VLT. A few studies have also taken measurements at periodic intervals during a VLT (Boucher, 2008; Buekers, 1998; Laukkenen et al., 2004; Xue, Kang, Hedberg, Zhang, & Jiang, 2019) which may provide more information of how vocal fatigue develops throughout a VLT. The other major limitation is that 20 many of the studies assume that vocal fatiguing is occurring as a result of the vocal loading which may not be the case. Vocal loading tasks are likely the most common way to investigate vocal fatigue because the researchers can control many of the factors in the experiment. However, others have taken a more ecologically valid approach to vocal loading by assessing vocal fatigue t hroughout a workday in real world environments . This has been done for teachers (Halpern et al., 2009; Jonsdottir et al., 2003; Laukkanen & Kankare, 2006; Rantala, Paavola, Körkkö, & Vilkman, 1998; Remacle et al., 2018) , call - center workers (Ben - David & Icht, 2016; Lehto et al., 2008) , and radio broadcasters (Cantor - Cutiva, Bottalico, et al ., 2018) . Other methods use surveys in populations where vocal fatigue is a common complaint (e.g. teachers and individuals with voice disorders) to study the prevalence of vocal fatigue and possible associated factors (Bastian & Thomas, 2016; dos Santos et al., 2018; Hunter & Banks, 2017; Munier & Kinsella, 2008) . Another methodology divides the subject pool into groups based on self - report ed vocal symptoms Geneid, 2017; La ukkanen et al., 2008) or measured level of overall fatigue (Cho et al., 2011) and studies the differences between th e defined groups. Throughout all of these studies, and in particular VLT, the only consistent measure associated with vocal fatigue has been perceived vocal effort. 2.2 Historical Background of Vocal Effort Perceived effort or exertion given a task or from the result of a task has been the focus of research in a broad range of fields such as exercise science, cognitive science and psychology , and audiology (listening effort) . Within the context of vocal fatigue, p erceived vocal effort w as the only consistent measurement of vocal fatigue in the vocal loading tasks. To better understand 21 vocal effort and its relationship to vocal fatigue, vocal load, and vocal loading, this section reviews the definitions , uses and measurements of vocal eff ort. 2. 2 .b Review of measurements of vocal effort The previous section concludes that vocal effort is more nuanced than vocal intensity and therefore requires other measurements. There are two main types of measurements of vocal effort: perceptual and phys iological. The perceptual measurements include either the perception of vocal effort from the talker or from a listener. Physiological estimates of vocal effort generally include aerodynamic and acoustic. Understanding the types of measurements that are ef fective in quantifying vocal effort will contribute to the understanding of the phenomenon. T he earliest type of measurements of vocal effort are from listeners. Many of these studies used scales developed for that particular study and were not used in an y standard way. One example of such a scale is a nine - point vocal effort rating of prerecorded harsh voices (Thomas - kersting & Casteel, 1989) . Eventually two main listener rating scales were developed, validated, and standardized for voice research. One is the Grade, Roughness, Breathiness, Asthenia, Strain (GRBAS) scale (Hirano, 1981 ) Voice (CAPE - V) scale (Kempster et al., 2009) which includes the same definition and application of the word strain. Beyond these two scales and others that were singularly developed, visual - analog scales (VAS) have been used (Eadie & Stepp, 2013) . This perspective of vocal effort is important clinically as clinicans are the listener - judges for the evaluation of vocal effort. For the quantification of self perception of effort , VAS have traditionally been the most common (Paes, 2017; Shewmaker, Hapner, Gilman, Klein, & Johns, 2010; Södersten, Granqvist, Hammarberg, & Szabo, 2002; Tanner et al., 2010; Warrick et al., 2000) . There has been recent 22 research that report vocal ef fort scales to match the development of standardized physical exertion scales, namely the Borg CR - 10 (van Leer & van Mersbergen, 2017) . The Borg CR - 10 lates more accurately to how physical exertion is perceived. This same design has been implemented into a single - question pictorial scale on perceived vocal effort/exertion called the OMNI Vocal Effort Scale (OMNI - VES; Shoffel - havakuk et al., 2019) . One problem with the Borg CR - 10 and OMNI - VES is that the majority of the scale is in the difficult exertion portion of the scale and the healthy voices tend to not be in that section of the scale resulting in a resolution problem. Visual analog scales do not have this problem because they are continuous. However, using VAS loses the be nefits of using the anchors in the Borg CR - 10 or OMNI - VES. One possible solution is to convert the Borg CR - 100 to a vocal effort scale. This scale would have the benefits of the anchors of the Borg CR - 10 with the resolution closer to the VAS. Additionally, the VAS is a linear scale while the Borg CR - 10 has logartihmic sp ac ing of the anchors that correlate best with human perception of exertion (Borg & Löllgen, 2001) . Here the vocal - effort adapted Borg CR - 100 scale is used to quantify percieved vocal effort. Self - rating surveys have been developed, standardized and widely used (similar to GRBAS or CAPE - V) that contain components related to vocal effort. The Voice - Handicap Index (VHI; Jacobson et al., 1997) has been used to measure self ratings of vocal effort (Sampaio & Jos, 2012) vocal effort. A similar scale to the VHI was developed for the use of vocal fatigue research, the Vocal Fatigue Index (VFI; Nanjundeswaran et al., 2015) . This survey ask 23 and vocal fatigue interact. These scales are very useful in collecting state levels of vocal effort. vocal effort across an experiment. In many cases, vocal effort is regarded as a perceptual conditi on, however attempts have been made to quantify vocal effort. One of the most common physiological measurements of vocal effort is phonation threshold pressure (PTP; Pearl Solomon & Stemmle DiMattia, 2000; Rosenthal et al., 2014; Solomon, Glaze, Arnold, & van Mersbergen, 2003) . Phonation threshold pressure is the minimum level of lung pressure required to produce sustained vocal fold oscillations. T ypically PTP cannot be measured directly (this would require a tracheal puncture) but can be well estimated by measuring the intraoral pressure of a bilabial plosive. It is unclear as to whether high PTP results in higher vocal effort or whether excessive vocal effort results in higher PTP. On one hand a lower PTP for a given task could represent improved vocal ability which would likely mean lower vocal effort on certain tasks. It could also be the case that elevated vocal effort could cause extra tension or miscoordination of the vocal mechanism which would result in increases in PTP. Either way, PTP seems to be a measure strongly linked to vocal effort. The major drawback is that the measurement is not as easy as acoustic measurements and its variability is affected by many other factors such as hydration (Verdolini et al., 1994) and fatigue (Chang & Karnell, 2004) . Many different acoustic measurements have been used to quantify change in vocal effort. Some have already been mentioned (fundamental frequency, formant amplitude and frequency, speech level ). Some of the acoustic measures are time based while others are spectral based. A promising time - based measure is relative fundamental frequency (RFF). Relative fundamental 24 frequency estimates differ ences in vocal fold vibration during voiced onsets. During regular speech, voicing starts (abducts) and stops (adducts) for different speech sounds. The offsets and onsets from the abduction and adduction result in changes to the rate of vibration of the v ocal folds until the vibration is steady. The rate of vocal fold vibration is the fundamental frequency. Relative fundamental frequency compares the fundamental frequencies of the individual glottal pulses after a voiceless consonant to the fundamental fre quency of the steady - state vowel. Lien and Stepp (2015) used RFF to track changes in healthy individuals as they changed their vocal effort. The authors also showed strong relationships between RFF and perceptual and aerodynamic measures of vocal effort. Relative fundamental frequency has been used to show differences in populations with characteristically high vocal effort: vocal hyperfunction (Stepp, Sawin, & Eadie, 2012) and ADSD (Eadie & Stepp, 2013) . Despite the promising nature of RFF , in practice it is subjective to estimate the glottal pulses relating to vocal offset or onset (e.g. i t is not always clear as to when the vowel begins or what exactly is the steady state ) . The analysis can also be computationally expensive. Despite these shortcomings, it offers a physiologically - minded acoustic measure of vocal effort that has been well validated. Other acoustic measures which relate to vocal ef fort include cepstral - peak prominence (CPP) and mel - frequency cepstral coefficients (MFCC). Both of these measures use the cepstrum which is the Fourier transform of the power spectrum of speech. Cepstral peak prominence is a measure of the decibel differe nce between the magnitude of the strongest cepstral peak and a regression of the average cepstral energy. The CPP often relates a measurement of the periodicity of the voice and has been widely used in voice disorder research (L eong et al., 2013) . Rosenthal, Lowell, and Colton (2014) showed higher values of CPP in maximal effort speech and significantly lower CPP during minimal effort speech. Although this seems promising, it was 25 noted that CPP often shares a linear relationship with vocal intensity and in this case, the maximal effort and minimal effort speech had a significant difference in vocal intensity. It is hard to discern whether the difference in CPP is a result of the change in intensity or the change in effort. Mel - frequency cepstral coefficients are bands of energy with the cepstrum. They are commonly used in speech signal processing, particularly in the area of speech recognition. Zelinka, Sigmund, and Schimmel (2012) used MFCC to t rain a computer to automatically classify categorical vocal effort levels from a database. The main drawback of MFCC is the lack of interpretation. Although the computer could correctly identify levels of vocal effort, the process it used to do that is not interpretable. 2.3 Theoretical Framework This section develops a theoretical framework to address vocal fatigue from the related concept of vocal effort . This is because the perce ived e ffort , compared to fatigue, is more universally studied and defin e d in a wide range of fields . This framework is partially modeled after the listening effort framework (Pichora - Fuller et al., 2016) . The proposed theoretical framework of vocal eff ort includes vocal fatigue and other related concepts from the voice literature such as vocal load, vocal loading, vocal ability, and vocal performance. Here they are provided with specific definitions to establish the framework. Several of these concepts are discussed in recent review s (Cantor - Cutiva et al., 2018; Hunter et al., 2020) . 2. 2 .a Review of definitions and use of vocal effort One of the early definitions of vocal effort focuses on communicative distance as the catalyst for vocal effort. Traunmüller and Eriksson (2000) ordinary speakers vary when they adapt their spe ech to the demands of an increased or decreased 26 talker - whether the listeners are actually rating the vocal effort and not just the loudness of the voice. If while using a loudspeaker the desire is to increase the acoustic radiation (in other words the communication distance), one needs to only increase the amplitude of t In the case of the voice, if the only goal is the increase of speech acoustic radiation, then, in theory, one only needs to be louder. However, this study accounted for that. The authors investigated a wide range (0.3 - 187.5 m) of c ommunication distance but also randomly modulated the amplitude of the speech samples. The listeners were still reliable in rating differences in the perceived talker - listener distance, which the authors use to quantify vocal effort. They conclude that voc al effort is a physiological exertion different from vocal intensity that accounts for changes in communication distance. This definition of vocal effort is widely used (being cited over 250 times) but it depends on subjective observations of listeners and vocal effort is arguably a sensation experienced and measurable only from the talker ( Hunter et al., 2020) . The notion of vocal ef fort being related to communication distance extends to other studies. However, instead of listeners rating a perceived talker - listener distance, the actual talker - listener distance was used as vocal effort. In these cases, the greater the distance, the gr eater the vocal effort. Liénard and Di Benedetto (1 999) used three different distances (close 0.4 m, normal 1.5 m, and far 6 m) and Pelegrín - García, Smits, Brunskog, and Jeong (2011) studied vocal effort at four distances (1.5, 3, 6, and 12 m). These studies found changes in fundamental frequency, formant frequencies and amplitudes, and sound pressure level of the speech to be related to this distance based vocal effort, or rather increases in communication distance. Again, the pattern of speech characteristics, other than the loudness, changed, with increasing talker - listener distance which i s consistent with Traunmüller and Eriksson. These studies illustrate one 27 possible component of vocal effort vocal adaptations to communication distance. It is straightforward to illustrate how vocal effort would increase with communicative distance. Recall any experience of trying to communicate with someone across a large room this requires greater vocal effort than talking to someone within close proximity. However, an observable example of trying to converse with a close neighboor in a noisey environ me nt refutes that this is the only component of vocal effort. In addition to communication distance, vocal effort may be needed to accommodate within an environment with excessive noise. It has been well understood that speech changes in response to background noise, this is often called the Lombard effect (Lombard, 1911) . Simply increased noise levels. Junqua (1993) explains the Lombard effect within the context of vocal effort as th vocal level. If the Lombard effect is truly a vocal effort modulation, then other speech characteristics should change. Similar changes in speech patterns (such as fundamental frequency and formant frequencies) that were found in increased communicative distances relate to vocal effort in response to background noise (Lu & Cooke, 2009; Vogel, Fletcher, Snyder, Fredrickson, & Maruff, 2011) . Another analogy can be drawn between Traunmüller and Eriksson (2000) and Södersten, Granqvist, Hammarberg, and Szabo (2002) who showed li stener judgements of vocal effort in increasing levels of background noise. Södersten, Granqvist, also increased with the noise. These studies show similarities between vocal effort in the context of communicative distance and in the context of excessive background noise. Both of these 28 conditions are environmental factors that affect the successful transmission of a spoken message. It can then be assumed that oth er environmental factors may similarly affect vocal effort. Another communication environment factor that could relate to increases in vocal effort is room reverberation. Reverberation presents a slightly different problem than static background noise. Her e the speech of the individual contributes to the noise in the form of reverberation. As has shown that room reverberation can increase vocal effort (Berar di, Hunter, & Leishman, 2015; Bottalico, Graetzer, et al., 2017; Hunter et al., 2015) . However, Bottalico (2017) demonstrates that the effect of reverberation on vocal effort is not linear and that some reverberation can be benifi c ial to the talker and reduce vocal effort. In this study, the participants spoke and then rated their vocal effort in three different reverberation conditions (anechoic, reverberant, and semi - reverberant). Additionally, the part icipants were recorded with and without acrylic glass panels 0.5 m from the talker to provide strong early reflections. If vocal effort in reverberation were similar to the Lombard effect then it would be the case that vocal effort would be the smallest in the anechoic condition and the most in the reverberant condition with the panels adding more of a response to vocal effort. Bottalico reported that vocal effort was the highest in the anechoic condition and that the presence of the panels reduced vocal ef fort. This implies that a certain amount of reverberation is actually preferred for support in the auditory feedback. When comparing the combination of noise and reverberation on vocal effort, Cipriano, Astolfi, and Pelegrín - García (2017) reported that vocal comfort related more with the reverberation is not as predictable as communication distance or bac kground noise, it still is an environmental factor in vocal effort. 29 In each of these cases, talker - listener distance, background noise, and room reverberation, have been shown to relate to the notion that vocal effort is an accommodation to the limitation s of the communication environment. These studies also illustrate that vocal effort is more than increasing vocal intensity. However, in many cases vocal effort is used analogously with vocal intensity and is measured as the sound pressure level of speech in decibels (dB). In fact the - weighted speech level at 1 m distance in front of (Cushing, Li, Cox, Worrall, & Jackson, 2011; Rosenthal et al., 2014) despite vocal effort being more nuanced than vocal intensity. Another example of the difference between vocal effort and vocal intensity is in the speech - loudness - e ffort hypothesis by Rosenblum et al (1991) . This effort of the talker and not the intensity of speech signal. Th e authors summarized research which showed perceived loudness to be more related to listener - estimated vocal effort than the SPL of the signal. Whether or not this hypothesis is completely accurate, it is the case that loudness perception and vocal effort are more than the amplitude of a speech signal. Vocal intensity alone is a limiting measure for vocal effort, but it has been used to differentiate a categorical definition of vocal effort. For example, Skinner et al (1997) defined speci fic dB values for different vocal effort levels: causal (56 dB), normal (60 dB), loud (74 dB), and shout (83 dB). Others have used this similar pattern to operationalize vocal effort (although not always with specific decibel values). Cushing, Li, Cox, Wor rall, and Jackson (2011) used (2012) used 30 Goldman (1995) definitions of vocal effort are straightforward to implement but can be problematic. With the exception of Skinner et al, there is no regulation for the vari ation of individual perception of similar vocal levels with different vocal efforts. Another body of literature suggest s that speaking style does have an impact on vocal effort. There is evidence that word stress (Mooshammer & Mooshammer, 2010; Sluijter & Van Heu ven, 1996) , hyper - articulation of fricatives (Meynadier, El Hajj, Pitermann, Legou, & Giovanni, 2018) , and even posture (Lagier et al., 2010) can affect vocal effort. These are not the only possible individual factors that could influence vocal effort. Vocal effort has also been defined as a symptom of other vocal conditions. Voice disorders hav e a higher prevalence and magnitude of reported vocal effort (Altman, Atkinson, & Lazarus, 2005; Roy, Merrill, Gray, & Smith, 2005; Smit h et al., 1998) . Some of the specific voice disorders associated with vocal effort include essential tremor (Warrick et al., 2000) , muscle - tension dysphonia (Roy, Smith, Allen, & Merrill, 2007) , and vocal nodules (Hillman, Holmberg, & Perkell, 1989) . Adductor spasmodic dysphonia (ADSD) has been the most studied in relation to excessive vocal effort (Cannito, Doiuchi, Murry, Woodson, & York, 2012 ; Eadie & Stepp, 2013; Roy et al., 2007; Shoffel - havakuk et al., 2019; M. E. Smith, Roy, Wilson, & Hypothesis, 2006) . In these cases, there are structural (vocal nodules) or neurological (ADSD) differences in these individuals that result in increased v ocal effort. Other physiological conditions have been shown to result in excessive vocal effort including vocal fold stiffness (Katherine Verdoli ni et al., 1994) , internal temperature (Sandage, Connor, & Pascoe, 2013) , and dehydration (Pearl 31 Solomon & Stem mle DiMattia, 2000) . A comprehensive definition of vocal effort would need to contain components of communication messages or objectives, environments, and capabilities. 2. 3 .a Theoretical framework for vocal effort and related terms The Fifth the Framework for Understanding Effortful Listening (FUEL). This framework defines effort as ng out a (Pichora - Fuller et al., 2016) . Although this definition was developed for listening effort, it can provide the scaffolding for a framework on vocal effort. There are three key parts to this a cognitive and active process and that the available resources play a role in effort. Second, esult of some limitation to clarity of - based which means that some sort of communication needs to be initiated for vocal effort to occur. This definition fits well with the previou sly mentioned components of a vocal effort definition. Incorporating these ideas with the definition of effort, vocal effort (as defined in this document) is then the deliberate allocation or exertion of cognitive or physiological resources to adapt existing communication capability to overcome internal (i.e. voice problems such as fa tigue or functional, structural, or neurological impairments) and external (i.e. communication environment s including talker - listener distance, background noise, room reverberation , or communciation intent ) obstacles in goal pursuit of a voice - related communication objective or 32 task. This proposed definition of vocal effort can be used to redefine and relate common terms used in voice research. The terms that will be described are vocal load, vocal loading, vocal ability , and vocal fatigue and they wil Within this framework, t he load(s) are the obstacles that are in opposition to the communication objective. The loading is the allocation of resources to adapt or overcome the load. In other words , lo ading is vocal response to the demand of the load. Ability is the existing voice - related communication capability to handle or endure the particular load. Performance is the general use of the vocal mechansim . Finally, fatigue is the loss of resources as a result of the loading. To try to put this all in context together, effort is the exertion in response to l oading in order to overcome a load towards a p erformance and /or effort as a res ult of loading. This is because an indiviudal can use more effort to maintain constant performance with loading , or lose performance with the absence of increased effort. A quick validity check of the framework is that an individual with prolonged vocal ef fort would eventually experience fatigue this is consistent with the previous results . Additionally, this framework supports the previously reported use of vocal effort. For quantity that ordinary speakers vary when they adapt their speech to the demands of Traunmüller and Eriksson (2000) . Here the loading is the adaptation of their speech to a load of inc reased or decreased communication distance. More importantly is that the framework accounts for the observations that individuals can greatly vary in their response to the same load. This framework takes into account the ability that may affect how much ef fort is needed to overcome a l oad. 33 2. 3 .b Summary of proposed defintions for voice - related terms For these terms two definitions are given. The first definition is within the framework of vocal effort as described above and the second is a more general term . Vocal Effort: (1) the deliberate allocation or exertion of cognitive or physiological resources to adapt existing communication capability to overcome internal (i.e. voice limitations such as fatigue or functional, structural, or neurological impairments) a nd external (i.e. communication environment demands such as talker - listener distance, background noise, and room reverberation) obstacles in goal pursuit of a voice - related communication objective or task (2) the perception of exertion as a result of vocal loa ding (see below) to a perceived vocal load ( see below ) Vocal Load: (1) the obstacles that are in opposition to the communication objective , (2) the vocal constraint including both internal (e.g. hydration, fatigue, vocie disorders, etc.) and external (e.g. acoustic environment, communication intent, etc.) to a particular voice - related communication task physiology to perform the task or their perception of the task Vocal Loading: (1) the allocation of resources to adapt or overcome the load (2) the change (or vocal performance) to accommodate the vocal load in a particular voice - related communication situation ( e.g. Lombard effect in the presence of noise, clear speech with hearing - impared 34 listeners, etc.) Vocal Ability : (1) the existing voice - related communication capability to handle or endure the particular load (2) for a particular voice - related communi cation situation Vocal Performance: (1) the general use of the vocal mechanism (2) the way an individual accomplishes a vocal task for a particular voice - related communication situation (i.e. fundamental frequency or amplitude of vocal fold vibration, vocal fold closure or posture, resonance, etc.) Vocal Fatigue: (1) t /or effort as a result of vocal loading (2) the physiological and/or perceptual manifestation of a change in the voi ce that vocal load for a particular voice - related communication situation which may be a result of vocal loading or vocal effort 2.4 Hypothesis and Research Questions The presented framework provides a novel and useful approach towards understanding vocal fatigue and related concepts . In particular, this framework allows for a dynamic system of underlying mechanisms of vocal fatigue as well as explaining the variability in previous experiments and accounting for individual vocal ability. Viewing vocal fatigue from the perspective of this framework, vocal fatigue could be measured as th e change in vocal 35 performance or vocal effort (or a combination of the two) with and wi thout a vocal load. This perspective allows for the study of the factors that contribute to vocal fatigue because it reduces the attempt to me asure the complex subsystems of vocal fatigue t o pra ctical measurements of vocal loading (change in vocal performa nce from a vocal load ) and vocal effort. 2. 4 .a Main Hypothesis The primary goal of this dissertation is to experimentally validate one of the primary assumptions of the proposed framework for vocal fatigue ( e.g. vocal fatigue is related to the change in vocal performance and/ or vocal effort when responding to a vocal load ) . Therefore, the overarching hypothesis of this research is: H 0 : The changes in vocal performance, vocal effort, and/or their interaction through a vocal load will implicate vocal fatigue. 2. 4 .b Research Questions Towards the purpose of testing H 0 , three sub - hypotheses and research questions were developed. Fundamental to the utility of the framework is the ability to reli ably measure vocal effort and show a direct relationship between effort and vocal performance. Thus, the first research question is: Q 1 : Can perceived vocal effort be measured reliably and if so, how does vocal performance in terms of vocal intensity change with vocal effort? Associated with this research question is Hypothesis 1 : H1: Vocal performance in terms of vocal intensity will be distinct for each vocal effort level and be consistent within and across participants. To test H1, Experiment 1 consists of participants performing various speech tasks at specific effort levels from the Borg CR - 100 scale. 36 With a reliable vocal effort scale and a relationship between vocal performance and vocal effort quantified (results of Experiment 1) , the next step (Experiment 2) is to test the effects of different levels of vocal loads . In other words, the relationship between vocal performance and vocal effort was tested for different vocal loads to examine the interaction between vocal load , vocal performance, and vocal effort . To control for the effects of different loads, three levels of three different equivalent loads were used ( t h ese are loads that theoretically have equivalent vocal loading, for more detail see 3 .2). Thus, the sec ond research question is: Q 2: To what degree are vocal performance and vocal effort related given three equivalent vocal load levels? Associated with this research question is Hypothesis 2: H2: The v ocal performance and vocal effort will be constant within equivalent load conditions. To test H 2 , Experiment 2 consists of participants performing communicative task s in three load levels of three different load source ( where each load source should be equivalent) and rate their vocal effort assoc iated with these tasks. Experiment 2 establishes relationships between measurements of vocal performance and vocal effort across a vocal loa d. Testing the main hypothesis (H0) requires investigation of changes in these measurements with in a controlled voca l load over time . Therefore, the third research question is: Q 3: To what degree do vocal performance, vocal effort, and/or their interaction change given a combined vocal load of excess background noise over time? Associated with this research question is Hypothesis 3, which is a more specific iteration of H0: 37 H3: The measured changes in vocal performance, vocal effort, and/or their interaction will change through a vocal load (background noise and prolonged speaking) . To test H3 and by extension H0, Experi ment 3 consists of participants performing a vocal loading task in the presence of excess background noise for an extended duration while rating their vocal effort levels throughout. 