THE EFFECT OF VOCAL FRY ON SPEECH INTELLIGIBILITY By Kaleigh Susan Cammenga A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Communicative Sciences and Disorders – Master of Arts 2018 ABSTRACT THE EFFECT OF VOCAL FRY ON SPEECH INTELLIGIBILITY By Kaleigh Susan Cammenga Speech intelligibility is a measurement of the interaction between three components: the speech signal, the transmission channel, and the listener. Anything that interferes with any of these components can affect intelligibility. Vocal fry, though sometimes associated with vocal pathology, is commonly used in non-disordered speech. Speech produced with vocal fry differs from typical (modal) voicing in terms of pitch, volume, and quality. These differences may negatively impact intelligibility. Currently, no direct evidence exists regarding the impact of vocal fry on speech intelligibility. The purpose of the current study was to answer the research question: Does vocal fry affect the intelligibility of spoken words? We hypothesized that single words produced with vocal fry would be less intelligible than single words spoken in modal voice due to the acoustic characteristics and perceptions of vocal fry. To test this hypothesis, words spoken in both vocal fry and modal voice were collected and compiled to produce a standard intelligibility test procedure. Data from 26 listeners who completed the intelligibility test were analyzed in terms of intelligibility score and listening difficulty rating. The mean intelligibility score as a percentage of words correctly identified was 62.08% for words spoken with fry and 64.56% for words spoken with no fry. This difference was statistically significant. The mean listening difficulty rating was 4.98 for words spoken with fry and 4.56 for words spoken with no fry. This difference was also statistically significant. These results suggest that vocal fry does negatively impact speech intelligibility at the single word level. Decreased speech intelligibility may have numerous possible linguistic, social, and economic implications. Copyright by KALEIGH SUSAN CAMMENGA 2018 This thesis is dedicated to my family and my fiancé. Thank you for your constant love, prayers, encouragement, and support. iv ACKNOWLEDGEMENTS A special thank you to Dr. Eric Hunter for serving as my thesis advisor. Thank you for sharing your knowledge and passion for voice research with me and for inspiring my newfound love for research. It has been an incredible experience learning under you. An additional thank you to the members of my committee, Dr. Peter Lapine and Dr. Amanda Hampton-Wray, for sharing your wisdom and perspectives with me. Thank you to Lady Catherine Cantor Cutiva and Pasquale Bottalico for your expertise and assistance in the development and execution of my thesis. Thank you to Mark Berardi and Callan Gavigan for assisting in the recording of the audio samples. Thank you to Kirsten Rockey and Russel Banks for your assistance in data collection. And finally, a special thank you to Jacob Morrison. Thank you for loving and encouraging me and for taking interest in my research, serving as my sounding board, and offering your experience and knowledge to assist me along the way. I love you. v TABLE OF CONTENTS LIST OF TABLES ........................................................................................................................ vii LIST OF FIGURES ....................................................................................................................... ix INTRODUCTION ...........................................................................................................................1 METHODS ......................................................................................................................................9 PART I: Recording and Preparation of Samples .......................................................................9 PART II: Intelligibility Listening Task....................................................................................10 PART III: Analysis Techniques ...............................................................................................14 RESULTS ......................................................................................................................................15 Participants ...............................................................................................................................15 Intelligibility Score ..................................................................................................................15 Listening Difficulty .................................................................................................................18 Intelligibility Score and Listening Difficulty: Worst Noise Conditions Removed .................21 Comparison of Male and Female Listeners ............................................................................24 DISCUSSION ................................................................................................................................28 CONCLUSION ..............................................................................................................................33 APPENDICES ...............................................................................................................................34 APPENDIX A: Words Recorded by the Male and Female Speakers ......................................35 APPENDIX B: Description of the Normalization Process ......................................................38 APPENDIX C: Lists of Words Used in Each Test Set ............................................................41 APPENDIX D: Order of Test Sets for Each Version of the Intelligibility Listening Task .....44 APPENDIX E: Image of IRB Approval Letter........................................................................47 APPENDIX F: Image of Consent Form ..................................................................................49 APPENDIX G: Test Set Response Form Example..................................................................53 APPENDIX H: Participant Intelligibility Score Data for All Test Conditions .......................56 APPENDIX I: Participant Listening Difficulty Data for All Test Conditions ........................61 REFERENCES ..............................................................................................................................66 vi LIST OF TABLES Table 1: Combination of conditions for each of the 24 test sets ....................................................11 Table 2: Comparison of mean intelligibility scores between the Fry and No Fry conditions .......16 Table 3: Comparison of mean intelligibility scores between the Fry and No Fry conditions at each level of signal-to-noise ratio ..................................................................................................16 Table 4: Comparison of mean listening difficulty ratings between the Fry and No Fry conditions .......................................................................................................................................19 Table 5: Comparison of mean listening difficulty ratings between the Fry and No Fry conditions at each level of signal-to-noise ratio ..............................................................................................19 Table 6: Comparison of mean intelligibility scores between Fry and No Fry excluding the 6 dB and 12 dB noise conditions ............................................................................................................22 Table 7: Comparison of mean listening difficulty ratings between Fry and No Fry excluding the 6 dB and 12 dB noise conditions ...................................................................................................23 Table 8: Comparison of mean intelligibility scores of the Fry subtests between the Male and Female listeners .............................................................................................................................25 Table 9: Comparison of mean listening difficulty ratings of the Fry subtests between the Male and Female listeners .......................................................................................................................25 Table 10: Words recorded by the Male speaker ............................................................................36 Table 11: Words recorded by the Female speaker .........................................................................37 Table 12: List of words used in each combination of conditions ..................................................42 Table 13: Order of test sets as presented in Version Ia and Version Ib of the intelligibility listening task. Each test set is identified by the sex of the speaker (Male or Female), voicing type (Fry or No Fry), level of signal-to-noise ratio (NaN, -12 dB, -6 dB, 0 dB, 6 dB, or 12 dB), and word list (A or B) that was presented in that test set .....................................................................45 Table 14: Order of test sets as presented in Version IIa and Version IIb of the intelligibility listening task. Each test set is identified by the sex of the speaker (Male or Female), voicing type (Fry or No Fry), level of signal-to-noise ratio (NaN, -12 dB, -6 dB, 0 dB, 6 dB, or 12 dB), and word list (A or B) that was presented in that test set .....................................................................46 vii Table 15: Intelligibility score as a percentage for each signal-to-noise ratio under the Male Fry condition for each subject ..............................................................................................................57 Table 16: Intelligibility score as a percentage for each signal-to-noise ratio under the Male No Fry condition for each subject ........................................................................................................58 Table 17: Intelligibility score as a percentage for each signal-to-noise ratio under the Female Fry condition for each subject ..............................................................................................................59 Table 18: Intelligibility score as a percentage for each signal-to-noise ratio under the Female No Fry condition for each subject ........................................................................................................60 Table 19: Mean listening difficulty rating for each signal-to-noise ratio under the Male Fry condition for each subject ..............................................................................................................62 Table 20: Mean listening difficulty rating for each signal-to-noise ratio under the Male No Fry condition for each subject ..............................................................................................................63 Table 21: Mean listening difficulty rating for each signal-to-noise ratio under the Female Fry condition for each subject ..............................................................................................................64 Table 22: Mean listening difficulty rating for each signal-to-noise ratio under the Female No Fry condition for each subject ..............................................................................................................65 viii LIST OF FIGURES Figure 1: Comparison of mean intelligibility scores between the Fry and No Fry conditions by speaker sex .....................................................................................................................................17 Figure 2: Comparison of mean intelligibility scores between the Fry and No Fry conditions by signal-to-noise ratio .......................................................................................................................18 Figure 3: Comparison of mean listening difficulty between the Fry and No Fry conditions by speaker sex .....................................................................................................................................20 