Running Head: ACOUSTIC PARAMETERS OF SPEECH 1 Acoustic Parameters of Speech and Attitudes Towards Speech in Childhood Stuttering: Predicting Persistence and Recovery

The relations between the acoustic parameters of jitter and fundamental frequency and children’s experience with stuttering were explored. Sixty-five children belonging to four talker groups will be studied. Children were categorized as stuttering (CWS) or non-stuttering (CWNS), and were grouped based on their diagnosis of stuttering/not stuttering at two time points in a longitudinal study: persistent stutterers (CWSàCWS), recovered stutterers (CWSàCWNS), borderline stutters (CWNSàCWS), and never stuttered (CWNSàCWNS). The children performed a social-communicative stress task during which they were audio-recorded to provide speech samples from which the acoustic parameters were measured. There were no significant relations between talker group and acoustic parameters, nor were children’s attitudes towards their speech different across talker groups. Therefore, acoustic parameters nor children’s attitudes towards their speech did not determining their prognosis with stuttering. ACOUSTIC PARAMETERS OF SPEECH 3 Stuttering is a type of speech disfluency in which words or phrases are repeated or sounds are prolonged. Yairi (1993) reported that stuttering affects about 1% of adults worldwide. The onset of stuttering typically begins in preschool children, with about 5% of children stuttering at some point in their lives. About three-fourths of these children recover, whereas the remaining children become the 1% of adults who stutter. There is a disproportionate gender difference, with stuttering affecting four times as many boys compared to girls. Some recent research concerned with stuttering has focused on identifying the different factors that may affect recovery, and identifying the differences between children who do (CWS) and children who do not stutter (CWNS). The results of research suggest that stuttering is related to emotion, temperament, and physiological measures of emotional arousal and regulation (such as skin conductance and respiratory sinus arrythmia) (e.g. Jones, Conture, & Walden, 2014; Jones, Conture, Walden, Buhr, Tumanova, & Porges, 2014; Walden, Frankel, Buhr, Johnson, Conture, & Karass, 2012). Walden et al. (2012) suggest that two diatheses (vulnerabilities or predispositions) affect stuttering: an emotion diathesis and a speech-language diathesis. An example of an emotion diathesis would be difficulty with emotional regulation or a difference in emotional reactivity in children who stutter compared to non-stuttering peers. Walden et al. propose that stuttering-type disfluencies occur in situations that are novel or familiar situations that change because of the emotional reaction to the situation. For a speech-language diathesis, an instance in which a person’s is on the spot, like an interview, speech planning, what they are planning on saying, and production, what they actually say, occurs simultaneously and would be a stressor on their expressive language abilities. Therefore, fluency problems in children who stutter may be related not only to the speech-language requirements in a speaking situation but also their emotions and how they feel. ACOUSTIC PARAMETERS OF SPEECH 4 Children’s Attitudes Toward their Speech The KiddyCAT is a questionnaire that asks children yes or no questions about their speech (Vanryckeghem, 2002). The KiddyCAT measures children’s attitudes toward their speech, particularly perceived difficulty with speech (Clark et al., 2012). This measure indicates their awareness of speech difficulty and their attitude towards it. Research has shown that CWS and CWNS differ in their KiddyCAT scores (Clark et al. 2012). KiddyCAT is a self-reported measure, so it shows how children feel about their speech rather than how the adults in their life perceive it. Vanryckeghem, Brutten, and Hernandez (2005) indicate the children as young as three are aware of disfluencies in speech and, by age six, children who stutter have a negative speech attitude compared to fluent peers. De Nil and Brutten (1991) also indicate a difference in attitude towards speech for stuttering and non-stuttering children. Acoustic Parameters of Speech Pittam & Scherer (1993) have shown that researchers can perceive emotion in a voice from recorded samples. Bachorowski & Owren (1995) induced positive or negative emotions through feedback on a task in which participants were supposed to name words from a string of letters; they were placed in either a punishment or reinforcement condition that gave them predetermined feedback regardless of their performance on the task and their voices were recorded in these tasks. In analysis of recorded samples, they found that acoustic parameters (i.e. jitter, fundamental frequency, shimmer) indicate