38 CHAPTER III: METHODOLICAL APPROACH Th is dissertation presents a theoretical framework for vocal fatigue and validates several aspects of it through experimentation. This validati on occur s through three experiments detailed in this chapter. These experiments directly test the research questions and hypotheses found in 2.4. First, descriptions of the measurement variables used th r oughout all of the experiments are provided. Then the experiment s are presented in sequential order. The experiments build in numerical order , therefore, methodological approaches described in earlier experiments will only be referenced in the later experiments. 3.1 Measurement Each of the experiments use si milar methodology for the measurement variables. This section provides the general details for these variables that are used through all of the experiments. 3.1.a Vocal Effort Measurement Perceived v ocal effort was measured using the Borg CR - 100 (Fig. 3. 1) . Borg scales have been used to quantify the perception of pain, exertion and effort generally (Borg & Löllgen, 2001; Fanchini et a l., 2015) . The Borg scales combine the mathematical precision of a direct magnitude estimation scale and the usability of a visual analog scale. Th e Borg CR - 10 ha s been successfully adapted for applications of vocal effort (Borg & Löllgen, 2001; Fanchini et al., 2015; van Leer & van Mersbergen, 2017) . Here the B org CR - 100 scale is modified from the previous Borg CR - 10 scale used for vocal effort but using the more granular intervals of the Borg CR - 100. The only change, other than the more granular 39 Figure 3. 1 Borg CR - 100 scale used to measure the self - perception of vocal effort . which is consistent from other applications of the Borg CR - 10 scale. In Experiment 1, this scale is used as an independent variable while i n Experiments 2 and 3 it is used as a dependent variable. 3.1.b Acoustic Measurement The specific speech samples elicited varied across the experiments, but they were processed in a similar manner. After segmentation, the speech samples were processed to h ave all of the non - voicing segments removed (Maryn, De Bodt, & Roy, 2010; Rubin et al., 2019) . While there are many acoustic speech parameters which could be used to reflect vocal performance , five were chosen here that represent basic performance qualities: pitch, pitch range, volume, dynamic range, and quality. Five acoustic parameters were derived from each voice concatenated speech segment : mean fundamental frequency ( F0 ), standard deviation of fundam ental frequency 40 ( F0sd ), speech level ( SL ), standard deviation of speech level ( S Lsd ), and smoothed cepstral peak prominence ( CPPS ). Fundamental Frequency The fundamental frequency is the lowest mode of v ocal fold oscillation during speech . While t his measu rement ( cycles per second or Hertz) is a cyclic quantity , it also relates to perception of pitch. The mean and standard deviation of F0 were measured as part of the vocal performance evaluation. The F0 was estimated using Praat built in pitch capabilitie s . For the male participants estimation, an F0 range of 65 Hz to 350 Hz was used. While for the female participants, an F0 range of 150 Hz to 800 Hz was used. Table 3.1 contains the other Praat parameters used. Praat estimates the F0 in small increments over time through the speech sample resulting in a time - based array of frequency values . In the results below, F0 represents the average measured F0 from a sample, and F0sd represents the standard deviation of the measured F0 from a sa mple. Since the perception of pitch is nonlinear (i.e. a doubl ing of pitch from 100 Hz to 200 Hz is fewer Hz than a doubling of pitch from 200 Hz to 400 Hz) , semitones (1/12 th of an octave or doubling of pitch) are used for F0. This allows for comparable m easures across all of the participants since habitual F0 significantly varies. The F0 measurements were converted to semitones as follows: Eq. 3. 1 where is the measured frequency in Hertz from sample and is the reference frequency. 41 Table 3. 1 Praat specifications for fundamental frequency calculation Praat Parameter Value Analysis Method Autocorrelation Octave Cost 0.01 Max number of candidates 15 Silences threshold 0.03 Voice threshold 0.45 Octave - jump cost 0.35 Voice/unvoiced cost 0.2 The reference frequency is slightly different for each experiment, but all represent a habitual or baseline frequency measurement of each participant. A semitone is 1/12 th of an octave where an octave is doubling of frequency. This is used to normalize the change in F0 across many participants with different baseline F0. Speech Level In general , t he speech level or sound pressure level (SPL ) is a measure ment of vocal intensity and relates to the perceived loudness of the talker and is calculated as follows : Eq. 3. 2 where is the root - mean - square (RMS) sound pressure and is the reference pressure of 20 micro p ascals ( µPA). Here a reference voltage was provided from a reference microphone and calibrator to convert the microphone voltage to sound pressure. The mean and standard deviation of the SPL were measured as part of the vocal performance evaluation. Here the mean speech level ( S L) is the average SPL of multiple windowed segments of the voice concatenated voice signal as follows: 42 Eq. 3.3 w here is the number of segments in the sample. The segments were 20 msec windows with 50% overlap. The standard deviation of the speech level ( S Lsd ) is the sample standard deviation of the SPL segments as follows: Eq. 3.4 Smoothed Cepstral Peak Prominence The smoothed ce pstral peak prominence (CPPS) is an acoustic measure that represents the relative strength or prominence of the periodicity of the voice signal. More specifically it is the distance (measure in decibels) between the peak of the first rahmonic (dominant quefrency) and a linear regression of the smooth cepstrum. The cepstrum is defined as the fast Fourier transform (FFT) the of log magnitude power spectrum as follows: Eq. 3.5 where is the Fourier operator and is the time - domain speech signal. Here the CPPS was calculated with Praat and used the voice - concatenated segments . This measure has been widely used in previous voice science work as an acoustic correlate to voice quality and therefore is included as one of the vocal performance measures (Hillenbrand, Getty, Clark, & Wheeler, 2005; Maryn et al., 2010) . 43 3.1.c Analysis Software The acoustic signal processing was compl eted in Matlab with custom functions for the voice concatenation , fundamental frequency estimation , and smoothed cepstral peak prominence measurement. Statistical analysis was completed using SPSS. 3. 2 Experiment 1 The purpose of this experiment wa s to test hypothesis 1 (H1). Q1: Can perceived vocal effort be measured reliably and if so, how does vocal performance in terms of vocal intensity change with vocal effort? H1: Vocal performance in terms of vocal intensity will be distinc t for each vocal effort level and be consistent within and across participants. The independent variable for this experiment was the vocal effort goal . The dependent variable for th is experiment was the vocal performance . In Experiment 1, the vocal effort level goal w as determined using the Borg CR - 100 scale (Fig 3.1) . This scale was also used in the other two experiments to measure perceive d vocal effort . Additionally, this experiment measured vocal performance in terms of the five voice acoustic pa rameters outlined above (fundamental frequency mean and standard deviation, speech level mean and standard deviation, and smoothed cepstral peak prominence) . In this experiment participants produced different speech samples at specific cued vocal effort le vels. 3.1.a Participants With protocol approval of the Michigan State University's Human Research Protection Programs Human Subject's Review Board , this e xperiment c onsisted of twenty college - age participants (10 male and 10 females). The participants wer e recruited through an online recruit ing and scheduling system at Michigan State University in the College of Communication 44 were screened for hearing limitations . The hearing screening required for inclusion consisted of pure - tone stimulation of at least 20 dB HL in both ears at 500 Hz, 1000 Hz, 2000 Hz, and 4000 Hz. Following informed consent (included in Appendix D) , the participants proceeded with the experimen t. 3.1.b Instrumentation Acoustic Measurement The participants were recorded with a head - mounted microphone (HMM; omni - directional; Countryman B3 - held recorder (ZOOM H5 Handy Recorder) in a sound - isolation booth ( m single - walled ) . A reference microphone ( Behringer ECM8000 mouth. The reference microphone was absolutely calibrated to 94 dB SPL (relative to 20 µ Pa) u sing a commercial calibrator that fit onto the microphone . This calibration was used to referentially calibrate the HMM. This was done through the participant producing a steady vowel 50 cm from the reference microphone. The voltage rati os between the cali bration signal of 94 dB and vowel productions were used to calibrate the HMM. The speech was sampled at 44.1 kHz with 16 - bit resolution. Stimulus Presentation Instrument The stimulus was present on a laptop using the open - source p ython - based program, P sy ch oPy (Peirce et al., 2019) . A dditionally, a - acoustic and non - survey responses were recor ded through this program. A schematic of the instrumentation used is contained in Fig. 3.2 45 Figure 3. 2 Schematic for the instrumentation used in experiment 1 3.1.c Stimuli Throughout the experiment the participants were instructed to read a particular stimulus at a specific vocal effort level from the Borg CR - 100 scale (Fig. 3.1). This study used three types of speech stimuli with three variations of each type. For a complet e list of the speech stimuli used, see Appendix A : Experiment 1 Stimuli. The speech stimuli were chosen such that it would take a typical talker 12 - 16 seconds to produce the utterance. The first type of speech stimuli was automatic speech. This is speech t hat it is generally accepted that any native English speaker could produce without a script. Here participants were asked to either state the alphabet (English), count to twenty - five, or say the names for days of the week and months of the year. The second type of speech stimuli was reading aloud standard sentences . These sentences were excerpts from standard speech acoustic reading passages, the Marvin Williams Passage 46 , the Rainbow Passage (Fairbanks, 1960) , and the Stella Passage (Weinberger & Kunath, 2011) . ( See Appendix A for the exact sentences used . ) The third type of speech stimuli used was a map description task. Here participants were asked to describe a specific route on a Portland, Oregon subway map (Fig. 3. 3 ). The image and a description of the desired path were shown to the participants. The rou tes were all in the form of map. There were three different routes used during the experiment (see Appendix A ). Figure 3. 3 Subway map used for the map descripti on speech task. 3.1.d Procedure Prior to the main experiment, the participants first went through a tutorial. The tutorial and experiment were narrated using artificial speech from a text script via WaveNet (Van Den Oord et al., 2016) . An artificial narrator was used to remove any possible experiment - administration bias and to keep all narration samples acoustically similar (e.g. later narration samples could be 47 generated to account for changes in the experiment without any perceptual change in when the sample was created). The tutorial consisted of h aving the participant first practicing each automatic and reading speech task (outlined in section 3.1.c and contained in Appendix A ). The tutorial and practice were implemented to reduce a novice and/or learning effect by the participants during the main experiment. The tutorial also explained how to properly describe the map (Fig. 3. 3 ) and provided an example different from the three routes used as stimuli. The example used was describe how to get from Hillsboro to the Airport via Gateway you would Starting at Hillsboro, I will take the blue line eastbound towards Beaverton. I will pass Beaverton, Pioneer Square, and the Rose Quarter. I will change at the Gateway station to the red line northbound to the Airport. And finally, I will ar rive at the Airport The participants then practiced describing a route on the map. Next the participants were introduced to the vocal effort scale (Fig. 3. 1). The scale was presented with anecdotal anchors for the extreme values of 1 and 100 as follows Following the tutorial, the participants started the main exper iment. Here the participants were instructed to speak a particular speech stimulus (see the pervious section 3.1.b and Appendix A for more information on the speech stimuli) and at specific vocal effort level. Four vocal effort levels were prompted : (1) 2 completed each speech stimulus (total of nine unique stimuli; three variations of three types) at each vocal effort level (four) for a total of thirty - six trials. The vocal effort scale (Fig. 3. 1), 48 speech stimulus, and map (if applicable) were shown to the participant for each trial (see Figs. 3. 4 and 3. 5 ). These tr ials were randomized for each participant. The ex periment c oncluded after the thirty - six trials. 3.1.e Statistical Analysis The reference used for the semitone calculation for each participant was the average fundamental frequency of the practice tasks. As mentioned in 3.1.c, SPSS statistical software was used for statistical analysis. First the statistical assumptions of normality, independence, and equal variance were checked. If these assumptions were met, the sample means for each of the five acoustic parameters (F0, F0sd, L, Lsd, CPPS) were compared across the VELs using one - way analysis of variance (ANOVA) tests w ith an alpha level of 0.05 . Post hoc Tukey HSD tests were used to compare the speech production of the VEL pairs. If the equal variance cou ld not be assumed, The test - retest reliability of the scale was measured of the measured speech levels ( S L) across the repeated vocal effort levels of the speech level . Outliers within the dataset were removed case - by - case that were either greater than the sum of the third quartile (Q3) and 1.5 times the interquartile range (IQR) or less than the difference of the first quartile (Q1) and 1.5 times the IQR as described in Equation 6 below. Eq. 6 49 Figure 3. 4 Example of the presentation of the vocal effort scale (left), target vocal effort level (upper - right) and speech stimulus (right) during an experimental trial. Figure 3. 5 Example of the presentation of the vocal effort scale (left) and speech stimulu s (right) for a map description task during an experimental trial. 50 3. 3 Experiment 2 The purpose of this experiment was to test hypothesis 2 (H2). Q2: To what degree are vocal performance and vocal effort related given three equivalent vocal load levels? H2: The vocal performance and vocal effort will be constant within equivalent load conditions. The independent variable s for this experiment w ere the vocal load types and levels . The dependent variables for this experiment were vocal effort ratings (VER) and vocal performance. Vocal performance was quantified through measurements of the mean and standard deviation of fundamental frequency (F0; F0sd) , the mean and standard deviation of speech level ( SL ; SLsd ), and the smoothed cepstral peak prominence (CPP S ) . Vocal effort was measured through self - ratings on the Borg CR - 100 scale. The vocal loads were communicative situations of talker - listener communicative distance, talker loudness goal, and excess background noise . These conditions were selected for this study because they have theoretical vocal intensity equivalences. In other words, a doubling of communicative distance, an increase of background noise of 9 dB (i.e. Lombard effect Bottalico, Passione, Graetzer, & Hunter, 2 017) , and an increase of a loudness target of 6 dB should all require an increase of 6 dB of vocal intensity to maintain acoustic energy equivalence . These theoretical equivalencies allowed for direct comparison of vocal performance and vocal effort rat ing s. In this experiment participants were asked to complete a communicative task with a particular vocal load ( communicative distance, loudness goal, or background noise ) and then rate their vocal effort level for that task . 51 3.2.a Participants With protocol approval of the Michigan State University's Human Research Protection Programs Human Subject's Review Board, this e xperiment c onsisted of forty college - age participants ( 20 male s and 20 females). The participants were recruited t hrough an online recruit and scheduling system at Michigan State University in the College of Communication Arts and screened for hearing limitations in the same man ner as Experiment 1 . Following informed consent, the participants proceeded with the experiment. 3.2.b Instrumentation The participants were recorded with a head - mounted microphone (HMM; omni - directional; Countryman B3) placed approximately 5 cm from the p digital hand - held recorder ( ZOOM , H5 Handy Recorder) . A reference microphone (Behringer ECM8000) calibrated to 94 dB SPL (relative to 20 µ Pa). Th is calibration was used to referentially calibrate the HMM using the same method from 3.2.b . The speech was sampled at 44.1 kHz with 16 - bit resolution. The experiment took place in an anechoic chamber ( m, IAC number 107840) a specialized resear ch facility designed to completely absorb the reflection of sound. The use of this facility was essential to this study as room reverberation plays a significant factor in the perception of background noise and communicative distance, as well as vocal effo rt (Berardi, Whiting, et al., 2015; Bottalico, 2017; Whiting, Leishman, Eyring, Berardi, & Rollins, 2015) . 52 The stimulus was present using PyschoPy via an external monitor connected to a laptop . All - acoustic and non - survey responses were recorded through this program. Schematic for instrumentation is shown in Fig. 3.6. Figure 3.6 Schematic for the instrumentation used in experiment 2 3.2.c Stimuli This experiment used a map description task to elicit speech in three different types of vocal load conditions with three variations of each condition (for a total of nine conditions). T he map description task is similar to the map description task outlined in Experiment 1. The difference in this experiment was that the researcher acts as a communication partner. This means that the participant was instructed that they needed to explain t he route on the map in such a way that the researcher could in real time trace the map. The participant was presented various maps via PsychoPy on a computer screen where all the routes are gray except for the route that must be described (see Fig. 3. 7). The research had a 53 stack of gray maps (no colored routes) and traced the described path as the participant explained the route (see Fig. 3.8) . There were nine different routes such that each route was repeated 3 times but never across the same condition (i.e. each map was not used for the same load condition more than once). (See Appendix B for the maps used and what vocal load conditions for which each was used.) The three vocal load conditions used in the experiment included communic ative distance , loudness goal , and background noise . During this experiment, the participant is seated in the corner of the anechoic chamber and the researcher is seated at 1 meter from the participant except for when communicative distance was manipulated . Figure 3. 7 Example map for map description task in Experiment 2. This map is instructing the participant to describe the route from Clackamas Town Center to Beaverton via Gateway. 54 Figure 3. 8 Researcher's template for Experiment 2. This template includes a color less map to be highlighted as the route i s described by the participant. The communicative distance consisted of the researcher moving to three different points in the room along a direct line from the participant (as shown in Fig. 3.9) . These three dista nces were as follows: D01; s hort distance: 1 meter D02; m oderate distance: 2 meters D04; l ong distance: 4 meters 55 Figure 3.9 Visual description of the communication distance vocal load. The graphic in the bottom left corner with the headset represents the participant, the ears represent the potential locations of the listener during the experiment. The jagged walls are reminders that the experiment was contained in an anechoic chamber. The loudness goal condition required the participant to speak such that their average voicing level was above a certain intensity threshold. The average voicing level was calculated using a reference microphone placed 1 meter in a direct line from the participant . The intensity thresholds (measured at 1 meters) were as follows: L54; l ow goal: 54 dB L60; m oderate goal: 60 dB L66; h igh goal: 66 dB 56 The participants were able to see their current average voicing level and target loudness level. A large red arrow pointing up would display on the screen to prompt the participant to increase their speaking intensity (see Fig. 3. 10 ). For the background noise condition, pink noise played from two loudspeakers placed 30 degrees off axis in a 2 - meter arc from the participant (as shown in Fig. 3.11) . The background ws: N53; l ow noise: 53 dBA N62; m oderate noise: 62 dBA N71; h igh noise: 71 dBA The interval of 9 dB was chosen to allow for a 6 - dB vocal intensity increase predicted by the Lombard effect (Bottalico, Passione, et al., 2017) . (Note that the Lombard effect has been shown Figure 3. 10 : Example of map stimulus for the participant in Experiment 2. This particular example would be for the loudness goal condition of 66 dB. The participant is shown their current level and if it is lower than the target, a red arrow (shown in the upper - righ t) reminds them to talk louder. 57 to change slope with different dB ranges, here the dB range was chosen such that the predicted Lombard effect would be consistent with a 6 - dB vocal intensity with a 9 - dB increase of background noise). Each of the three vocal load conditions had three variations for a total of nine conditions. Each condition was repeated three times for a total of 27 trials. The trials were presented to each participant randomly. Figure 3.11 Visual description of the background noise vocal load. The graphic in the bottom left corner with the headset represents the participant, the ear represents the location of the listener (1 m). The loudspeakers are shown to be in a 2 m arc from the participant 30 degrees off axis. T he jagged walls are reminders that the experiment was contained in an anechoic chamber. 3.2.d Procedure The participants started with a tutorial to familiarize them with stimuli in the experiment. First the participants were introduced to the vocal effort scale ( Borg CR - 100). This scale used an anecdotal anchor similar to Experiment 1: 58 someone while standing on an Then the scale was experientially anchored by having the participants state the days of the week and months of the year at vocal effort levels of their own sense of vocal effort for four distinct vocal effort levels. Next the participants were trained on how to perform the map task. This consis ted of a demonstration of an example map route. The example used was the same as described in Experiment 1. The participants then practiced describing a map route. Corrective instructions were made by the researcher to ensure clarity and consistency of tas k performance. Following the training, the participants were presented with the randomized twenty - seven trials. Each trial consisted of a map route and rating of vocal effort . For each trial, the researcher had a template (Fig. 3.8 ) to actively record the route description. This was implemented to create a realistic communication scenario. Following the twenty - seven trials, the experiment concluded. 3.2.e Statistical Analysis The reference used for the semitone calculation was the average fundamental freque ncy of the baseline condition for a participant which had no loudness goal or background noise and the communication partner was at 1 meter. As mentioned in 3.1.c, SPSS statistical software was used for statistical analysis. First the s tatistical assumptions of normality, independence, and equal variance were checked. If these assumptions were met, the sample means for the self - reported vocal e ffort level (VE R ) and each of the five acoustic parameters ( F0 , F0sd , L, Lsd , CPPS) were compared across the vocal load levels using one - way analysis of variance (ANOVA) tests w ith an alpha level of 0. 05 . Post hoc Tukey HSD tests were used to provide pair - wise comparison of e ach load level ( e.g. 1 m, 2 m, 4 59 m in the case of communication distance). Outliers within the dataset were removed case - by - case that w ere either greater than the sum of the third quartile (Q3) and 1.5 times the interquartile range (IQR) or less than the difference of the first quartile (Q1) and 1.5 times the IQR as described in Equation 6 in 3.1.e. 3. 4 Experiment 3 The purpose of this ex periment was to test hypothesis 3 (H3). Q3: To what degree do vocal performance, vocal effort, and/or their interaction change given a combined vocal load of excess background noise over time? H3: The measured changes in vocal performance, vocal effort, an d/or their interaction will change through a vocal load (background noise and prolonged speaking) . The independent variable for this experiment was the point in time across the vocal loading task . The dependent variables for this experiment were vocal effort ratings (VER) and vocal performance . Vocal performance was quantified through measurements of the mean and standard deviation of fundamental frequency (F0; F0sd), the mean and standard deviation of speech level (SL; SLsd), and the smoothed cepstral peak prominence (CPPS). Vocal effort was measured through the Borg CR - 100 scale. The vocal load s w ere background noise and prolonged speaking. While the previous experiment measured vocal loading in va ried equivalent vocal loads, this experiment investigated the effect s of the background noise load and time. 60 3.3.a Participants With protocol approval of the Michigan State University's Human Research Protection Programs Human Subject's Review Board, this experiment c onsisted of forty participants ( 20 male and 20 females). The participants were recruited through an online recruit and scheduling system at Michigan State University in the College of Communication Arts and Sciences. The compensated with course credit. The participants were screened for hearing limitations in the same manner as Experiments 1 and 2 . Following informed consent, the participants proceeded with the experiment. 3.3.b Instrumentation The participants were record ed with a head - mounted microphone (HMM; omni - directional; Countryman B3) . The microphone signal went through a pre - amp lifier (Millen n ia HV - 3 D) and A/D converter (RME ADI - 8 DS ) then was recorded using REAPER, a digital audio workstation. A reference sound level meter ( SLM; IEC 60 651 Type 2 microphone of the SLM was absolutely calibrated to 94 dB SPL (relative to 20 µ Pa). This cali bration was used to referentially calibrate the HMM using the same method from 3.2.b. The speech was sampled at 44.1 kHz with 16 - bit resolution. The stimulus was present ed using PyschoPy using an external monitor - acoustic and non - sur vey responses were recorded through this program. Schematic for instrumentation is shown in Fig. 3.6. 61 Figure 3.12 Schematic for instrumentation used in experiment 3 3.3.c Stimuli This experiment had pretest and posttest voice tasks, a vocal loading tas k, and an intermittent vocal effort rating throughout the vocal loading task. The pretest and posttest voice tasks consisted of the participant read ing the Rainbow Passage (Fairbanks, 1960) . During the vocal loading task, the participants described routes on maps . Here the participants were shown a series of maps with a start, path, and end point (see Fig. 3.1 3 ). They were asked to describe the map such that someone else could recreate the exact route. T he advancement of the maps was self - paced. That is to say, the participant advanced the next page or map when needed. 62 Figure 3. 13 : Example of a map for Experiment 2. Each map had a compass in one of the corners showing north. Additionally, each map had a starting point denoted by a red circle, an ending point denoted by a red "X" , and a dotted line between the two denoting the route to be described . During the vocal loading task, every five minutes the reading or maps description was interrupted to allow for vocal effor t ratings using the Borg CR - 100 scale . T hen the participants were returned to the last seen map. Additionally, during these breaks in the vocal loading task, participants were cued to drink 3 0 mL from a small measured cup . This was implemented to prevent d ry - mouth sensations from influencing the feeling of vocal effort or fatigue. 3.3.d Procedure Prior to the main experiment, participants completed the same screenings and tutorials outlined in Experiment 2 (i.e. hearing screening and vocal effort scale trai ning ). The participants received additional instruction on how to advance the maps for the vocal loading test. Following the screenings and tutorials, the participants did the pretest voice tasks , including rating their current vocal effort level ( Borg CR - 100) . 