Figure 4: Comparison of mean listening difficulty between the Fry and No Fry conditions by signal-to-noise ratio .......................................................................................................................21 Figure 5: Comparison of adjusted mean intelligibility scores between the Fry and No Fry conditions by speaker sex ..............................................................................................................22 Figure 6: Comparison of adjusted mean listening difficulty between the Fry and No Fry conditions by speaker sex ..............................................................................................................24 Figure 7: Comparison of mean intelligibility scores between the Male and Female listeners by speaker sex .....................................................................................................................................26 Figure 8: Comparison of mean listening difficulty rating between the Male and Female listeners by speaker sex ................................................................................................................................27 Figure 9: Normalization process example .....................................................................................39 Figure 10: Image of IRB approval letter ........................................................................................48 Figure 11: Image of consent form ..................................................................................................50 ix INTRODUCTION The term vocal register refers to the “perceptually distinct regions of vocal quality that can be maintained over some ranges of pitch and loudness” (Titze, 2000, pp. 282). Three vocal registers are typically most recognizable: modal, falsetto, and vocal fry. These registers differ based on the vibratory patterns, degree of tension, compression, and adductive force of the vocal folds. Modal phonation is used most frequently in conversational speech. It typically ranges in fundamental frequency from 150 to 500 Hz in adult women and from 80 to 450 Hz in adult men. Vocal fold vibration using modal voice is highly periodic, and it is characterized by moderate tension, compression, and adductive force. Falsetto typically occurs at frequencies above modal phonation and is most commonly used in performance voice rather than normal speech. Falsetto voice production is characterized by moderately high tension and compression and high adductive force of the vocal folds. The vocal folds are typically long, thin, stiff, and often bow-shaped during falsetto voicing (Boone, McFarlane, Von Berg, & Zraick, 2014). The bursts of acoustic energy that occur following each closure of the vocal folds are not individually perceived in the modal and falsetto registers. Therefore, voice produced in these registers is perceived as continuous sound. However, when the fundamental frequency drops below approximately 70 Hz, referred to as the crossover frequency, the individual bursts of acoustic energy and subsequent periods of silence that are produced by each cycle of opening and closing of the vocal folds can begin to be perceived by the auditory system. This cycle of pulses and gaps is what is perceived as vocal fry (Titze, 2000). Michel (1968) determined that vocal fry occurred at fundamental frequencies significantly lower than both modal register and harsh voice, providing further evidence that vocal fry exists as a distinct phonatory register. Previous authors have reported a mean range of 24-52 Hz (mean of 36.4 Hz) for males speaking 1 in vocal fry and a mean range of 18-46 Hz for females speaking in vocal fry (Hollien & Michel, 1968; Michel, 1968). Blomgren, Chen, Ng, and Gilbert (1998) compared the aerodynamic, physiologic, and perceptual characteristics of speech produced in the modal and vocal fry registers in 10 female and 10 male normal speakers. The results indicated that speech in vocal fry was produced at significantly lower fundamental frequencies than speech produced in the modal register for both males and females. It was also found that measurements of jitter and shimmer were significantly higher in vocal fry compared to modal. Signal to noise ratio was found to be significantly lower in vocal fry compared to modal. Subglottic air pressure was approximately 1.5-3 times higher in modal voice compared to vocal fry. From the physiological point of view, vocal fry is associated with short, thick vocal folds with a lax covering and mild adductive force. Specifically, the thyroarytenoid muscles and vocal ligaments are shortened, which reduces the rate of vocal fold vibration (Boone, McFarlane, Von Berg, & Zraick, 2014). In order to maintain the pattern of vocal fold vibration associated with vocal fry, a minimum subglottal pressure of approximately 2 cm H2O is needed. The average airflow rate associated with vocal fry ranges from 12 to 20 cc/s (Boone, McFarlane, Von Berg, & Zraick, 2014). Because vocal fry occurs with less subglottal air pressure, it more frequently occurs at the end of sentences or phrases when air pressure naturally drops off. Because of this, some researchers have considered vocal fry to be a linguistic marker of paragraph and sentence boundaries (Wolk, Abdelli-Beruh, & Slavin, 2012); the contextual aspects of vocal fry will be discussed below. Whitehead, Metz, and Whitehead (1984) investigated the vibratory patterns of the vocal folds during production of vocal fry. The researchers assessed the vibratory patterns of the vocal 2 folds by using high speech laryngeal films during phonation of the English schwa vowel in pulse register by a female speaker. The speaker produced the vocal fry pattern with both a single and double opening and closing pattern. Both patterns produced the perceptual and acoustic characteristics associated with vocal fry (low pitched, popping quality; fundamental frequency within range typically associated with vocal fry). This suggests that vocal fry can be produced with either a single or multiple opening and closing pattern per cycle; however, the crackling sound that typically distinguishes vocal fry is often thought to be associated with this syncopated pattern and secondary beat (Boone, McFarlane, Von Berg, & Zraick, 2014). Vocal fry was traditionally classified as an abnormal phonation pattern associated with vocal pathology or was thought to be vocally harmful at the least. Because vocal fry is produced with decreased subglottic air pressure, the resulting sound is typically at a lower sound pressure level (dB). Therefore, in order to compensate for this decrease in volume, speakers using vocal fry will typically increase vocal fold tension in order to increase loudness. Because of this increased tension, consistent production of fry has been shown to be harmful to the vocal fold mucosa (Boone, McFarlane, Von Berg, & Zraick, 2014) and could be damaging to the voice (Colton, Casper, & Leonard, 2006). Although once classified as abnormal and potentially pathological, such classification is now considered inaccurate (Hollien, Moore, Wendahl, & Michel, 1966) as multiple studies have shown that vocal fry is commonly used in non-disordered speech and may be associated with the biological sex of the individual. For example, Wolk, Abdelli-Beruh, and Slavin (2012) examined the prevalence of vocal fry use in young adult female native Standard American- English speakers where they found that a majority of participants used vocal fry during a sentence reading task. Abdelli-Beruh, Wolk, and Slavin (2014) examined the prevalence of vocal 3 fry in young adult males and compared it to the results of Wolk, Abdelli-Beruh, and Slavin (2012). The results indicated that the prevalence of vocal fry in sentence reading was significantly lower for male speakers compared to female speakers. However, Abdelli-Beruh, Drugman, and Red Owl (2016) examined the speech of 25 male and 29 female subjects. Using an algorithm to detect the “creak segments” associated with vocal fry, they found that the frequency with which vocal fry is used is not significantly related to the individual’s biological sex. This contradicts previous research suggesting that vocal fry is used more commonly by females. The prevalence of vocal fry use is thought to be increasing during recent years. Borrie and Delfino (2017) examined a possible explanation for this increased used. They studied the potential role of conversational entrainment in the use of vocal fry during conversation. Conversational entrainment describes the tendency of communication partners to modify their own behaviors to match up with those of the other person during a conversation. The results indicated that speakers used vocal fry significantly more frequently when conversing with a person who used a substantial amount of vocal fry than when conversing with a person who used less vocal fry. In addition, greater similarity between partners in terms of their use of vocal fry was associated with greater communicative efficiency and enjoyment, which may act as positive reinforcement towards the use of vocal fry in conversation. Not only is vocal fry frequently used in non-disordered speech, a certain style of vocal fry production has been used as a vocal training technique and has even been utilized in vocal therapy. Because this type of vocal fry requires a relaxed larynx, it can be a useful technique for decreasing tension in the larynx allowing speech to be produced with considerably less tension in the vocal folds. Therefore, it has been used in the treatment of hyperfunctional problems in the 4 cases of vocal nodules, functional dysphonia, and spasmodic dysphonia (Boone, McFarlane, Von Berg, & Zraick, 2014). Vocal fry has also been shown to be useful in the treatment of both puberphonia and ventricular phonation because it requires the client to relax and lower the larynx (Boone, McFarlane, Von Berg, & Zraick, 2014). Vocal fry also seems to serve a linguistic role. For example, vocal fry is most often exhibited at the end of phrases when subglottal air pressure naturally drops. Abdelli-Beruh, Drugman, and Red Owl (2016) found that vocal fry occurred with the greatest frequency at the end of sentences, regardless of the length of the sentence. This suggests that vocal fry may act as a syntactic marker signaling paragraph and sentence boundaries (Wolk, Abdelli-Beruh, & Slavin, 2012). Lee (2015) submits that vocal fry may be used to mark parenthetical segments, such as elaborating on previous expressions or inserting background information into a narrative structure. Thus, vocal fry is a normal register that can be used selectively by speakers who do not have vocal pathologies. Its uses may include organizing information into different levels of importance, signaling meaning, or indicating a level of distance or detachment. The use of fry in daily communication may come with a social cost. For example, vocal fry has been associated with negative connotations in relation to success in the labor market (Anderson, Klofstad, Mayew, & Venkatachalam, 2014). However, little is known about why the use of vocal fry, with its unique vibratory pattern and sound quality, is identifiable by a listener in a way that would lead to social implications. Since vocal fry is associated with perceptual differences in pitch, volume, and quality, the use of vocal fry may affect the intelligibility of speech. This may be perceived by the listener in terms of increased or decreased listening effort which may then affect the listener’s perception of the speaker. To explore this possibility, the relationship between production of vocal fry and speech intelligibility should be examined. 