there is a relation between emotion and acoustic parameters in fluent people. Rothkranz, Wiggers, van Wees, & van Vark (2004) found that listeners can hear the level of the stress response a person feels, because of psychological and physiological reactions to stress. These physiological changes include effects on the speech organs, specifically in respiration and muscle tension in the vocal cords for adults (Rothkranz et ACOUSTIC PARAMETERS OF SPEECH 5 al., 2004). Fuller & Horrii (1986) identify that the difference in respiration, the change in diameter of the airway, and the tightening of the vocal cords changes the pitch, called fundamental frequency (F0 ), and causes perturbations in the pitch itself, called jitter, in infants. F0 and jitter are not only useful for adult vocal analysis, but also can be used to measure children’s voices. Protopopas & Lieberman (1997) found a correlation between mean F0 and the amount of stress a person is under for fluent adults—the higher the mean F0 the more stressful the situation. The stressful situation in this instance is fear—the vocal samples were taken from real life helicopter pilot in neutral conversations with the control tower and stressful situations like when the helicopter was going to crash. Additionally, Mendoza and Carballo (1999) found that adults in stressful situations, a situation in which their cognitive load was high because of difficult speaking tasks, experienced an increase in mean F0 and a decrease in jitter. Researchers have found differences in F0 and jitter in children and adults who stutter (e.g. Hall & Yairi, 1992; Subramanian, Yairi, & Amir 2003; Yaruss & Conture, 1993; Zebrowski, Conture, & Cudahy, 1985). Stuttering individuals have smaller increases in mean F0 when speaking than those who do not stutter who experience greater changes in F0 in speaking tasks (Bosshardt, Sappok, Knipschild, & Hölscher, 1997). Adults who stutter show less pitch variability in more emotional circumstances than fluent peers—emotional circumstances in these studies being ones in which extreme states of nervousness are induced through questioning, blindfolding, firing a gun, and administering electrical shocks or by embarrassing the speaker as they were speaking (Travis, 1926; Bryngelson, 1932). Similarly, children who stutter have been found to have lower mean F0 than their fluent peers (Healey, 1982; Natke, Grosser, and Kalveram, 2001). ACOUSTIC PARAMETERS OF SPEECH 6 The results of studies such as those conducted by Subramanian et al. (2003) and Yaruss et al. (1993) have suggested there are differences in acoustic parameters (i.e. jitter, shimmer, F0) of preschool children who do and do not stutter. Hall et al. (1992) only identified differences between CWS and CWNS to be significant in regards to shimmer, which is a random change in the amplitude of speech over time and affects the roughness or harshness of the voice, but identified close to significant differences in jitter and fundamental frequency between these two talker groups, fundamental frequency for stutterers being lower than those who do stutter. Subramanian et al. (2003) took those outcomes a step further and identified a longitudinal difference in the formant frequencies, how much energy is concentrated around a certain frequency in a speech wave, in the children, studying differences second formant transitions in children who persisted in their stuttering and children who recovered from stuttering over the longitudinal study in fluent speech samples, collecting data every six months for several years. This study focuses on the intersection of children’s stress or emotional arousal, children’s attitudes towards their speech, and the acoustic qualities of speech to identify relations between arousal, attitudes, and the speech itself over time. Data were collected during a socialcommunicative stress tasks to measure emotional arousal. This study looked at developmental stuttering through a longitudinal lens. The study focuses on what acoustic parameters might indicate for persistent stutterers (CWSàCWS), recovered stutterers (CWSàCWNS), borderline stutterers (CWNSàCWS), and children who have never stuttered (CWNSàCWNS), as well as children’s attitude towards their stuttering. The hypotheses are as follows: 1. There is a significant difference between the recovered stutterers (CWSàCWNS) and persistent stutterers (CWSàCWS) for mean F0 and jitter during social stress task. ACOUSTIC PARAMETERS OF SPEECH 7 a. Directional: Recovered stutterers (CWSàCWNS) would have higher mean F0 than persistent stutterers (CWSàCWS) at T2. b. Directional: Persistent stutters’ (CWSàCWS) jitter will be lower than recovered stutters (CWSàCWNS) at T2. 