63 Following the pretest voice task, the participants were instructed on how to properly complete the reading or map task. Additionally, they were instructed that they could signal to the researcher to end the vocal loading task if vocal effort or pain in the throat became too elevated. When the vocal loading task started, background noise in the form of multi - talker speech babble (made of six female and six male North American speakers) was played . Speech babble was chosen because it is quasisteady with a spectrum identical to normal speech creating a more realistic communication situation. Additionally, the presence of this type of noise replicates the design of a clinical VLT by Whitling, Rydell, & Lyberg Åhlander (2015) . Th e background noise started at 45 dBA and gradually increased to 75 dBA over a period of 30 seconds at a rate of 10 dB every 10 seconds (this matches a doubling of perceived loudness every 10 seconds) . The background noise persisted through the vocal loadin g task at 75 dBA until the experiment was either voluntarily terminated or six five - minute intervals (a total of 30 minutes) had been completed. As mentioned above, after each five - minute interval the participants completed an intermittent vo cal effort rat ing and a drink of water. Following the end of the vocal loading task , t he participants did the posttest voice tasks including rating their current vocal effort level ( VER ) us ing the Borg CR - 100 . Following this posttest, the experiment concluded. 3.3.e Sta tistical Analysis The reference used for the semitone calculation for each individual was the average fundamental frequency of the Rainbow Passage reading recorded prior to the vocal loading task. As mentioned in 3.1.c, SPSS statistical software was used f or statistical analysis. First the statistical assumptions of normality, independence, and equal variance were checked. If these assumptions were met, the sample means for the self - reported vocal effort level (VER) and each 64 of the five acoustic parameters (F0, F0sd, L, Lsd, CPPS) were compared across each time point of the VLT using one - way analysis of variance (ANOVA) tests w ith an alpha level of 0.05 . Post hoc Tukey HSD tests were used to provide pair - wise comparison of each time point ( pre, post, and the six five - minute increments during the loading task ). If the equal variance could not be Outliers within the dataset were removed case - by - case that were either greater than the sum of th e third quartile (Q3) and 1.5 times the interquartile range (IQR) or less than the difference of the first quartile (Q1) and 1.5 times the IQR as described in Equation 6 in 3.1.e. In order to investigate the interaction of vocal effort and vocal performanc e as well as investigate subgroups within the data as predicted by the framework, clustering of vocal effort ratings and acoustic measurements were performed . First , a set of ten features were derived from the vocal effort ratings (shown in Table 3.2 ) . The n the data were clustered into two groups using k - means clustering in SPSS . The clustering models were iterated until a sufficient model was developed with the minimum features. The features were systemically excluded based on feature importance and statis tical significance within the model. The two groups were labeled based on the framework s assumptions with how changes in vocal effort would relate to vocal fatigue through vocal loading . One group with a clus ter center suggesting vocal effort changes associated with the vocal loading task was labeled the The vocal performance acoustic metrics were also split into groups. Based on previous literature it is possible that not all of the participants will have a significant change in vocal performan ce as a result of the vocal loading. Additionally, from the background review (Chapter 2) it is u nclear w hich acoustic measures are most likely to change from the vocal loading . To 65 cluster the participants, a general linear model (GLM) with time (pre and post vocal loading task) as the dependent variable and covariates as the five acoustic parameters was fit for each participant. Participants with both a significant model and at least one significant change in acoustic feature within the model others into a Following both the VER and aco ustics clustering, f our groups were formed from the cross - section of these groups . The vocal performance of the groups was compared pre - and post - vocal loading u sing independent - samples t - tests for each variable. 66 Table 3. 2 Fea tures used for VER clustering Feature Formula VER Linear Slope Slope of the linear fit of the six vocal effort ratings during the vocal loading task VER Linear Fit The goodness of fit coefficient of the linear fit of the six vocal effort ratings during the vocal loading task VER Linear Offset The intercept value of the linear fit of the six vocal effort ratings during the vocal loading task Noise Load Response ; Difference between vocal effort level after five minutes of vocal loading and the vocal effort level prior to the loading task Temporal Load Response ; Difference between vocal effort level after thirty minutes of vocal loading and the vo cal effort level after five minutes of vocal loading Noise Recovery Response ; Difference between vocal effort level after vocal loading task but no background noise and the vocal effort level after thirty minutes of vocal loading VER Difference ; Difference between vocal effort level after vocal loading task but no background noise and the vocal effort level prior to the loading task VER Maximum ; Maximum vocal effort rating during the vocal loading test VER Minimum ; Minimum vocal effort rating during the vocal loading test VER Range ; Difference between the maximum and minimum vocal effort levels during the vocal loading task 67 3.3.f Distribution for Collaborative Work As mentioned in Chapter 2, one limitation in the area of vocal fatigue is the inability to compare across studies. While vocal loading tasks (VLT) are the most commonly used, they vary widely in their methodological approach. Experiment 3 provides an experimental protocol that can be easily reproduced and appropriately varied to support collaborative work in v ocal fatigue. Towards this end, the code and relevant examples and explanations of the experimental protocol are available through GitHub . 68 CHAPTER IV: RESULTS This chapter provides the results of the three outlined experiments. The results of each experi ment are presented separately. Then , f or each experiment is included a table of abbreviations and acronyms used for that section, the demographics of the population studied, and the results of the statistical analyses outlined in Chapter III. 4 . 1 Experiment 1 Table 4. 1 Abbreviations for results of experiment 1 Abbreviation Meaning F0 Fundamental frequency F0sd Standard deviation of fundamental frequency SL Speech level SLsd Standard deviation of speech level CPPS Smoothed cepstral peak prominence VEL Vocal effort level VEL02 Vocal effort level of 2 VEL13 Vocal effort level of 13 VEL25 Vocal effort level of 25 VEL50 Vocal effort level of 50 ST Semitones dB Decibels R 4 . 1 .a Demographics Twenty - two participa n ts consented and completed Experiment 1 . One participant was not included d ue to equipment failure resulting in loss of data. Another participant was excluded due to failure to properly complete the instructions of the protocol. The remaining 20 participants , 10 males and 10 females, were included in these analyses presented below. All of these participants 69 were within normal hearing limits. The participants were all college age and most received course credit as compensation for their participation. 4 . 1 .b Results The goal of this experiment was to show how vocal performance relat ed with subjective vocal effort from the Borg CR - 100 scale. Five acoustic parameters were selected to measure vocal performance, mean fundamental frequency ( F0 ), standard deviation of fundamental frequency ( F0sd ), speech level ( S L), standard deviation of s peech level ( S Lsd ), and smoothed cepstral peak prominence ( CPPS ). There were four vocal effort levels (2, minimal vocal effort; 13, slight vocal effort; 25 moderate vocal effort; 50 sever e vocal effort) that the acoustic measures were compared across. Anal ysis of variance (ANOVA) tests were used to compare the mean speech acoustic measures across the four vocal effort levels were used to investigate the pair - wise differences across the four vocal effort levels. If equal variance could The results are separated by acoustic measure. Fundamental Frequency T he measured F0 met the assumptions for normality and independence. However, equal variance cannot be assume d . For each vocal effort level (VEL) the F0 (M; SD) is as follows (summarized in Table 4 . 2 with 95% con fidence interval, minimum , and maximum ): VEL 0 2 ; minimal vocal effort (M = - 0 .0 9 ST ; SD = 1. 44 ST) , VEL13 ; slight vocal effort (M = 0.4 3 ST ; SD = 1. 29 ST ), VEL25 ; moderate vocal effort (M = 1 .3 2 ST ; SD = 1.7 2 ), and VEL50 ; severe vocal effort (M = 2.93 ST ; SD = 1.93 ). Figure 4 .1 shows boxplots of the F0 across the vocal effort levels. 70 Table 4. 2 Descriptive statistics for F0 and VEL in e xperiment 1 Fundamental Frequency (ST) VEL Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound 2 - 0.09 1.44 - 0.31 0.14 - 4.16 5.25 13 0.43 1.29 0.24 0.62 - 3.01 5.34 25 1.32 1.72 1.07 1.57 - 2.92 6.11 50 2.93 1.93 2.62 3.24 - 1.33 6.56 Figure 4. 1 Boxplot of F0 and VEL for e xperiment 1 There was a significant (p < 0.001) main effect of F0 on VEL. The post hoc tests show that F0 was different across each VEL (summarized in Table 4.3 ) . There was an increase (p = 0.004) in F0 of 0.52 ST from VER02 to VER13, an increase (p < 0 . 001) of 1.41 S T from VER13 to VER25, and an increase (p < 0.001) of 3.02 ST from VER25 to VER50. 71 Table 4. 3 Multiple comparison statistics for F0 and VEL for experiment 1 (I) VE L (J) VE L Mean Difference (I - J) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound 2 13 - 0.52 0.15 0.004 - 0.91 - 0.12 25 - 1.41 0.17 < 0.00 1 - 1.86 - 0.96 50 - 3.02 0.19 < 0.00 1 - 3.54 - 2.51 13 2 0.52 0.15 0.004 0.12 0.91 25 - 0.89 0.16 < 0.00 1 - 1.32 - 0.47 50 - 2.51 0.18 < 0.00 1 - 3.00 - 2.02 25 2 1.41 0.17 < 0.00 1 0.96 1.86 13 0.89 0.16 < 0.00 1 0.47 1.32 50 - 1.61 0.20 < 0.00 1 - 2.15 - 1.08 50 2 3.02 0.19 < 0.00 1 2.51 3.54 13 2.51 0.18 < 0.00 1 2.02 3.00 25 1.61 0.20 < 0.00 1 1.08 2.15 Fundamental Frequency Standard Deviation The measured F0sd met the assumptions for normality and independence. However, equal variance cannot be assumed. For each VEL the F0sd (M; SD) is as follows (summarized in Table 4 . 4 with 95% confidence interval, minimum , and maximum): VE L02; minimal vocal effort (M = 1.83 ST; SD = 0.83 ST), VEL13; slight vocal effort (M = 2.13 ST; SD = 1.02 ST), VEL13; moderate vocal effort (M = 2.03 ST; SD = 0.89 ST), VEL50; severe vocal effort (M = 2.16 ST; SD = 0.82 ST). Figure 4.2 shows boxplots of the F0sd across the vocal effort levels. 72 Table 4. 4 Descripti v e s tatistics for F0s d and VEL for e xperiment 1 Fundamental Frequency Standard Deviation (ST) VEL Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound 2 1.83 0.83 1.70 1.96 0.74 4.91 13 2.13 1.02 1.98 2.29 0.69 4.88 25 2.03 0.89 1.89 2.17 0.76 4.98 50 2.16 0.82 2.03 2.29 0.93 4.82 Figure 4. 2 Boxplot of F0sd and VEL for e xperiment 1 There was a significant (p < 0.001) main effect of F0sd on VEL. The post hoc tests results are summarized in Table 4. 5 . There was an increase (p = 0.0 19 ) in F0 sd of 0. 31 ST from VER02 to VER13 and an increase (p = 0.00 2 ) of 0.34 ST from VER 02 to VER 50. 73 Tab le 4 . 5 Multiple comparison statistics for F0sd and VEL for experiment 1 (I) VE L (J) VE L Mean Difference (I - J) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound 2 13 - 0.31 0.10 0.019 - 0.58 - 0.03 25 - 0.20 0.10 0.208 - 0.46 0.05 50 - 0.34 0.09 0.002 - 0.58 - 0.09 13 2 0.31 0.10 0.019 0.03 0.58 25 0.11 0.11 0.902 - 0.18 0.39 50 - 0.03 0.10 1.000 - 0.30 0.24 25 2 0.20 0.10 0.208 - 0.05 0.46 13 - 0.11 0.11 0.902 - 0.39 0.18 50 - 0.13 0.10 0.654 - 0.39 0.12 50 2 0.34 0.09 0.002 0.09 0.58 13 0.03 0.10 1.000 - 0.24 0.30 25 0.13 0.10 0.654 - 0.12 0.39 Speech Level The measured SL met the assumptions for normality, independence, and equal variance. For each VEL the SL (M; SD) is as follows (summarized in Table 4 . 6 with 95% confidence interval, minimum , and maximum): VEL02; minimal vocal effort (M = 52.39 dB; SD = 5.4 6 dB), VEL13; slight vocal effort (M = 56.95 dB; SD = 6.28 dB), VEL25; moderate vocal effort (M = 60.54 dB; SD = 6.94 dB), VEL50; severe vocal effort (M = 66.89 dB; SD = 6.89 dB). Figure 4 . 3 shows boxplots of the SL across the VELs . 74 Table 4. 6 Descriptive statistics for speech level and VEL for experiment 1 Speech Level (dB) VEL Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound 2 52.39 5.46 51.5 6 53.2 3 39.87 63.32 13 56.94 6.28 56.0 2 57.8 8 45.35 72.13 25 60.53 6.94 59.5 2 61.5 6 47.63 79.50 50 66.8 9 6.8 9 65.8 7 67.9 1 51.01 82.52 Figure 4. 3 Boxplot for SL and VEL for experiment 1 75 There was a significant (p < 0.001) main effect of SL on VEL. The post hoc tests show that SL was different across each VEL (summarized in Table 4. 7 ). There was an increase (p < 0.001) in SL of 4.56 dB from VE L 02 to VE L 13, an increase (p < 0.001) of 3.59 dB from VE L 13 to VE L 25, and an increase (p < 0.001) of 6.35 dB from VE L 25 to VE L 50. Table 4. 7 Multiple comparison statistics for SL and VEL for experiment 1 (I) VE L (J) VE L Mean Difference (I - J) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound 2 13 - 4.56 0.69 < 0.00 1 - 6.34 - 2.77 25 - 8.14 0.69 < 0.00 1 - 9.93 - 6.36 50 - 14.50 0.70 < 0.00 1 - 16.29 - 12.70 13 2 4.56 0.69 < 0.00 1 2.77 6.34 25 - 3.59 0.68 < 0.00 1 - 5.34 - 1.84 50 - 9.94 0.68 < 0.00 1 - 11.70 - 8.18 25 2 8.14 0.69 < 0.00 1 6.36 9.93 13 3.59 0.68 < 0.00 1 1.84 5.34 50 - 6.35 0.68 < 0.00 1 - 8.11 - 4.60 50 2 14.50 0.70 < 0.00 1 12.70 16.29 13 9.94 0.68 < 0.00 1 8.18 11.70 25 6.35 0.68 < 0.00 1 4.60 8.11 Standard Deviation of Speech Level The measured SLsd met the assumptions for normality, independence, and equal variance. For each VEL the SLsd (M; SD) is as follows (summarized in Table 4. 8 with 95% confidence interval, minimum , and maximum): VEL02; minimal vocal effort (M = 3.43 dB; SD = 0.44 dB), VEL13; slight vocal effort (M = 3.47 dB; SD = 0.4 dB), VEL13; moderate vocal effort (M = 3.51 dB; SD = 0.44 dB), VEL50; severe vocal effort (M = 3 .64 dB; SD = 0.39 dB). Figure 4 . 4 shows boxplots of the SLsd across the VELs. 76 T able 4. 8 Descriptive statistics for SLsd and VEL for experiment 1 Speech Level Standard Deviation (dB) VEL Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound 2 3.43 0.44 3.36 3.49 2.46 4.67 13 3.47 0.40 3.41 3.53 2.60 4.66 25 3.51 0.44 3.44 3.57 2.52 4.63 50 3.64 0.39 3.58 3.69 2.68 4.55 Figure 4. 4 Boxplot for SLsd and VEL for experiment 1 There was a significant (p < 0.001) main effect of SLsd on VEL. The post hoc tests show that SLsd was only different in the pair - wise comparison s with VEL50 (summarized in Table 4. 9 ). There was an increase (p < 0.001) in SLsd of 0.21 dB from VE L 02 to VE L50 , an increase (p = 77 0.001 ) of 0.17 dB from VE L 13 to VE L50 , and an increase (p =0. 021 ) of 0.13 dB from VE L 25 to VE L 50. Table 4. 9 Multiple comparison statistics for SLsd and VEL for experiment 1 (I) VE L (J) VE L Mean Difference (I - J) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound 2 13 - 0.04 0.05 0.767 - 0.16 0.07 25 - 0.08 0.05 0.256 - 0.20 0.03 50 - 0.21 0.05 < 0.00 1 - 0.33 - 0.09 13 2 0.04 0.05 0.767 - 0.07 0.16 25 - 0.04 0.04 0.812 - 0.15 0.08 50 - 0.17 0.04 0.001 - 0.28 - 0.05 25 2 0.08 0.05 0.256 - 0.03 0.20 13 0.04 0.04 0.812 - 0.08 0.15 50 - 0.13 0.04 0.021 - 0.24 - 0.01 50 2 0.21 0.05 < 0.00 1 0.09 0.33 13 0.17 0.04 0.001 0.05 0.28 25 0.13 0.04 0.021 0.01 0.24 Smoothed Cepstral Peak Prominence The measured CPPS met the assumptions for normality and independence. However, equal variance cannot be assumed. For each VEL the CPPS (M; SD) is as follows (summarized in Table 4 . 10 with 95% confidence interval, minimum , and maximum): VEL02; minimal vocal effort (M = 11.99 dB; SD = 2.45 dB), VEL13; slight vocal effort (M = 13.35 dB; SD = 2.44 dB), VEL13; moderate vocal effort (M = 14.56 dB; SD = 2.05 dB), VEL50; severe vocal effort (M = 16.04 dB; SD = 1.53 dB). Figure 4 . 5 shows boxplots of the CPPS across the VELs. 78 Table 4. 10 Descriptive statistics for CPPS and VEL for experiment 1 Smoothed Cepstral Peak Prominence VEL Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound 2 11.99 2.45 11.59 12.40 7.07 17.09 13 13.35 2.44 12.99 13.71 7.36 18.71 25 14.56 2.05 14.25 14.86 7.65 19.31 50 16.04 1.53 15.82 16.27 12.31 19.18 Figure 4. 5 Boxplot for CPPS and VEL for experiment 1 There was a significant (p < 0.001) main effect of CPPS on VEL. The post hoc tests show that CPPS was different across each VEL (summarized in Table 4.1 1) . There was an increase (p < 0.001) in CPPS of 1.36 dB fro m VEL02 to VEL13, an increase (p < 0.001) of 2.56 dB from VEL13 to VEL25, and an increase (p < 0.001) of 4.05 dB from VEL25 to VEL50. 79 Table 4. 11 Multiple comparison statistics for CPPS and VEL for experiment 1 (I) VE L (J) VE L Mean Difference (I - J) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound 2 13 - 1.36 0.28 < 0.00 1 - 2.09 - 0.63 25 - 2.56 0.26 < 0.00 1 - 3.24 - 1.89 50 - 4.05 0.23 < 0.00 1 - 4.67 - 3.43 13 2 1.36 0.28 < 0.00 1 0.63 2.09 25 - 1.21 0.24 < 0.00 1 - 1.84 - 0.57 50 - 2.69 0.22 < 0.00 1 - 3.27 - 2.12 25 2 2.56 0.26 < 0.00 1 1.89 3.24 13 1.21 0.24 < 0.00 1 0.57 1.84 50 - 1.49 0.19 < 0.00 1 - 1.99 - 0.98 50 2 4.05 0.23 < 0.00 1 3.43 4.67 13 2.69 0.22 < 0.00 1 2.12 3.27 25 1.49 0.19 < 0.00 1 0.98 1.99 Test - Retest Reliability of Speech Level Each participant had significant regression (p < 0.001) for S L across the VELs. Descriptive (R) . The mean R was 0.9 0 (SD = 0.3 0 ). Table 4.1 2 below shows the descriptive statistics for R. Table 4. 12 Descriptive statistics for Pearson's R for linear regression of SL and VEL for experiment 1 Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound R 0.90 0.30 0.87 0.93 0.77 0.96 80 4 .1 Experiment 2 Table 4. 13 Abbreviations for results of experiment 2 Abbreviation Meaning F0 Fundamental frequency F0sd Standard deviation of fundamental frequency L Speech level Lsd Standard deviation of speech level CPPS Smoothed cepstral peak prominence VER Vocal effort rating ST Semitones dB Decibels dBA A - weighted decibels D01 Vocal load of communication distance at 1 meter D02 Vocal load of communication distance at 2 meters D04 Vocal load of communication distance at 4 meters L54 Vocal load of loudness target of 54 dB L60 Vocal load of loudness target of 60 dB L66 Vocal load of loudness target of 66 dB N53 Vocal load of background noise of 53 dBA N62 Vocal load of background noise of 62 dBA N71 Vocal load of background noise of 71 dBA 4 .2.a Demographics Forty - eight participants consented for the study. Eleven participants were not included in the analyses. Five were not included for not being native English speakers. One was not included for not passing the hearing screening (HL > 20 dB for 200 Hz and 400 Hz in left ear) . The other five participants were not included for either not finishing protocol or not prope rly following the instructions. A total of 37 participants, 19 males and 18 females, were included in the analyses presented below. All of these participants were within normal hearing limits. The participants were all college age and most received course credit as compensation for their participation. 81 4 .2.b Results The purpose of this experiment was to show how vocal performance and subjective vocal effort changed with different types of vocal loads. Vocal effort rating (VER) was measured using the Borg CR - 100 following each speech task. Five acoustic parameters were selected to measure vocal performance, mean fundamental frequency (F0), standard deviation of fundamental frequency (F0sd), speech level (L), standard deviation of speech level (Lsd), and sm oothed cepstral peak prominence (CPPS). Nine vocal load conditions were used including communication distances of 1 meter (D01), 2 meters (D02), and 4 meters (D04), loudness goals of 54 dB (L54), 60 dB (L60), and 66 dB (L66), and background noise levels of 53 dBA (N53), 62 dBA (N62), and 71 dBA (N71). Analysis of variance (ANOVA) tests were used to compare the mean VEL and speech acoustic measures across the variations of each vocal load for each HSD tests were used to investigate the pair - wise differences across the experimental conditions. If equal variance could not be separated by acoustic measure and type of voca l load. Vocal Effort Rating and Communication Distance The measured VER met the assumptions for normality and independence. However, equal variance cannot be assumed. For each vocal load the VER (M; SD) is as follows (summarized in Table 4. 1 4 with 95% confidence interval, minimum, and maximum): D01; communicat ion distance at 1 meter/no vocal load (M = 25.13 ; SD = 16.43 ), D02; communication distance at 2 meters (M = 28.23 ; SD = 14.67 ), D04; communication distance at 4 meters (M = 34.59 ; SD = 18.16 ), Figure 4. 6 shows boxplots of the VER across the vocal loa d s. 82 Table 4. 14 Descriptive statistics for VER and communication distance vocal load for experiment 2 Vocal Effort Rating Load Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 25.1 16.4 22.0 28.3 2 65 D02 28.2 14.7 25.4 31.0 2 65 D04 34.6 18.2 31.1 38.1 2 75 Figure 4. 6 Boxplot for VER and communication distance load for experiment 2 There was a significant (p < 0.001) main effect of VER on the vocal load of communication distance. The post hoc tests show that VER was only different in the pair - wise comparisons with D04 (summarized in Table 4. 1 5 ). There was an increase (p < 0.001) in VER of 9.5 from D01 to D04 and an increase (p = 0.017) of 6.4 from D0 2 to D04. 83 Table 4. 1 5 Multiple comparison statistics for VER and communication distance vocal load for experiment 2 (I) Load (J) Load Mean Difference (I - J) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound D01 D02 - 3.1 2.1 0.378 - 8.2 2.0 D04 - 9.5 2.4 < 0.00 1 - 15.2 - 3.7 D02 D01 3.1 2.1 0.378 - 2.0 8.2 D04 - 6.4 2.3 0.017 - 11.8 - 0.9 D04 D01 9.5 2.4 < 0.00 1 3.7 15.2 D02 6.4 2.3 0.017 0.9 11.8 Fundamental Frequency and Communication Distance The measured F0 met the assumptions for normality, independence, and equal variance. For each vocal load the F0 (M; SD) is as follows (summarized in Table 4. 16 with 95% confidence interval, minimum, and maximum): D01; communication distance at 1 meter/no v ocal load (M = - 0.02 Hz; SD = 0.62 Hz), D02; communication distance at 2 meters (M = 0.37 Hz; SD = 0.74 Hz), D04; communication distance at 4 meters (M = 0.75 Hz; SD = 0.82 Hz), Figure 4. 7 shows boxplots of the F0 across the vocal load s. Table 4. 16 Descrip tive statistics for F0 and communication distance vocal load for experiment 2 F undamental Frequency (ST) Load Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 - 0.02 0.62 - 0.14 0.10 - 1.76 2.09 D02 0.37 0.74 0.22 0.51 - 1.33 2.25 D04 0.75 0.82 0.58 0.91 - 1.75 2.53 84 Figure 4. 7 Boxplot for F0 and communication distance load for experiment 2 There was a significant (p < 0.001) main effect of F0 on the vocal load of communication distance. The post hoc tests show that F0 was only different in the pair - wise comparisons with D04 (summarized in Table 4. 1 7 ). There was an increase (p < 0.001) in F0 of 0.77 ST from D01 to D04 and an increase (p = 0.001) of 0.38 ST f rom D02 to D04. 85 Table 4. 17 Multiple comparison statistics and F0 and communication distance vocal load for experiment 2 (I) Load (J) Load Mean Difference (I - J) (ST) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound D01 D02 - 0.39 0.10 < 0.00 1 - 0.62 - 0.15 D04 - 0.77 0.10 < 0.00 1 - 1.01 - 0.53 D02 D01 0.39 0.10 < 0.00 1 0.15 0.62 D04 - 0.38 0.10 0.001 - 0.62 - 0.14 D04 D01 0.77 0.10 < 0.00 1 0.53 1.01 D02 0.38 0.10 0.001 0.14 0.62 Fundamental Frequency Standard Deviation and Communication Distance The measured F0sd met the assumptions for normality, independence, and equal variance. For each vocal load the F0sd (M; SD) is as follows (summarized in Table 4. 18 with 95% confidence interval, minimum, and maximum): D0 1; communication distance at 1 meter/no vocal load (M = 2.04 Hz; SD = 0.83 Hz), D02; communication distance at 2 meters (M = 2.23 Hz; SD = 0.92 Hz), D04; communication distance at 4 meters (M = 2.21 Hz; SD = 0.77 Hz), Figure 4. 8 shows boxplots of the F0sd across the vocal load s. Table 4. 18 Descriptive statistics for F0sd and communication distance vocal load for experiment 2 Fundamental Frequency Standard Deviation (ST) Load Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 2.04 0.83 1.87 2.20 0.73 4.52 D02 2.23 0.92 2.05 2.41 0.20 4.65 D04 2.21 0.77 2.06 2.37 0.89 4.47 86 Figure 4. 8 Boxplot for F0sd and communication distance load for experiment 2 There was no significant main effect of F0sd on the vocal load of communication distance. Speech Level and Communication Distance The measured SL met the assumptions for normality, independence, and equal variance. For each vocal load the SL (M; SD) is as follows (summarized in Table 4. 1 9 with 95% confidence interval, minimum, and maximum): D01; communication distance at 1 meter/no vocal load (M = 65.54 dB; SD = 0.33 dB), D02; communication distance at 2 meters (M = 66.86 dB; SD = 0.33 dB), D04; communication distance at 4 meters (M = 68.3 6 dB; SD = 0.36 dB), Figure 4. 9 shows boxplots of the SL across the vocal load s. 87 Table 4. 19 Descriptive statistics for SL and communication distance vocal load for experiment 2 S peech L evel (dB) Load Mean Std. Error 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 65.54 0.33 64.89 66.19 57.29 74.08 D02 66.86 0.33 66.20 67.52 58.67 73.32 D04 68.36 0.36 67.64 69.07 57.83 76.51 Figure 4. 9 Boxplot for SL and communication distance load for experiment 2 There was a significant (p < 0.001) main effect of SL on the vocal load of communication distance. The post hoc tests show that SL was different across each voc al load (summarized in Table 4. 20 ). There was an increase (p = 0.018) in SL of 1.32 dB from D01 to D02 and an increase (p = 0.006) of 1.49 dB from D02 to D04. 88 Table 4. 20 Multiple comparison statistics for SL and communication distance vocal load for exper iment 2 (I) Load (J) Load Mean Difference (I - J) (dB) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound D01 D02 - 1.32 0.48 0.018 - 2.45 - 0.19 D04 - 2.81 0.48 < 0.00 1 - 3.95 - 1.68 D02 D01 1.32 0.48 0.018 0.19 2.45 D04 - 1.49 0.48 0.006 - 2.63 - 0.36 D04 D01 2.81 0.48 < 0.00 1 1.68 3.95 D02 1.49 0.48 0.006 0.36 2.63 Speech Level Standard Deviation and Communication Distance The measured SLsd met the assumptions for normality, independence, and equal variance. For each vocal load the SLsd (M; SD) is as follows (summarized in Table 4. 21 with 95% confidence interval, minimum, and maximum): D01; communication distance at 1 meter/no vocal load (M = 3.4 dB; SD = 0.28 dB), D02; communication distance at 2 meters (M = 3.43 dB; SD = 0.28 dB), D04; communication distance at 4 meters (M = 3.47 dB; SD = 0.27 dB), Figure 4. 1 0 shows boxplots of the SLsd across the vocal load s. Table 4. 21 Descriptive statist ics for SLsd and communication distance vocal load for experiment 2 Speech Level Standard Deviation (dB) Load Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 3.40 0.28 3.35 3.45 2.76 4.14 D02 3.43 0.28 3.37 3.48 2.84 4.14 D04 3.47 0.27 3.41 3.52 2.79 4.05 89 Figure 4. 10 Boxplot for SLsd and communication distance load for experiment 2 There was no significant main effect of SLsd on the vocal load of communication distance. Smoothed Cepstral Peak Prominence and Communication Distance The measured CPPS met the assumptions for normality, independence, and equal variance. For each vocal loa d the CPPS (M; SD) is as follows (summarized in Table 4. 2 2 with 95% confidence interval, minimum, and maximum): D01; communication distance at 1 meter/no vocal load (M = 15.3 dB; SD = 1.25 dB), D02; communication distance at 2 meters (M = 15.62 dB; SD = 1. 25 dB), D04; communication distance at 4 meters (M = 16.08 dB; SD = 1.23 dB), Figure 4. 1 1 shows boxplots of the CPPS across the vocal load s. 90 Table 4.2 2 Descriptive statistics for CPPS and communication distance vocal load for experiment 2 Smoothed Cepstral Peak Prominence (dB) Load Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 15.30 1.25 15.06 15.54 12.08 18.12 D02 15.62 1.25 15.38 15.86 12.06 18.26 D04 16.08 1.23 15.85 16.32 12.49 19.17 Figure 4. 11 Boxplot for CPPS and communication distance load for experiment 2 There was a significant (p < 0.001) main effect of CPPS on the vocal load of communication distance. The post hoc tests show that CPPS was only different in the pair - wise comparisons with D04 (summarized in Table 4. 2 3 ). There was an increase (p < 0.001) in CPPS of 0.78 dB from D01 to D04 and an increase (p = 0.019) of 0.47 dB from D02 to D04. 91 Table 4. 23 Multiple comparison statistics for CPPS and communication distance vocal load for experiment 2 (I) Load (J) Load Mean Difference (I - J) (dB) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound D01 D02 - 0.32 0.17 0.155 - 0.72 0.09 D04 - 0.78 0.17 < 0.00 1 - 1.19 - 0.38 D02 D01 0.32 0.17 0.155 - 0.09 0.72 D04 - 0.47 0.17 0.019 - 0.87 - 0.06 D04 D01 0.78 0.17 < 0.00 1 0.38 1.19 D02 0.47 0.17 0.019 0.06 0.87 Vocal Effort Rating and Loudness Goal The measured VER met the assumptions for normality, independence, and equal variance. For each vocal load the VER (M; SD) is as follows (summarized in Table 4. 2 4 with 95% confidence interval, minimum, and maximum): D01; no vocal load (M = 25.13; SD = 16.43 ), L54; loudness goal of 54 dB (M = 35.43; SD = 16.03), L60; loudness goal of 60 dB (M = 39.06; SD = 15.75), L66; loudness goal of 66 dB (M = 49.81; SD = 16.81). Figure 4. 1 2 shows boxplots of the VER across the loads. Table 4. 