5 Speech intelligibility has been defined as the degree to which the intended message of the speaker matches the response of the listener across various transmission systems (Schiavetti, 1992). Thus speech intelligibility is a measurement of the interaction between three components: the speaker, the transmission channel, and the listener. Anything that interferes with any of these components can impact the intelligibility of the communication process. Thus, because the production of vocal fry alters the speech signal produced by the speaker, intelligibility of speech might also be altered. Schiavetti (1992) identified two types of tasks to measure intelligibility: word identification and scaling procedures. Word identification provides a frequency count of the number of correctly identified words which is typically converted into a percentage. This is the most widely used measure in the assessment of speech intelligibility in the speech sciences (Munro, 2013). Scaling procedures involve the listener making a judgment about the talker’s intelligibility by identifying their perception of the intelligibility on a rating scale. This provides a direct measurement of the degree to which aspects of the speech signal may impact listening. For the purpose of this study, we will be assessing intelligibility in terms of both percent words correct and listening effort. Currently, little direct evidence exists regarding the effects of vocal fry on speech intelligibility. However, previous research has examined the effect of various acoustical characteristics of speech and voice on speech intelligibility that may provide insight into the possible relationship between vocal fry and intelligibility. Ryalls and Lieberman (1982) performed a study in which the subjects were asked to identify vowels differing in fundamental frequency. They found that subjects identified vowels produced with a low or average F0 significantly more accurately than vowels produced with a high F0. However, there was no 6 significant difference between vowels produced with an average F0 compared to a low F0 (Ryalls & Lieberman, 1982). Vocal fry is produced with a low fundamental frequency. This suggests that perhaps intelligibility may not differ between vocal fry and modal voice, at least not as a result of a decreased F0. Bradlow, Torretta, and Pisoni (1996) evaluated the difference in intelligibility between sentences spoken by 20 speakers in terms of various acoustic characteristics. Their findings relevant to the current study showed that mean F0 was not correlated with intelligibility. However, range of F0 was positively correlated with intelligibility so that a wider range of F0 was associated with higher overall intelligibility. The findings of Watson and Schlauch (2008) were consistent with these results. They compared the intelligibility of resynthesized naturally produced sentences in terms of percent words correctly identified between the unmodified signals and low-flattened and median-flattened signals. Sentences were acoustically modified to have flattened intonation at either the median F0 of the original signal or at the average low F0 of the original signal. They found that the unmodified signal was more intelligible than the low-flattened signal and the low-flattened signal was more intelligible than the median-flattened. However, the differences between the low-flattened and median-flattened signals, though statistically significant, were small. This suggests that the specific value of F0 contributes relatively little to the intelligibility of speech. Lack of variation, however, appears to offer a greater contribution to intelligibility differences (Watson & Schlauch, 2008). This may suggest that low F0 associated with vocal fry may not greatly impact intelligibility. However, according to Colton, Casper, & Leonard (2006), lack of flexibility during the production of vocal fry typically results in a monotone voice. This lack of variation in F0 may negatively impact intelligibility. 7 Ramig (1992) suggested that decreased intelligibility of disordered phonation may be less related to the specific characteristics of the voice and more related to its perception. Disordered phonation is often negatively perceived and considered less effective and acceptable or distracting. This perception of disordered voice then interferes with the listener’s ability to receive and comprehend the speech signal. Vocal fry may also be perceived in this way which could lead to decreased intelligibility. Currently, no direct connection between speech produced with vocal fry and intelligibility has been identified. The purpose of the present study is to answer the question: does speech produced with vocal fry impact the intelligibility of spoken words? We hypothesized that words spoken in vocal fry would be less intelligible due the acoustic characteristics and perceptions of vocal fry. We did this by presenting listeners with words spoken in both fry and no fry. Words were spoken by both a male and female speaker and presented with varying levels of signal-to-noise ratio. Intelligibility was measured in terms of the percentage of words correctly identified and the listener’s perception of listening difficulty. These results allowed us to quantify the effect of vocal fry on speech 8 PART I: Recording and Preparation of Samples METHODS In preparation for the intelligibility listening task, one male and one female subject were recorded while speaking a list of predetermined words both with and without vocal fry. Both subjects were native English speakers with no reported history of any communication disorders or differences. The recorded words were randomly selected from the Single Word Intelligibility subtest of the Assessment of Intelligibility of Dysarthric Speech (Yorkston & Beukelman, 1984). The Assessment of Intelligibility of Dysarthric Speech (AIDS) is typically used to assess the speech intelligibility of persons with dysarthria. The Single Word Intelligibility subtest contains 50 sets of 12 phonetically similar words. A total of 400 words were randomly selected from the AIDS for this task. The 400 words were then divided into two separate lists, each comprised of 200 different words. One list was recorded by the male subject, and the other list was recorded by the female subject. See Appendix A for the lists of words recorded by each subject. The participants recorded their assigned list twice through, the first time using vocal fry and the second time using modal voice without vocal fry. This produced a total of 800 recorded words. The subjects were instructed to read each word at the end of the carrier phrase, “The next word is.” The purpose of this was to simulate the natural production of vocal fry which typically occurs at the end of phrases when subglottic pressure naturally decreases. The recording procedure was conducted in a sound isolation booth with background noise less than 25 dBA and reverberation time equal to 0.05 seconds. The recordings were acquired with two microphones, a head-mounted Shure B53 microphone and a NTI M2211 Sound Level Meter (NTI Instruments) used as a microphone at a distance of 30 cm. Both microphones were connected to a PC via a Tascam UH 7000 soundboard. Audacity 2.1.2 was utilized as the audio 9 recording and editing software. Once recorded, each word was individually segmented and saved as its own audio file. To ensure that audio quality would not interfere with intelligibility testing, each recorded word was confirmed to have a signal-to-noise ratio (the ration of the speech signal to any background noise present in the recording) higher than 15 dB in all of the octave bands between 125 Hz and 8000 Hz. In addition, the energetic content of each sample was normalized using the voiced portion of each word. This is similar to what has been done in other intelligibility research (Tjaden, Sussman, & Wilding, 2014). See Appendix B for a more detailed description of the normalization process. Each recorded word was also assessed to ensure that words of poor quality (i.e. poorly recorded samples, samples with interfering background noise, samples with poor vocal fry or intonation) were not included in the intelligibility listening task to minimize the impact of outside factors on intelligibility scores. Three samples were determined to have poor intelligibility due to the segmentation of the sample from the original audio recording (i.e. the sample was cut too close to the beginning of the word). These samples were recut to improve intelligibility and included in the intelligibility listening task. PART II: Intelligibility Listening Task The prepared samples were then used to create the intelligibility listening task. Two different versions (Version I and Version II) were created. Each version of the task included a total of 24 test sets. A test set consisted of 16 words spoken by the same speaker (Male or Female), with the same type of voicing (Fry or No Fry), and with the same level of signal-to- noise ratio (No Added Noise-NaN, -12 dB, -6 dB, 0 dB, 6 dB, or 12 dB). There are 24 possible combinations of these three conditions (speaker biological sex, voicing type, and signal-to-noise 10 ratio). Each combination of conditions corresponded to an individual test set. Table 1 contains a list of the conditions for each of the 24 test sets. Table 1: Combination of conditions for each of the 24 test sets. Fry Female NaN -12 -6 0 6 12 No Fry Male NaN -12 -6 0 6 12 Female NaN -12 -6 0 6 12 Male NaN -12 -6 0 6 12 Two different versions of each test set were created with each containing a different list of 16 words (List A and List B). For each combination of speaker sex and signal-to-noise ratio, List A and List B consisted of the same list of words in both the Fry and No Fry conditions. The two different versions of the intelligibility listening task were created such that if the Fry version of List A was used in one version, then the No Fry version of List A was used in the other version, and vice versa. In addition, if the Fry version of List A was used in one version, then the No Fry version of List B would be used in that same version, and vice versa. For example, for the test set consisting of the conditions of Male speaker, Fry, and signal-to-noise ratio of -6 dB, List A was used in Version I and List B was used in Version II. For the test set consisting of the same conditions (Male speaker, -6 dB), but No Fry instead of Fry, List B was used in Version I and List A was used in Version II. See Appendix C for the list of words presented in each test set. This was done so that every word was presented in both Fry and No Fry in order to counterbalance for the possible effect of the phonological aspects of each word on its intelligibility based on the assumption that any possible impact on intelligibility should affect both Fry and No Fry relatively similarly. 