2. There is a significant difference between the borderline stutterers (CWNSàCWS group) and the never stuttered (CWNSàCWNS group) for mean F0 and jitter during social stress task. a. Directional: Borderline stutterers (CWNSàCWS) lower mean F0 than never stuttered (CWNSàCWNS) at T2. b. Directional: Jitter lower for borderline stutterers (CWNSàCWS) than never stuttered (CWNSàCWNS) at T2. 3. CWS with higher KiddyCAT scores will have lower F0 and lower jitter for the group. a. Persistent stutterers (CWSàCWS) will have the highest scores. b. Recovered stutterers (CWSàCWNS) will have scores lower than the persisting group. Method Participants Participants were paid participants who are naïve to the hypotheses of this study. They were all participants in a longitudinal study focusing on developmental stuttering, emotion reactivity, and emotion regulation. Besides the participants who stutter, the participants had no known other speech-language, neurological, emotional, hearing, developmental, or intellectual problems. All are monolingual, native speakers of Standard American English. There were 65 participants. At Timepoint 1 the participants ages were 3;0 to 6;4 (years;months; mean, M=4;11, ACOUSTIC PARAMETERS OF SPEECH 8 standard deviation, SD=0;7). At Timepoint 2 the participants ages were 3;8 to 7;0 (years;months, mean, M=4;8, standard deviation, SD=0;7).There were 41 females and 24 males. Participants were referred to the Vanderbilt Bill Wilkerson Center for participation by their caregiver. Caregivers were informed of the study through a free, widely read parent-oriented magazine, local health care provider, or self/professional referral to the Vanderbilt Bill Wilkerson Hearing and Speech Center. The original data set contained 148 participants. 26 were excluded because they were too old. 43 were excluded because they had no workable audio files. 6 were excluded because they did not complete the card stressor task. 8 were excluded because they did not complete the KiddyCAT. Groups. There were 13 participants in three groups, and 26 in one group. These groups are based on classifications of stuttering or non-stuttering at two different time points over the longitudinal study. Children are either categorized as stuttering (CWS) or non-stuttering (CWNS) based on the number of stuttered speech-language disfluencies in a 300-word conversation. Stuttered speech-language disfluencies, according to our lab’s paradigm, are single, whole, and partial word repetitions, phrase repetitions, interjections, revisions, and audible and inaudible sound prolongations. Children with more than three stuttered speechlanguage disfluencies per one hundred words are considered CWS. Depending on their classification at the two timepoints, the children are placed into one group either as persistent stutterers (CWSàCWS), recovered stutterers (CWSàCWNS), borderline stutterers (CWNSàCWS), or never stuttered (CWNSàCWNS). Persistent Stutters (CWSàCWS) Participants were 13 preschool-age children who stuttered at timepoint 1 and at timepoint 2. At timepoint 1, participants were between the ages of 3;0 and 5;1 (years;months, mean, M=3;11, standard deviation, SD=0;6). At timepoint 2, ACOUSTIC PARAMETERS OF SPEECH 9 participants were between the ages of 3;8 and 5;9 (years;months, mean, M=4;8, standard deviation, SD=0;7). The persistent stutters consisted of 10 boys and 3 girls. Recovered Stutters (CWSàCWNS) Participants were 13 preschool-age children who stuttered at timepoint 1 and at timepoint 2. At timepoint 1, participants were between the ages of 3;0 and 5;3 (years;months; mean, M=4;1, standard deviation, SD=0;8). At timepoint 2, participants were between the ages of 3;8 and 5;11 (years;months; mean, M=4;8, standard deviation, SD=0;7). The persistent stutters consisted of 11 boys and 2 girls. Borderline Stutters (CWNSàCWS) Participants were 13 preschool-age children who did not stutter at timepoint 1 and stuttered at timepoint 2. This distinction is more challenging than other groups, because 9 participants eventually recovered, whereas 4 participants are unknown, because they did not complete the study. However, these children experienced the onset of stuttering in the lab, regardless of whether they recovered. Therefore, they are borderline stutterers. Participants at timepoint 1 were between the ages of 3;0 and 6;4 (years;months; mean, M=3;11, standard deviation, SD=0;11). At timepoint 2, participants were between the ages of 3;8 and 7;0 (years;months; mean, M=4;8, standard deviation, SD=0;10). The borderline stutterers consisted of 5 boys and 8 girls. Never Stuttered (CWNSàCWNS) Participants were 26 preschool-age children who did not stutter at timepoint 1 or timepoint 2. At timepoint 1, participants were between the ages of 3;1 and 4;11 (years;months; mean, M=3;11, standard deviation, SD=6). Participants were between the ages of 3;9 and 5;8 (years;months; mean, M=4;8, standard deviation, SD=6). The never stuttered group consisted of 15 boys and 11 girls. ACOUSTIC PARAMETERS OF SPEECH 10