24 Descriptive statistics for VER and loudness goal vocal load for experiment 2 Vocal Effort Rating Load Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 25.1 16.4 22.0 28.3 2 65 L54 35.4 16.0 32.4 38.5 7 80 L60 39.1 15.7 36.1 42.1 10 70 L66 49.8 16.8 46.6 53.0 3 85 92 Figure 4. 12 Boxplot of VER and loudness goal vocal load for experiment 2 There was a significant (p < 0.001) main effect of VER on the vocal load of loudness goal. The post hoc tests show that VER was different in the pair - wise comparisons with D01 and L66 (summarized in Table 4. 2 5 ). There was an increase (p < 0.001) in VER of 10.3 from D01 to L54 and an increase (p < 0.001) of 10.7 from L60 to L66. 93 Table 4. 25 Multiple comparisons statistics for VER and loudness goal vocal load for experiment 2 (I) Load (J) Load Mean Difference (I - J) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound D01 L54 - 10.3 2.2 < 0.00 1 - 16.0 - 4.6 L60 - 13.9 2.2 < 0.00 1 - 19.7 - 8.2 L66 - 24.7 2.2 < 0.00 1 - 30.4 - 18.9 L54 D01 10.3 2.2 < 0.00 1 4.6 16.0 L60 - 3.6 2.2 0.355 - 9.3 2.1 L66 - 14.4 2.2 < 0.00 1 - 20.1 - 8.7 L60 D01 13.9 2.2 < 0.00 1 8.2 19.7 L54 3.6 2.2 0.355 - 2.1 9.3 L66 - 10.7 2.2 < 0.00 1 - 16.5 - 5.0 L66 D01 24.7 2.2 < 0.00 1 18.9 30.4 L54 14.4 2.2 < 0.00 1 8.7 20.1 L60 10.7 2.2 < 0.00 1 5.0 16.5 Fundamental Frequency and Loudness Goal The measured F0 met the assumptions for normality and independence. However, equal variance cannot be assumed. For each vocal load the F0 (M; SD) is as follows (summarized in Table 4. 2 6 with 95% confidence interval, minimum, and maximum): D01; no vocal load (M = - 0.01 ST; SD = 0.69 ST), L54; loudness goal of 54 dB (M = 0.65 ST; SD = 0.98 ST), L60; loudness goal of 60 dB (M = 0.8 ST; SD = 0.92 ST), L66; loudness goal of 66 dB (M = 1.14 ST; SD = 1.18 ST). Figure 4. 1 3 shows boxplots of the F0 across the loads. 94 Table 4. 26 Descriptive statistics for F0 and loudness goal vocal load for experiment 2 F undamental Frequency (ST) Load Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 - 0.01 0.69 - 0.14 0.12 - 1.93 2.74 L54 0.65 0.98 0.47 0.84 - 2.00 2.89 L60 0.80 0.92 0.62 0.98 - 1.53 3.23 L66 1.14 1.18 0.91 1.38 - 1.76 3.62 Figure 4. 13 Boxplot of F0 and loudness goal vocal load for experiment 2 There was a significant (p < 0.001) main effect of F0 on the vocal load of loudness goal. The post hoc tests show that F0 was different in the pair - wise comparisons with D01 and between L54 and L66 (summarized in Table 4. 2 7 ). There was an increase (p < 0.001) in F0 of 0.67 ST from D01 to L54 and an increase (p = 0.008) of 0.49 ST from L54 to L66. 95 Table 4. 27 Multiple comparison statistics for F0 and loudness goal vocal load for experiment 2 (I) Load (J) Load Mean Difference (I - J) (ST) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound D01 L54 - 0.67 0.11 < 0.00 1 - 0.97 - 0.36 L60 - 0.81 0.11 < 0.001 - 1.11 - 0.52 L66 - 1.16 0.14 < 0.001 - 1.52 - 0.80 L54 D01 0.67 0.11 < 0.001 0.36 0.97 L60 - 0.14 0.13 0.842 - 0.49 0.20 L66 - 0.49 0.15 0.008 - 0.89 - 0.09 L60 D01 0.81 0.11 < 0.00 1 0.52 1.11 L54 0.14 0.13 0.842 - 0.20 0.49 L66 - 0.35 0.15 0.117 - 0.74 0.05 L66 D01 1.16 0.14 < 0.00 1 0.80 1.52 L54 0.49 0.15 0.008 0.09 0.89 L60 0.35 0.15 0.117 - 0.05 0.74 Fundamental Frequency Standard Deviation and Loudness Goal The measured F0sd met the assumptions for normality, independence, and equal variance. For each vocal load the F0sd (M; SD) is as follows (summarized in Table 4. 2 8 with 95% confidence interval, minimum, and maximum): D01; no vocal load (M = 2.01 ST; SD = 0.79 ST), L54; loudness goal of 54 dB (M = 2.08 ST; SD = 0.74 ST), L60; loudness goal of 60 dB (M = 2.21 ST; SD = 0.8 ST), L66; loudness goal of 66 dB (M = 2.11 ST; SD = 0.78 ST). Figure 4. 1 4 shows boxplots of the F0sd across the loads. 96 Table 4. 28 Descriptive statistics for F0sd and loudness goal vocal load for experiment 2 Fundamental Frequency Standard Deviation (ST) Load Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 2.01 0.79 1.85 2.17 0.73 4.45 L54 2.08 0.74 1.93 2.22 0.73 4.37 L60 2.21 0.80 2.05 2.37 0.85 4.18 L66 2.11 0.78 1.97 2.26 0.93 4.38 Figure 4. 14 Boxplot of F0st and loudness goal vocal load for experiment 2 There was no significant main effect of F0sd on the vocal load of loudness goal. Speech Level and Loudness Goal The measured SL met the assumptions for normality, independence, and equal variance. For each vocal load the SL (M; SD) is as follows (summarized in Table 4. 2 9 with 95% confidence interval, minimum, and maximum): D01; no vocal load (M = 65.54 dB; SD = 3.39 dB), L54; 97 loudness goal of 54 dB (M = 6 7.4 dB; SD = 3.72 dB), L60; loudness goal of 60 dB (M = 67.99 dB; SD = 3.6 dB), L66; loudness goal of 66 dB (M = 69.9 dB; SD = 4.27 dB). Figure 4. 1 5 shows boxplots of the SL across the loads. Table 4. 29 Descriptive statistics for SL and loudness goal vocal load for experiment 2 Speech Level (dB) Load Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 65.54 3.39 64.89 66.19 57.29 74.08 L54 67.40 3.72 66.70 68.10 56.93 74.29 L60 67.99 3.60 67.31 68.67 57.41 75.34 L66 69.90 4.27 69.08 70.71 57.48 77.56 98 Figure 4. 15 Boxplot of SL and loudness goal vocal load for experiment 2 There was a significant (p < 0.001) main effect of SL on the vocal load of loudness goal. The post hoc tests show that SL was different in all pair - wise comparisons with D01 and L66 (summarized in Table 4. 30 ). There was an increase (p = 0.002) in SL of 1.8 6 dB from D01 to L54 and an increase (p = 0.001) of 1.91 dB from L60 to L66. 99 Table 4. 30 Multiple comparison statistics for SL and loudness goal vocal load for experiment 2 (I) Load (J) Load Mean Difference (I - J) (dB) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound D01 L54 - 1.86 0.51 0.002 - 3.17 - 0.54 L60 - 2.45 0.51 < 0.00 1 - 3.76 - 1.13 L66 - 4.35 0.51 < 0.00 1 - 5.68 - 3.03 L54 D01 1.86 0.51 0.002 0.54 3.17 L60 - 0.59 0.51 0.648 - 1.89 0.71 L66 - 2.50 0.51 < 0.00 1 - 3.81 - 1.19 L60 D01 2.45 0.51 < 0.00 1 1.13 3.76 L54 0.59 0.51 0.648 - 0.71 1.89 L66 - 1.91 0.51 0.001 - 3.22 - 0.59 L66 D01 4.35 0.51 < 0.00 1 3.03 5.68 L54 2.50 0.51 < 0.00 1 1.19 3.81 L60 1.91 0.51 0.001 0.59 3.22 Standard Deviation of Speech Level and Loudness Goal The measured SLsd met the assumptions for normality, independence, and equal variance. For each vocal load the SLsd (M; SD) is as follows (summarized in Table 4. 31 with 95% confidence interval, minimum, and maxim um): D01; no vocal load (M = 3.41 dB; SD = 0.3 dB), L54; loudness goal of 54 dB (M = 3.5 dB; SD = 0.32 dB), L60; loudness goal of 60 dB (M = 3.48 dB; SD = 0.32 dB), L66; loudness goal of 66 dB (M = 3.47 dB; SD = 0.27 dB). Figure 4. 1 6 shows boxplots of the SLsd across the loads. 100 Table 4. 31 Descriptive statistics for SLsd and loudness goal vocal load for experiment 2 Speech Level Standard Deviation (dB) Load Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 3.41 0.30 3.35 3.46 2.63 4.21 L54 3.50 0.32 3.44 3.56 2.66 4.22 L60 3.48 0.32 3.42 3.54 2.79 4.26 L66 3.47 0.27 3.42 3.52 2.98 4.19 Figure 4. 16 Boxplot of SLsd and loudness goal vocal load for experiment 2 There was no significant main effect of SLsd on the vocal load of loudness goal. 101 Smoothed Cepstral Peak Prominence and Loudness Goal The measured CPPS met the assumptions for normality, independence, and equal variance. For each vocal load the CPPS (M; SD) is as follows (summarized in Table 4. 3 2 with 95% confidence interval, minimum, and maximum): D01; no vocal load (M = 15.33 dB; SD = 1.22 dB), L54; loudness goal of 54 dB (M = 15.83 dB; SD = 1.33 dB), L60; loudness goal of 60 dB (M = 15.89 dB; SD = 1.32 dB), L66; loudness goal of 66 dB (M = 16.39 dB; SD = 1.36 dB). Figure 4. 1 7 shows boxplots of the CPPS across the loads. Table 4 .32 Descriptive statistics for CPP S and loudness goal vocal load for experiment 2 Smoothed Cepstral Peak Prominence (dB) Load Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 15.33 1.22 15.09 15.57 12.40 18.12 L54 15.83 1.33 15.58 16.08 12.43 19.20 L60 15.89 1.32 15.64 16.14 12.46 19.01 L66 16.39 1.36 16.13 16.64 12.20 19.17 102 Figure 4. 17 Boxplot of CPPS and loudness goal vocal load for experiment There was a significant (p < 0.001) main effect of CPPS on the vocal load of loudness goal. The post hoc tests show that CPPS was different in all pair - wise comparisons with D01 and L66 (summarized in Table 4. 3 3 ). There was an increase (p = 0.029) in CPPS of 0.50 dB from D01 to L54 and an increase (p = 0.025) of 0.50 dB from L60 to L66. 103 Table 4. 33 Multiple comparison sta tistics for CPPS and loudness goal vocal load for experiment 2 (I) Load (J) Load Mean Difference (I - J) (dB) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound D01 L54 - .50 .18 .0289 - .96 - .04 L60 - .56 .18 .0099 - 1.02 - .10 L66 - 1.06 .18 < 0.00 1 - 1.52 - .60 L54 D01 .50 .18 .0289 .04 .96 L60 - .06 .18 .9832 - .52 .39 L66 - .56 .18 .0080 - 1.02 - .11 L60 D01 .56 .18 .0099 .10 1.02 L54 .06 .18 .9832 - .39 .52 L66 - .50 .18 .0253 - .95 - .04 L66 D01 1.06 .18 < 0.00 1 .60 1.52 L54 .56 .18 .0080 .11 1.02 L60 .50 .18 .0253 .04 .95 V ocal E ffort R ating and B ackground N oise The measured VER met the assumptions for normality, independence, and equal variance. For each vocal load the VER (M; SD) is as follows (summarized in Table 4. 3 4 with 95% confidence interval, minimum, and maximum): D01; no vocal load (M = 25.13 ; SD = 16.43 ), N53; background noise of 53 dBA (M = 31.91 ; SD = 15.68 ), N62; background noise of 62 dBA (M = 40.19 ; SD = 18.52 ), N71; backgrou nd noise of 71 dBA (M = 52.62 ; SD = 17.99 ). Figure 4. 1 8 shows boxplots of the VER across the loads. 104 Table 4. 34 Descriptive statistics for VER and background noise vocal load for experiment 2 Vocal Effort Rating Load Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 25.1 16.4 22.0 28.3 2 65 N53 31.9 15.7 28.9 34.9 2 75 N62 40.2 18.5 36.7 43.7 2 90 N71 52.6 18.0 49.2 56.1 13 100 Figure 4. 18 Boxplot for VER and background noise vocal load for experiment 2 T here was a significant (p < 0.001) main effect of VER on the vocal load of background noise. The post hoc tests show that VER was different in all pair - wise comparisons (summarized in Table 4. 3 5 ). There was an increase (p = 0.021) in VER of 6.8 from D01 to N53, an increase (p = 0.003) of 8.3 from N53 to N62, and an increase (p < 0.001) of 12.4 from N62 to N71. Table 4. 35 Multiple comparison statistics for VER and background noise vocal load for experiment 2 105 (I) Load (J) Load Mean Difference (I - J) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound D01 N53 - 6.8 2.3 0.021 - 12.8 - 0.7 N62 - 15.1 2.3 < 0.00 1 - 21.1 - 9.0 N71 - 27.5 2.4 < 0.00 1 - 33.5 - 21.4 N53 D01 6.8 2.3 0.021 0.7 12.8 N62 - 8.3 2.3 0.003 - 14.3 - 2.2 N71 - 20.7 2.3 < 0.00 1 - 26.8 - 14.7 N62 D01 15.1 2.3 < 0.00 1 9.0 21.1 N53 8.3 2.3 0.003 2.2 14.3 N71 - 12.4 2.3 < 0.00 1 - 18.5 - 6.4 N71 D01 27.5 2.4 < 0.00 1 21.4 33.5 N53 20.7 2.3 < 0.00 1 14.7 26.8 N62 12.4 2.3 < 0.00 1 6.4 18.5 Fundamental Frequency and Background Noise The measured F0 met the assumptions for normality and independence. However, equal variance cannot be assumed. For each vocal load the F0 (M; SD) is as follows (summarized in Table 4. 3 6 with 95% confidence inte rval, minimum, and maximum): D01; no vocal load (M = - 0.01 ST; SD = 0.69 ST), N53; background noise of 53 dBA (M = 0.65 ST; SD = 0.78 ST), N62; background noise of 62 dBA (M = 1.69 ST; SD = 1.17 ST), N71; background noise of 71 dBA (M = 2.53 ST; SD = 1.36 ST). Figure 4. 19 shows boxplots of the F0 across the loads. 106 Table 4. 36 Descriptive statistics for F0 and background noise vocal load for experiment 2 Fundamental Frequency (ST) Load Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 - 0.01 0.69 - 0.14 0.12 - 1.93 2.74 N53 0.65 0.78 0.50 0.80 - 1.02 2.81 N62 1.69 1.17 1.46 1.91 - 0.54 4.40 N71 2.53 1.36 2.24 2.81 - 0.32 5.07 Figure 4. 19 Boxplot for F0 and background noise vocal load for experiment 2 There was a significant (p < 0.001) main effect of F0 on the vocal load of background noise. The post hoc tests show that F0 was different in all pair - wise comparisons (summarized in Table 4. 3 7 ). There was an increase (p < 0.0 0 1) in F0 of 0.66 ST from D01 to N53, an increase (p < 0.00 1 ) of 1.04 ST from N53 t o N62, and an increase (p < 0.001) of 0.84 ST from N62 to N71. 107 Table 4. 37 Multiple comparison statistics for F0 and background noise vocal load for experiment 2 (I) Load (J) Load Mean Difference (I - J) (ST) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound D01 N53 - .66 .10 < .00 01 - .93 - .40 N62 - 1.70 .13 < .00 01 - 2.05 - 1.35 N71 - 2.54 .16 < .00 01 - 2.96 - 2.12 N53 D01 .66 .10 < .00 01 .40 .93 N62 - 1.04 .14 < .00 01 - 1.40 - .68 N71 - 1.87 .16 < .00 01 - 2.30 - 1.45 N62 D01 1.70 .13 < .00 01 1.35 2.05 N53 1.04 .14 < .00 01 .68 1.40 N71 - .84 .18 < .00 01 - 1.32 - .36 N71 D01 2.54 .16 < .00 01 2.12 2.96 N53 1.87 .16 < .00 01 1.45 2.30 N62 .84 .18 < .00 01 .36 1.32 Standard Deviation of Fundamental Frequency and Background Noise The measured F0sd met the assumptions for normality, independence, and equal variance. For each vocal load the F0sd (M; SD) is as follows (summarized in Table 4. 3 8 with 95% confidence interval, minimum, and maximu m): D01; no vocal load (M = 1.99 ST; SD = 0.76 ST), N53; background noise of 53 dBA (M = 1.99 ST; SD = 0.63 ST), N62; background noise of 62 dBA (M = 2.08 ST; SD = 0.72 ST), N71; background noise of 71 dBA (M = 1.92 ST; SD = 0.7 ST). Figure 4. 2 0 shows boxp lots of the F0sd across the loads. 108 Table 4. 38 Descriptive statistics for F0sd and background noise vocal load for experiment 2 Fundamental Frequency Standard Deviation (ST) Load Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 1.99 0.76 1.84 2.14 0.73 4.08 N53 1.99 0.63 1.87 2.12 0.88 3.65 N62 2.08 0.72 1.94 2.22 0.91 4.05 N71 1.92 0.70 1.78 2.05 0.21 3.99 Figure 4. 20 Boxplot for F0sd and background noise vocal load for experiment 2 There was no significant main effect of F0sd on the vocal load of background noise. 109 Speech Level and Background Noise The measured SL met the assumptions for normality, independence, and equal variance. For each vocal load the SL (M; SD) is as follows (summarized in Table 4. 3 9 with 95% confidence interval, minimum, and maximum): D01; no vocal load (M = 65.54 dB; SD = 3.39 dB), N53; background no ise of 53 dBA (M = 68.21 dB; SD = 3.75 dB), N62; background noise of 62 dBA (M = 71.09 dB; SD = 3.57 dB), N71; background noise of 71 dBA (M = 75.24 dB; SD = 3.41 dB). Figure 4. 2 1 shows boxplots of the SL across the loads. Table 4. 39 Descriptive statistics for SL and background noise vocal load for experiment 2 S peech L evel (dB) Load Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 65.54 3.39 64.89 66.19 57.29 74.08 N53 68.21 3.75 67.51 68.92 58.07 75.54 N62 71.09 3.57 70.42 71.76 61.47 80.56 N71 75.24 3.41 74.60 75.89 65.49 82.78 110 Figure 4. 21 Boxplot for SL and background noise vocal load for experiment 2 T here was a significant (p < 0.001) main effect of SL on the vocal load of background noise. The post hoc tests show that SL was different in all pair - wise comparisons (summarized in Table 4. 40 ). There was an increase (p < 0.0 0 1) in SL of 2.67 dB from D01 to N53, an increase (p < 0.00 1 ) of 2.88 dB from N53 to N62, and an increase (p < 0.001) of 4.15 dB from N62 to N71. 111 Table 4. 40 Multiple comparison statistics for SL and background noise vocal load for experiment 2 (I) Load (J) Load Mean Difference (I - J) (dB) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound D01 N53 - 2.67 0.48 < 0.00 1 - 3.90 - 1.44 N62 - 5.55 0.48 < 0.00 1 - 6.78 - 4.31 N71 - 9.70 0.48 < 0.00 1 - 10.94 - 8.46 N53 D01 2.67 0.48 < 0.00 1 1.44 3.90 N62 - 2.88 0.47 < 0.00 1 - 4.10 - 1.65 N71 - 7.03 0.48 < 0.00 1 - 8.26 - 5.80 N62 D01 5.55 0.48 < 0.00 1 4.31 6.78 N53 2.88 0.47 < 0.00 1 1.65 4.10 N71 - 4.15 0.48 < 0.00 1 - 5.38 - 2.93 N71 D01 9.70 0.48 < 0.00 1 8.46 10.94 N53 7.03 0.48 < 0.00 1 5.80 8.26 N62 4.15 0.48 < 0.00 1 2.93 5.38 Standard Deviation of Speech Level and Background Noise The measured SLsd met the assumptions for normality, independence, and equal variance. For each vocal load the SLsd (M; SD) is as follows (summarized in Table 4. 41 with 95% confidence interval, minimu m, and maximum): D01; no vocal load (M = 3.41 dB; SD = 0.3 dB), N53; background noise of 53 dBA (M = 3.43 dB; SD = 0.31 dB), N62; background noise of 62 dBA (M = 3.43 dB; SD = 0.31 dB), N71; background noise of 71 dBA (M = 3.32 dB; SD = 0.34 dB). Figure 4. 2 2 shows boxplots of the SLsd across the loads. 112 Table 4. 41 Descriptive statistics for SLsd and background noise vocal load for experiment 2 Speech Level Standard Deviation (dB) Load Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 3.41 0.30 3.35 3.46 2.63 4.21 N53 3.43 0.31 3.37 3.49 2.57 4.22 N62 3.43 0.31 3.37 3.49 2.59 4.04 N71 3.32 0.34 3.25 3.39 2.61 4.08 Figure 4. 22 Boxplot for SLsd and background noise vocal load for experiment 2 There was no main effect of SLsd on the vocal load of background noise. 113 Smoothed Cepstral Peak Prominence and Background Noise The measured CPPS met the assumptions for normality, independence, and equal variance. For each vocal load the CPPS (M; SD) is as follows (summarized in Table 4. 42 with 95% confidence interval, minimum, and maximum): D01; no vocal load (M = 15.27 dB; SD = 1.29 dB), N53; background noise of 53 dBA (M = 15.43 dB; SD = 1.44 dB), N62; background noise of 62 dBA (M = 15.07 dB; SD = 1.34 dB), N71; background noise of 71 dBA (M = 14.42 dB; SD = 1.13 dB). Figure 4. 2 3 shows boxplots of the CPPS across the loads. Table 4. 42 Descriptive statistics for CPPS and backgroun d noise vocal load for experiment 2 Smoothed Cepstral Peak Prominence (dB) Load Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound D01 15.27 1.29 15.02 15.52 11.80 18.12 N53 15.43 1.44 15.16 15.70 11.69 18.24 N62 15.07 1.34 14.82 15.33 11.92 18.09 N71 14.42 1.13 14.20 14.65 11.46 17.35 114 Figure 4. 23 Boxplot for CPPS and background noise vocal load for experiment 2 There was a significant (p < 0.001) main effect of CPPS on the vocal load of background noise. The post hoc tests show that SL was different in all pair - wise comparisons with N71 (summarized in Table 4. 43 ). There was a de crease (p < 0.0 0 1) in CPPS of 0.84 dB from D01 to N 71 , a de crease (p < 0.00 1 ) of 1.01 dB f rom N53 to N 71 , and a de crease (p = 0.00 2 ) of 0.65 dB from N62 to N71. 115 Table 4. 43 Multiple comparison statistics for CPPS and background noise vocal load for experiment 2 (I) Load (J) Load Mean Difference (I - J) (d B ) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound D01 N53 - 0.17 0.18 0.789 - 0.62 0.29 N62 0.20 0.18 0.693 - 0.27 0.66 N71 0.84 0.18 < 0.00 1 0.37 1.31 N53 D01 0.17 0.18 0.789 - 0.29 0.62 N62 0.36 0.18 0.174 - 0.09 0.82 N71 1.01 0.18 < 0.00 1 0.55 1.47 N62 D01 - 0.20 0.18 0.693 - 0.66 0.27 N53 - 0.36 0.18 0.174 - 0.82 0.09 N71 0.65 0.18 0.002 0.18 1.12 N71 D01 - 0.84 0.18 < 0.00 1 - 1.31 - 0.37 N53 - 1.01 0.18 < 0.00 1 - 1.47 - 0.55 N62 - 0.65 0.18 0.002 - 1.12 - 0.18 116 4 .3 Experiment 3 Table 4. 44 Abbreviations for results of experiment 3 Abbreviation Meaning F0 Fundamental frequency F0sd Standard deviation of fundamental frequency L Speech level Lsd Standard deviation of speech level CPPS Smoothed cepstral peak prominence VER Vocal effort rating ST Semitones dB Decibels VLT Vocal loading task PRE Time before vocal loading task POST Time after vocal loading task VL05 Time at 5 minutes after start of vocal loading task VL10 Time at 10 minutes after start of vocal loading task VL15 Time at 15 minutes after start of vocal loading task VL20 Time at 20 minutes after start of vocal loading task VL25 Time at 25 minutes after start of vocal loading task VL30 Time at 30 minutes after start of vocal loading task 4 .3.a Demographics Forty - two participants consented for the study. F ive participants were not included for either not finishing protocol or not properly following the instructions. A total of 37 participants, 1 8 males and 1 9 females, were included in the analyses presented below. All of these participants were within normal hearing limits. The participants were all college age and most received course credit as compensation for their participation. 117 4 .3.b Results The purpose of this experiment was to measure changes in vocal performance and vocal effort through a vocal load of background noise over time to implicate vocal fatigue from the vocal loading. Vocal effort rating ( VER ) was measured using the Borg CR - 100 b efore (PRE) the vocal loading task (VLT) , every five minutes during the vocal loading test (LT05, LT10, LT15, LT20, LT25, LT30) , and after (POST) the VLT . The same f ive acoustic parameters were selected to measure vocal performance, mean fundamental frequency (F0), standard deviation of fundamental frequency (F0sd), speech level (L), standard deviation of speech level (Lsd), and smoothed cepstral peak prominence (CPPS). One - way ANOVA tests were used to compare VER and each vocal performanc e measure throughout the VLT. used to investigate the pair - wise differences across each measurement in time. If equal variance Additionally, the data was clustered in to high and low vocal effort groups, voice change and no voice change groups, and four groups representing the intersections of these groups. The PRE and POST vocal performance of each group was compared using independent - samples t - tes ts for each acoustic variable. The results in this section are separated by VER results, the vocal performance results, the VER clustering results, and the final group acoustic results. Vocal Effort Rating and Vocal Loading The measured VER met the assumpt ions for normality and independence. However, equal variance cannot be assumed. For each time point of the vocal loading the VER (M; SD) is as follows (summarized in Table 4. 4 5 with 95% confidence interval, minimum, and maximum): PRE; before VLT (M = 17.11 ; SD = 13.88 ), VL05; after 5 minutes of VLT (M = 34.68 ; SD = 17.73 ), VL10; after 10 minutes of VLT (M = 36.86 ; SD = 19.15 ), VL15; after 15 minutes of 118 VLT (M = 40.27 ; SD = 20.51 ), VL20; after 20 minutes of VLT (M = 42.57 ; SD = 22.48 ), VL25; after 25 minutes of VLT (M = 45.35 ; SD = 24.69 ), VL30; after 30 minutes of VLT (M = 47.62 ; SD = 28.22 ), POST; after the complete VLT (M = 39.84 ; SD = 27.76 ), F igure 4. 24 shows a line graph of the VER across the load ing . Table 4. 45 Descriptive statistics for VER over time for experiment 3 V ocal Effort Ratings Time Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound PRE 17.1 13.9 12.5 21.7 2 60 VL05 34.7 17.7 28.8 40.6 7 80 VL10 36.9 19.1 30.5 43.2 8 79 VL15 40.3 20.5 33.4 47.1 10 85 VL20 42.6 22.5 35.1 50.1 7 85 VL25 45.4 24.7 37.1 53.6 7 90 VL30 47.6 28.2 38.2 57.0 2 100 POST 39.8 27.8 30.6 49.1 2 100 119 Figure 4. 24 Graph of VER over time for experiment 3 There was a significant (p < 0.001) main effect of VER across the time of the VLT. Only the VER at time PRE was significantly different the other time points of the VLT (summarized in Table 4.4 6 ) . There was a significant increase (p < 0.001 ) of 17.6 VER between VL05 and PRE . The were no significant changes of VER after VL05. The re was a significant increase (p = 0.001) of 22.7 VER between POST and PRE. 120 Table 4. 46 Multiple comparison statistics for VER over time for experiment 3 (I) Time (J) Time Mean Difference (I - J) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound PRE VL05 - 17.6 3.7 < 0.00 1 - 29.6 - 5.6 VL10 - 19.8 3.9 < 0.00 1 - 32.4 - 7.1 VL15 - 23.2 4.1 < 0.00 1 - 36.4 - 9.9 VL20 - 25.5 4.3 < 0.00 1 - 39.6 - 11.3 VL25 - 28.2 4.7 < 0.00 1 - 43.5 - 13.0 VL30 - 30.5 5.2 < 0.00 1 - 47.5 - 13.5 POST - 22.7 5.1 0.001 - 39.5 - 6.0 VL30 PRE 30.5 5.2 < 0.00 1 13.5 47.5 VL05 12.9 5.5 0.454 - 4.9 30.8 VL10 10.8 5.6 0.821 - 7.5 29.0 VL15 7.4 5.7 0.998 - 11.3 26.0 VL20 5.1 5.9 1.000 - 14.2 24.3 VL25 2.3 6.2 1.000 - 17.7 22.2 POST 7.8 6.5 0.999 - 13.3 28.8 POST PRE 22.7 5.1 0.001 6.0 39.5 VL05 5.2 5.4 1.000 - 12.5 22.8 VL10 3.0 5.5 1.000 - 15.1 21.0 VL15 - 0.4 5.7 1.000 - 18.9 18.0 VL20 - 2.7 5.9 1.000 - 21.8 16.3 VL25 - 5.5 6.1 1.000 - 25.3 14.3 VL30 - 7.8 6.5 0.999 - 28.8 13.3 Fundamental Frequency and Vocal Loading The measured F0 met the assumptions for normality and independence. However, equal variance cannot be assumed. For each time point of the vocal loading the F0 (M; SD) is as follows (summarized in Table 4. 4 7 with 95% confidence interval, minimum, and maximum): PRE; before VLT (M = - 0.01 ST; SD = 0.78 ST), VL05; after 5 minutes of VLT (M = 2.72 ST; SD = 1.9 ST), VL10; after 10 minutes of VLT (M = 3 ST; SD = 1.96 ST), VL15; after 15 121 minutes of VLT (M = 2.97 ST; SD = 1.92 ST), VL20; after 20 minutes of VLT (M = 3.06 ST; SD = 1.92 ST), VL25; after 25 minutes of VLT (M = 3.17 ST; SD = 2.04 ST), VL30; after 30 minutes of VLT (M = 3.04 ST; SD = 2.02 ST), POST; after the complete VLT (M = 0.83 ST; SD = 1.34 ST), Figur e 4. 25 shows a line graph of the F0 across the load ing . Table 4. 47 Descriptive statistics for F0 over time for experiment 3 F undamental Frequency (ST) Time Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound PRE - 0.01 0.78 - 0.11 0.09 - 1.84 3.31 VL05 2.72 1.90 2.41 3.04 - 1.08 7.52 VL10 3.00 1.96 2.68 3.32 - 0.42 8.14 VL15 2.97 1.92 2.64 3.29 - 1.00 8.28 VL20 3.06 1.92 2.74 3.38 - 0.41 7.94 VL25 3.17 2.04 2.83 3.51 - 0.23 7.82 VL30 3.04 2.02 2.70 3.38 - 0.89 8.09 POST 0.83 1.34 0.65 1.01 - 2.90 7.29 122 Figure 4. 25 Graph of F0 over time for experiment 3 There was a significant (p < 0.001) main effect of F0 across the time of the VLT. There was a significant increase (p < 0.001) in F0 of 2.74 ST from PRE to VL05 (summarized in Table 4.4 8 ) . There were no significant changes of F0 across the VLT. There was a significant increase (p < 0.001) of 0. 84 ST from PRE to POST. There was a significant decrease (p < 0.001) of 2. 21 ST from VL30 to POST. 123 Table 4. 48 Multiple comparison statistics for F0 over time for experiment 3 (I) Time (J) Time Mean Difference (I - J) (ST) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound PRE VL05 - 2.74 0.17 < 0.00 1 - 3.26 - 2.21 VL10 - 3.01 0.17 < 0.00 1 - 3.55 - 2.47 VL15 - 2.98 0.17 < 0.00 1 - 3.52 - 2.44 VL20 - 3.07 0.17 < 0.00 1 - 3.61 - 2.54 VL25 - 3.19 0.18 < 0.00 1 - 3.76 - 2.62 VL30 - 3.05 0.18 < 0.00 1 - 3.62 - 2.48 POST - 0.84 0.11 < 0.00 1 - 1.17 - 0.51 VL30 PRE 3.05 0.18 < 0.00 1 2.48 3.62 VL05 0.31 0.23 0.997 - 0.42 1.05 VL10 0.04 0.24 1.000 - 0.71 0.78 VL15 0.07 0.24 1.000 - 0.68 0.82 VL20 - 0.03 0.24 1.000 - 0.77 0.72 VL25 - 0.14 0.24 1.000 - 0.90 0.63 POST 2.21 0.19 < 0.00 1 1.59 2.82 POST PRE 0.84 0.11 < 0.00 1 0.51 1.17 VL05 - 1.90 0.18 < 0.00 1 - 2.48 - 1.32 VL10 - 2.17 0.19 < 0.00 1 - 2.76 - 1.58 VL15 - 2.14 0.19 < 0.00 1 - 2.73 - 1.55 VL20 - 2.23 0.19 < 0.00 1 - 2.82 - 1.65 VL25 - 2.35 0.20 < 0.00 1 - 2.96 - 1.73 VL30 - 2.21 0.19 < 0.00 1 - 2.82 - 1.59 Standard Deviation of Fundamental Frequency and Vocal Loading The measured F0sd met the assumptions for normality, independence, and equal variance. For each time point of the vocal loading the F0sd (M; SD) is as follows (summarized in Table 4. 4 9 with 95% confidence interval, minimum, and maximum): PRE; before VLT (M = 1.91 ST; SD = 0.83 ST), VL05; after 5 minutes of VLT (M = 2.3 ST; SD = 0.76 ST), VL10; after 10 minutes of VLT (M = 2.44 ST; SD = 0.77 ST), VL15; after 15 minutes of VLT (M = 2.49 ST; 124 SD = 0.73 ST), VL20; after 20 minutes of VLT (M = 2.52 ST; SD = 0.81 ST), VL25; after 25 minutes of VLT (M = 2.53 ST; SD = 0.78 ST), VL30; after 30 minutes of VLT (M = 2.51 ST; SD = 0.76 ST), POST; after the complete VLT (M = 2.16 ST; SD = 0.81 ST), Figure 4. 26 shows a line graph of the F0sd across the load ing . Table 4. 49 Descriptive statistics for F0sd over time for experiment 3 F undamental Frequency Standard Deviation (ST) Time Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound PRE 1.91 0.83 1.79 2.03 0.55 4.61 VL05 2.30 0.76 2.17 2.42 0.99 4.55 VL10 2.44 0.77 2.31 2.57 1.08 4.47 VL15 2.49 0.73 2.36 2.61 1.18 4.73 VL20 2.52 0.81 2.38 2.66 1.01 4.57 VL25 2.53 0.78 2.40 2.66 1.17 4.68 VL30 2.51 0.76 2.39 2.64 1.10 4.76 POST 2.16 0.81 2.04 2.28 0.88 4.68 125 Figure 4. 26 Graph of F0sd over time for experiment 3 There was a significant (p < 0.001) main effect of F0sd across the time of the VLT. There was an increase (p < 0.001) in F0sd of 0.6 ST from PRE to VEL05, no change across the VLT, and a decrease (p = 0.002) of 0.36 ST from VEL30 to POST (summarized in Tab le 4. 50 ) . There was also an increase (p = 0. 043) of 0.25 ST from PRE to POST. 126 Table 4. 50 Multiple comparison statistics for F0sd over time for experiment 3 (I) Time (J) Time Mean Difference (I - J) (ST) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound PRE VL05 - 0.39 0.09 < 0.00 1 - 0.65 - 0.12 VL10 - 0.53 0.09 < 0.00 1 - 0.79 - 0.27 VL15 - 0.58 0.09 < 0.00 1 - 0.84 - 0.31 VL20 - 0.61 0.09 < 0.00 1 - 0.87 - 0.35 VL25 - 0.62 0.09 < 0.00 1 - 0.88 - 0.35 VL30 - 0.60 0.09 < 0.00 1 - 0.87 - 0.34 POST - 0.25 0.08 0.043 - 0.49 0.00 VL30 PRE 0.60 0.09 < 0.00 1 0.34 0.87 VL05 0.22 0.09 0.285 - 0.07 0.50 VL10 0.08 0.09 0.993 - 0.21 0.36 VL15 0.03 0.09 1.000 - 0.26 0.32 VL20 - 0.01 0.09 1.000 - 0.29 0.28 VL25 - 0.01 0.09 1.000 - 0.30 0.27 POST 0.36 0.09 0.002 0.09 0.63 POST PRE 0.25 0.08 0.043 0.00 0.49 VL05 - 0.14 0.09 0.773 - 0.41 0.13 VL10 - 0.28 0.09 0.032 - 0.55 - 0.01 VL15 - 0.33 0.09 0.006 - 0.60 - 0.06 VL20 - 0.36 0.09 0.001 - 0.63 - 0.09 VL25 - 0.37 0.09 0.001 - 0.64 - 0.10 VL30 - 0.36 0.09 0.002 - 0.63 - 0.09 Speech Level and Vocal Loading The measured SL met the assumptions for normality, independence, and equal variance. For each time point of the vocal loading the SL (M; SD) is as follows (summarized in Table 4. 51 with 95% confidence interval, minimum, and maximum): PRE; before VLT (M = 56.23 dB; SD = 5.26 dB), VL05; after 5 minutes of VLT (M = 66.99 dB; SD = 4.91 dB), VL10; after 10 minutes of VLT (M = 67.45 dB; SD = 4.79 dB), VL15; after 15 minutes of VLT (M = 67 .57 dB; 127 SD = 4.79 dB), VL20; after 20 minutes of VLT (M = 67.53 dB; SD = 4.67 dB), VL25; after 25 minutes of VLT (M = 67.42 dB; SD = 5.11 dB), VL30; after 30 minutes of VLT (M = 66.86 dB; SD = 4.99 dB), POST; after the complete VLT (M = 57.46 dB; SD = 5.13 dB), Figure 4. 27 shows a line graph of the SL across the load ing . Table 4. 