11 For each version (Version I and Version II), the 24 tests sets were presented in two different pseudorandomization orders. The first order was created utilizing a random sequence generator, and the second order was created by reversing the first order. A pseudorandomization order was only included if no more than two of the 6 dB and/or 12 dB signal-to-noise ratio conditions were presented consecutively. The purpose of this was to minimize the possible effect of listener fatigue on intelligibility scores and listening difficulty ratings by spacing out the tests sets containing the two greatest levels of background noise. Including two pseudorandomization orders in which the second order was a reverse of the first was intended to minimize the effect of listener fatigue as the intelligibility listening task progressed. A decrease in intelligibility of the test sets presented at the end of the intelligibility task due to listener fatigue should be balanced by having those test sets presented at the beginning of the task in the other pseudorandomization order. This created four total intelligibility listening tasks which were labeled Version Ia, Version Ib, Version IIa, and Version IIb. See Appendix D for the order of test sets presented in each version of the intelligibility listening task. Prior to recruiting participants, IRB approval was obtained from the Michigan State University Human Research Protection Program (Protocol # MCDCR00000104, 17-1218- Intelligibility; approval letter located in Appendix E). Participants were recruited via the Michigan State University College of Communication Arts and Sciences SONA system to participate in the intelligibility listening task. Participants provided informed consent prior to participation (consent form located in Appendix F). Participants were included in the task if they were at least 18 years of age and spoke English as their native language as verified by each participant’s SONA prescreen data and verbal confirmation from each participant. The participant’s data were only included in the analysis if they had normal hearing, as verified by a 12 hearing screening (using a Beltone Model 120 audiometer) in which hearing thresholds were verified to be within the normal range (≤ 20 dB) in each ear at 250 Hz, 500 Hz, 1000 Hz, 2000 Hz, 4000 Hz, and 8000 Hz. Prior to completing the intelligibility listening task, participants were asked to complete a training test set consisting of 16 words in the Male, No Fry condition at a signal-to-noise ratio of -6 dB. The participants then completed the intelligibility listening task. The intelligibility listening task was presented to each participant via Sennheiser HD-205 headphones. Each test set was presented as an individual audio file on a computer. A separate folder was created for each version of the test containing the audio files for each of the 24 subtests arranged in the order which they appeared in the packet of response forms (audio files and response forms were presented in the randomization orders previously discussed). Participants were instructed to play each audio file in the order presented. Within each file, each of the 16 words were presented at the end of the carrier phrase, “The next word is.” The carrier phrase was identical for all presented words. The words were spaced with approximately 8 seconds in between each word to allow participants time to mark their answers. For each test set, participants were asked to indicate the word they believe they heard from a field of 12 phonetically similar words presented on the corresponding test form. Each grouping of 12 words was taken from the Single Word Intelligibility subtest of the AIDS and was presented to the participants in 3 by 4 tables as they appear in the AIDS assessment book. Participants were also asked to rate their perceived listening difficulty on a scale ranging from 1 (extremely easy) to 10 (extremely difficult) for each word. See Appendix G for an example of a test set response form. The total participation time for each participant was between 50 to 60 minutes. 13 PART III: Analysis Techniques An intelligibility score and listening difficulty rating were calculated for each test set per participant. The intelligibility score was calculated as a percentage of words correctly identified in each test set. The listening difficulty rating was calculated as the mean of the ratings for each of the 16 words in each test set. The mean and standard deviation of the intelligibility score and listening difficulty rating were calculated and compared for both the Fry and No Fry conditions overall, for the Male and Female speakers, and at each level of signal-to-noise ratio. Comparisons were analyzed using a paired sample T-test. Differences were considered significant if the P value was < .05. 14 Participants RESULTS Forty-four participants were recruited via the Michigan State University College of Communication Arts and Sciences SONA system. Of the 44 recruited participants, 29 completed the intelligibility task. Three participants did not meet the inclusion criteria for normal hearing, each having at least one hearing threshold greater than 20 dB as identified on the hearing screening. Therefore, data from 26 participants were included in the analysis. All participants were students of Michigan State University, at least 18 years of age, and native speakers of English. Five participants were male, and 21 were female. Seven participants completed Version Ia of the intelligibility listening task, seven completed Version Ib, six completed Version IIa, and six completed Version IIb. The version of the intelligibility task that each participant completed was assigned based on their order of enrollment such that Participant 1 received Version Ia, Participant 2 received Version IIa, Participant 3 received Version Ib, Participant 4 received Version IIb, Participant 5 received Version Ia, and so forth. Intelligibility Score A general comparison of intelligibility scores between the Fry and No Fry (modal) conditions for all of the tests are shown in Table 2. Overall, the mean intelligibility score for the Fry condition was significantly lower than that of the No Fry condition (62.08% to 64.56% respectively, P < .001). The mean intelligibility scores of the Fry and No Fry conditions were also compared separately for each speaker sex. For both the Male spoken samples and Female spoken samples, the mean intelligibility score for the Fry condition was lower than that of the No Fry condition. This difference was not significant for the Male spoken samples (63.66% to 64.63% respectively, P = .255), but it was significant for the Female spoken samples (60.50% to 15 64.26% respectively, P < .001). Figure 1 shows the comparison between the Fry and No Fry conditions for both the Male and Female speakers. Table 2: Comparison of mean intelligibility scores between the Fry and No Fry conditions. Male speaker Female speaker Overall *Significant difference Fry SD 31.15 31.73 31.48 Mean 63.66 60.50 62.08 No Fry Mean 64.63 64.26 64.56 SD 30.91 30.50 30.76 P value .255 < .001* < .001* Table 3 shows the mean intelligibility scores for all of the tests under the Fry and No Fry conditions at each level of signal-to-noise ratio (NaN, -12 dB, -6 dB, 0 dB, 6 dB, and 12 dB). At each level, the mean intelligibility score for the Fry condition was lower than that of the No Fry condition. However, this difference was only significant at the levels of -12 dB (84.13% to 87.02%, P = .011) and 12 dB (8.17% to 12.62%, P = .002). Figure 2 shows the comparison of mean intelligibility scores between the Fry and No Fry conditions at each level of signal-to-noise ratio. Table 3: Comparison of mean intelligibility scores between the Fry and No Fry conditions at each level of signal-to-noise ratio. Fry No Fry Mean 95.91 87.02 82.69 67.55 41.59 12.62 SD 8.70 7.13 10.37 12.44 13.97 9.53 P value .052 .011* .366 .783 .231 .002* NaN -12 dB -6 dB 0 dB 6 dB 12 dB NaN = No added noise *Significant difference Mean 93.87 84.13 80.53 66.95 38.82 8.17 SD 7.09 7.60 12.89 12.46 12.09 7.99 Differences in mean intelligibility scores were also assessed between the Male and Female speakers. Overall (including both Fry and No Fry conditions), the mean intelligibility 16 score was significantly greater for the Male speaker than for the Female speaker (64.26% to 62.38% respectively, P = .018). Under just the Fry condition, the mean intelligibility score was also significantly greater for the Male speaker compared to the Female speaker (63.66% to 60.50% respectively, P = .007). However, under the No Fry condition, the mean intelligibility score was only marginally greater for the Male speaker compared to the Female speaker, and this difference was not significant (64.63% to 64.26% respectively, P = .576). These differences can also be visualized in Figure 1. See Appendix H for the mean intelligibility score data from all participants. 66% 65% 64% 63% 62% 61% 60% 59% 58% 57% 56% E R O C S Y T I L I B G I L L E T N I I N A E M Fry No Fry Fry No Fry MALE FEMALE Figure 1: Comparison of mean intelligibility scores between the Fry and No Fry conditions by speaker sex. The first two bars represent the mean intelligibility scores for the Male speaker with the dotted bar representing the Fry condition and the striped bar representing the No Fry condition. The difference between Fry and No Fry for the Male speaker was not statistically significant (63.66% to 64.56%, P = .255). The second two bars represent the mean intelligibility scores for the Female speaker with the dotted bar representing Fry and the striped bar representing No Fry. This difference between Fry and No Fry for the Female speaker was statistically significant (60.50% to 64.26%, P < .001). Between the male and female speaker, differences were significant for the Fry condition (63.66% to 60.50%, P = .007) but not for the No Fry condition (64.63% to 64.26%, P = .576). Error bars represent standard error. 17 E R O C S Y T I L I B G I L L E T N I I N A E M 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Fry No Fry NaN -12 -6 0 6 12 SIGNAL-TO-NOISE RATIO (dB) Figure 2: Comparison of mean intelligibility scores between the Fry and No Fry conditions by signal-to-noise ratio. At each level of signal-to-noise ratio the mean intelligibility score was lower for Fry (represented by the dotted bars) compared to No Fry (represented by the striped bars). These differences were only significant at the -12 dB and 12 dB levels (84.13% to 87.02%, P = .011; 8.17% to 12.62%, P = .002, respectively). Error bars represent standard error. Listening Difficulty A comparison of mean listening difficulty ratings between the Fry and No Fry conditions is shown in Table 4. Overall, the mean listening difficulty rating was significantly greater for the Fry condition compared to the No Fry condition (4.98 to 4.56 respectively, P < .001). The mean listening difficulty of the Fry and No Fry conditions were also compared separately for each speaker sex. For the Male spoken samples, the mean listening difficulty was significantly greater for the Fry condition compared to the No Fry condition (4.97 to 4.74 respectively, P = .003). For the Female spoken samples, the mean listening difficulty was also significantly greater for the Fry condition compared to the No Fry condition (5.00 to 4.42 respectively, P < .001). Figure 3 shows the comparison between the Fry and No Fry conditions for both the Male and Female speakers. 18 Table 4: Comparison of mean listening difficulty ratings between the Fry and No Fry conditions. Male speaker Female speaker Overall *Significant difference Fry SD 3.30 3.20 3.25 Mean 4.97 5.00 4.98 No Fry Mean 4.74 4.42 4.56 SD 3.23 3.24 3.24 P value .003* < .001* < .001* Table 5 shows the mean listening difficulty ratings for all of the tests in the Fry and No Fry conditions for each level of signal-to-noise ratio. At each level, the mean listening difficulty rating for the Fry condition was greater than that of the No Fry condition. This difference was significant at the NaN (1.65 to 1.34, P = .012), -12 dB (2.36 to 1.92, P = .003), 0 dB (5.09 to 4.43, P = .011), and 6 dB levels (7.89 to 7.19, and P < .001), but not at the -6 dB (3.05 to 2.74, P = .110) or 12 dB levels (9.86 to 9.75, P = .180). Figure 4 shows the comparison of mean listening difficulty ratings between the Fry and No Fry conditions at each level of signal-to-noise ratio. Table 5: Comparison of mean listening difficulty ratings between the Fry and No Fry conditions at each level of signal-to-noise ratio. Fry NaN -12 dB -6 dB 0 dB 6 dB 12 dB NaN = No added noise *Significant difference Mean 1.65 2.36 3.05 5.09 7.89 9.86 No Fry SD .84 1.15 1.31 1.59 1.73 .34 Mean 1.34 1.92 2.74 4.43 7.19 9.75 SD .53 .84 1.33 1.71 1.58 .60 P value .012* .003* .110 .011* < .001* .180 The mean listening difficulty ratings were also compared between the Male and Female speakers. Overall (including both Fry and No Fry conditions), there was not a significant difference in mean listening difficulty ratings between the Male and Female speakers (4.83 to 19 4.71, P = .096). In just the Fry condition, the difference between the Male and Female speakers was also not significant (4.97 to 5.00, P= .756). However, in the No Fry condition, the mean listening difficulty for the Male speaker was significantly greater than the Female speaker (4.74 to 4.42, P = .008). These differences can also be visualized in Figure 3. See Appendix I for the mean listening difficulty data from all participants. 