51 Descriptive statistics for SL over time for experiment 3 S peech Level (dB) Time Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound PRE 56.23 5.26 55.55 56.91 43.55 68.16 VL05 66.99 4.91 66.18 67.80 55.21 83.28 VL10 67.45 4.79 66.67 68.24 57.13 81.93 VL15 67.57 4.79 66.78 68.37 57.45 83.02 VL20 67.53 4.67 66.77 68.30 55.76 82.62 VL25 67.42 5.11 66.58 68.26 55.54 82.74 VL30 66.86 4.99 66.04 67.69 54.73 81.65 POST 57.46 5.13 56.75 58.17 42.93 70.54 128 Figure 4. 27 Graph of SL over time for experiment 3 There was a significant (p < 0.001) main effect of SL across the time of the VLT. There was a significant increase in S L of 10. 6 dB (p < 0.001) from PRE to VL 0 5 (summarized in Table 4.5 2 ) . There were no significant changes of SL across the VLT or from PRE and POST . There was a significant decrease in S L of 9.4 dB (p < 0.001) from VL30 to POST . 129 Table 4. 52 Multiple comparison statistics for SL over time for experiment 3 (I) Time (J) Time Mean Difference (I - J) (dB) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound PRE VL05 - 10.76 0.53 < 0.00 1 - 12.37 - 9.16 VL10 - 11.22 0.53 < 0.00 1 - 12.83 - 9.62 VL15 - 11.35 0.53 < 0.00 1 - 12.96 - 9.73 VL20 - 11.31 0.53 < 0.00 1 - 12.91 - 9.70 VL25 - 11.19 0.53 < 0.00 1 - 12.80 - 9.59 VL30 - 10.63 0.53 < 0.00 1 - 12.24 - 9.02 POST - 1.23 0.48 0.170 - 2.69 0.23 VL30 PRE 10.63 0.53 < 0.00 1 9.02 12.24 VL05 - 0.13 0.59 1.000 - 1.92 1.66 VL10 - 0.59 0.59 0.974 - 2.38 1.19 VL15 - 0.71 0.59 0.930 - 2.51 1.08 VL20 - 0.67 0.59 0.947 - 2.46 1.12 VL25 - 0.56 0.59 0.981 - 2.35 1.23 POST 9.40 0.55 < 0.00 1 7.75 11.06 POST PRE 1.23 0.48 0.170 - 0.23 2.69 VL05 - 9.53 0.54 < 0.00 1 - 11.19 - 7.88 VL10 - 9.99 0.54 < 0.00 1 - 11.64 - 8.34 VL15 - 10.12 0.55 < 0.00 1 - 11.77 - 8.46 VL20 - 10.07 0.54 < 0.00 1 - 11.73 - 8.42 VL25 - 9.96 0.54 < 0.00 1 - 11.61 - 8.31 VL30 - 9.40 0.55 < 0.00 1 - 11.06 - 7.75 Speech Level Standard Deviation and Vocal Loading The measured SLsd met the assumptions for normality and independence. However, equal variance cannot be assumed. For each time point of the vocal loading the SLsd (M; SD) is as follows (summarized in Table 4. 53 with 95% confidence interval, minimum, and maximum): PRE; befo re VLT (M = 3.56 dB; SD = 0.32 dB), VL05; after 5 minutes of VLT (M = 3.78 dB; SD = 0.22 dB), VL10; after 10 minutes of VLT (M = 3.77 dB; SD = 0.19 dB), VL15; after 15 130 minutes of VLT (M = 3.82 dB; SD = 0.21 dB), VL20; after 20 minutes of VLT (M = 3.81 dB; SD = 0.21 dB), VL25; after 25 minutes of VLT (M = 3.8 dB; SD = 0.21 dB), VL30; after 30 minutes of VLT (M = 3.79 dB; SD = 0.23 dB), POST; after the complete VLT (M = 3.63 dB; SD = 0.33 dB), Figure 4. 28 shows a line graph of the SLsd across the load ing . Table 4. 53 Descriptive statistics for SLsd over time for experiment 3 Speech Level Standard Deviation (dB) Time Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound PRE 3.56 0.32 3.51 3.61 3.00 4.31 VL05 3.78 0.22 3.74 3.81 3.22 4.37 VL10 3.77 0.19 3.74 3.81 3.11 4.20 VL15 3.82 0.21 3.79 3.85 3.26 4.45 VL20 3.81 0.21 3.78 3.84 3.29 4.38 VL25 3.80 0.21 3.77 3.84 3.33 4.32 VL30 3.79 0.23 3.75 3.83 3.25 4.50 POST 3.63 0.33 3.58 3.68 3.01 4.38 131 Figure 4. 28 Graph of SLsd over time for experiment 3 There was a significant (p < 0.001) main effect of SLsd across the time of the VLT. There was a significant increase in SLsd of 0.23 dB (p < 0.001) from PRE to VL05 (summarized in Table 4.5 4 ) . There were no significant changes of SLsd across the VLT or from PRE and POST. There was a significant decrease in SL sd of 0.16 dB (p < 0.001) from VL30 to POST. 132 Table 4. 54 Multiple comparison statistics for SLsd over time for experiment 3 (I) Time (J) Time Mean Difference (I - J) (dB) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound PRE VL05 - 0.21 0.03 < 0.00 1 - 0.31 - 0.12 VL10 - 0.21 0.03 < 0.00 1 - 0.30 - 0.12 VL15 - 0.26 0.03 < 0.00 1 - 0.35 - 0.16 VL20 - 0.25 0.03 < 0.00 1 - 0.34 - 0.15 VL25 - 0.24 0.03 < 0.00 1 - 0.34 - 0.15 VL30 - 0.23 0.03 < 0.00 1 - 0.32 - 0.13 POST - 0.07 0.04 0.838 - 0.18 0.05 VL30 PRE 0.23 0.03 < 0.00 1 0.13 0.32 VL05 0.01 0.03 1.000 - 0.07 0.10 VL10 0.01 0.02 1.000 - 0.06 0.09 VL15 - 0.03 0.03 0.999 - 0.11 0.05 VL20 - 0.02 0.03 1.000 - 0.10 0.06 VL25 - 0.02 0.03 1.000 - 0.10 0.07 POST 0.16 0.03 < 0.00 1 0.06 0.26 POST PRE 0.07 0.04 0.838 - 0.05 0.18 VL05 - 0.15 0.03 < 0.00 1 - 0.25 - 0.04 VL10 - 0.15 0.03 < 0.00 1 - 0.24 - 0.05 VL15 - 0.19 0.03 < 0.00 1 - 0.29 - 0.09 VL20 - 0.18 0.03 < 0.00 1 - 0.28 - 0.08 VL25 - 0.18 0.03 < 0.00 1 - 0.28 - 0.07 VL30 - 0.16 0.03 < 0.00 1 - 0.26 - 0.06 Smoothed Cepstral Peak Prominence and Vocal Loading The measured CPPS met the assumptions for normality and independence. However, equal variance cannot be assumed. For each time point of the vocal loading the CPPS (M; SD) is as follows (summarized in Table 4. 5 5 with 95% confidence interval, minimum, and ma ximum): PRE; before VLT (M = 13.39 dB; SD = 1.37 dB), VL05; after 5 minutes of VLT (M = 13.57 dB; SD = 1 dB), VL10; after 10 minutes of VLT (M = 13.53 dB; SD = 1.08 dB), VL15; after 15 133 minutes of VLT (M = 13.51 dB; SD = 1.07 dB), VL20; after 20 minutes of VLT (M = 13.49 dB; SD = 1.03 dB), VL25; after 25 minutes of VLT (M = 13.7 dB; SD = 0.99 dB), VL30; after 30 minutes of VLT (M = 13.61 dB; SD = 0.94 dB), POST; after the complete VLT (M = 13.44 dB; SD = 1.42 dB), Figure 4. 29 shows a line graph of the CPPS a cross the load ing . Table 4. 55 Descriptive statistics for CPPS over time for experiment 3 Smoothed Cepstral Peak Prominence (CPPS) Time Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound PRE 13.39 1.37 13.20 13.58 10.28 16.26 VL05 13.57 1.00 13.40 13.74 10.85 15.53 VL10 13.53 1.08 13.35 13.70 10.51 15.43 VL15 13.51 1.07 13.34 13.69 10.57 15.42 VL20 13.49 1.03 13.32 13.66 10.41 15.68 VL25 13.70 0.99 13.53 13.87 10.38 15.80 VL30 13.61 0.94 13.45 13.77 10.97 15.21 POST 13.44 1.42 13.23 13.65 10.38 16.41 134 Figure 4. 29 Graph of CPPS over time for experiment 3 There was no main effect of CPPS across the time of the VLT. Vocal Load Response Clustering In order to detect potential participants groups that differ in vocal effort as a response to the vocal loading, k - means clustering was used on the feature set from 3.4.e. Two features were determined to be sufficient to cluster the data. These features were the n oise l oad r esponse (NLR) and the t emporal l oad r esponse (TLR). Both features were statistically significant ( NLR: p = 0.003; TLR: p < 0.001) features in the k - means clustering (Table 4. 5 6 ). This significan ce was obtained through F tests and is used only for descriptive purpose s as the clusters are chosen to maximize the differences across the cases in the clusters. Independent samples t - tests confirm that means of th e two clusters for NLR (p < 0.001) and TLR (p = 0.001) are significantly different (Table 4. 5 7 ). These two features were not correlated meeting the assumptions of orthogonality (Table 4. 5 8 ) . There were 14 participants in cluster 1 (CL1) and 24 participants in 135 cluster 2 (CL2). The center of CL1 was NLR of 26.4 and TLR of 32.9. The center of CL2 was NLT of 12.2 and TLR of 0.8. Since both features relate to vocal load responses and are higher in CL1, CL1 was relabeled to high vocal load response (HVLR) and CL2 was relabeled to low vocal load response ( LVLR). Table 4. 5 9 summarizes the count and centers of the two clusters and Figure 4. 30 shows the data separated by cluster including cluster centers. Table 4. 56 F tests for feature s ignificance in k - means clustering of VER for experiment 3 Cluster Error F Sig. Mean Square df Mean Square df NLR 1739.954 1 164.146 35 10.600 .003 TLR 8993.050 1 213.338 35 42.154 < 0.00 1 Table 4. 57 Independent samples t - test for comparison of means between LVLR and HVLR clusters in experiment 3 Levene's Test for Equality of Variances t - test for Equality of Means F Sig. t df Sig. (2 - tailed) Mean Difference Std. Error Difference 95% Confidence Interval of the Difference Lower Upper NLR .940 .339 5.902 35 .000 20.461 3.467 13.423 27.498 TLR 1.512 .227 3.469 35 .001 21.730 6.264 9.014 34.447 136 Table 4. 58 Bivariate correlation between features NLR and TLR for experiment 3 TLR NLR Pearson Correlation .083 Sig. (2 - tailed) .625 Table 4. 59 Cluster information including number of cases and centers for NLR and TLR for experiment 3 Group Cluster Center N NLR TLR LVLR 23 12.2 0.8 HVLR 14 26.4 32.9 Figure 4. 30 Scatter plot of clustered data with cluster centers. Square markers are the LVLR group, triangle markers are the HVLR group, and the circle markers are the cluster centers as labeled. 137 The analyses for VER and vocal loadings were repeated with the two gro ups. The measured VER for HVLR met the assumptions for normality and independence. However, equal variance cannot be assumed. For each time point of the vocal loading the VER for HVLR (M; SD) is as follows (summarized in Table 4. 60 with 95% confidence inte rval, minimum, and maximum): PRE; before VLT (M = 15.43 ; SD = 9.9 ), VL05; after 5 minutes of VLT (M = 41.79 ; SD = 12.4 ), VL10; after 10 minutes of VLT (M = 45.07 ; SD = 16.14 ), VL15; after 10 minutes of VLT (M = 53.93 ; SD = 17.23 ), VL20; after 10 mi nutes of VLT (M = 60.86 ; SD = 15.82 ), VL25; after 10 minutes of VLT (M = 67.43 ; SD = 14.84 ), VL30; after 10 minutes of VLT (M = 74.71 ; SD = 16.05 ), POST; after the complete VLT (M = 61.86 ; SD = 29 ), Figure 4. 31 shows a line graph of the VER across the loading. The measured VER for LVLR met the assumptions for normality, independence, and equal variance. For each time point of the vocal loading the VER for LVLR (M; SD) is as follows (summarized in Table 4. 60 with 95% confidence interval, minimum, and maximum): PRE; before VLT (M = 18.13 ; SD = 15.95 ), VL05; after 5 minutes of VLT (M = 30.35 ; SD = 19.28 ), VL10; after 10 minutes of VLT (M = 31.87 ; SD = 19.42 ), VL15; after 10 minutes of VLT (M = 31.96 ; SD = 17.95 ), VL20; after 10 minutes of VLT (M = 31.43 ; SD = 18.35 ), VL25; after 10 minutes of VLT (M = 31.91 ; SD = 19.2 ), VL30; after 10 minutes of VLT (M = 31.13 ; SD = 19.99 ), POST; after the complete VLT (M = 26.43 ; SD = 16.37 ), Figure 4. 31 shows a line graph of the VER across the loading separated by the cluster groups. 138 Table 4. 60 Descriptive statistics for VER across time for HVLR and LVLR clusters for experiment 3 V ocal Effort Rating Cluster Time Mean Std. Deviation 95% Confidence Interval for Mean Minimum Maximum Lower Bound Upper Bound HVLR PRE 15.4 9.9 9.7 21.1 3.0 30.0 VL05 41.8 12.4 34.6 48.9 13.0 65.0 VL10 45.1 16.1 35.8 54.4 12.0 70.0 VL15 53.9 17.2 44.0 63.9 35.0 85.0 VL20 60.9 15.8 51.7 70.0 25.0 85.0 VL25 67.4 14.8 58.9 76.0 45.0 90.0 VL30 74.7 16.0 65.5 84.0 55.0 100.0 POST 61.9 29.0 45.1 78.6 2.0 100.0 LVLR PRE 18.1 15.9 11.2 25.0 2.0 60.0 VL05 30.3 19.3 22.0 38.7 7.0 80.0 VL10 31.9 19.4 23.5 40.3 8.0 79.0 VL15 32.0 17.9 24.2 39.7 10.0 79.0 VL20 31.4 18.4 23.5 39.4 7.0 83.0 VL25 31.9 19.2 23.6 40.2 7.0 84.0 VL30 31.1 20.0 22.5 39.8 2.0 85.0 POST 26.4 16.4 19.4 33.5 2.0 75.0 139 Figure 4. 31 Line graph of VER over time separated clusters HVLR and LVLR for experiment 3 There was a significant (p < 0.001) main effect of VER across the VLT for HVLR . There was no significant effect of VER across the VLT for LVLR . For HVLR , there was an increase of VER of 26.4 (p < 0.00 1 ) from PRE to VL05, an increase of 32.9 ( p < 0.001) from VEL30 to VEL 05, and an increase of 46.4 (p = 0.001) between PRE and POST (summarized in Table 4. 61 ) . 140 Table 4. 61 Multiple comparison statistics for VER across time for HVLR for experiment 3 (I) Time (J) Time Mean Difference (I - J) Std. Error Sig. 95% Confidence Interval Lower Bound Upper Bound PRE VL05 - 26.4 4.2 < 0.001 - 41.2 - 11.6 VL10 - 29.6 5.1 < 0.001 - 47.6 - 11.7 VL15 - 38.5 5.3 < 0.001 - 57.5 - 19.5 VL20 - 45.4 5.0 < 0.001 - 63.1 - 27.7 VL25 - 52.0 4.8 < 0.001 - 68.8 - 35.2 VL30 - 59.3 5.0 < 0.001 - 77.2 - 41.4 POST - 46.4 8.2 0.001 - 77.0 - 15.9 VEL30 PRE 59.3 5.0 < 0.001 41.4 77.2 VL05 32.9 5.4 < 0.001 14.0 51.9 VL10 29.6 6.1 0.001 8.5 50.7 VL15 20.8 6.3 0.075 - 1.1 42.6 VL20 13.9 6.0 0.570 - 7.0 34.8 VL25 7.3 5.8 0.999 - 13.0 27.6 POST 12.9 8.9 0.993 - 18.9 44.6 POST PRE 46.4 8.2 0.001 15.9 77.0 VL05 20.1 8.4 0.558 - 10.8 51.0 VL10 16.8 8.9 0.879 - 15.0 48.5 VL15 7.9 9.0 1.000 - 24.2 40.0 VL20 1.0 8.8 1.000 - 30.7 32.7 VL25 - 5.6 8.7 1.000 - 37.0 25.8 VL30 - 12.9 8.9 0.993 - 44.6 18.9 There was not a significant difference in VER between HVLR and LVLR at PRE or VL05. There were significant differences between HVLR and LVLR for the other times (summarized in Table 4.6 2 ) . In each of these cases, the VER of HVLR were higher than LVLR . Ther e was a 141 VER difference of 13.2 (p = 0.04) at VL10, 22.0 at VL15 (p = 0.001), 29.4 at VL20 (p < 0.001), 3 5.5 at VL25 (p < 0.001), 43.6 at VL30 (p < 0.001) , and 35.4 at POST (p < 0.001). Table 4. 62 Independent sample comparison of VER across time between c luster groups HVLR and LVLR for experi men t 3 Mean Difference ( HVLR - LVLR) Std. Error Difference 95% Confidence Interval of the Difference Sig. (2 - tailed) Lower Upper PRE - 2.7 4.7 - 12.3 6.9 0.573 VL05 11.4 5.8 - 0.3 23.2 0.056 VL10 13.2 6.2 0.6 25.8 0.040 VL15 22.0 6.0 9.8 34.1 0.001 VL20 29.4 5.9 17.4 41.4 < 0.00 1 VL25 35.5 6.0 23.3 47.7 < 0.00 1 VL30 43.6 6.3 30.8 56.4 < 0.00 1 POST 35.4 7.4 20.3 50.5 < 0.00 1 Acoustic Voice Change Clustering The re were 16 participants with significant (p < 0.05 for the model and for at least one acoustic covariate) general linear models (GLM) comparing the PRE - POST differences of the five vocal performance measures. The 21 participants without significant models were classi fied as a no voice change group (NVC) and the other 16 participants were classified as a voice change group (YVC) . The average goodness - of - (SD = 0.27) , since t he GLM of NVC were not significant , therefore no g oodness - of - fit coefficients ar e present ed (summarized in Table 4. 63 ) 142 Table 4. 63 Descriptive statistics for Pearson's R for general linear model for PRE - POST acoustic changes in voice change group (YVC) f or experiment 1 Mean Std. Deviation Minimum Maximum R 0.93 0.27 0.84 0.98 Vocal Load Response and Acoustic Voice Change Clustering The cross - sections of the clusters developed for vocal load response and acoustic voice change formed four groups, low vocal load response and no voice change (LV L R - NVC), low vocal load response and voice changes (LV L R - YVC), high vocal load response and no voice change ( HV L R - NVC), and high vocal load response and voice changes (HVLR - YVC). There wer e 15 participants (10 males and 5 females) in LVLR - NVC , 8 participants (2 males and 6 females) in LVLR - YVC , 6 participants (3 males and 3 females) in HVLR - NVC, and 8 participants ( 3 males and 5 females) in HVLR - YVC (summarize in Figure 4.32) . Voice Change Yes (YVC) 8 8 No (NVC) 15 6 Low (LVLR) High (HVLR) Vocal Load Response Figure 4. 32 Number of participants in each of the four vocal load response - voice change groups 143 The group descriptive statistics for PRE and POST the vocal loading task for vocal effort rating (VER), fundamental frequency (F0), fundamental frequency standard deviation (F0sd), speech level (SL), speech level standard deviation (SLsd), and smoothed cepstral peak prominence (CPPS) are summarized in Table 4. 6 4 . Table 4. 64 Descriptive statistics for all five acoustic voice measures across PRE - POST for each vocal load response - voice change groups for experiment 3 Time Group VER F0 (ST) F0sd (ST) SL (dB) SLsd (dB) CPPS (dB) PRE LVLR - NVC Mean 17.3 - 0.01 2.01 57.12 3.54 13.15 Std. Deviation 16.2 0.71 0.90 5.28 0.40 1.38 Minimum 2 - 1.31 0.90 45.68 2.44 9.66 Maximum 60 2.03 4.61 68.16 4.76 16.75 LVLR - YVC Mean 19.8 - 0.09 1.86 53.87 3.36 13.78 Std. Deviation 16.4 0.76 1.04 3.91 0.39 1.48 Minimum 3 - 1.33 0.55 44.79 2.51 9.44 Maximum 55 2.12 6.14 64.24 4.19 16.26 HVLR - NVC Mean 11.3 - 0.01 1.88 53.20 3.35 12.31 Std. Deviation 10.1 0.65 0.99 4.52 0.44 2.09 Minimum 3 - 1.20 0.76 44.26 2.32 8.61 Maximum 30 1.73 5.01 59.98 4.27 15.75 HVLR - YVC Mean 18.5 0.00 2.03 57.44 3.55 13.54 Std. Deviation 9.1 0.86 0.88 3.70 0.43 1.25 Minimum 3 - 1.84 0.69 51.90 2.70 10.16 Maximum 30 2.21 5.61 65.73 4.70 16.09 144 Table 4. 64 POST LVLR - NVC Mean 26.1 0.83 1.98 57.92 3.56 13.55 Std. Deviation 19.2 0.91 0.74 4.57 0.40 1.14 Minimum 2 - 1.53 0.88 44.83 2.54 10.07 Maximum 75 2.60 4.46 66.18 4.60 16.10 LVLR - YVC Mean 27.1 0.04 2.22 55.56 3.39 13.32 Std. Deviation 10.1 0.95 1.02 4.31 0.49 2.28 Minimum 7 - 1.84 0.92 45.38 2.50 8.99 Maximum 35 2.10 5.32 64.28 4.56 17.56 HVLR - NVC Mean 56.5 1.08 2.37 55.53 3.55 13.03 Std. Deviation 28.0 0.86 0.98 5.77 0.41 2.20 Minimum 10 - 0.70 1.00 45.38 2.49 9.53 Maximum 89 2.48 4.83 66.52 4.18 16.18 HVLR - YVC Mean 65.9 0.79 2.46 58.91 3.82 13.00 Std. Deviation 31 1.13 0.88 2.75 0.49 1.35 Minimum 2 - 2.22 1.26 54.84 2.84 9.73 Maximum 100 2.59 5.48 67.52 5.04 14.92 Only F0 was significantly different from PRE to POST for LVLR - NVC, HVLR - NVC , and HVLR - YVC . Group LVLR - NVC had a POST F0 increase (p < 0.001) of 0.8 5 ST, group HVLR - NVC had a POST increase (p < 0.001) of 1. 09 S T , and group HVLR - YVC had a post increase (p < 0.001) of 0. 78 ST . In addition to F0, group HVLR - YVC had a significant POST F0s d increase (p = 0.022) of 0.42 ST, a POST SL increase (p = 0.029) of 1.48 dB, a POST S Lsd increase of 0. 28 dB ( p = 0.0 04 ), and a POST CPPS decrease of 0.54 dB (p = 0.0 46 ). These comparisons are summarized in Table 4.6 5 . 145 Table 4. 6 5 Independent samples t - test comparing each of the five acoustic measures across PRE - POST for each vocal load response - voice change group for experiment 3 Sig. (2 - tailed) Mean Difference (POST - PRE) Std. Error Difference 95% Confidence Interval of the Difference Lower Upper LVLR - NVC F0 < 0.00 1 0.85 0.13 - 1.10 - 0.60 F0sd 0.866 - 0.02 0.13 - 0.24 0.29 SL 0.284 0.80 0.75 - 2.28 0.67 SLsd 0.752 0.02 0.06 - 0.15 0.11 CPPS 0.056 0.39 0.20 - 0.80 0.01 LVLR - YVC F0 0.482 0.13 0.18 - 0.49 0.23 F0sd 0.112 0.36 0.22 - 0.80 0.09 SL 0.057 1.70 0.88 - 3.45 0.05 SLsd 0.771 0.03 0.09 - 0.21 0.16 CPPS 0.250 - 0.46 0.40 - 0.33 1.25 HVLR - NVC F0 < 0.00 1 1.09 0.19 - 1.48 - 0.70 F0sd 0.066 0.50 0.26 - 1.02 0.03 SL 0.081 2.33 1.31 - 4.95 0.29 SLsd 0.118 0.20 0.13 - 0.45 0.05 CPPS 0.190 0.73 0.55 - 1.83 0.37 HVLR - YVC F0 < 0.00 1 0.78 0.19 - 1.17 - 0.40 F0sd 0.022 0.42 0.18 - 0.79 - 0.06 SL 0.029 1.48 0.67 - 2.80 - 0.15 SLsd 0.004 0.28 0.09 - 0.46 - 0.09 CPPS 0.045 - 0.54 0.27 0.01 1.07 146 CHAPTER V: DISCUSSION This chapter provides the interpretation, implications, limitations, and future reco mmendations from the findings presented in Chapter IV. The chapter is outlined as follows, for each experiment a review of the hypotheses and research questions will be followed by a discussion of the interpretation of the significant findings, implications of how th ese suppor t or do not support the research hypotheses, the limitations of the experimental methods, and future recommendations for further exploration. Following the discussions for the individual experiment s is a general discussion from the perspective of all of th e experimental results and the central hypothesis , H0 . H0: The changes in vocal performance, vocal effort, and/or their interaction through a vocal load will implicate vocal fatigue. 5.1 Experiment 1 The purpose of this experiment was to test hypothesis 1 (H1). Q1: Can perceived vocal effort be measured reliably and if so, how does vocal performance in terms of vocal intensity change with vocal effort? H1: Vocal performance in terms of vocal intensity will be distinct for each vocal effort level and be consistent within and across participants. In context of the central hypothesis (H0) , Experiment 1 aims to validate the use of the Borg CR - 100 scale as an instrument to measure vocal effort to be used in future work of vocal effort thro ugh vocal loading. Since vocal effort is a psychophysical phenomenon, vocal production as the physical manifestation of vocal effort should directly relate to the psychological sensations of 147 vocal effort. The refore, the instrument will be considered valid if the voice production for different vocal effort levels are distin c t and repeatable. 5.1.a Interpretation s and Implication s There were significant increase s of fundamental frequency (F0), speech level (SL), and smoothed cepstral peak prominence (CPPS) a cross the four cued vocal effort levels ( VEL02 - minimal vocal effort , VEL13 - slight vocal effort, VEL 2 5 - moderate vocal effort , VEL 50 - severe vocal effort ; from Borg CR - 100 scale on Fig. 3.1 ). It was expected that SL would increase with vocal effort l evel . This expectation comes from p revious experiments reporting increases of SL with vocal effort (Cushing et al., 2011; Rosenthal et al., 2014; Skinner et al., 1997) . Additionally, the acoustic definition for vocal effort is the speech level at 1 meter (ISO, 2002) , which supports the expectation of a direct connection between SL and vocal effort . Since vocal intensity measured by the speech level is distinct across the VELs, the first part of h ypothesis 1 (H1 , ocal pe rformance in terms of vocal intensity will be distinct for each vocal effort level across the participants is accepted. The significant changes of F0 and CPPS suggest that vocal intensity is not the only way talkers adjust their voices in response to increased vocal effort. Previous wor k (Jessen, Köster, & Gfroerer, 2005; McKenna & Stepp, 2018) has shown similar differences in F0 between conversation al and raised vocal effort speech. Likewise, CPPS has been shown to increase with raised vocal effort (McKenna & Stepp, 2018; Rosenthal et al., 2014) . These results further support that the effort levels elicited using the four levels from the Borg CR - 100 scale relates directly with voice production associated with vocal effort. There were significant changes in F0sd and SLsd with VEL50. This finding suggests that the speech changes asso ciated with higher levels of vocal effort may go beyond changes due to 148 slight and moderate vocal effort. This is a reasonable assumption as these measures have been shown to differ across severe conditions of the voice (V. Wolfe & Martin, 1997) . The test - retest reliability of SL was strong (R = 0.90) . This finding supports the second part of h ypothesis 1 (H1 ocal performance in terms of vocal intensity will be consistent within participants . Th is implies that a talker who self - calibrate s to the scale , as was done in the experiment, would be able to reliably repeat VEL at about the same segmentation distance . As a result, a talker should be able to experientially anchor themselves to the scale to reliably rate their vocal effort for a range of voice task s . The results of this experiment validate the use of the Borg CR - 100 scale to measure perceptual vocal effort . The VELs from the scale were distinct and repeatable in vo ice production which supports the psychophysical nature of vocal effort . 5.1.b Limitations First, the study is limited thro ugh having only four vocal effort levels elicited . Although they were distinct, they only represent half the scale. Additionally, the VELs in between the ones in the study may be interesting. Second, the population consists of college - age adults . Although this allows for a more homogenous population to study, the generalizability of the results may be limited. While the differences in t he groups are significantly different, this study cannot provide normative or expected values of vocal production with these levels of vocal effort due to the small sample size (N = 20). Lastly, the vocal effort scale was anecdotally anchored, in other wor ds the participants were provided with an ex ample of the extremes of the scale. Presu mably if the participant had not experienced or could not properly imagine the example presented, then the scale would be different for that individual. 149 5.1.c Future Reco mmendations The first recommendation for future applications of these findings is to scale up the experiment by the hundreds. As stated above (5.1.b) the study does not provide any normative values and does not provide differences within different populati ons. Future work can explore the various population factor s (biological sex, age, voice impairment, hearing impairment, etc.) that could affect vocal production and vocal effort. Additionally, population nor mative values could benefit clinical assessment of vocal effort , as evaluation of normative deviation is a common clinical instrument in diagnosis and therapy progress. Since multiple acoustic measures of vocal performance related to VEL, future analyses c ould investigate the relative relatio nship between these measures and their contribution to the differences across the VELs. 5.2 Experiment 2 The purpose of this experiment was to test hypothesis 2 (H2). Q2: To what degree are vocal performance and vocal effort related given three equivalent vocal load levels? H2: The vocal performance and vocal effort will be constant within equivalent load conditions. Experiment 2 contributes to understanding of how vocal effort and vocal performance are affected by vocal load ing from different vocal loads. More specifically, this experiment tests whether three equivalent loads (vocal loads that should maintain acoustic energy equivalence) will elicit th r ee distinct and equivalent vocal effort levels and vocal performance metrics . 5.2.a Interpretations and Implications Communication Distance Vocal Load 150 For the condition of communication distance vocal load there were significant increases with distance for both F0 and SL. Only the most extreme condition, D04, saw increases in VER and CPPS . The changes in F0 were not necessarily expected, but the effects are small (about 0.4 ST per doubling of distances) . There were expected differences in SL since it is natural to speak louder to someone further away , but the effects are much smaller than anticipated ( about 1.4 dB per doubling of distance). The inverse - square law states that a doubling of distance from the sound source results in an attenuati on of 6 dB of sound intensity. Similarly, by adjusting the measured SL for the distances (the calibration was at 50 cm, so t he distances are 6 dB reduction at each doubling) it would be expected to see that the SL at 1 meter is 59.5 dB, at 2 meters is 54.9 dB, and at 4 meters is 50.4 dB. The 60 dB SPL at 1 meter is consistent with the ISO standard of normal vocal effort (ISO, 2002) . Additionally, the VER at 1 meter was 25.1 which is - 100 scale. This connection further illustrates the utility of the Borg CR - 100 scale in measurements of vocal effort as the normal vocal effort from the ISO is equivalent to the moderate vocal effort of the Borg CR - 100 scale. The lower distance - adjusted SL values suggest that individuals adjust beyond the normal (D01 is this cas e) enough to meet the needs of the new communication situation instead of making each situation equally intelligible. In other words, the perceived loudness would have decreased with distance (since the SL increased less than 6 dB per doubling of distance) . While there was an SL difference for the distance conditions, talkers did not experience the effort of the production the same way as t here was only a significant change in VER between the baseline D01 and the extreme D04 conditions. However, the distance load response may have been surpassed by , a social phenomenon ; Traunmüller and Eriksson (2000, 2011) in speech production where differ ences between 0.3 m and 1.5 151 Talkers appear to retain their habitual vocal effort for this [ close ] range of distances [ between a communication partner ] and This effect was also observed for CPPS which suggests that voice quality improve ments are made to communicate at longer distances beyond the habitual floor. Loudness Goal Vocal Load For the condition of loudness goal vocal load there were significant increases from baseline to the three goal levels for VER, F0, SL, and CPPS. The first two goal levels (L54 and L60) were not significantly different in any parameter. The most extreme case of L66 has significant increases from L54 a nd L60 for VER, F0, SL, and CPPS. These findings suggest that there is a habitual floor effect with loudness goal of around L60 . As participants needed to meet a loudness goal that was excessive to their habitual loudness , there were voice production accom modations beyond increase of SL. Since the average SL for the baseline condition was 59.5 dB at 1 meter , the L54 and L60 should not require more effort than baseline. However, there is a difference between the baseline condition of D01 and the lower loudne ss goal conditions (L54 and L60) of VER, F0, SL, and CPPS . This suggests that although more vocal effort was not needed to meet the acoustic intensity requirements of the goal, participants perceived a vocal load associated with the goal. This has implicat ions that individuals may adapt to a perceived vocal demand whether or not actual voice production changes are needed to communicate. One example of this is in telecommunication settings. If two individuals are communicating orally through an online medium (e.g. phone call, video conferencing) , one may increase vocal effort and change their vocal production because they perceive a vocal load in the communication situation whether or 152 not the other person is having difficulties hearing them. If the inverse is also true (individuals not reducing effort when a load is removed) then this has implications in amplification use in schoolteachers. Schoolteachers that use amplification but continue to perceive a vocal load would likely not benefit from the gains provi ded by the amplification system. Background Noise Vocal Load For the vocal load condition of background noise there were significant differences between all load conditions for VER, F0, and SL . The parameters of VER, F0, and SL followed similar patterns as the previous loads of increasing with load severity. However, the background noise did not have a habitual floor effect. Changes of F0 and SL to accommodate for background noise is called the Lombard effect and has been well studied . There was a prediction of 6 dB differences across the three vocal loads per the Lombard slope found in Bottalico 2017. This was not the case for this data. The average Lombard slope was 0.39 ( dB of SL per 1 dBA of noise level ) as opposed to the 0. 65 per 1 dBA of noi se level previously reported. However, the slope in the present study is between the two slopes identified by Bottalico (0.24, 0.65) and also fall within the range of Lombard slopes reported by Lazarus ( 0.3 0.6 dB per 1 dB of noise level). T here t wo possib le explanation s of the se difference . (1) T he listener was 1 meter away as opposed to 2.5 m which might flatten the Lombard slope. (2) The speech tasks in the present study were map descriptions and not reading tasks. This additional cognitive load may have resulted in a dampened Lombard effect. There was also a Lombard effect of F0 resulting in a n average change of 0.10 ST per dBA of noise level. Another finding is that CPPS had significantly decrease d in the extreme noise condition (N71). This is remarkabl e because in all previous scenarios (including experiment 1) CPPS has increased with vocal effort and loads. This suggests that there are vocal adaptations inherently 153 different in voice response to background noise load than other loads. It could be the ca se that a goal of vocal quality is a ssociated with increased vocal effort . This would be supported by experiment 1, where CPPS increased with effort when there was no vocal load present. Additionally, the vocal loads of communication distance and loudness goal could have innate vocal quality goal implications . This is opposed to background noise where the vocal quality goal is superseded by the Lombard effect. The lack of significant diffe rences in CPPS for the lower background noise conditions could be a result of competing demands of voice quality and overcoming the noise interference, which is overcome when the noise becomes excessive. It reasonable that this is a learned reflex to optim ize effort while communicating in background noise. All Vocal Load Conditions The findings support that vocal effort and vocal performance change with vocal load when the load is beyond habitual voice use. This validates the use of the Borg CR - 100 scale an d the acoustic measures to be used in experiment 3 to show changes in vocal effort and vocal production as a result of changing vocal load level . However, evidence was not found that supports hy pothesis 2 (H2 The vocal performance and vocal effort will b e constant within equivalent load conditions . An explanation for why the communication distance was not equivalent with the other vocal loads is that it was too limited in extent (room dimensions) to be at the same level as the other two . The inequivalen ce between loudness goal and background noise may relate to vocal ability, specifically previous experience with a vocal load. The loudness goal vocal load is not common and therefore most participants do not have trained reflex to the load. Conversely, sp eaking in background noise is extremely common resulting in an establish reflex (Lombard effect ). This leads to the conclusion that experience with a vocal 154 load may result in more consistent response s to the vocal loading (response to the vocal load) as ev idenced by the differences in variance for the vocal effort ratings and acoustic measurements for the background noise load. Therefore, it is recommended to use a vocal load of background noise (and not loudness goal) for the most consistent vocal load responses, supporting the use of this load in experiment 3 . 5.2.b Limitations The most obvious limitation is that communication distance was limited to an extent of 4 meters. This was a space limitation. The other three vocal load conditions only had three variations and, in the case of loudness goal, the first two approached equivale nce due to the habitual floor. Similar to experiment 1, the population consists of college - age adults limiting the generalization of the res ults. Another possible limitation is that the room was an anechoic chamber. This is not a typical acoustic space and it could have made the participants disoriented . In particular, the background noise from loudspeakers sound s different in a space without reverberance due to the lack of surface reflections. As a result, the loudspeakers sound like headphones (only hearing the direct sound) which could have affected the perception of the noise (some participants commented on this effect). 5.2.c Future Recomm endations In general, the study accomplished its design. Further work could be done to test larger extents of the vocal load conditions or investigate various factors that contributed to the results . For example, a future study in a much large r anechoic or hemi - anechoic chamber could test the effect of additional distances to see if the expected 6 dB change in SL would start at a larger distance. Another future iteration could include different speaking tasks (such as reading) or more complicated descriptio n tasks to observe how the cognitive load could influence the 155 Lombard response. Finally, testing different types of background noise (pink, talker babble, etc.) or acoustic environment s and their effect on vocal loading would be beneficial. The results sug gest that experience with a vocal load may influence the response of the vocal load. This should be further tested through investigating the effect of familiarity and / or a training effect of the vocal load response. 5.3 Experiment 3 The purpose of this experiment was to test hypothesis 3 (H3). Q3: To what degree do vocal performance, vocal effort, and/or their interaction change given a combined vocal load of excess background noise over time? H3: The measured changes in vocal performance, vocal effort, and/or their interaction will change through a vocal load (background noise and prolonged speaking) . Using the tools developed in experiments 1 and 2, experiment 3 investigates the changes in vocal e ffort and vocal performance from prolonged speaking in background noise. If the o bserved changes in vocal effort , v ocal performance and their interaction implicate vocal fatigue, then the central hypothesis can be accepted validating the proposed framework . 5.3.a Interpretations and Implications The clearest changes in vocal effort and vocal performance are between the time before (PRE) the vocal loading task (VLT) and after 5 minutes (VL05) of the VLT. Here there are significant increases in each of the ou tcome variables except CPPS ( VER, F0, F0sd, SL, SLsd). This is consistent with the Lombard effect observed in experiment 2 . In that experiment there was an increase of 71 dBA of background noise , while here there was an increase of 75 dBA of background noi se. However, there were no observed changes throughout the VLT. This was not expected. Th is suggest s that there is a change in the manner of voicing to accommodate the noise 156 and that does not change until the noise is removed. VER trended upward through th e VLT but was not significant until clustering was performed. There were only PRE - POST differences of VER and F0 ( a nd a small difference in F0sd ). The PRE - POST increase of VER was expected and consistent with a majority of previous VLT studies. For example, t he increase d F0 has been shown in some studies and not in others. One possible explanation is that t he increase could be from a warm - up effect. This effect has been shown in college students as a voice change throughout the day (Ben - David and Icht 2016) , it has also been shown in schoolteachers across the workday (Rantala, Vilkman, Bloigu 2002). More changes in vocal production were expected between P RE and POST. Prior to clustering, the changes in VER and F0 are consistent with previously reported responses to vocal loading and vocal warmup but do not implicate vocal fatigue. The VER clustering provided two distinct groups that had significantly diff erent responses in VER to the vocal load. The two features that resulted from the clustering were the noise demand response and temporal demand response. These features are interesting because they directly relate to individual responses to the two differe nt vocal loads presented (background noise load and prolonged speaking load ). The fact that the group with high vocal load responses (HVLR) had large changes in VER, while the other group with low vocal load responses (LVLR) had no significant changes in V ER suggests that there is a strong individual component in voice change from loading and therefore vocal fatigue. This is consistent with previous work that showed individual differences in vocal fatigue throughout vocal loading. This also provides insight in the contradictory nature of previous attempts for measuring vocal fatigue associated with vocal loading. 157 Although the VER clustering provided useful information, the second stage of acoustic clustering provides further clarity. Four groups of particip ants were formed and compared PRE and POST. Three of the four groups showed the same changes in F0 a s seen by the aggregate subject pool. It is interesting to note that the group with low vocal load response and significant voice changes (LVLR - YVC) had no significant acoustic voice changes as a group. Further investi gation reveals that the individual variation is very high in this group with the direction of the voice changes not being consistent (which would result in an aggregate of no c hange, e.g. an increase in F0 averaged with a decrease in F0 would result in a group average of no change in F0). It was the case that the acoustic changes for the group associated with high vocal load response and significant voice changes (HVLR - YVC) were all similar in direction resulting in statistically significant changes in all acoustic measures comparing PRE and POST. This finding he measured changes in vocal performance, vocal effort, and/or their interaction will change through a vocal load , background noise load and temporal load ) as the interaction between vocal effort and vocal performance created a group with significant changes in vocal performance and vocal effort. 5.3.b Limitations Similar to the previous two ex periments , the population consists of college - age adults limiting the generalization of the results. Although this study had more participants than most other vocal loading studies (Fujiki & Sivasankar, 2017) , s egmenting the population into four groups greatly lowers the power of the statistics . Still seeing statistical differences with the small groups is notable. 158 5.3.c Future Recommendations As stated in the previous section (5.3.b Limitations), the study would benefit from more subjects. The study de sign and implementation were developed such that this study could be greatly scaled up for m any more participants . One element of this is that the code was developed using the free python - based platform PsychoPy. This allows for the study to be run with id entical instructions anywhere with a computer and the necessary hardware (microphones, speakers, etc.) . Additionally, the presentation program was developed to have auto segmentation protocols allowing for fast processing of the data thus greatly reducing computation cost. In order to allow for this work to be done at a much larger scale, the scripts used to present the VLT are available on GitHub for the download and use of others ( https://github.com/markolopolis/sVLT ) . Using similar VLT designs w ill allow for comparable work and an effective increase in sample size. In addition to scaling up the protocol, psycholog ical and physical measurements should be made of the participants to start to investigate potential correlations of the effect of vocal loading and individual characteristics (e.g. personality, vocal experience, etc.). Identifying these features of the participants could reveal potential risk factors for vocal fatigu e leading to an enhanced understanding of vocal fatigue , providing the fo undation to reduce its prevalence and impact. 5.4 Validation and Application of the Theoretical Framewo rk The primary motivation of the dissertation was to propose and test a theoretical framework for vocal fatigue. The framework was developed based on a literature review of vocal fatigue and the related concept of vocal effort. The framework is built around the concept that vocal fatigue (more specifically state fatigue) is the physiological and/or perceptual manifestation of a change in the voice that in of vocal load for a particular voice - related communication situation which may be a result of 159 vocal loading or vocal effort . Practically this framework models the concept that t he vocal fatigue cannot be necessarily determined, but the changes in vocal performance, vocal effort, and/or their interaction through a vocal load can be determined and will implicate vocal fatigue. This is related to the central hypothesis (H0) of the dissertation. The primary support of this framework is provided in part by H3. Here significant changes of vocal effort and vocal performance were measured through vocal loading as a result of a classifier from the inter action o f vocal effort and vocal performance. The second part of validating the framework is whether these changes implicate vocal fatigue. The HVLR - YVC group showed significant changes in vocal effort and vocal performance that are consistent with possibl e changes associated with vocal fatigue (e.g. increase in variability of F0 and SL, decrease in CPPS). It reasonable to suggest that the participants in this group experience vocal fatigue. Additionally, this framework provi des theoretical relationships b etween vocal performance and vocal fatigue that provide a possible explanation for the previous work on vocal fatigue that have reported inconclusive or contradictory results. If multiple groups of individuals exist with varying levels of fatigability, the n averaging all of these groups would result in null findings. Here subgroups ( whose existence s upports the framework ) reveal a si ngle group consisting of a little over 20% of the total participants to have significant differences in vocal effort and vocal performance. In other words, future applications will benefit from this classif ication of participants within a VLT that are most likely to have experienced vocal fatigue . This classification c an be used as an independent variable to determine possible covariates associated with fatigability. The other groups from the classification may also be interesting. Typical VLT studies have the assumption that the participant s all belong in either LVLR - NVC or HVLR - YVC. In other 160 words, there is a linear relationship assumed between voice production and vocal effort changes across vocal loading . The framework demonstrates that there are other subgroups that could be present in the data. One of these subgroups is a group th at had significant voice changes but did not have a high vocal load response (LVLR - YVC). This group is interesting because it may be the case that these individuals are experiencing vocal fatigue but not feeling effort changes in their voice. As a result, they may be the group of individuals that do not take proper vocal rest when experiencing vocal fatigue (like the HVLR - YVC group may). This is consistent with the theory provided by Whitling et al. (2015) that saw a group of particip ants with extraordinary endurance in the VLT. They concluded that this over endurance group possibly share traits with patients in voice clinics. In other words, the repeated overuse of the voice without regulatory measures could be a risk factor for voice disorders. This may be the more important group to study. As stated above, application of this framework would include using the classification as an independent variable to study the possib le differences in the groups. Although the study lacks statistic al power due to a low sample size for four groups, examples are presented below to show the application of the framework. The f irst example is the biological sex distribution in the four groups (Figure 5.1 ). The biggest differences are that there are 10 : 5 males to females in the LVLR - NVC group and 6 - 2 females to males in the LVLR - YVC group. This is interesting because, as stated above, LVLR - YVC is a group that is possibly associated with a higher risk of voice problems and the biological sex distribution here matches the fact that many more females report chronic voice disorders than males (Hunter, Tann er, & Smith, 2011; Roy et al., 2005) . 161 Voice Change Yes (YVC) Male 2 3 Female 6 5 No (NVC) Male 10 3 Female 5 3 Low (LVLR) High (HVLR) Vocal Load Response Figure 5. 1 Distribution of males and female participants across the vocal load response - voice change groups A final example of applying the framework is an investigation of ratings of vocal fatigue. Prior to completing experiment 3, the participants completed the vocal fatigue index (VFI). Additionally, the participants also rated their perceived vocal fatigue on a visual analog scale (VAS) before and afte r the VLT. Looking at this data in the context of the groups , it is the case that the groups with voice change (LVLR - YVC and HVLR - YVC) have significantly higher (p = 0.01) score s of the second component of the VFI as compared to the group with no voice cha nge (LVLR - NVC and HVLR - NVC). The voice change groups had a mean VFI - 2 score of 3.64 (SD = 2.56) , while the no voice change group had a mean score of 1.52 (SD = 2.04) . The VFI - 2 is the groups with high vocal load response ( HVLR - NVC and HVLR - YVC) had a significantly higher (p < 0.001) POST vocal fating rating ( POST - VFR ) tha n the group s with a low vocal load response ( LVLR - NVC and LVLR - HVC). The HVLR groups had a mean VFR of 0.70 (SD = 0.15) and the LVLR group had a mean POST - VFR of 0.39 (SD = 0.25). Additionally, there were no differences comparing PRE - VFR. These two findings show that the HVLR - YVC is contained in two different 162 approaches to quantify vocal fatigue. It als o informs how vocal fatigue is experienced differently between the four groups. Future applications of the framework could provide distinct groups of fatigable individuals to study the factors associated with vocal fatigue. Additionally, the framework mode ls the concept that an individual potential to fatigue is based on their vocal load response . Using the methods of this dissertation, the changes in vocal load response associated with a vocal load could be measured to test p otential benefits of therapeu tic interventions for individuals with chronic vocal fatigue . 163 CHAPTER VI: CONCLUSION In order to better understand vocal fatigue, a framework was proposed, and several underlying assumptions tested through experimentation. This framework models the conce pt th at vocal fatigue can be implicated through measured changes in vocal effort, vocal performance, and/or their interaction through vocal loading (response to a vocal load). Towards supporting this framework, t hree experiments were conducted . Experiment 1 illustrated how the Borg CR - 100 scale could be used as a tool to measure perceived vocal effort ratings (VER) and quantify the relationships between vocal performance and vocal effort. Th e results of this experiment illustrated connections between fundamental frequency (F0), speech level (SL), and smoothed cepstral peak prominence (CPPS) and vocal effort level . Additionally, this experiment showed that the speech levels produced by the participants were reliably repeatable across elicited vocal effort levels. Experiment 2 explored the relationships between vocal effort, vocal performance, and vocal loads. Using three different types of vocal loads, communication distance, loudness goal, and background noise, it was found that sign ificant changes in VER, F0, SL, and CPPS existed for vocal loads that were beyond habitual communication experiences. In particular, the background noise vocal load showed the largest changes in VER and the acoustic parameters, suggesting a more refined vo cal load response than the other vocal loads. Experiment 3 investigated the effects of VER and vocal performance throughout a vocal loading test (VLT) consisting of a temporal load (prolonged speech task of describing complex routes on maps and a backgroun d noise load) . I nitially not many changes could be detected as a result of the VLT , which is consistent with previous research with VLT. The data was then cluster ed into four groups based on the effort ratings associated with the responses to the vocal 164 loa ds of noise and time and t he acoustic vocal performance changes resulted . This clustering revealed that one group of participants that experienced both high vocal load responses and significant voice changes had [as a group] significant changes in VER and each acoustic parameter of vocal performance (F0 mean and standard deviation, SL mean and standard deviation, and CPPS). This finding suggests that this group of individuals, as opposed to the other groups, experienced the most vocal fatigue, validating th e framework. Th e proposed f ramework for the study of vocal fatigue can be applied in future studies to examine potential risks associated with vocal fatigue. Th is framework provides an analytical approach not previously used to determine fatiguability in individuals. However, future work must be done to expand the capabilities and further test the potential applications of the framework. 165 APPENDICES 166 APPEND I CES APPENDI X A : Experiment 1 Stimuli Automatic Speech Segments: The alphabet: A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U, V, W, X, Y, Z Count from one to twenty - five: One, two, three, four, five, six, seven, eight, nine, ten, eleven, twelve, thirteen, fourteen, fifteen, sixteen, seventeen, eighteen, nineteen, twenty, twenty - one, twenty - two, twenty - three, twenty - four, twenty - five Days of the week and months of the year: Monday, Tuesday, Wednesday, Thursday, Friday, Saturday, Sunday. January, February, March, April, May, June, July, August, Se ptember, October, November, December Reading Speech Segments: Marvin Williams is only nine. Marvin lives with his mother on Monroe Avenue in Vernon Valley. Marvin loves all movies. Whenever a new movie is in the area, Marvin is in row one, along the aisle. When the sunlight strikes raindrops in the air, they act as a prism and form a rainbow. The rainbow is a division of white light into many beautiful colors. These take the shape of a long round arch with its path high above. Please call Stella. Ask her to bring these things with her from the store: Six spoons of fresh snow peas, five thick slabs of blue cheese, and maybe a snack for her brother Bob. Route Descriptions: Route A: Describe how to get from Gresham to the Expo Center via Pioneer Square. Route B: Describe how to get from Union Station to the Airport via the Rose Quarter. Route C: Describe how to get from the Clackamas Towncenter to Beaverton via Gateway. 167 APPENDIX B: Experiment 2 Stimuli B elow are the maps and routes used for the communic ation task in experiment 2 (Figures B.1 through B.12) . Figure B. 1 Map used as example during tutorial Figure B. 2 Map used as practice during the tutorial Figure B. 3 Map used for D01, L60, and N71 Figure B. 4 Map used for D04, L60, and N53 Figure B. 5 Map used for D02, L66, and N53 Figure B. 6 Map used for D04, L54, and N62 168 Figure B. 7 Map used for D01, L60, and N71 Figure B. 8 Map used for D01, L66, and N62 Figure B. 9 Map used for D02, L54, N71 Figure B. 10 Map used for D04, L54, N62 Figure B. 11 Map used for D02, L66, and N53 Figure B. 12 Map used by communication partner 169 APPENDIX C: Experiment 3 Stimuli Rainbow Passage: When the sunlight strikes raindrops in the air, they act as a prism and form a rainbow. The rainbow is a division of white light into many beautiful colors. These take the shape of a long round arch, with its path high above, and its two e nds apparently beyond the horizon. There is, according to legend, a boiling pot of gold at one end. People look, but no one ever finds it. When a man looks for something beyond his reach, his friends say he is looking for the pot of gold at the end of the rainbow . Maps: The images used for the map description tasks are below (Figures C.1 through C .25) Figure C. 1 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 2 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 3 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 4 Image used in the map description task during the vocal loading task for experiment 3 170 Figure C. 5 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 6 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 7 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 8 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 9 Image used in the map description task during the vocal loading task for e xperiment 3 Figure C. 10 Image used in the map description task during the vocal loading task for experiment 3 171 Figure C. 11 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 12 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 13 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 14 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 15 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 16 Image used in the map description task during the vocal loading task for experiment 3 172 Figure C. 17 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 18 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 19 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 20 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 21 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 22 Image used in the map description task during the vocal loading task for experiment 3 173 Figure C. 23 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 24 Image used in the map description task during the vocal loading task for experiment 3 Figure C. 25 Image used in the map description task during the vocal loading task for experiment 3 174 APPENDIX D : Informed Consent Forms For experiments 1 and 2: Research Participant Information and Consent Form You are being asked to participate in a research study. Researchers are required to provide a consent form to inform you about the research study, to convey that participation is voluntary, to explain risks and benefits of participation, and to empower you to make an informed decision. You should feel free to ask the researchers any questions you may have. Study Title: Acoustic Recording of Voice and Speech Production Researcher and Title: Dr. Eric Hunter, Professor Department and Institution: Department of Communicative Sciences and Disorders at Michigan State University Address and Contact Information: 113 Oyer, East Lansing 48823, 517.353.8641 Sponsor: Michigan State University 1. PURPOSE OF RESEARCH You are being asked to participate in a study examining measures of speech function as well as how speech production is affected by a variety of factors. This project is to examine voice production mechanisms with a three part goal of: (1) training students on general voice analysis procedures, (2) co llecting pilot speech production data for potentially new ideas in understanding speech production, and (3) testing speech production hypothesis. This study is voluntary and you can stop at any time without penalty. 2. ELIGIBILITY CRITERIA To participate, it is expected that: you are above the age of 18 years old you are in good physical and mental health you have no previous history of persistent and significant speech or voice complaints You may be asked to confirm these criteria. 2. ALTERNATIVE OPTIONS There are no alternative procedures, but you have the option not to participate in this research study. 3. WHAT YOU WILL DO Participants will be asked to produce a range of voice and speech tasks while being recorded using a variety of sensors to investigate the coordination of speech production mechanisms (like breathing or mouth movement) and the final acoustic output (what can be heard). You may be asked to do a short hearing screen to ensure that your hearing is within normal lim its. If you agree to participate, you will perform several speech related tasks during a session lasting less than an hour. These tasks will be performed in a quiet room that may be unique in that it is built for recording speech. At the beginning of your participation, you will be asked to complete questionnaires about your voice use, vocal health, and personality: these may including the Voice 175 Handicap Index (VHI), the Vocal - Related Quality of Life (V - RQOL), the Vocal Fatigue Index (VFI), the Reflux Sympt om Index (RSI), and the Big Five Inventory (BFI - 10). You may also be administered a brief cognitive screening. Next, we will take audio recordings and make measurements of your voice and speech. You will be asked to perform various speech/voice and breathi ng tasks. These speech tasks will be performed with recording equipment to ensure that high quality information is obtained from the observation, which will later be used for our analysis of your speech. During the observation, we will ask you to perform several voice and speech tasks; some of these may be similar to what you would do in normal life while others may be more unique to you (creating funny sounds with your voice). Tasks may include reading printed materials, describing pictures, or problem - so lving puzzles. Some of these tasks may be performed while wearing earphones through which you may hear your own voice; your voice may be unprocessed, mixed with noise, or processed to simulate a reverberant room. We may ask you to rate your vocal effort, v ocal discomfort, quality of your voice, or general preference for a task. None of these tasks are designed to make you uncomfortable in anyway, though they may be unique and not what you would normally do with your voice in everyday life. These are used me asure your speech range profile (which measures your usual speech loudness and pitch) or other speech production parameters. 