5.5 5 Y T L U C I I F F I D G N N E T S I L N A E M 4.5 4 3.5 3 Fry No Fry Fry No Fry MALE FEMALE Figure 3: Comparison of mean listening difficulty between the Fry and No Fry conditions by speaker sex. The first two bars represent the mean listening difficulty for the Male speaker, with the dotted bar representing Fry and the striped bar representing No Fry. This difference was statistically significant (4.97 to 4.74, P = .003). The second two bars represent the mean listening difficulty for the female speaker, with the dotted bar representing Fry and the striped bar representing No Fry. This difference was also statistically significant (5.00 to 4.42, P < .001). Between the male and female speaker, differences were not significant for the Fry condition (4.97 to 5.00, P= .756) but were significant for the No Fry condition (4.74 to 4.42, P = .008). Error bars represent standard error. 20 Y T L U C I I F F I D G N N E T S I L N A E M 10 9 8 7 6 5 4 3 2 1 0 NaN -12 -6 0 6 12 SIGNAL-TO-NOISE RATIO (dB) Fry No Fry Figure 4: Comparison of mean listening difficulty between the Fry and No Fry conditions by signal-to-noise ratio. At each level of signal-to-noise ratio the mean listening difficulty was greater for Fry (represented by the dotted bars) compared to No Fry (represented by the striped bars). These differences were significant at the NaN (1.65 to 1.34, P = .012), -12 dB (2.36 to 1.92, P = .003), 0 dB (5.09 to 4.43, P = .011), and 6 dB levels (7.89 to 7.19, and P < .001). Error bars represent standard error. Intelligibility Score and Listening Difficulty: Worst Noise Conditions Removed Because of the sharp decrease in intelligibility for the test sets presented with the two worst noise conditions (signal-to-noise ratio levels 6 dB and 12 dB), the results of these tests sets were removed and the mean intelligibility scores and listening difficulty ratings were recalculated using the remaining tests. This dramatically reduced standard deviations indicating less variability between each participant’s data. Table 6 shows a comparison of the adjusted overall mean intelligibility scores. The adjusted overall mean intelligibility score was significantly lower for the Fry condition compared to the No Fry condition (81.37% to 83.29%, P = .030). For the Male samples, the Fry condition was lower than the No Fry condition; however, this difference was not significant (83.23% to 84.56%, P = .285). For the Female 21 samples, the Fry condition was lower than the No Fry condition, and this difference was significant (79.51% to 82.03%, P = .049). This is the same pattern of significance that was observed in the original analysis. Figure 5 shows the comparison between the adjusted mean intelligibility scores between the Fry and No Fry conditions for both the Male and Female speakers. Table 6: Comparison of mean intelligibility scores between Fry and No Fry excluding the 6 dB and 12 dB noise conditions. Male speaker Female speaker Overall *Significant difference Fry SD 12.05 15.78 14.16 Mean 83.23 79.51 81.37 No Fry Mean 84.56 82.03 83.29 SD 12.44 14.93 13.80 P value .285 .049* .030* E R O C S Y T I L I B G I L L E T N I I N A E M 90% 85% 80% 75% 70% 65% 60% Fry No Fry Fry No Fry MALE FEMALE Figure 5: Comparison of adjusted mean intelligibility scores between the Fry and No Fry conditions by speaker sex. The first two bars represent the mean intelligibility scores for the male speaker with the dotted bar representing Fry and the striped bar representing No Fry. This difference was not statistically significant (83.23% to 84.56%, P = .285). The second two bars represent the mean intelligibility scores for the female speaker with the dotted bar representing Fry and the striped bar representing No Fry. This difference was statistically significant (79.51% to 82.03%, P = .049). Error bars represent standard error. 22 Table 7 shows a comparison of the adjusted overall mean listening difficulty ratings. Overall, the adjusted mean listening difficulty rating was significantly greater for the Fry condition compared to the No Fry condition (3.03 to 2.62, P < .001). For the Male samples, the Fry condition was greater than the No Fry condition; however, this difference was not significant (2.90 to 2.74, P = .176). For the Female samples, the Fry condition was greater than the No Fry condition, and this difference was significant (3.18 to 2.48, P <.001). This pattern of significance differs from the original analysis in that the difference between Fry and No Fry for the Male spoken samples is no longer significant when the worst two noise conditions are removed. Figure 6 shows the comparison between the adjusted mean listening difficulty ratings between the Fry and No Fry conditions for both the Male and Female speakers. Table 7: Comparison of mean listening difficulty ratings between Fry and No Fry excluding the 6 dB and 12 dB noise conditions. Male speaker Female speaker Overall *Significant difference Mean 2.90 3.18 3.03 Fry SD 1.66 1.90 1.79 No Fry Mean 2.74 2.48 2.61 SD 1.70 1.62 1.67 P value .176 < .001* < .001* 23 Y T L U C I I F F I D G N N E T S I L N A E M 4 3.5 3 2.5 2 1.5 1 0.5 0 Fry No Fry Fry No Fry MALE FEMALE Figure 6: Comparison of adjusted mean listening difficulty between the Fry and No Fry conditions by speaker sex. The first two bars represent the mean listening difficulty for the Male speaker with the dotted bar representing Fry and the striped bar representing No Fry. This difference was not statistically significant (2.90 to 2.74, P = .176). The second two bars represent the mean listening difficulty for the female speaker with the dotted bar representing Fry and the striped bar representing No Fry. This difference was statistically significant (3.18 to 2.48, P <.001). Error bars represent standard error. Comparison of Male and Female Listeners The Fry subtests were also compared between the Male and Female listeners. The comparison of mean intelligibility scores between the Male and Female listeners is shown in Table 8. Overall, there was no significant difference in mean intelligibility score of the Fry subtests between the Male and Female listeners (60.94% to 62.35% respectively; P = .756). Mean intelligibility scores were also compared between the Male and Female listeners for the Male speaker and Female speaker conditions separately. For the Male speaker, the mean intelligibility score was greater for the Male listeners compared to the Female listeners; however, this difference was not significant (65.83% to 63.14%, P = .673). For the Female speaker, the mean intelligibility score was lower for the Male listeners compared to the Female listeners, but 24 this difference was also not significant (56.04% to 61.56%, P = .395). These differences can be visualized in Figure 7. Table 8: Comparison of mean intelligibility scores of the Fry subtests between the Male and Female listeners. Male listeners Male speaker Female speaker Overall Mean 65.83 56.04 60.94 SD 28.58 31.56 30.50 Female listeners SD Mean 63.14 61.56 62.35 31.71 31.68 31.71 P value .673 .395 .756 Table 9 shows the comparison of mean listening difficulty ratings of the Fry subtests between the Male and Female listeners. Overall, there was no significant difference in mean listening difficulty rating of the Fry subtests between the Male and Female listeners (4.781 to 5.031 respectively; P = .593). Mean listening difficulty ratings were also compared between the Male and Female listeners for the Male speaker and Female speaker conditions separately. For the Male speaker, the mean listening difficulty rating was lower for the Male listeners compared to the Female listeners; however, this difference was not significant (4.881 to 4.988, P = .875). For the Female speaker, the mean listening difficulty rating was also lower for the Male listeners compared to the Female listeners, and this difference was also not significant (4.681 to 5.075, P = .547). These differences can be visualized in Figure 8. Table 9: Comparison of mean listening difficulty ratings of the Fry subtests between the Male and Female listeners. Male listeners Female listeners SD Mean 4.988 5.075 5.031 3.295 3.197 3.247 P value .875 .547 .593 Male speaker Female speaker Overall Mean 4.881 4.681 4.781 SD 3.325 3.192 3.261 25 E R O C S Y T I L I B G I L L E T N I I N A E M 80% 70% 60% 50% 40% 30% 20% 10% 0% MALE SPEAKER FEMALE SPEAKER Male listeners Female listeners Figure 7: Comparison of mean intelligibility scores between the Male and Female listeners by speaker sex. The first two bars represent the mean intelligibility scores for the Male speaker with the dotted bar representing the Male listeners and the dashed bar representing the Female listeners. This difference was not statistically significant (65.83% to 63.14%, P = .673). The second two bars represent the mean intelligibility scores for the Female speaker with the dotted bar representing the Male listeners and the dashed bar representing the Female listeners. This difference was also not statistically significant (56.04% to 61.56%, P = .395). Error bars represent standard error. 26 Y T L U C I I F F I D G N N E T S I L N A E M 5.2 5.1 5 4.9 4.8 4.7 4.6 4.5 4.4 4.3 MALE SPEAKER FEMALE SPEAKER Male listeners Female listeners Figure 8: Comparison of mean listening difficulty ratings between the Male and Female listeners by speaker sex. The first two bars represent the mean listening difficulty ratings for the Male speaker with the dotted bar representing the Male listeners and the dashed bar representing the Female listeners. This difference was not statistically significant (4.881 to 4.988, P = .875). The second two bars represent the mean listening difficulty ratings for the Female speaker with the dotted bar representing the Male listeners and the dashed bar representing the Female listeners. This difference was also not statistically significant (4.681 to 5.075, P = .593). Error bars represent standard error. 27 DISCUSSION Overall, the mean intelligibility score of all of the tests was significantly lower for the Fry condition compared to the No Fry condition. This difference remained significant when the two worst noise conditions (signal-to-noise ratio levels 6 dB and 12 dB) were removed and the mean intelligibility scores were recalculated. This suggests that single words spoken in vocal fry are less intelligible than words spoken without vocal fry (modal voicing). In addition, the overall mean listening difficulty rating was significantly greater for the Fry condition compared to the No Fry condition. This difference also remained significant when the 6 dB and 12 dB noise conditions were removed. This suggests, that not only are words spoken in vocal fry less intelligible, the perceived difficulty in understanding these words is also greater. These results confirm our hypothesis. Differences in intelligibility between vocal fry and modal voicing may have possible social and vocational implications. A study by Anderson, Klofstad, Mayew, and Venkatachalam (2014) compared the perceptions of vocal fry and modal voicing when used in the context of the labor market. Their results suggested that, especially when used by young females, vocal fry was perceived more negatively than modal voicing. In particular, job candidates using vocal fry were perceived as less competent, educated, trustworthy, and hirable when compared to candidates using a normal speaking voice. This leads us to question if perhaps differences in intelligibility or listening difficulty may play a role in the negative social and vocational perceptions of vocal fry. Another possible implication of the intelligibility differences associated with vocal fry include the potential role vocal fry may play in language processing. Morton and Watson (2001) compared language processing between modal voicing and dysphonic voicing in the classroom 28 setting. Their results showed that recall and inferencing of spoken passages were better when presented in a modal voice compared to a dysphonic voice. Participants also indicated that they perceived the dysphonic voice as unpleasant which resulted in increased listening demands. Although vocal fry is not a dysphonia, it may share some commonalities with dysphonia