4. POTENTIAL BENEFITS While the program in which you are being asked to participate may have no immediate benefit for you, it ma y benefit others by increasing our knowledge of factors affecting measures of speech and vocal function. 5. POTENTIAL RISKS There is no know medical risk involved in this research program and the procedures should not cause you any undue discomfort. Perha ps your voice will experience some fatigue at the end of participation. Equipment which may be used are those found in singing studios, linguistic laboratories, and speech production laboratories. They include such items as microphones, vibration sensors o n the neck, straps around the waist/torso to measure breath, and tubes to measure airflow in the mouth. Any tube that would go in your mouth is discarded at the end of your participation. If a sensor is placed on your neck, it will not use adhesive and is low - voltage (the voltage is so low that they cannot be felt). There are no known risks for these devices. 6. PRIVACY AND CONFIDENTIALITY The data for this study are being collected confidentially. Neither the researchers nor anyone else will be able to l ink data to you. Data from this study will be stored in a secured location with limited access (locked cabinet in a locked room or a password protected computer in the locked laboratory). All information will be kept for at least three years after the clos e of the study. Only trained researchers under the jurisdiction of this project and Human Research Protection Program will have access to the data collected in the study. Information about you will be kept confidential to the maximum extent allowable by la w. Although we will make every effort to keep your data confidential there are certain times, such as a court order, where we may have to disclose your data . Identifying information will not be attached to any of your individual 176 responses when reporting re sults from the recordings or surveys. You will not be asked to give your name or any other information during the recording that will allow you or your place of employment to be identified. The results of this study may be published or presented at profess ional meetings, but the identities of all research participants will remain anonymous. By participating you agree to allow audio recordings which will be used for analysis. We would like to ask your permission to use your recordings in other ways outside of what is presented above. Please mark below if you allow us to use your recordings: (1) to be presented, usually as an example in a scientific reports or presentations; and/or (2) to allow your recordings to be part of a larger dataset that researchers o utside of the research team could access (e.g. public recording repository). In both cases, the recordings would be anonymous. (1) I agree to allow my anonymous voice recordings to be presented in reports and presentations. Yes No Initials____________ (2) I agree to allow my anonymous voice recordings to be part of a larger dataset for others to use. Yes No Initials____________ 7. YOUR RIGHTS TO PARTICIPATE, SAY NO, OR WITHDRAW Part icipation is voluntary. Refusal to participate will involve no penalty or loss of benefits to which you are otherwise entitled. You may discontinue participation at any time during the process without penalty or loss of benefits to which you are otherwise entitled. You may change your mind at any time and withdraw. You may choose not to answer any question, or to not complete a specific task, or choose to stop participating at any time. Whether you choose to participate or not will have no effect on your g rade or evaluation. 8. COSTS AND COMPENSATION FOR BEING IN THE STUDY As an incentive to participate, subjects/students who participate in this research will be allowed to earn CAS SONA credit for participation. Otherwise, no compensation or remuneration is implied. For those enrolled in courses that allow for CAS SONA credit, you may also find alternative assignments to earn extra credit if you choose not to participate in this research study. The CAS SONA system awards 1 credit per 1 hour of research par ticipation. Neither researchers nor individual instructors will know what studies participants are involved in. 9. THE RIGHT TO GET HELP IF INJURED If you are injured as a result of your participation in this research project, Michigan State University will assist you in obtaining emergency care, if necessary, for your research related injuries. If you have insurance for medical care, your insurance carrier will be billed in the ordinary manner. As with any medical insurance, any costs that are not cover ed or in excess of what are is not to provide financial compensation for lost wages, disability, pain or discomfort, unless required by law to do so. This d oes not mean that you are giving up any legal rights you may have. You may contact Dr. Eric Hunter at 517.353.8641 with any questions or to report an injury. 10. Contact INFORMATION If you have concerns or questions about this study, such as scientific issues, how to do any part of it, or to report an injury, please contact Dr. Eric Hunter, Michigan State Univ, 113 Oyer, East Lansing, MI 48823, 517.353.8641, ejhunter@msu.edu . 177 If you have questions or concerns about your role and rights as a research participant, would like to obtain information or offer input, or would like to register a complaint about this study, you may contact, anonymously Protection Program at 517 - 355 - 2180, Fax 517 - 432 - 4503, or e - mail irb@msu.edu or regular mail at 4000 Collins Rd, Suite 136, Lansing, MI 48910. 11. D OCUMENTATION OF INFORMED CONSENT Your signature below means that you voluntarily agree to participate in this research study. ___________________________ ______ ___________________ Signature Date You will be given a copy of this form to keep. 178 Consent form for Experiment 3: Research Participant Information and Consent Form You are being asked to participate in a research study. Researchers are required to provide a consent form to inform you about the research study, to convey that participation is voluntary, to explain risks and bene fits of participation, and to empower you to make an informed decision. You should feel free to ask the researchers any questions you may have concerning this project. Study Title: Gender, Age and Vocal Effort Researcher and Title: Dr. Eric Hunter, Associate Professor Department and Institution: Department of Communicative Sciences & Disorders at Michigan State University Address and Contact Information: 113 Oyer, East Lansing 48823, 517.353.8641 Sponsor: Michigan State University 1. PURPOSE OF RESEARCH You are being asked to participate in this study to help researchers gain a better understanding of how fast the voice gets tired due to a long reading task. In this task you will speak until you are tired of speaking (less than 30 minutes). This study is voluntary. 2. ELIGIBILITY CRITERIA It is expected that you have no significant vocal complaints and are in good physical and mental health. Persons with a history of recent hospitalization or suffering from any respiratory or ora l infections will be excluded from participating. Additionally: You must be between 18 - 70 years of age. You must be a native English speaker. You will be asked about items which might affect your voice (e.g. hearing, heartburn, smoker) 3. ALTERNATIV E OPTIONS There are no alternative procedures, but you have the option not to participate in this research study. 4. WHAT YOU WILL DO We expect that full participation in the study will take between 60 - 90 minutes. After a brief introduction to the stud y, you will be asked to participate in a screening process (15 - 25 minutes) followed by a prolonged speaking task (no more than 30 minutes). Since it is common for your mouth to dry out while speaking for a long time, during the prolonged speaking task, you will be given the opportunity to take regular small drinks of water. However, so that all participants drink the same amount of water, we will use small measured cups and have you drink at a regular intervals. The total amount of water to drink is about t he same amount as in a can of soda (less than 20 oz). During the screening process, you may be asked to do some or all of the following: Complete questionnaires about your voice and voice use. Answer questions about your vocal habits, hydration levels, and history of vocal fatigue. 179 Asked about current medications which may affect voice use (e.g. asthma, allergies, heartburn). Complete a short hearing screen to ensure that your hearing is within the normal age appropriate ranges. Complete a breathing test us ing a spirometer to measure your lung function; this will be repeated three to five times (our goal is three similar breaths). You will be asked to breathe in deeply and blow into the spirometer followed by an inhalation. How you breathe while speaking wi ll be observed, this is done by observation. We may additionally ask you to where a strap around your waist (outside your clothing) which detects when you breath. We may ask you to use a scale specifically designed to calculate body water percentage. The scale will also provide other readings such as body weight, body mass, etc. These measurements will only be used as they relate to hydration levels. After the screening process, you will do a prolonged speech reading task in a soundproof room used for rec ordings. You will wear a microphone that goes loosely around your neck and a microphone that goes on your head (similar to headphones). We may additionally ask you to wear a strap around your waist (outside your clothing) which detects when you breathe. Before and after the long speaking task, you will be asked to perform some simple vocal tasks, such as a you will be asked to read out loud for 30 minutes while noise is played in the room. This will likely result in you speaking at a loud volume. We expect that your voice will get tired before the 30 minutes is up. When you get tired of speaking, you may quit. Most people will go longer than 15 minutes and less t han 30 minutes. We are interested to know when people get tired of speaking. Every few minutes you will be reminded to drink a small cup (less than 30 mL) of water. 5. POTENTIAL BENEFITS While the program in which you are being asked to participate may h ave no immediate benefit for you, it may benefit others by increasing our knowledge of factors affecting measures of speech and vocal function. 6. POTENTIAL RISKS There is minimal risk involved in this research program and the procedures should cause you no undue discomfort. The noise level, while annoying, is less than the occupational safety limits. Our goal is to have you speak at a louder volume until your voice is tired. Likely your voice will experience some vocal fatigue, but this should resolve with some nominal vocal rest. In rare cases, extended speaking can result in hoarseness. If you think your voice is getting too tired or if you become uncomfortable with the tasks, you may quit at any time. Except for the spirometer, other devices to be u sed are similar to those found in singing studios, linguistic laboratories, and speech production laboratories. They include such items as microphones and surface microphones that go on the neck (to detect speech use in noise). An ear microphone and record er may be used to record the sound you are surrounded by. While the pulmonary function test is unlikely to cause injury, breathing hard may cause some discomfort. If there is anything in the screening that does not make you a good subject for our study, y ou will be remunerated for your time (see below) and no further participation is needed. If you are not healthy enough to participate (for example, if you have a cold or are hoarse from cheering at a 180 sports activity), or if one of the screening procedures indicates that you might not match the level of communication function we are looking for (for example, your hearing is limited compared to peers age matched adults), you may be asked to not participate further. The testing performed in this project is no t intended to find abnormalities, the protocol does not diagnose illness and we do not refer to health care providers. Data collected do not comprise a diagnostic or clinical study. Undetected vocal abnormalities are rare but it is possible that the invest igators may perceive a vocal abnormality during the initial screening. If this occurs, you will be advised to consult with a licensed physician to determine whether a health examination would be prudent. 7. PRIVACY AND CONFIDENTIALITY The data recorded for this study will be collected confidentially. Neither the researchers nor anyone else will be able to link data to you. The data for this project will be kept confidential. Data from this study will be stored in a locked cabinet in a locked roo m or a password protected computer in the locked laboratory. All information will be kept for at least three years after the close of the study. Only trained researchers under the jurisdiction of this project and Human Research Protection Program will have access to the data collected in the study. Information about you will be kept confidential to the maximum extent allowable by law. Although we will make every effort to keep your data confidential there are certain times, such as a court order, where we m ay have to disclose your data. Identifying information will not be attached to any of your individual responses or recordings when reporting results from the surveys. You will not be asked to give your name or any other information during the recording tha t will allow you or your place of employment to be identified. All results will be kept in a secure location accessible only to those involved in the study. The results of this study may be published or presented at professional meetings, but the identitie s of all research participants will remain anonymous. By participating, you agree to allow audio recordings of your speech. 8. YOUR RIGHTS TO PARTICIPATE, SAY NO, OR WITHDRAW Participation is voluntary. Refusal to participate will involve no penalty or lo ss of benefits to which you are otherwise entitled. You may discontinue participation at any time without penalty or loss of benefits to which you are otherwise entitled. You have the right to say no. You may change your mind at any time and withdraw. You may choose not to answer specific questions or to stop participating at any time. Whether you choose to participate or not will have no effect on your grade or evaluation. 9. COSTS AND COMPENSATION FOR BEING IN THE STUDY As an incentive to participate, s ubjects/students who participate in this research will be offered $15 per hour of participation (up to 2 hours or $30) or, if applicable, you can choose to earn extra credit through the MSU SONA software system. If participants are enrolled in a course tha t allows them to participate in a research study for credit, and the course accepts SONA credit, participants will have the option to receive MSU SONA credit instead of the cash remuneration. For those enrolled in such courses, students can also find alte rnative assignments to earn extra credit if they choose not to participate in this research study but wish to earn extra credit. The SONA system awards 1 credit per 1 hour of research participation with a bonus of 0.25 for participating in person (up to 2. 25 credits total). Within the SONA system, neither researchers nor individual instructors will know what studies participants are involved in. If your participation is over an hour, which it will likely be, we will compensate you in half hour increments (r ounding up) for up to two hours total. 181 10. THE RIGHT TO GET HELP IF INJURED In the unlikely event that you are injured as a result of participation in this project, Michigan State University will assist you in obtaining emergency care, if necessary, for your research related injuries. If you have insurance for medical care, your insurance carrier will be billed in the ordinary manner. As with any medical insurance, any costs that are not covered or in excess of what are paid by your insurance, including d policy is not to provide financial compensation for lost wages, disability, pain or discomfort, unless required by law to do so. This does not mean that you are giving up any legal rights you may have. Please contact Eric Hunter at 517 - 353 - 8641 with questions or to report an injury. 11. CONTACT INFORMATION If you have concerns or questions about this study, such as scientific issues, how to do any part of it, or to report an injury, please contact the researc her(s): Dr. Eric Hunter, Michigan State Univ, 113 Oyer, East Lansing, MI 48823, 517 - 353 - 8641, ejhunter@msu.edu Mark Berardi, Michigan State Univ, 110 Oyer , East Lansing, MI 48823, 517 - 353 - 8641, mberardi@msu.edu If you have questions or concerns about your role and rights as a research participant, would like to obtain information or offer input, or would like to regi ster a complaint about this study, you Protection Program at 517 - 355 - 2180, Fax 517 - 432 - 4503, or e - mail irb@msu.edu or regular mail at 4000 Collins Rd, Suite 136, Lansing, MI 48910. 12. DOCUMENTATION of Informed consent Your signature below means that you voluntarily agree to participate in this research study. ________________________________________ _____________________________ Signature Date You will be given a copy of this form to keep. At times, it is useful to use recordings in teaching, presenting research, or future analysis. Therefore, we would like to ask for special permission to use your recordings in those contexts. Your identificatio n would not be associated with the recording. If you do not give permission, it will not affect your ability to participate in the research. If you agree to allow your voice recordings (audio or video) in this way, please indicate: Yes No Initials______ ______ 182 BIBLIOGRAPHY 183 BIBL I O GRAPHY Altman, K. W., Atkinson, C., & Lazarus, C. (2005). Current and emerging concepts in muscle tension dysphonia: A 30 - month review. Journal of Voice . https://doi.org/10.1016/j.jvoice.2004.03.007 Bastian, R. W., Keidar, A., & Verdolini - Marston, K. (1990). Simp le vocal tasks for detecting vocal fold swelling. Journal of Voice . https://doi.org/10.1016/S0892 - 1997(05)80144 - 4 Bastian, R. W., & Thomas, J. P. (2016). Do Talkativeness and Vocal Loudness Correlate With Laryngeal Pathology? A Study of the Vocal Overdoer/ Underdoer Continuum. Journal of Voice , 30 (5), 557 562. https://doi.org/10.1016/j.jvoice.2015.06.012 Ben - David, B. M., & Icht, M. (2016). Voice Changes in Real Speaking Situations during a Day, with and Without Vocal Loading: Assessing Call Center Operators. Journal of Voice , 30 (2), 247e1 - 247e11. https://doi.org/10.1016/j.jvoice.2015.04.002 Berardi, M. L. , Hunter, E. J., & Leishman, T. W. (2015). Using the stability of vocal onsets to evaluate vocal effort in response to changing acoustical conditions. J Acoust Soc Am , 137 (4), 2433. Berardi, M. L., Whiting, J. K., Eyring, N. G., Rollins, M. K., Jensen, Z. R., Hunter, E. J., & Leishman, T. W. (2015). Acuity to noise and occupational voice risks in simulated acoustical environments. The Journal of the Acoustical Society of America , 137 (4), 2393. in scales. Deutsche Zeitschrift Fur Sportmedizin . Bottalico, P. (2017). Speech Adjustments for Room Acoustics and Their Effects on Vocal Effort. Journal of Voice . https://doi.org/10.1016/j.jvoice.2016.10.001 Bottalico, P., Cantor Cutiva, L. C., & Hunter, E . J. (2017). Vocal fatigue in virtual acoustics scenarios. The Journal of the Acoustical Society of America , 141 (5), 3541. Bottalico, P., Graetzer, S., & Hunter, E. J. (2016). Effects of speech style, room acoustics, and vocal fatigue on vocal effort. The Journal of the Acoustical Society of America , 139 (5), 2870 2879. https://doi.org/10.1121/1.4950812 Bottalico, P., Graetzer, S., Hunter, E. J., Bottalico, P., Graetzer, S., & Hunter, E. J. (2017). Effects of speech style , room acoustics , and vocal fatigue on vocal effort, 2870 (2016). Bottalico, P., Passione, I. I., Graetzer, S., & Hunter, E. J. (2017). Evaluation of the starting point of the lombard effect. Acta Acustica United with Acustica . 184 https://doi.org/10.3813/AAA.919043 Boucher, V. J. (2008). Acoustic correlates of fatigue in la ryngeal muscles: findings for a criterion - based prevention of acquired voice pathologies. Journal of Speech Language and Hearing Research . Boucher, V. J., Ahmarani, C., & Ayad, T. (2006). Physiologic features of vocal fatigue: Electromyographic spectral - co mpression in laryngeal muscles. Laryngoscope , 116 (6), 959 965. https://doi.org/10.1097/01.MLG.0000216824.07244.00 Boucher, V. J., & Ayad, T. (2010). Physiological Attributes of Vocal Fatigue and Their Acoustic Effects: A Synthesis of Findings for a Criteri on - Based Prevention of Acquired Voice Disorders. Journal of Voice , 24 (3), 324 336. https://doi.org/10.1016/j.jvoice.2008.10.001 Boyas, S., & Guével, A. (2011). Neuromuscular fatigue in healthy muscle: Underlying factors and adaptation mechanisms. Annals of Physical and Rehabilitation Medicine . https://doi.org/10.1016/j.rehab.2011.01.001 Buekers, R. (1998). Are voice endurance tests able to assess vocal fatigue? Clinical Otolaryngology and Allied Sciences . https://doi.org/10.1046/j.1365 - 2273.1998.2360533.x C alas, M., Verhulst, J., Lecoq, M., Dalleas, B., & Seilhean, M. (1989). Vocal pathology of teachers. Revue de Laryngologie - Otologie - Rhinologie . Cannito, M. P., Doiuchi, M., Murry, T., Woodson, G. E., & York, N. (2012). Perceptual Structure of Adductor S pasmodic Dysphonia and Its Acoustic Correlates. Journal of Voice , 26 (6), 818.e5 - 818.e13. https://doi.org/10.1016/j.jvoice.2012.05.005 Cantor - Cutiva, L. C., Banks, R., Berardi, M., Johnson, B., Clawson, R., Martinez, S., & Hunter, E. (2018). From vocal effo rt to vocal fatigue. What does the literature say? In 11th International Conference on Voice Physiology and Biomechanics (pp. 21 22). Cantor - Cutiva, L. C., Bottalico, P., & Hunter, E. (2018). Work - related communicative profile of radio broadcasters: a case study. Logopedics Phoniatrics Vocology , 0 (0), 1 14. https://doi.org/10.1080/14015439.2018.1504983 Cantor Cutiva, L. C., & Burdorf, A. (2015). Medical Costs and Productivity Costs Related to Voice Symptoms in Colombian Teachers. Journal of Voice . https://d oi.org/10.1016/j.jvoice.2015.01.005 Carroll, T., Nix, J., Hunter, E., Emerich, K., Titze, I., & Abaza, M. (2006). Objective measurement of vocal fatigue in classical singers: A vocal dosimetry pilot study. Otolaryngology - Head and Neck Surgery , 135 (4), 59 5 602. https://doi.org/10.1016/j.otohns.2006.06.1268 Chang, A., & Karnell, M. P. (2004). Perceived phonatory effort and phonation threshold pressure across a prolonged voice loading task: A study of vocal fatigue. Journal of Voice , 18 (4), 185 454 466. https:// doi.org/10.1016/j.jvoice.2004.01.004 Childers, D. G., Hicks, D. M., Moore, G. P., Eskenazi, L., & Lalwani, A. L. (1990). Electroglottography and Vocal Fold Physiology. Journal of Speech Language and Hearing Research . https://doi.org/10.1044/jshr.3302.245 C ho, S. W., Yin, C. S., Park, Y. B., & Park, Y. J. (2011). Differences in self - rated, perceived, and acoustic voice qualities between high - and low - fatigue groups. Journal of Voice , 25 (5), 544 552. https://doi.org/10.1016/j.jvoice.2010.07.006 Cipriano, M., Astolfi, A., & Pelegrín - García, D. (2017). Combined effect of noise and room acoustics on vocal effort in simulated classrooms. The Journal of the Acoustical Society of America . https://doi.org/10.1121/1.4973849 Claassen, H., & Werner, J. A. (1992). Fiber differentiation of the human laryngeal muscles using the inhibition reactivation myofibrillar ATPase technique. Anatomy and Embryology . https://doi.org/10.1007/BF00185983 Colton, R. H., Casper, J. K., & Leonard, R. (2011). Understanding voice problem: A physiological perspective for diagnosis and treatment: Fourth edition . Understanding Voice Problem: A Physiological Perspective for Diagnosis and Treatment: Fourth Edition . https://doi.org/10.1006/bbrc.2000.3788 Cooper, D. S., & Rice, D. H. (1990). Fatigue resistance of canine vocal fold muscle. Annals of Otology, Rhinology and Laryngology . Cooper, Donald S., & Titze, I. R. (1985). Generation and Dissipation of Heat in Vocal Fold Tissue. Journal of Speech Language and Hearing Research . https://doi.org/10.1044/jshr.2802.207 Cushing, I. R., Li, F. F., Cox, T. J., Worrall, K., & Jackson, T. (2011). Vocal effort levels in anechoic conditions. Applied Acoustics . https://doi.org/10.1016/j.apacoust.2011.02.011 Reggiani, C. (2002). Contractile properties and myosin heavy chain isoform composition in single fibre of human laryngeal muscles. Journal of Muscle Research & Cell Motili ty , 23 (3), 187 195. Lierde, K. (2016). Factors Involved in Vocal Fatigue: A Pilot Study. Damsté, P. H. (1970). The phonetogram. Practica Oto - Rhino - Laryngologica , 32 ( 3), 185 187. Davis, J. M. (1995). Central and peripheral factors in fatigue. Journal of Sports Sciences . https://doi.org/10.1080/02640419508732277 De Bodt, M. S., Wuyts, F. L., Van De Heyning, P. H., Lambrechts, L., & Abeele, D. Vanden. 186 (1998). Predicting vocal outcome by means of a vocal endurance test: A 5 - year follow - up study in female teachers. Laryngoscope . https://doi.org/10.1097/00005537 - 199809000 - 00020 Deliyski, D. D. (1993). Acoustic Model And Evaluation of Pathological Voice Production, (September ). Doellinger, M., & Berry, D. A. (2006). Visualization and Quantification of the Medial Surface Dynamics of an Excised Human Vocal Fold During Phonation. Journal of Voice , 20 (3), 401 413. https://doi.org/https://doi.org/10.1016/j.jvoice.2005.08.003 Doelli nger, M., Lohscheller, J., McWhorter, A., & Kunduk, M. (2009). Variability of Normal Vocal Fold Dynamics for Different Vocal Loading in One Healthy Subject Investigated by Phonovibrograms. Journal of Voice , 23 (2), 175 181. https://doi.org/10.1016/j.jvoice. 2007.09.008 dos Santos, A. P., Silverio, K. C. A., Dassie - Leite, A. P., Costa, C. da C., & Siqueira, L. T. D. (2018). Relation Between Musculoskeletal Pain and Voice Self - Assessment in Tele - Operators. Journal of Voice , 1 11. https://doi.org/10.1016/j.jvoic e.2018.07.006 Drechsel, J. S., & Thomson, S. L. (2008). Influence of supraglottal structures on the glottal jet exiting a two - layer synthetic, self - oscillating vocal fold model. The Journal of the Acoustical Society of America , 123 (6), 4434 4445. https://d oi.org/10.1121/1.2897040 Eadie, T. L., & Stepp, C. E. (2013). Acoustic Correlate of Vocal Effort in Spasmodic Dysphonia, 122 (3), 169 176. https://doi.org/10.1177/000348941312200305 Edstrom, L., Lindquist, C., & Martensson, A. (1974). Correlation between fu nctional and histochemical properties of the intrinsic laryngeal muscles in the cat. Ventilatory and Phonatory Control Systems , 392 404. Enflo, L., Sundberg, J., & McAllister, A. (2013). Collision and Phonation Threshold Pressures Before and After Loud, Pr olonged Vocalization in Trained and Untrained Voices. Journal of Voice , 27 (5), 527 530. https://doi.org/https://doi.org/10.1016/j.jvoice.2013.03.008 Erickson - Levendoski, E., & Sivasankar, M. (2011). Investigating the Effects of Caffeine on Phonation. Journ al of Voice , 25 (5), e215 e219. https://doi.org/https://doi.org/10.1016/j.jvoice.2011.02.009 Eustace, C. S., Stemple, J. C., & Lee, L. (1996). Objective measures of voice production in patients complaining of laryngeal fatigue. Journal of Voice , 10 (2), 146 154. https://doi.org/10.1016/S0892 - 1997(96)80041 - 5 Fabron, E. M. G., Regaçone, S. F., Marino, V. C. de C., Mastria, M. L., Motonaga, S. M., & Sebastião, L. T. (2015). Self - perception, complaints and vocal quality among undergraduate students enrolled i n a Pedagogy course. CoDAS , 27 (3), 285 291. https://doi.org/10.1590/2317 - 1782/20152014178 187 Faham, M., Jalilevand, N., Torabinezhad, F., Silverman, E. P., Ahmadi, A., Anaraki, Z. G., & Jafari, N. (2017). Relationship between Voice Complaints and Subjective a nd Objective Measures of Vocal Function in Iranian Female Teachers. Journal of Voice , 31 (4), 507.e1 - 507.e6. https://doi.org/10.1016/j.jvoice.2016.10.011 Fairbanks, F. (1960). The Rainbow Passage. In Voice and Articulation Drillbook . Fanchini, M., Ferraresi , I., Modena, R., Schena, F., Coutts, A. J., & Impellizzeri, F. M. (2015). Use of the CR100 Scale for Session Rating of Perceived Exertion in Soccer and Its Interchangeability with the CR10. International Journal of Sports Physiology and Performance . https ://doi.org/10.1123/ijspp.2015 - 0273 Fisher, K. V, & Swank, P. R. (1997). Estimating Phonation Threshold Pressure. Journal of Speech, Language, and Hearing Research , 40 (5), 1122 1129. Retrieved from http://dx.doi.org/10.1044/jslhr.4005.1122 Ford Baldner, E., Doll, E., & van Mersbergen, M. R. (2015). A Review of Measures of Vocal Effort With a Preliminary Study on the Establishment of a Vocal Effort Measure. Journal , 29 (5), 530 541. https://doi.org/10.1016/j. jvoice.2014.08.017 Fujiki, R. B., Chapleau, A., Sundarrajan, A., McKenna, V., & Sivasankar, M. P. (2017). The Interaction of Surface Hydration and Vocal Loading on Voice Measures. Journal of Voice , 31 (2), 211 217. https://doi.org/10.1016/j.jvoice.2016.07.0 05 Fujiki, R. B., & Sivasankar, M. P. (2017). A Review of Vocal Loading Tasks in the Voice Literature. Journal of Voice , 31 (3), 388.e33 - 388.e39. https://doi.org/10.1016/j.jvoice.2016.09.019 Gelfer, M. P., Andrews, M. L., & Schmidt, C. P. (1991). Effects of prolonged loud reading on selected measures of vocal function in trained and untrained singers. Journal of Voice . https://doi.org/10.1016/S0892 - 1997(05)80179 - 1 Gorham - Rowan, M., Berndt, A., Carter, M., & Morris, R. (2016). The effect of a vocal loading task on vocal function before and after 24 hours of thickened liquid use. J. Speech Pathol. Ther , 1 , 1 5. Gotaas, C., & Starr, C. D. (1993). Vocal Fatigue Among Teachers. Folia Phoniatrica et Logopaedica . https://doi.org/10.1159/000266237 Guzmán, M., Malebrán, M. C., Zavala, P., Saldívar, P., & Muñoz, D. (2013). Acoustic Changes of the Voice as Signs of Vocal Fatigue in Radio Broadcasters: Preliminary Findings. Acta Otorrinolaringologica (Engli sh Edition) . https://doi.org/10.1016/j.otoeng.2013.06.011 Halpern, A. E., Spielman, J. L., Hunter, E. J., & Titze, I. R. (2009). The Inability to Produce Soft Voice (IPSV): A Tool to Detect Vocal Changes in School Teachers. Logopedics Phoniatrics Vocology . https://doi.org/10.1080/14015430903062712 188 Hamdan, A L, Sibai, A., & Rameh, C. (2006). Effect of Fasting on Voice in Women [epub ahead of print] [Record Supplied By Publisher]. Journal of Voice . https://doi.org/10.1016/j.jvoice.2006.01.009 Hamdan, Abdul La tif, Ziade, G., Kasti, M., Akl, L., Bawab, I., & Kanj, N. (2017). Phonatory Symptoms and Acoustic Findings in Patients with Asthma: A Cross - Sectional Controlled Study. Indian Journal of Otolaryngology and Head and Neck Surgery . https://doi.org/10.1007/s120 70 - 016 - 1035 - 8 Hast, M. H. (1969). The primate larynx: A comparative physiological study of intrinsic muscles. Acta Oto - Laryngologica . https://doi.org/10.3109/00016486909124371 Hillenbrand, J., Getty, L. A., Clark, M. J., & Wheeler, K. (2005). Acoustic characteristics of American English vowels. The Journal of the Acoustical Society of America . https://doi.org/10.1121/1.411872 Hillman, R. E., Holmberg, E. V. A. B., & Perkell, J. S. (1989). OBJECTIVE ASSESSMENT OF INITIAL RESULTS, 32 (June), 373 392. phonation. Clinical Examination of Voice , 5 , 83 84. Hoh, J. F. Y. (2005). Laryngeal muscle fibre types. Acta Physiologica Scandinavica . https://doi.org/10.1111/j.1365 - 201X.2004.01402.x Holmberg, E. B., Hillman, R. E., Perkell, J. S., Guiod, P. C., & Goldman, S. L. (1995). Comparisons among aerodynamic, electroglottographic, and acoustic spectral measures of female voice. Journal of Speech, Language, and Hearing Research , 38 (6), 1212 1223. Holmberg, E., Ihre, E., & Södersten, M. (2007). Phonetograms as a tool in the voice clinic: Changes acro ss voice therapy for patients with vocal fatigue. Logopedics Phoniatrics Vocology , 32 (3), 113 127. https://doi.org/10.1080/14015430701305685 - Acceleration and Impact Stress as Possible Loading Factors in Phonation: A Computer Modeling Study. Folia Phoniatrica et Logopaedica , 61 (3), 137 145. https://doi.org/10.1159/000219949 Hunter, E. J., & Banks, R. E. (2017). Gender Differences in the Reporting of Vocal Fatigue in Te achers as Quantified by the Vocal Fatigue Index. Annals of Otology, Rhinology and Laryngology , 126 (12), 813 818. https://doi.org/10.1177/0003489417738788 Whiting, J. K. (2015). Teachers and teaching: speech production accommodations due to changes in the acoustic environment. Energy Procedia , 78 , 3102. 