in how it is perceived (as negative and unpleasant), thus leading to similar implications. In addition, the decreased intelligibility of vocal fry as indicated in the current study may also contribute to differences in language processing. Further research may be warranted to examine the effect of vocal fry on language processing. When broken down by the speaker’s biological sex, a greater difference in mean intelligibility scores between the Fry and No Fry conditions existed for Female speaker compared to the Male speaker. In addition, this difference in mean intelligibility score was only significant for the Female condition and not for the Male condition. This suggests that vocal fry has a greater impact on speech intelligibility when used by a female speaker compared to a male speaker. One possible explanation for this phenomenon is that a greater difference in fundamental frequency exists between vocal fry and modal voicing for female speakers compared to male speakers. Fundamental frequency typically ranges from 150 to 500 Hz in adult women and 80 to 450 Hz in adult men (Boone, McFarlane, Von Berg, & Zraick, 2014). Vocal fry typically occurs at frequencies below 70 Hz (Titze, 2000) with one group of researchers suggesting 24-52 Hz as the typical range for male speakers and 18-46 Hz for female speakers in vocal fry (Hollien & Michel, 1968; Michel, 1968). Thus the difference between vocal fry and modal voicing in terms of fundamental frequency is typically greater for female speakers compared to male speakers. This, perhaps, decreases the naturalness and familiarity of the voice, with a greater impact on the female voice. This greater difference in naturalness of 29 vocal fry for female speakers may contribute to the greater difference in intelligibility as well as the greater likelihood of vocal fry being perceived as negative when used by females as demonstrated in previous studies (Anderson et al., 2014). When intelligibility scores of vocal fry were compared between the Male and Female listeners for the Male speaker, the mean intelligibility score was greater for the Male listeners compared to the Female listeners. For the Female speaker, the mean intelligibility score was lower for the Male listeners compared to the Female listeners. Although these differences were not statistically significant, this trend may suggest that the intelligibility of vocal fry may be more greatly affected if the speaker and listener are of opposite sex. These results may be limited by sample size as this analysis only included data from 5 Male listeners. Future research should include a greater number of participants to further understand any differences in the interaction between the sex of the speaker and the sex of the listener in terms of speech intelligibility of vocal fry. Differences in mean intelligibility scores and listening difficulty ratings were not significant at all levels of signal-to-noise ratio. This may suggest that intelligibility may be affected differently by various levels of background noise. This could have potential implications for certain applications of vocal fry, such as individuals using vocal fry in noisy environments, (e.g. teachers speaking in noisy classrooms or employees speaking on the floor of a factory). Competing noise varying in intensity or frequency might affect intelligibility of vocal fry and modal voicing differently due to the different acoustic characteristics of each. To better understand vocal fry and its effect on communication, future research should explore the differences between intelligibility of male and female speakers and between different levels of noise and different types of noise (e.g. spectral differences, modulated, talker babble). 30 As is the case of all research, there are several potential limiting factors. For example, during the normalization process of the audio samples, the voiced portion of each word was normalized. Since vocal fry voicing is generally at a lower dB level than modal voicing, normalizing the voicing part of the word resulted in an overemphasis of the consonants of the Fry samples. This could have impacted intelligibility by aiding listeners in the identification of words spoken in vocal fry. Despite this, vocal fry was still less intelligible than modal voicing. Perhaps this difference would have been greater had the consonants of the Fry words not been emphasized more than in the No Fry words. Sample size presents another possible limitation. The analysis was intended to be completed with data from at least 40 participants. Due to a high rate of participant no show, the analysis could only be completed with data from 26 participants. Participants were also all of typical college age, an age group more often associated with the use of vocal fry. Intelligibility of vocal fry may vary for different generations of listeners. Related, since only two talkers provided the recorded samples, dialectal differences unique to these two talkers could also present a possible limitation. For example, one of the words spoken by the male speaker was the word “bag.” However, the word “beg” was one of the options on the response form for this sample. Due to dialectal differences in both production and perception, “beg” could be reasonably chosen as a response. Because each word was included in both the Fry and No Fry conditions, differences in intelligibility due to dialectal differences should have affected both the Fry and No Fry conditions relatively similarly. However, the possibility remains that dialectal differences may have impacted intelligibility scores. Therefore, future research should include more samples from a wider range of talkers and with more listeners. 31 Finally, the current study examined intelligibility at the single-word level only. Intelligibility in expanded contexts such as at the conversational level may not be as greatly affected by vocal fry due to the added context of the surrounding phrase/sentence. Further research should examine the effect of vocal fry on speech intelligibility in these contexts to better understand the impact of vocal fry on speech intelligibility. 32 CONCLUSION The results of this study indicate that speech intelligibility at the single word level is reduced for words spoken in vocal fry compared to words spoken in modal voicing. Reduced intelligibility was greater for the female speaker compared to the male speaker. In addition, perceived listening difficulty was also greater for vocal fry compared to modal voicing, with this difference again greater for the female speaker compared to the male speaker. Reduced intelligibility may contribute to the negative perception of vocal fry. Further research should explore if vocal fry is associated with reduced intelligibility and increased listening difficulty at the phrase or conversational level when context is introduced or if vocal fry affects language processing. 33 APPENDICES 34 APPENDIX A: Words Recorded by the Male and Female Speakers 35 Table 10: Words recorded by the Male speaker Arch Ark Author Bait Ball Ban Barn Bat Beam Beat Beep Beer Beg Below Bender Best Bill Bin Bitten Bitter Blown Bone Bought Bowl Bread Groan Brother Design Grove Bumper Divide Heart Doe Burn Hide Dusk But Impress Dust By Intact Eye Cane Judge Faint Cape Ladle Fall Car Late Fell Carp Law File Cart Leaf Flicker Case Leak Floor Cave Leap Flow City Least Fool Claw Lecture Compress Fork Limb Fort Contact Lip Full Coupe List Crawl Gag Loss Game Cross Map Gang Dark Gauge Day Mate Gesture Meat Dead Mere Globe Debate Mist Glow Deer Define Grain Mixture Mother Naval Near Neat Occur Offer Often Option Owl Pace Page Pale Par Paste Pay Pierce Planner Porch Praise Preach Quart Quicker Quit Rage Rain Rake Ramp Rate Reach Red Refer Renter Rest Root Rupture Said Sane Sap Scorch Scrap Scratch Screech Scrub Seer Sender Serve Shade Shark Shear Shed Table Sheet Tanner Ship Tarred Shock Tart Shoe Thread Short Told Side Tore Sit Tort Sitter Trace Sitting Trade Slid Train Slim Troop Sold Twine Sort Warm Soul Waste Sprain We’re Stable Weep Stall Whip Start Wide Store Storm Wife Strange Wine Stream Wipe Wire Suit Super Witty Swarm Wore 36 Table 11: Words recorded by the Female speaker Absurd After Ape Art Awful Bank Banner Bark Bash Beast Bed Beef Big Bit Blow Boast Boat Book Boot Booth Born Boss Bother Brace Brain Bull Cable Cage Cake Call Care Carve Center Cheer Closure Cloth Coal Cold Conserve Form Contract Contrast Cork Corn Cough Court Creature Grab Dart Grace Grade Decide Grange Defeat Defend Grape Pave Live Peach Loop Pear Low Phone Make Pier Mall Pole Manner Port March Pour Mat Preserve May Pretty Member Race Mold Raid Mole Range More Rattle Musk Ray Nap Nest Reef Obscure Ride Office Rust Saddle Other Sat Paid Sauce Pain Paint Screen Seam Panel Sharp Park Part Shelf Denture Grave Deny Great Detract Green Dread Grind Gross Dress Group Each Ember Grow Harp Fable Hear Fear High Fierce Final Honest Horn Fitting Jade For Just Kitten Knit Lease Leave Leer Left Less Lie Lisp Lit Litter Fruit Funnel Gain Glass Gleam Glitter Teacher Shell Term Shore Texture Shot Threat Shred Timber Shut Tired Sicker Torn Sigh Trait Sketch Twice Slip Twist So Vague Space Wait Spade Wake Spark Wall Sport Wart Steer Weave Stiff Street Weird Stretch While Suppress White Swift Take Tall Tap Tape Taste Why Wicked Win Wise Wives Written 37 APPENDIX B: Description of the Normalization Process 38 Figure 9: Normalization process example. The blue dots in the center graph represent the estimated vowel area of the sample word “bait.” A first attempt at normalizing all of the word files involved equalizing the average amount of acoustic energy of the entire word. Because the words differed in vowel length, number of vowels (e.g. bait, divide), and number of consonants (e.g. quit, contact), equalizing the average acoustic energy of each word resulted in words that were perceptually unequal in energy. Instead, the average acoustic energy of only the voicing portion of each word (vowels and voiced consonants) was normalized as opposed to the entire word. This is similar to was has been done in other research in intelligibility (Tjaden, Sussman, & Wilding, 2014). To do this, a custom MATLAB script was used to obtain the dB contour, fundamental frequency contour, and pitch strength contour for each word. Next, voicing segments of the word were identified using 39 the time information from the fundamental frequency contour. From this, the average dB of the voicing segment was estimated. An example of this is shown by the voicing area of the word “bait” in Figure 9. The voicing area is indicated by the blue circles along the dB contour shown in the second graph. For each word, the average acoustic energy for this voicing area was then normalized to -22 dB. This resulted in a potential advantage for words in the Fry condition. Vocal fry is less intense by nature, therefore increasing the energy of the entire word during the normalization process also increased the intensity of consonants in words in the Fry condition more than the words in the No Fry condition. This could have increased in the intelligibility of words spoken in vocal fry by overemphasizing the consonants which could have assisted listeners in correctly identifying the word. However, the results indicated that the Fry condition was still less intelligible than the No Fry condition despite this effect. 