189 Hunter, E. J., Maxfield, L., & Graetzer, S. (2019). The Effect of Pulmonary Function on the Incidence of Vocal Fat igue Among Teachers. Journal of Voice . https://doi.org/10.1016/j.jvoice.2018.12.011 Hunter, E. J., Tanner, K., & Smith, M. E. (2011). Gender differences affecting vocal health of women in vocally demanding careers. Logopedics Phoniatrics Vocology . https:// doi.org/10.3109/14015439.2011.587447 Hunter, E. J., & Titze, I. R. (2009). Quantifying vocal fatigue recovery: Dynamic vocal recovery trajectories after a vocal loading exercise. Annals of Otology, Rhinology and Laryngology , 118 (6), 449 460. https://doi.or g/10.1177/000348940911800608 Ilomäki, I., Kankare, E., Tyrmi, J., Kleemola, L., & Geneid, A. (2017). Vocal Fatigue Symptoms and Laryngeal Status in Relation to Vocal Activity Limitation and Participation Restriction. Journal of Voice , 31 (2), 248.e7 - 248.e10 . https://doi.org/10.1016/j.jvoice.2016.07.025 ISO. (2002). ISO 9921:2002(E), Ergonomics - Assessment of Speech Communication. Geneva, Switzerland: ISO. J., H. E., Catherine, C. - Susanna, W. (2020). Toward a Consensus Description of Vocal Effort, Vocal Load, Vocal Loading, and Vocal Fatigue. Journal of Speech, Language, and Hearing Research , 63 (2), 509 532. https://doi.org/10.1044/2019_JSLHR - 19 - 00057 Jacobson, B. H., Johnson, A., Grywalski, C. , Silbergleit, A., Jacobson, G., Benninger, M. S., & Newman, C. W. (1997). The voice handicap index (VHI): development and validation. American Journal of Speech - Language Pathology , 6 (3), 66 70. Jessen, M., Köster, O., & Gfroerer, S. (2005). Influence of v ocal effort on average and variability of fundamental frequency. International Journal of Speech, Language and the Law . https://doi.org/10.1558/sll.2005.12.2.174 Jonsdottir, V., Laukkenen, A. - Working Day with and without Electric Sound Amplification. Folia Phoniatrica e Logopaedica , 601 , 282 287. https://doi.org/10.1159/000066149 Junqua, J. (1993). The Lombard reflex and its role on human listeners and automatic speech recognizers. The Journal of the Acoustical Society of America . https://doi.org/10.1121/1.405631 Kagan, L. S., & Heaton, J. T. (2017). The Effectiveness of Low - Level Light Therapy in Attenuating Vocal Fatigue. Journal of Voice , 31 (3), 384.e15 - 384.e23. https://doi.org/10.1016/j.jvoice.2016.09.004 Katch KI, K. V. M. W. (2009). Exercise Physiology: Nutrition, Energy, and Human Performance . 7th Edition . https://doi.org/10. 1161/01.STR.25.4.793 190 Kelchner, L., Toner, M. M., & Lee, L. (2006). Effects of prolonged loud reading on normal adolescent male voices. Language, Speech & Hearing Services in Schools . https://doi.org/http://dx.doi.org/10.1044/0161 - 1461(2006/012) Kempster, G . B., Gerratt, B. R., Abbott, K. V., Barkmeier - Kraemer, J., & Hillman, R. E. (2009). Consensus auditory - perceptual evaluation of voice: Development of a standardized clinical protocol. American Journal of Speech - Language Pathology . https://doi.org/10.1044/ 1058 - 0360(2008/08 - 0017) Kitch, J. A., & Oates, J. (1994). The perceptual features of vocal fatigue as self - reported by a group of actors and singers. Journal of Voice , 8 (3), 207 214. https://doi.org/10.1016/S0892 - 1997(05)80291 - 7 Kitch, J. A., Oates, J., & Greenwood, K. (1996). Performance effects on the voices of 10 choral Journal of Voice . https://doi.org/10.1016/S0892 - 1997(96)80002 - 6 Kostyk, B. E., & Rochet, A. P. (1998). Laryngeal airway resistance in teachers w ith vocal fatigue: A preliminary study. Journal of Voice , 12 (3), 287 299. https://doi.org/10.1016/S0892 - 1997(98)80019 - 2 Koufman, J. A., & Blalock, D. (1988). Vocal fatigue and dysphonia in the professional voice user: Bogart - Bacall Syndrome. Laryngoscope . https://doi.org/10.1288/00005537 - 198805000 - 00003 Kristiansen, J., Lund, S. P., Persson, R., Shibuya, H., Nielsen, P. M., & Scholz, M. (2014). A time during teaching, and the effects on vocal and mental fatigue development. International Archives of Occupational and Environmental Health , 87 (8), 851 860. https://doi.org/10.1007/s00420 - 014 - 0927 - 8 Lagier, A., Vaugoyeau, M., Ghio, A., Legou, T., Giovanni, A., & Assaiante, C . (2010). Coordination between Posture and Phonation in Vocal Effort Behavior, 195 202. https://doi.org/10.1159/000314264 Laukkanen, A. M., Ilomäki, I., Leppänen, K., & Vilkman, E. (2008). Acoustic Measures and Self - reports of Vocal Fatigue by Female Teach ers. Journal of Voice , 22 (3), 283 289. https://doi.org/10.1016/j.jvoice.2006.10.001 Laukkanen, A. M., & Kankare, E. (2006). Vocal loading - voices investigated before and after a working day. Folia Phoniatrica et Logopaedica , 58 (4), 229 239. https://doi.org/10.1159/000093180 Laukkanen, A. M., Mäki, E., & Leppänen, K. (2009). Electroglottogram - based estimation of - output - Folia Phoniatrica et Logopaedica , 61 (6), 316 322. https://doi.org/10.1159 /000252847 191 Laukkenen, A. - M., Jarvinen, K., Artkoski, M., Waaramaa - Maki - Kulmala, T., Kankare, E., - min Vocal Loading Test in Female Subjects with Vocal Training. Folia Pho niatrica e Logopaedica . Leanderson, R., & Sundberg, J. (1988). Breathing for singing. Journal of Voice . https://doi.org/10.1016/S0892 - 1997(88)80051 - 1 Lehto, L., Laaksonen, L., Vilkman, E., & Alku, P. (2006). Occupational voice complaints and objective acoustic measurements - Do they correlate? Logopedics Phoniatrics Vocology , 31 (4), 147 152. https://doi.org/10.1080/14015430600654654 Lehto, L., Laaksonen , L., Vilkman, E., & Alku, P. (2008). Changes in Objective Acoustic Measurements and Subjective Voice Complaints in Call Center Customer - Service Advisors During One Working Day. Journal of Voice , 22 (2), 164 177. https://doi.org/10.1016/j.jvoice.2006.08.010 Leong, K., Hawkshaw, M. J., Dentchev, D., Gupta, R., Lurie, D., & Sataloff, R. T. (2013). Reliability of objective voice measures of normal speaking voices. Journal of Voice , 27 (2), 170 176. https://doi.org/10.1016/j.jvoice.2012.07.005 Lien, Y. - A. S., Mic hener, C. M., Eadie, T. L., & Stepp, C. E. (2015). Individual Monitoring of Vocal Effort With Relative Fundamental Frequency: Relationships With Aerodynamics and Listener Perception. Journal of Speech, Language, and Hearing Research . https://doi.org/10.104 4/2015_jslhr - s - 14 - 0194 Liénard, J. - S., & Di Benedetto, M. - G. (1999). Effect of vocal effort on spectral properties of vowels. The Journal of the Acoustical Society of America , 106 (1), 411 422. Linville, S. E. (1995). Changes in glottal configuration in wom en after loud talking. Journal of Voice , 9 (1, Mar), 57 65. https://doi.org/https://doi.org/10.1016/S0892 - 1997(05)80223 - 1 Annales Des . https: //doi.org/10.1145/1168987.1169028 Lourenço, B. M., Costa, K. M., & Da Silva Filho, M. (2014). Voice disorder in cystic fibrosis patients. PLoS ONE . https://doi.org/10.1371/journal.pone.0096769 Lu, Y., & Cooke, M. (2009). Speech production modifications pro duced in the presence of low - pass and high - pass filtered noise. The Journal of the Acoustical Society of America . https://doi.org/10.1121/1.3179668 Maryn, Y., De Bodt, M., & Roy, N. (2010). The Acoustic Voice Quality Index: Toward improved treatment outcom es assessment in voice disorders. Journal of Communication Disorders . https://doi.org/10.1016/j.jcomdis.2009.12.004 Mascarello, F., & Veggetti, A. (1979). A comparative histochemical study of intrinsic laryngeal 192 muscles of ungulates and carnivores. Basic and Applied Histochemistry . Matsushita, H. (1975). The Vibratory Mode of the Vocal Folds in the Excised Larynx. Folia Phoniatrica et Logopaedica , 27 (1), 7 18. https://doi.org/10.1159/000263963 Maxfield, L. M., Hunter, E., & Greatzer, S. (2016). The effect of compromised pulmonary function on speech production among female school teachers. The Journal of the Acoustical So ciety of America , 139 (4), 2105. https://doi.org/10.1121/1.4950255 Mccabe, D. J., & Titze, I. R. (2002). Chant Therapy for Treating Vocal Fatigue Among Public School Teachers (2), 11 (November). McKenna, V. S., & Stepp, C. E. (2018). The relationship between acoustical and perceptual measures of vocal effort. The Journal of the Acoustical Society of America , 144 (3), 1643 1658. https://doi.org/10.1121/1.5055234 McKenzie, D. C. (2012). Respiratory physiology: Adaptations to high - level exercise. British Journal of Sports Medicine . https://doi.org/10.1136/bjsports - 2011 - 090824 Meynadier, Y., El Hajj, A., Pitermann, M., Legou, T., & Giovanni, A. (2018). Estimating Vocal Effort from the Aerodynamics of Labial Fricatives: A Feasibility Study. Journal of Voice . https://doi.org/10.1016/j.jvoice.2017.08.010 Mooshammer, C., & Mooshammer, C. (2010). Acoustic and laryngographic measures of the laryngeal reflexes of linguistic prominence and vocal effort in German Acoustic and laryngographic measures of the laryngeal r eflexes of linguistic prominence and vocal effort in German a ), 1047 . https://doi.org/10.1121/1.3277160 Munier, C., & Kinsella, R. (2008). The prevalence and impact of voice problems in primary school teachers. Occupational Medicine , 58 (1), 74 76. https:/ /doi.org/10.1093/occmed/kqm104 Naghibolhosseini, M., Deliyski, D. D., Zacharias, S. R. C., de Alarcon, A., & Orlikoff, R. F. (2018). Temporal Segmentation for Laryngeal High - Speed Videoendoscopy in Connected Speech. Journal of Voice , 32 (2), 256 -- e1. Nanjun deswaran, C., Jacobson, B. H., Gartner - Schmidt, J., & Verdolini Abbott, K. (2015). Vocal Fatigue Index (VFI): Development and Validation. Journal of Voice , 29 (4), 433 440. https://doi.org/10.1016/j.jvoice.2014.09.012 Nanjundeswaran, C., van Mersbergen, M., & Morgan, K. (2017). Restructuring the Vocal Fatigue Index Using Mokken Scaling: Insights Into the Complex Nature of Vocal Fatigue. Journal of Voice . https://doi.org/10.1016/j.jvoice.2017.09.008 Neils, L. R., & Yairi, E. (1987). Effects of speaking in noi se on vocal fatigue and vocal recovery. Folia Phoniatrica et Logopaedica . https://doi.org/10.1159/000265846 193 Niebudek - - Kowalska, M. (2007). Evaluation of voice acoustic parameters related to the vocal - loading test in profe ssionally active teachers with dysphonia. International Journal of Occupational Medicine and Environmental Health , 20 (1), 25 30. https://doi.org/10.2478/v10001 - 007 - 0001 - 9 sphonic and 29 (1), 1 9. https://doi.org/10.1590/2317 - 1782/20172016048 Pearl Solomon, N., & Stemmle DiMattia, M. (2000). Effects of a vocally fatiguing task and systemic hydration on phonation threshold pressure. Journal of Voice , 14 ( 3), 341 362. https://doi.org/10.1016/S0892 - 1997(00)80080 - 6 K. (2019). PsychoPy2: Experiments in behavior made easy. Behavior Research Methods . https://doi.org/1 0.3758/s13428 - 018 - 01193 - y Pelegrín - García, D., Smits, B., Brunskog, J., & Jeong, C. - H. (2011). Vocal effort with changing talker - to - listener distance in different acoustic environments. The Journal of the Acoustical Society of America , 129 (4), 1981 1990. P ellicani, A. D., Fontes, A. R., Santos, F. F., Pellicani, A. D., & Aguiar - Ricz, L. N. (2018). Fundamental Frequency and Formants Before and After Prolonged Voice Use in Teachers. Journal of Voice , 32 (2), 177 184. https://doi.org/10.1016/j.jvoice.2017.04.01 1 Pellicani, A. D., Ricz, H. M. A., & Ricz, L. N. A. (2015). Phonatory function after prolonged voice use in brazilian woman. CoDAS , 27 (4), 392 399. https://doi.org/10.1590/2317 - 1782/20152014201 Pette, D., & Staron, R. S. (1990). Cellular and Molecular Div ersities of Mammalian Skeletal Muscle Fibers. Rev. Physiol. Biochem. Pharmacol. https://doi.org/10.1007/s00198 - 012 - 2236 - y Pichora - Fuller, M. K., Kramer, S. E., Eckert, M. A., Edwards, B., Hornsby, B. W. Y., Humes, L. airment and cognitive energy: The framework for understanding effortful listening (FUEL). In Ear and Hearing . https://doi.org/10.1097/AUD.0000000000000312 Poburka, B. J., Patel, R. R., & Bless, D. M. (2017). Voice - Vibratory Assessment With Laryngeal Imagin g (VALI) form: reliability of rating stroboscopy and high - speed videoendoscopy. Journal of Voice , 31 (4), 513 -- e1. Potvin, J. R., & Fuglevand, A. J. (2017). A motor unit - based model of muscle fatigue. PLoS Computational Biology . https://doi.org/10.1371/journal.pcbi.1005581 Rantala, L., Paavola, L., Körkkö, P., & Vilkman, E. (1998). Working - day effects on the spectral characteristics of teaching voice. Folia Phoniatrica et Logopaedica , 50 (4), 205 211. https://doi.org/10.1159/000021 462 194 Remacle, A., Garnier, M., Gerber, S., David, C., & Petillon, C. (2018). Vocal Change Patterns During a Teaching Day: Inter - and Intra - subject Variability. Journal of Voice , 32 (1), 57 63. https://doi.org/10.1016/j.jvoice.2017.03.008 Rosen, C. A., & Simpson, C. B. (2008). Operative Techniques in Laryngology . Springer. Summerfield, Q. (1991). Audiovisual Investigation of the Loudness - Effort Effect fo r Speech and Nonspeech Events, 17 (4), 976 985. Rosenthal, A. L., Lowell, S. Y., & Colton, R. H. (2014). Aerodynamic and acoustic features of vocal effort. Journal of Voice . https://doi.org/10.1016/j.jvoice.2013.09.007 Roy, N., Merrill, R. M., Gray, S. D., & Smith, E. M. (2005). Voice Disorders in the General 1995. https://doi.org/10.1097/01.mlg.0000179174.32345.41 Roy, N., Smith, M. E., Allen, B., & Merrill, R. M. (2007). Add uctor Spasmodic Dysphonia Laryngeal Nerve Lidocaine Block, 116 (3), 161 168. Rubin, A. D., Jackson - Menaldi, C., Kopf, L. M., Marks, K., Skeffington, J., Skowronski, M. D., r, E. J. (2019). Comparison of Pitch Strength With Perceptual and Other Acoustic Metric Outcome Measures Following Medialization Laryngoplasty. Journal of Voice . https://doi.org/10.1016/j.jvoice.2018.03.019 Rubin, A. D., Praneetvatakul, V., Heman - Ackah, Y. , Moyer, C. A., Mandel, S., & Sataloff, R. T. (2005). Repetitive phonatory tasks for identifying vocal fold paresis. Journal of Voice , 19 (4), 679 686. https://doi.org/10.1016/j.jvoice.2004.11.001 Sahgal, V., & Hast, M. H. (1974). Histochemistry of primate laryngeal muscles. Acta Oto - Laryngologica . https://doi.org/10.3109/00016487409126356 Sampaio, C., & Jos, E. (2012). Vocal Effort and Voice Handicap Among Teachers, 26 (6), 15 18. https://doi.org/10.1016/j.jvoice.2012.06.003 Sandage, M. J., Connor, N. P., & Pascoe, D. D. (2013). Voice function differences following resting breathing versus submaximal exercise. Journal of Voice , 27 (5), 572 578. Sapienza, C. M., Crandell, C. C., & Curtis, B. (1999). Effects of sound - field frequency modulation amplification on r Journal of Voice , 13 (3), 375 381. https://doi.org/10.1016/S0892 - 1997(99)80042 - 3 Scherer, R. C., Titze, I. R., Raphael, B. N., Wood, R. P., Ramig, L. A., & Blager, R. F. (1987). Vocal fatigue in a tra ined and an untrained voice user. Laryngeal Function in Phonation and Respiration , 533 544. 195 Searl, J., & Knollhoff, S. (2018). Sense of Effort and Fatigue Associated With Talking After Total Laryngectomy. American Journal of Speech - Language Pathology . http s://doi.org/10.1044/2018_AJSLP - 17 - 0218 Shewmaker, M. B., Hapner, E. R., Gilman, M., Klein, A. M., & Johns, M. M. (2010). Analysis Journal of Voice , 24 (3), 308 313. https://doi.org/10.1 016/j.jvoice.2008.09.002 Shiotani, A., Westra, W. H., & Flint, P. W. (1999). Myosin heavy chain composition in human laryngeal muscles. Laryngoscope . https://doi.org/10.1097/00005537 - 199909000 - 00030 Shoffel - havakuk, H., Marks, K. L., Morton, M., Iii, M. M. J., & Hapner, E. R. (2019). Validation of the OMNI Vocal Effort Scale in the Treatment of Adductor Spasmodic Dysphonia, (February), 448 453. https://doi.org/10.1002/lary.27430 Sivasankar, M. (2002). Effects of vocal fatigue on voice parameters of Indian teachers. Indian Journal of Otolaryngology and Head and Neck Surgery , 54 (3), 245 247. https://doi.org/10.1007/BF02993116 Sivasankar, Mahalakshmi, Erickson, E., Schneider, S., & H awes, A. (2008). Phonatory Effects of Airway Dehydration: Preliminary Evidence for Impaired Compensation to Oral Breathing in Individuals With a History of Vocal Fatigue. Journal of Speech Language and Hearing Research . https://doi.org/10.1044/1092 - 4388(20 08/07 - 0181) Skinner, M. W., Holden, L. K., Holden, T. A., Demorest, M. E., Fourakis, M. S., Skinner, M. simulated soft , conversational , and raised - to - loud vocal efforts by adults with cochlear implants, 3766 . https://doi.org/10.1121/1.418383 Sluijter, A. M. C., & Van Heuven, V. J. (1996). Spectral balance as an acoustic correlate of linguistic stress. The Journal of the Acoustical Society of America , 100 (4), 2471 2485. Sm ith, E., Mendoza, M., Barkmeier, J., Lemke, J., Hoffman, H., Smith, E., & Al, E. T. (1998). - Related Functioning, 12 (2), 223 232. Smith, M. E., Roy, N., Wilson, C., & Hypothes is, O. (2006). Lidocaine Block of the Recurrent (April), 591 595. https://doi.org/10.1097/01.MLG.0000205588.04450.AC Smith, N. R., Rivera, L. A., Dietrich, M., Shyu, C. R., Pa ge, M. P., & DeSouza, G. N. (2016). Detection of Simulated Vocal Dysfunctions Using Complex sEMG Patterns. IEEE Journal of Biomedical and Health Informatics , 20 (3), 787 801. https://doi.org/10.1109/JBHI.2015.2490087 Södersten, M., Granqvist, S., Hammarberg , B., & Szabo, A. (2002). Vocal behavior and vocal loading factors for preschool teachers at work studied with binaural DAT recordings. 196 Journal of Voice , 16 (3), 356 371. https://doi.org/10.1016/S0892 - 1997(02)00107 - 8 Solomon, N. P. (2006). What is orofacial fatigue and how does it affect function for swallowing and speech? Seminars in Speech and Language . https://doi.org/10.1055/s - 2006 - 955117 Solomon, N. P. (2008). Vocal fatigue and its relation to vocal hyperfunction. International Journal of Speech - Languag e Pathology , 10 (4), 254 266. https://doi.org/10.1080/14417040701730990 Solomon, N. P., Glaze, L. E., Arnold, R. R., & van Mersbergen, M. (2003). Effects of a vocally Journal of Voice , 17 (1), 31 46. Spencer, M., Siegmund, T., & Mongeau, L. (2008). Determination of superior surface strains and stresses, and vocal fold contact pressure in a synthetic larynx model using digital image correlation. The Journal of the Acoustical Society of America , 123 (2), 1089 1103. https://doi.org/10.1121/1.2821412 Stemple, J. C., Stanley, J., & Lee, L. (1995). Objective measures of voice production in normal subjects following prolonged voice use. Journal of Voice . https://doi.org/10.1016/S0892 - 1997(05)80245 - 0 Stepp, C. E., Sawin, D. E., & Eadie, T. L. (2012). The Relationship Between Perceptio n of Vocal Effort and Relative Fundamental Frequency During Voicing Offset and Onset. Journal of Speech, Language, and Hearing Research . https://doi.org/10.1044/1092 - 4388(2012/11 - 0294) pressure levels of voiced speech from skin vibration of the neck. The Journal of the Acoustical Society of America . https://doi.org/10.1121/1.1850074 Tanner, K., Merrill, R. M., Houtz, D. R., Sauder, C., Elstad, M., & Wright - costa, J. (2010). Nebulized Iso tonic Saline Versus Water Following a Laryngeal Trained Sopranos, 53 (December). Tao, C., & Jiang, J. J. (2007). Mechanical stress during phonation in a self - oscillating finite - element vocal fold model. Journal of Biomechanics , 40 (10), 2191 2198. https://do i.org/10.1016/j.jbiomech.2006.10.030 Tepe, E. S., Deutsch, E. S., Sampson, Q., Lawless, S., Reilly, J. S., & Sataloff, R. T. (2002). A pilot survey of vocal health in young singers. Journal of Voice . https://doi.org/10.1016/S0892 - 1997(02)00093 - 0 Ternström, S., Bohman, M., & Södersten, M. (2006). Loud speech over noise: Some spectral attributes, with gender differences. The Journal of the Acoustical Society of America , 119 (3), 1648 1665. Thomas - kersting 197 PERCEPTUAL RATINGS AND SPECTRAL NOISE LEVELS OF HEARING - IMPAIRED CHILDREN, 22 , 125 135. Timmermans, B., De Bodt, M., Wuyts, F., & Van de Heyning, P. (2003). Vocal hygiene in radio students and in ra dio professionals. Logopedics Phoniatrics Vocology , 28 (3), 127 132. https://doi.org/10.1080/14015430310018333 Titze, I., Lemke, J., & Montequin, D. (1997). Populations in the U.S. workforce who rely on voice as a primary tool of trade: A preliminary report . Journal of Voice . https://doi.org/10.1016/S0892 - 1997(97)80002 - 1 Titze, I. R. (1992). Acoustic interpretation of the voice range profile (phonetogram). Journal of Speech, Language, and Hearing Research , 35 (1), 21 34. Titze, I. R. (1994). Mechanical stress in phonation. Journal of Voice . https://doi.org/10.1016/S0892 - 1997(05)80302 - 9 Titze, I. R. (1999). Toward occupational safety criteria for vocalization. Logopedics Phoniatrics Vocology . https://doi.org/10.1080/140154399435110 Titze, I. R., Hunter, E. J., vocalizations of teachers. The Journal of the Acoustical Society of America . https://doi.org/10.1121/1.2390676 Traunmüller, H., & Eriksson, A. (2000). Acoustic effects of variation in vo cal effort by men, women, and children. The Journal of the Acoustical Society of America . https://doi.org/10.1121/1.429414 Traunmüller, H., & Eriksson, A. (2011). Acoustic effects of variation in vocal effort by men , women , and children Acoustic effects of variation in vocal effort by men , women , and children, 3438 (2000). https://doi.org/10.1121/1.429414 Kavukcuoglu, K. (2016). WaveNet: A generative model for raw audio. SS W , 125 . van Leer, E., & van Mersbergen, M. (2017). Using the Borg CR10 Physical Exertion Scale to Measure Patient - perceived Vocal Effort Pre and Post Treatment. Journal of the Voice Foundation , 31 (3), 389.e19 - 389.e25. https://do i.org/10.1016/j.jvoice.2016.09.023 Verdolini, K., & Ramig, L. O. (2001). Review: Occupational risks for voice problems. Logopedics Phoniatrics Vocology . https://doi.org/10.1080/14015430119969 Verdolini, Katherine, Chan, R., Titze, I. R., Hess, M., & Bierha ls, W. (1998). Correspondence of electroglottographic closed quotient to vocal fold impact stress in excised canine larynges. Journal of Voice , 12 (4), 415 423. https://doi.org/https://doi.org/10.1016/S0892 - 1997(98)80050 - 7 198 Verdolini, Katherine, Titze, I. R. , & Fennell, A. (1994). Dependence of phonatory effort on hydration level. Journal of Speech, Language, and Hearing Research , 37 (5), 1001 1007. Verstraete, J., Forrez, G., Mertens, P., & Debruyne, F. (1993). The effect of sustained phonation at high and lo w pitch on vocal jitter and shimmer. Folia Phoniatrica et Logopaedica . https://doi.org/10.1159/000266266 Vilkman, E. (2004). Occupational safety and health aspects of voice and speech professions... 28th World Congress of the International Association of L ogopedics and Phoniatrics, Brisbane, Australia, 29th August to 2nd September 2004. Folia Phoniatrica et Logopaedica . Vilkman, Erkki, Lauri, E. - R., Alku, P., Sala, E., & Sihvo, M. (1999). Effects of prolonged oral reading on F0, SPL, subglottal pressure and amplitude characteristics of glottal flow waveforms. Journal of Voice , 13 (2), 303 312. https://doi.org/https://doi.o rg/10.1016/S0892 - 1997(99)80036 - 8 Vintturi, J., Alku, P., Lauri, E. R., Sala, E., Sihvo, M., & Vilkman, E. (2001). The effects of post - loading rest on acoustic parameters with special reference to gender and ergonomic factors. Folia Phoniatrica et Logopaedica . https://doi.org/10.1159/00005 2687 Vogel, A. P., Fletcher, J., Snyder, P. J., Fredrickson, A., & Maruff, P. (2011). Reliability, stability, and sensitivity to change and impairment in acoustic measures of timing and frequency. Journal of Voice . https://doi.org/10.1016/j.jvoice.2009.09. 003 Warrick, P., Dromey, C., Irish, J. C., Durkin, L., Pakiam, A., & Lang, A. (2000). Botulinum Crossover Design Study of Unilateral Versus Bilateral Injection, (August), 1366 1374. Weinberger, S. H., & Kunath, S. A. (2011). The Speech Accent Archive: Towards a typology of English accents. In Language and Computers . https://doi.org/10.1163/9789401206884_014 Welham, N. V., & Maclagan, M. A. (2003). Vocal fatigue: Current kno wledge and future directions. Journal of Voice , 17 (1), 21 30. https://doi.org/10.1016/S0892 - 1997(03)00033 - X Wenke, R. J., Goozee, J. V, Murdoch, B. E., & LaPointe, L. L. (2006). Dynamic assessment of articulation during lingual fatigue in myasthenia gravis . Journal of Medical Speech - Language Pathology , 14 (1), 13 32. Whiting, J. K., Leishman, T. W., Eyring, N. G., Berardi, M. L., & Rollins, M. K. (2015). Evaluation of a real - time convolution system for perception of self - generated speech in simulated rooms. The Journal of the Acoustical Society of America , 138 (3), 1900. Whitling, S., Lyberg - Åhlander, V., & Rydell, R. (2017). Recovery From Heavy Vocal Loading in Women With Different Degrees of Functional Voice Problems. Journal of Voice , 31 (5), 645.e1 - 645.e14. https://doi.org/10.1016/j.jvoice.2016.12.012 199 Whitling, S., Rydell, R., & Lyberg Åhlander, V. (2015). Design of a clinical vocal loading test with long - time measurement of voice. Journal of Voice . https://doi.org/10.1016/j.jvoice.2014.07.012 Willis, J., Mi chael, D. D., Boyer, H., & Misono, S. (2015). Prevalence and severity of dysphonia in patients with cystic fibrosis: A pilot study. Otolaryngology - Head and Neck Surgery (United States) . https://doi.org/10.1177/0194599815581832 Wingate, J. M., Brown, W. S ., Shrivastav, R., Davenport, P., & Sapienza, C. M. (2007). Treatment Outcomes for Professional Voice Users. Journal of Voice , 21 (4), 433 449. https://doi.org/10.1016/j.jvoice.2006.01.001 Wolfe, V. I., Long, J., Youngblood, H. C., Williford, H., & Olson, M . S. (2002). Vocal parameters of aerobic instructors with and without voice problems. Journal of Voice , 16 (1), 52 60. https://doi.org/10.1016/S0892 - 1997(02)00072 - 3 Wolfe, V., & Martin, D. (1997). Acoustic correlates of dysphonia: Type and severity. Journal of Communication Disorders . https://doi.org/10.1016/S0021 - 9924(96)00112 - 8 Wu, Y. Z., Crumley, R. L., Armstrong, W. B., & Caiozzo, V. J. (2000). New perspectives about human laryngeal muscle: Single - fiber analyses and interspecies comparisons. Archives of Otolaryngology - Head and Neck Surgery . https://doi.org/10.1001/archotol.126.7.857 Xue, C., Kang, J., Hedberg, C., Zhang, Y., & Jiang, J. J. (2019). Dynamically Monitoring Vocal Fatigue and Recovery Using Aerodynamic, Acoustic, and Subjective Self - Rating M easurements. Journal of Voice , 33 (5), 809.e11 - 809.e18. https://doi.org/https://doi.org/10.1016/j.jvoice.2018.03.014 Yiu, E. M. L., & Chan, R. M. M. (2003). Effect of hydration and vocal rest on the vocal fatigue in amateur karaoke singers. Journal of Voice . Journal of Voice. https://doi.org/10.1016/S0892 - 1997(03)00038 - 9 Yiu, E. M. L., Wang, G., Lo, A. C. Y., Chan, K. M. K., Ma, E. P. M., Kong, J., & Barrett, E. A. (2013). Quantitative high - speed laryngoscopic analysis of vocal fold vibration in fatigued voi ce of young karaoke singers. Journal of Voice . https://doi.org/10.1016/j.jvoice.2013.06.010 Zacharias, S. R. C., Deliyski, D. D., & Gerlach, T. T. (2018). Utility of laryngeal high - speed videoendoscopy in clinical voice assessment. Journal of Voice , 32 (2), 216 220. Zealer, D. L. (1983). The contractile properties and functions of single motor uni ts of thyroartenoid muscle in the cat. Vocal Fold Physiology: Contemporary Research and Clinical Issues. College Hill Press, San Diego . Zelinka, P., Sigmund, M., & Schimmel, J. (2012). Impact of vocal effort variability on automatic speech recognition. Spe ech Communication , 54 (6), 732 742. https://doi.org/10.1016/j.specom.2012.01.002