40 APPENDIX C: Lists of Words Used in Each Test Set 41 Table 12: List of words used in each combination of conditions. Male Fry NaN -12 dB -6 dB List A Arch Beg Bread Cave Define Flow Grain Leak Mixture Par Rake Sap Shed Sold Super Train Ark Below Brother City Design Fool Groan Leap Mother Paste Ramp Scorch Sheet Sort Swarm Troop Author Bender Bumper Claw Divide Fork Grove Least Naval Pierce Rate Scrap Ship Soul Table Twine List B Barn Bitten Cane Crawl Eye Game Intact List Often Preach Renter Seer Side Start Thread Weep Bat Bitter Cape Cross Fall Gang Judge Loss Option Quart Rest Sender Sit Store Told Whip Male No Fry List B List A Arch Beg Bread Cave Define Flow Grain Leak Mixture Par Rake Sap Shed Sold Super Train Ark Below Brother City Design Fool Groan Leap Mother Paste Ramp Scorch Sheet Sort Swarm Troop Barn Bitten Cane Crawl Eye Game Intact List Often Preach Renter Seer Side Start Thread Weep Bat Bitter Cape Cross Fall Gang Judge Loss Option Quart Rest Sender Sit Store Told Whip Female Fry List A Absurd Big Brain Coal Defeat Final Grade High Lisp Member Paint Preserve Sauce Sketch Swift Torn List B Banner Boot Care Corn Dress Gain Grind Lease Make Nest Pear Ray Shore Sport Teacher Wake Female No Fry List B List A Absurd Big Brain Coal Defeat Final Grade High Lisp Member Paint Preserve Sauce Sketch Swift Torn Banner Boot Care Corn Dress Gain Grind Lease Make Nest Pear Ray Shore Sport Teacher Wake After Bit Bull Cold Defend Fitting Grange Honest Lit Mold Panel Pretty Screen Slip Take Trait Bark Booth Carve Cough Each Glass Gross Leave Mall Obscure Phone Reef Shot Steer Term Wall After Bit Bull Cold Defend Fitting Grange Honest Lit Mold Panel Pretty Screen Slip Take Trait Bark Booth Carve Cough Each Glass Gross Leave Mall Obscure Phone Reef Shot Steer Term Wall Beam Blown Car Day Fell Gauge Ladle Map Owl Quicker Root Serve Sitter Storm Tore Wide Author Bender Bumper Claw Divide Fork Grove Least Naval Pierce Rate Scrap Ship Soul Table Twine Beam Blown Car Day Fell Gauge Ladle Map Owl Quicker Root Serve Sitter Storm Tore Wide 42 Ape Blow Cable Conserve Denture For Grape Horn Litter Mole Park Race Seam So Tall Twice Bash Born Center Court Ember Gleam Group Leer Manner Office Pier Ride Shred Stiff Texture Wart Ape Blow Cable Conserve Denture For Grape Horn Litter Mole Park Race Seam So Tall Twice Bash Born Center Court Ember Gleam Group Leer Manner Office Pier Ride Shred Stiff Texture Wart Table 12 (cont’d) Male Fry Male No Fry List B List A 0 dB 6 dB 12 dB List A Bait Best Burn Compress Doe Fort Heart Lecture Neat Planner Reach Scratch Shock Sprain Tanner Warm Ball Bill But Contact Dusk Full Hide Limb Occur Porch Red Screech Shoe Stable Tarred Waste Ban Bin By Coupe Dust Gag Impress Lip Offer Praise Refer Scrub Short Stall Tart We're List B Beat Bone Carp Dead File Gesture Late Meat Pace Quit Rupture Shade Sitting Strange Tort Wife Beep Bought Cart Debate Flicker Globe Law Mere Page Rage Said Shark Slid Stream Trace Wine Bait Best Burn Compress Doe Fort Heart Lecture Neat Planner Reach Scratch Shock Sprain Tanner Warm Ball Bill But Contact Dusk Full Hide Limb Occur Porch Red Screech Shoe Stable Tarred Waste Female Fry List A Art Boast Cage Contract Deny Form Grave Jade Live More Part Raid Sharp Space Tap Twist List B Beast Boss Cheer Creature Fable Glitter Grow Left March Other Pole Rust Shut Street Threat Weird Female No Fry List B List A Art Boast Cage Contract Deny Form Grave Jade Live More Part Raid Sharp Space Tap Twist Beast Boss Cheer Creature Fable Glitter Grow Left March Other Pole Rust Shut Street Threat Weird Beat Bone Carp Dead File Gesture Late Meat Pace Quit Rupture Shade Sitting Strange Tort Wife Beep Bought Cart Debate Flicker Globe Law Mere Page Rage Said Shark Slid Stream Trace Wine Awful Boat Cake Contrast Detract Fruit Great Just Loop Musk Pave Range Shelf Spade Tape Vague Bed Bother Closure Dart Fear Grab Harp Less Mat Paid Port Saddle Sicker Stretch Timber While Awful Boat Cake Contrast Detract Fruit Great Just Loop Musk Pave Range Shelf Spade Tape Vague Bed Bother Closure Dart Fear Grab Harp Less Mat Paid Port Saddle Sicker Stretch Timber While Beer Bowl Case Deer Floor Glow Leaf Mist Pale Rain Sane Shear Slim Suit Trade Wipe Ban Bin By Coupe Dust Gag Impress Lip Offer Praise Refer Scrub Short Stall Tart We're Beer Bowl Case Deer Floor Glow Leaf Mist Pale Rain Sane Shear Slim Suit Trade Wipe Bank Book Call Cork Dread Funnel Green Knit Low Nap Peach Rattle Shell Spark Taste Wait Beef Brace Cloth Decide Fierce Grace Hear Lie May Pain Pour Sat Sigh Suppress Tired White Bank Book Call Cork Dread Funnel Green Knit Low Nap Peach Rattle Shell Spark Taste Wait Beef Brace Cloth Decide Fierce Grace Hear Lie May Pain Pour Sat Sigh Suppress Tired White 43 APPENDIX D: Order of Test Sets for Each Version of the Intelligibility Listening Task 44 Table 13: Order of test sets as presented in Version Ia and Version Ib of the intelligibility listening task. Each test set is identified by the sex of the speaker (Male or Female), voicing type (Fry or No Fry), level of signal-to-noise ratio (NaN, -12 dB, -6 dB, 0 dB, 6 dB, or 12 dB), and word list (A or B) that was presented in that test set. Version Ia Speaker Sex Voicing Fry Fry Fry No Fry Fry Fry 1. Female 2. Female No Fry 3. Female 4. Female 5. Female No Fry 6. Female No Fry 7. Male 8. Female No Fry 9. Male No Fry 10. Male 11. Male 12. Female 13. Male 14. Male 15. Male 16. Male 17. Male No Fry 18. Male No Fry 19. Female No Fry 20. Male 21. Male 22. Female 23. Female No Fry 24. Female No Fry Fry Fry Fry Fry Fry Fry Fry No Fry SNR 0 dB -12 dB -12 dB 6 dB NaN 12 dB 0 dB 6 dB -6 dB 6 dB -12 dB -6 dB NaN 12 dB -12 dB 12 dB NaN 6 dB 0 dB 0 dB -6 dB NaN 12 dB 6 dB Word List A B A A B A B B B A B A A B A A B B B A A A B B Speaker Version Ib Sex No Fry Fry Fry Fry Fry Fry Voicing Female No Fry Female No Fry Female Male Male Female No Fry Male No Fry Male No Fry Male Male Male Male Female Male Male Male No Fry Female No Fry Male No Fry Female Female No Fry Female Female Female No Fry Female Fry Fry Fry Fry No Fry Fry Fry Fry SNR 6 dB 12 dB NaN -6 dB 0 dB 0 dB 6 dB NaN 12 dB -12 dB 12 dB NaN -6 dB -12 dB 6 dB -6 dB 6 dB 0 dB 12 dB NaN 6 dB -12 dB -12 dB 0 dB Word List B B A A A B B B A A B A A B A B B B A B A A B A SNR = Signal-to-noise ratio NaN = No added noise 45 Table 14: Order of test sets as presented in Version IIa and Version IIb of the intelligibility listening task. Each test set is identified by the sex of the speaker (Male or Female), voicing type (Fry or No Fry), level of signal-to-noise ratio (NaN, -12 dB, -6 dB, 0 dB, 6 dB, or 12 dB), and word list (A or B) that was presented in that test set. Speaker Sex Version IIa Voicing No Fry No Fry Fry 1. Male 2. Male 3. Female 4. Female No Fry 5. Female 6. Female No Fry 7. Female No Fry 8. Male 9. Female No Fry Fry Fry Fry Fry Fry Fry No Fry No Fry 10. Female 11. Female 12. Female 13. Male 14. Male 15. Male 16. Female No Fry 17. Male No Fry 18. Male 19. Male 20. Female No Fry 21. Male 22. Female No Fry 23. Male 24. Male Fry Fry Fry Fry Fry SNR -6 dB -12 dB NaN NaN -12 dB 0 dB 12 dB -12 dB 6 dB -6 dB 6 dB 12 dB 0 dB 6 dB 12 dB -6 dB 12 dB NaN 6 dB 0 dB NaN -12 dB 0 dB -6 dB Word List A A B A B A A B A B B B A A B A A B B B A A B B Speaker Version IIb Fry Voicing Fry Fry Fry Fry Fry Sex Male Male Female No Fry Male No Fry Female Male Male Male No Fry Female No Fry Male Male Male Female Female Female Female No Fry Male Female No Fry Female No Fry Female Female No Fry Female Male Male No Fry No Fry No Fry No Fry Fry Fry Fry Fry Fry Fry SNR -6 dB 0 dB -12 dB NaN 0 dB 6 dB NaN 12 dB -6 dB 12 dB 6 dB 0 dB 12 dB 6 dB -6 dB 6 dB -12 dB 12 dB 0 dB -12 dB NaN NaN -12 dB -6 dB Word List B B A A B B B A A B A A B B B A B A A B A B A A SNR = Signal-to-noise ratio NaN = No added noise 46 APPENDIX E: Image of IRB Approval Letter 47 Figure 10: Image of IRB approval letter. 48 APPENDIX F: Image of Consent Form 49 Figure 11: Image of consent form. 50 Figure 11 (cont’d) 51 Figure 11 (cont’d) 52 APPENDIX G: Test Set Response Form Example 53 Word #1 Word #5 Please circle the word you hear: Please circle the word you hear: Art Bark Lark Heart March Dark Arch Tart Park Ark Dart Spark Divide Design Defeat Devise Debate Defend Define Deny Decide Divine Decal Defy Please rate your difficulty in understanding the previous word: Please rate your difficulty in understanding the previous word: Easy 1 2 3 4 Word #2 5 6 7 8 9 10 1 2 3 Difficult Easy 4 5 6 7 8 9 10 Difficult Word #6 Please circle the word you hear: Please circle the word you hear: Big Bog Boat Bag Ban Bought Bank Bat Bash Beg Boot Beat Glow Flow Coal Blow Sew Bowl Below Blown Pole Low Doe Soul Please rate your difficulty in understanding the previous word: Please rate your difficulty in understanding the previous word: Easy 1 2 3 4 Word #3 5 6 7 8 9 10 1 2 3 Difficult Easy 4 5 6 7 8 9 10 Difficult Word #7 Please circle the word you hear: Please circle the word you hear: Dread Bread Tread Dress Bed Said Rest Best Ned Red Nest Dead Grain Rain Brain Grange Pain Sane Range Rage Gauge Train Strange Grade Please rate your difficulty in understanding the previous word: Please rate your difficulty in understanding the previous word: Easy 1 2 3 4 Word #4 5 6 7 8 9 10 1 2 3 Difficult Easy 4 5 6 7 8 9 10 Difficult Word #8 Please circle the word you hear: Please circle the word you hear: Cave Cane Tape Cake Cape Pave Cage Take Page Case Taste Pace Leap Leak Least Leaf Limp Lean Lease Leash Lip Leave Lit Live Please rate your difficulty in understanding the previous word: Please rate your difficulty in understanding the previous word: Easy 1 2 3 4 5 6 7 8 9 10 1 2 3 Difficult Easy 4 5 6 7 8 9 10 Difficult 54 Word #9 Word #13 Please circle the word you hear: Please circle the word you hear: Texture Gesture Denture Lecture Rapture Preacher Mixture Closure Creature Rupture Fixture Teacher Stretch Sketch Fed Threat Said Shed Dread Red Shred Bread Thread Dead Please rate your difficulty in understanding the previous word: Please rate your difficulty in understanding the previous word: Easy 1 2 3 4 Word #10 5 6 7 8 9 10 1 2 3 Difficult Easy 4 5 6 7 8 9 10 Difficult Word #14 Please circle the word you hear: Please circle the word you hear: Carve Tired Carp Care Tear Pear Car Tore Pour Tarred Cart Par Full Mold Mole Bull Pole Cold Sold Fold Bowl Pull Told Soul Please rate your difficulty in understanding the previous word: Please rate your difficulty in understanding the previous word: Easy 1 2 3 4 Word #11 5 6 7 8 9 10 1 2 3 Difficult Easy 4 5 6 7 8 9 10 Difficult Word #15 Please circle the word you hear: Please circle the word you hear: Rate Brace Trait Grace Trace Wait Rake Pace Waste Race Praise Wake Occur Reserve Super Absurd Refer Obscure Observe Preserve Converge Conserve Deserve Serve Please rate your difficulty in understanding the previous word: Please rate your difficulty in understanding the previous word: Easy 1 2 3 4 Word #12 5 6 7 8 9 10 1 2 3 Difficult Easy 4 5 6 7 8 9 10 Difficult Word #16 Please circle the word you hear: Please circle the word you hear: Vamp Bat Sap Map Sat Tap Mat Rat Ramp Vat Nap Damp Raid Rain Train Rate Rake Trade Range Race Trait Rage Rave Trace Please rate your difficulty in understanding the previous word: Please rate your difficulty in understanding the previous word: Easy 1 2 3 4 5 6 7 8 9 10 1 2 3 Difficult Easy 4 5 6 7 8 9 10 Difficult 55 APPENDIX H: Participant Intelligibility Score Data for All Test Conditions 56 Table 15: Intelligibility score as a percentage for each signal-to-noise ratio under the Male Fry condition for each subject. Male Fry NaN dB 81.25 93.75 87.5 87.5 93.75 93.75 100 93.75 100 81.25 93.75 100 100 81.25 81.25 87.5 100 93.75 87.5 93.75 87.5 87.5 81.25 68.75 87.5 93.75 89.90385 Sub01 Sub02 Sub03 Sub04 Sub05 Sub06 Sub07 Sub09 Sub10 Sub11 Sub12 Sub13 Sub14 Sub15 Sub16 Sub17 Sub19 Sub20 Sub22 Sub23 Sub24 Sub25 Sub27 Sub28 Sub29 Sub30 Mean -12 dB 93.75 81.25 93.75 81.25 93.75 87.5 87.5 93.75 68.75 93.75 93.75 93.75 81.25 100 81.25 87.5 93.75 81.25 81.25 75 93.75 68.75 87.5 87.5 93.75 93.75 87.25962 -6 dB 93.75 81.25 75 93.75 87.5 93.75 87.5 93.75 81.25 81.25 87.5 87.5 68.75 75 68.75 93.75 81.25 81.25 93.75 93.75 87.5 75 100 68.75 100 87.5 85.33654 0 dB 62.5 81.25 81.25 62.5 81.25 75 62.5 50 81.25 43.75 81.25 50 75 62.5 75 56.25 81.25 62.5 87.5 87.5 56.25 75 81.25 68.75 75 75 70.43269 6 dB 31.25 31.25 31.25 68.75 31.25 56.25 18.75 25 43.75 50 37.5 43.75 56.25 18.75 50 18.75 37.5 25 43.75 62.5 25 6.25 37.5 37.5 37.5 56.25 37.74038 12 dB 6.25 18.75 6.25 18.75 25 18.75 6.25 12.5 6.25 6.25 18.75 12.5 18.75 12.5 6.25 6.25 0 12.5 0 18.75 0 0 31.25 6.25 25 0 11.29808 57 Table 16: Intelligibility score as a percentage for each signal-to-noise ratio under the Male No Fry condition for each subject Male No Fry NaN dB 100 81.25 93.75 87.5 100 100 93.75 100 100 81.25 93.75 93.75 100 100 87.5 93.75 100 93.75 100 93.75 87.5 87.5 100 93.75 100 100 94.71154 Sub01 Sub02 Sub03 Sub04 Sub05 Sub06 Sub07 Sub09 Sub10 Sub11 Sub12 Sub13 Sub14 Sub15 Sub16 Sub17 Sub19 Sub20 Sub22 Sub23 Sub24 Sub25 Sub27 Sub28 Sub29 Sub30 Mean -12 dB 87.5 81.25 100 93.75 100 93.75 93.75 93.75 93.75 87.5 87.5 93.75 81.25 100 81.25 93.75 93.75 81.25 93.75 93.75 87.5 87.5 87.5 81.25 93.75 93.75 90.625 -6 dB 75 62.5 93.75 75 87.5 68.75 87.5 87.5 68.75 87.5 100 87.5 62.5 93.75 68.75 81.25 87.5 81.25 81.25 81.25 87.5 87.5 62.5 75 87.5 100 81.49038 0 dB 68.75 43.75 75 75 93.75 62.5 75 75 81.25 62.5 68.75 75 62.5 93.75 75 68.75 81.25 68.75 68.75 81.25 68.75 62.5 62.5 68.75 56.25 81.25 71.39423 6 dB 25 18.75 18.75 43.75 43.75 50 18.75 50 37.5 18.75 43.75 37.5 50 37.5 37.5 25 43.75 62.5 37.5 56.25 50 18.75 31.25 43.75 18.75 25 36.29808 12 dB 12.5 6.25 12.5 25 12.5 18.75 0 6.25 18.75 0 12.5 18.75 18.75 18.75 37.5 6.25 0 25 0 18.75 18.75 0 25 25 18.75 25 14.66346 58 Table 17: Intelligibility score as a percentage for each signal-to-noise ratio under the Female Fry condition for each subject Female Fry NaN dB 93.75 100 93.75 93.75 100 93.75 100 100 100 100 93.75 100 87.5 100 100 100 100 100 100 100 100 93.75 100 100 93.75 100 97.83654 Sub01 Sub02 Sub03 Sub04 Sub05 Sub06 Sub07 Sub09 Sub10 Sub11 Sub12 Sub13 Sub14 Sub15 Sub16 Sub17 Sub19 Sub20 Sub22 Sub23 Sub24 Sub25 Sub27 Sub28 Sub29 Sub30 Mean -12 dB 81.25 75 81.25 81.25 87.5 81.25 81.25 75 87.5 75 81.25 93.75 81.25 81.25 87.5 81.25 75 75 81.25 81.25 81.25 68.75 93.75 81.25 75 81.25 81.00962 -6 dB 81.25 25 68.75 75 75 87.5 75 50 93.75 68.75 87.5 62.5 75 68.75 87.5 75 75 87.5 75 75 100 87.5 75 81.25 75 81.25 75.72115 0 dB 62.5 50 68.75 56.25 68.75 68.75 37.5 50 68.75 43.75 62.5 62.5 68.75 93.75 50 68.75 68.75 62.5 68.75 62.5 50 68.75 62.5 87.5 68.75 68.75 63.46154 6 dB 37.5 37.5 43.75 31.25 31.25 50 43.75 25 43.75 37.5 37.5 37.5 37.5 37.5 37.5 56.25 50 37.5 31.25 50 43.75 37.5 37.5 31.25 31.25 62.5 39.90385 12 dB 6.25 0 12.5 18.75 0 12.5 6.25 12.5 6.25 0 12.5 6.25 0 0 0 0 0 0 0 6.25 6.25 0 12.5 0 0 12.5 5.048077 59 Table 18: Intelligibility score as a percentage for each signal-to-noise ratio under the Female No Fry condition for each subject Female No Fry Sub01 Sub02 Sub03 Sub04 Sub05 Sub06 Sub07 Sub09 Sub10 Sub11 Sub12 Sub13 Sub14 Sub15 Sub16 Sub17 Sub19 Sub20 Sub22 Sub23 Sub24 Sub25 Sub27 Sub28 Sub29 Sub30 Mean NaN dB 93.75 100 87.5 93.75 93.75 100 100 93.75 100 87.5 100 93.75 93.75 100 100 100 93.75 100 100 100 100 93.75 100 100 100 100 97.11538 -12 dB 75 93.75 87.5 87.5 81.25 87.5 75 81.25 93.75 75 81.25 81.25 68.75 81.25 93.75 81.25 81.25 81.25 81.25 93.75 87.5 75 87.5 87.5 87.5 81.25 83.41346 -6 dB 93.75 75 93.75 87.5 100 87.5 62.5 100 93.75 87.5 81.25 87.5 68.75 81.25 68.75 81.25 81.25 87.5 87.5 81.25 68.75 93.75 75 75 87.5 93.75 83.89423 0 dB 68.75 62.5 56.25 62.5 75 56.25 68.75 68.75 43.75 50 56.25 81.25 56.25 81.25 37.5 81.25 81.25 56.25 81.25 50 81.25 56.25 43.75 75 62.5 62.5 63.70192 6 dB 50 50 43.75 50 56.25 62.5 31.25 56.25 43.75 50 43.75 37.5 43.75 56.25 37.5 50 62.5 62.5 68.75 50 25 43.75 6.25 43.75 37.5 56.25 46.875 12 dB 6.25 0 6.25 12.5 6.25 12.5 12.5 6.25 0 6.25 18.75 18.75 6.25 18.75 18.75 18.75 0 0 0 25 25 0 6.25 12.5 6.25 31.25 10.57692 60 APPENDIX I: Participant Listening Difficulty Data for All Test Conditions 61 Table 19: Mean listening difficulty rating for each signal-to-noise ratio under the Male Fry condition for each subject Male Fry NaN dB 3.3125 1 1.5 1.75 1.125 1.5 1.125 1.5 1 1.125 1.875 2 1.5625 1 2.0625 1.125 3.125 1.6875 2.1875 1.3125 1.6875 3 1.0625 3 1.8125 1.375 1.723558 Sub01 Sub02 Sub03 Sub04 Sub05 Sub06 Sub07 Sub09 Sub10 Sub11 Sub12 Sub13 Sub14 Sub15 Sub16 Sub17 Sub19 Sub20 Sub22 Sub23 Sub24 Sub25 Sub27 Sub28 Sub29 Sub30 Mean -12 dB 4.4375 3.0625 1.4375 1.3125 1.4375 1.875 1.4375 1.4375 1.3125 1 2.3125 1.533333 1 1 3.8125 1.4375 3.1875 2.4 2.3125 3.125 1.25 3.3125 1.5 1.75 1.625 1.5 1.992628 -6 dB 3.8125 6.4375 1.6875 3.1875 3.5625 2.8125 1.25 1.5625 1.875 1.625 2.875 2.75 3.5625 3.6875 4.5625 1.875 4.375 3.9375 2.9375 2.5625 2.5 3.9375 1.3125 4.375 1.75 1.5625 2.9375 0 dB 5.625 8 2.4375 4.875 7 4.8125 3.0625 2.75 5.5625 2.9375 5.6875 5.0625 6.375 4.25 4.9375 6.6875 5.5 5.6875 4.25 5.375 4.3125 5.6875 5.1875 5.875 3.5625 2.625 4.927885 6 dB 9.6875 10 6.9375 7.875 9.875 8.4375 7.375 9 9.9375 4.25 8.125 7.5625 8.1875 10 6.25 9.6875 10 8.6875 8.875 8.6875 8.25 8.875 7.5 9.4375 8.1875 8.125 8.454327 12 dB 10 10 10 10 10 10 8.5625 10 10 9.875 9.25 8.5625 10 10 8.8125 10 10 10 10 10 10 10 10 10 9.375 9.5 9.766827 62 Table 20: Mean listening difficulty rating as a percentage for each signal-to-noise ratio under the Male No Fry condition for each subject Male No Fry NaN dB 1.25 2 1.25 1.125 1 1.1875 1.25 1.3125 1 1.0625 1.8125 1.375 1.25 1.5 3.125 1 1.625 1.5 1 1.0625 1.375 3 1 1 1.125 1.125 1.396635 Sub01 Sub02 Sub03 Sub04 Sub05 Sub06 Sub07 Sub09 Sub10 Sub11 Sub12 Sub13 Sub14 Sub15 Sub16 Sub17 Sub19 Sub20 Sub22 Sub23 Sub24 Sub25 Sub27 Sub28 Sub29 Sub30 Mean -12 dB 2.4375 3.733333 1.375 1.125 1.25 2.25 1 1.5 1.0625 1.1875 2.9375 1.230769 2.375 1.875 3.4375 1.375 2.625 1.625 2.375 2.8125 2.5 4.1875 2.25 1.6875 1.4375 1.0625 2.027465 -6 dB 3.5 3.375 2.25 1.125 2.6875 5.25 1.25 1.8125 1.8125 1.625 4.25 2.5625 2.5625 3.9375 3.75 3.25 8.0625 1.9375 3.125 4.1875 2.25 4 2.625 2.75 1.5 2.25 2.987981 0 dB 5.5625 7.9375 3.9375 3.4375 7.0625 7.25 2.875 3.5625 3.375 1.9375 2.9375 3.8125 5.125 7.0625 5.8125 4.875 6.625 3.6875 4.0625 4.25 4.6875 3.6875 2.5 6.8125 3.125 2.4375 4.555288 6 dB 9.0625 9.6875 5.133333 5.375 8.6875 9 6.0625 8.6875 8.6875 5.0625 6.6875 7.5625 6.1875 7.5625 7.3125 9.75 9.3125 6.375 6.375 8.75 7.4375 8.5625 5.1875 7.625 7.75 5.666667 7.444231 12 dB 9.9375 9.9375 9.928571 10 10 9.8125 9.75 10 9.625 10 8.75 9.6875 9.6875 10 8.1875 10 10 10 10 9.9375 10 10 9.875 10 9.6875 9.875 9.79533 63 Table 21: Mean listening difficulty rating as a percentage for each signal-to-noise ratio under the Female Fry condition for each subject Female Fry NaN dB 2.9375 2.125 5.5625 1.125 1 2 1.1875 1.0625 1.0625 1 1.875 1 1.625 1 2.5625 1 1 1.0625 1 1.1875 1.125 2.1875 1.0625 2 1.25 1.0625 1.579327 Sub01 Sub02 Sub03 Sub04 Sub05 Sub06 Sub07 Sub09 Sub10 Sub11 Sub12 Sub13 Sub14 Sub15 Sub16 Sub17 Sub19 Sub20 Sub22 Sub23 Sub24 Sub25 Sub27 Sub28 Sub29 Sub30 Mean -12 dB 4.875 5.75 1.25 2.1875 3.0625 4.375 1.375 2.3125 1.4375 1.625 3.9375 1.75 2.0625 1.375 3.5625 3 2.8125 3 3.3125 5.1875 2 2.75 1.9375 2.4375 1.125 2.3125 2.723558 -6 dB 4.6875 7.6875 1.9375 3.1875 4.25 2.5 2.5 3.4375 1.625 1.4375 2.4375 2.6875 2.3125 4.3125 3.6875 4.375 5.4375 3.3125 3.0625 2.625 2.9375 3.875 1.625 2 2.25 1.8125 3.153846 0 dB 6.6875 8 4.785714 4.875 5.5625 6.266667 4.5625 4.3125 7.5625 2.3125 5.375 4.9375 6.3125 8.466667 5.25 5.125 7.0625 3.625 7.25 6.1875 5.6875 3.3125 2.1875 6.1875 1.875 2.8125 5.253136 6 dB 8.5 9.3125 6.0625 6.1875 8.9375 9 6.75 3.875 9.5625 3.125 5.3125 6.1875 6 10 7.6875 8.125 9.375 5.375 9.6875 9.25 7.625 5.5625 7.0625 9.3125 4.8 7.875 7.328846 12 dB 10 10 9.75 10 10 10 10 9.9375 10 10 9.8125 10 10 10 10 10 10 10 10 10 9.9375 10 10 10 9.6875 9.75 9.956731 64 Table 22: Mean listening difficulty rating as a percentage for each signal-to-noise ratio under the Female No Fry condition for each subject Female No Fry NaN dB 2.1875 3.125 1.125 1 1.0625 1 1.125 1 1 1.125 1 1.5 1.1875 1 1.6875 1.1875 1.3125 1 1.0625 1.1875 1.0625 2.375 1.125 1 1.0625 1.0625 1.290865 Sub01 Sub02 Sub03 Sub04 Sub05 Sub06 Sub07 Sub09 Sub10 Sub11 Sub12 Sub13 Sub14 Sub15 Sub16 Sub17 Sub19 Sub20 Sub22 Sub23 Sub24 Sub25 Sub27 Sub28 Sub29 Sub30 Mean -12 dB 2.0625 5 1.375 2.25 1.625 1.1875 1 1.5 1.125 1.25 1.875 1.5625 2.0625 1.75 2.1875 1.25 2.8125 2.0625 2.75 1.375 1.375 2.3125 1.1875 1.5 1.0625 1.625 1.8125 -6 dB 1.625 6.5625 4.8125 1.875 2.4375 2.25 1.75 1.375 2.1875 1.9375 2.0625 1.75 2.375 3 4.4375 3 3.5 2.0625 1.875 1.25 2.5625 2.6875 1.9375 2.125 1.25 2.1875 2.495192 0 dB 4.875 6.375 2.8 3.1875 5.6875 8.375 2.25 2.4375 5.25 2.875 3.375 4.0625 3.9375 3.9375 5.8125 4.6875 5.3125 3.9375 3.4375 8.5 2.0625 5.75 5.5 2.875 2.625 2.1875 4.312019 6 dB 7.8125 9.5 5.0625 7.125 8.25 8.8125 5.4375 5.625 7.75 4.75 7.25 5 7.1875 10 6.5625 8.3125 8.1875 5.9375 8.125 9.0625 5 5.25 8 5.8125 3.6875 6.625 6.927885 12 dB 10 9.125 10 10 10 10 6.375 10 10 10 9.4375 9.4375 9.4375 10 10 10 10 10 10 10 8.875 10 10 10 9.4375 10 9.697115 65 REFERENCES 66 REFERENCES Abdelli-Beruh, N.B., Drugman, T., & Red Owl, R.H. (2016). Occurrence frequencies of acoustic patterns of vocal fry in American English speakers. Journal of Voice, 30(6), 11-20. Abdelli-Beruh, N.B., Wolk, L., & Slavin, D. (2014). Prevalence of vocal fry in young adult male American English speakers. Journal of Voice, 28(2), 185-190. Anderson, R.C., Klofstad, C.A., Mayew, W.J., & Venkatachalam, M. (2014). Vocal fry may undermine the success of young women in the labor market. PLOS ONE, 9(5). Blomgren, M., Chen, Y., Ng, M.L, & Gilbert, H.R. (1998). Acoustic, aerodynamic, physiologic, and perceptual properties of modal and vocal fry registers. The Journal of the Acoustical Society of America, 103, 2649-2658. Boone, D.R., McFarlane, S.C., Von Berg, S.L., & Zraick, R.I. (2014). The voice and voice therapy (9th ed.). Boston, MA: Pearson Education. Borrie, S.A. & Delfino, C.R. (2017). Conversational entrainment of vocal fry in young adult female American English speakers. Journal of Voice, 31(4), 513.e25-513.e32. Bradlow, A.R., Torretta, G.M., & Pisoni, D.B. (1996). Intelligibility of normal speech I: Global and fine-grained acoustic-phonetic talker characteristics. Speech Communication, 20(3), 255-272. Colton, R.H., Casper, J.K., & Leonard, R. (2006). Understanding voice problems: A physiological perspective for diagnosis and treatment (3rd ed.). Baltimore, MD: Lippincott Williams & Wilkins. Hollien, H. & Michel, J.F. (1968). Vocal fry as a phonational register. Journal of Speech and Hearing Research, 11, 600-604. Hollien, H., Moore, P., Wendahl, R.W., & Michel, J.F. (1966). On the nature of vocal fry. Journal of Speech and Hearing Research, 9, 245-247. Lee, S. (2015). Creaky voice as a phonational device marking parenthetical segments in talk. Journal of Sociolinguistics, 19(3), 275-302. Morton, V. & Watson, D.R. (2001). The impact of impaired vocal quality on children’s ability to process spoken language. Log Phon Vocol, 26, 17-25. Munro, M.J. (2013). Intelligibility. In C.A. Chapelle (Ed.), The encyclopedia of applied linguistics (pp. 1-7). Hoboken, NJ: Blackwell Publishing. 67 Ramig, L.O. (1992). The role of phonation in speech intelligibility: A review and preliminary data from patients with Parkinson’s disease. In R.D. Kent (Ed.) Intelligibility in speech disorders (pp. 119-155). Philadelphia, PA: John Benjamins. Ryalls, J.H. & Liberman, P. (1982). Fundamental frequency and vowel perception. The Journal of the Acoustical Society of American, 72, 1631-1634. Shiavetti, N. (1992). Scaling procedures for the measurement of speech intelligibility. In R.D. Kent (Ed.), Intelligibility in speech disorders (pp. 11-34). Philadelphia, PA: John Benjamins. Titze, I.R. (2000). Principles of voice production (2nd ed.). Salt Lake City, UT: National Center for Voice and Speech. Tjaden, K., Sussman, J.E., & Wilding, G.E. (2014). Impact of clear, loud, and slow speech on scaled intelligibility and speech severity in Parkinson’s disease and multiple sclerosis. Journal of Speech, Language, and Hearing Research, 57(3), 779-792. Watson, P.J. & Schlauch, R.S. (2008). The effect of fundamental frequency on the intelligibility of speech with flattened intonation contours. American Journal of Speech-Language Pathology, 17, 348-355. Whitehead, R.L., Metz, D.E., & Whitehead, B.H. (1984). Vibratory patterns of the vocal folds during pulse register phonation. The Journal of the Acoustical Society of America, 75, 1293-1297. Wolk, L., Abdelli-Beruh, N.B., & Slavin, D. (2012). Habitual use of vocal fry in young adult female speakers. Journal of Voice, 26(3), 111-116. Yorkston, K.M. & Beukelman, D.R. (1984). Assessment of the intelligibility of dysarthric speech. Austin, TX: Pro-Ed. 68