Articulatory–acoustic vowel space: Application to clear speech in individuals with Parkinson's disease

Journal of Communication Disorders 51 (2014) 19–28

Contents lists available at ScienceDirect

Journal of Communication Disorders

Articulatory–acoustic vowel space: Application to clear speech in individuals with Parkinson’s disease Jason A. Whitfield *, Alexander M. Goberman Department of Communication Sciences and Disorders, 200 Health and Human Services Building, Bowling Green State University, Bowling Green, OH 43403, USA

A R T I C L E I N F O

A B S T R A C T

Article history: Received 9 April 2014 Received in revised form 16 June 2014 Accepted 30 June 2014 Available online 12 July 2014

Background/purpose: Individuals with Parkinson disease (PD) often exhibit decreased range of movement secondary to the disease process, which has been shown to affect articulatory movements. A number of investigations have failed to find statistically significant differences between control and disordered groups, and between speaking conditions, using traditional vowel space area measures. The purpose of the current investigation was to evaluate both between-group (PD versus control) and within-group (habitual versus clear) differences in articulatory function using a novel vowel space measure, the articulatory–acoustic vowel space (AAVS). Methods: The novel AAVS is calculated from continuously sampled formant trajectories of connected speech. In the current study, habitual and clear speech samples from twelve individuals with PD along with habitual control speech samples from ten neurologically healthy adults were collected and acoustically analyzed. In addition, a group of listeners completed perceptual rating of speech clarity for all samples. Results: Individuals with PD were perceived to exhibit decreased speech clarity compared to controls. Similarly, the novel AAVS measure was significantly lower in individuals with PD. In addition, the AAVS measure significantly tracked changes between the habitual and clear conditions that were confirmed by perceptual ratings. Conclusions: In the current study, the novel AAVS measure is shown to be sensitive to disease-related group differences and within-person changes in articulatory function of individuals with PD. Additionally, these data confirm that individuals with PD can modulate the speech motor system to increase articulatory range of motion and speech clarity when given a simple prompt. Learning outcomes: The reader will be able to (i) describe articulatory behavior observed in the speech of individuals with Parkinson disease; (ii) describe traditional measures of vowel space area and how they relate to articulation; (iii) describe a novel measure of vowel space, the articulatory–acoustic vowel space and its relationship to articulation and the perception of speech clarity. ß 2014 Elsevier Inc. All rights reserved.

Keywords: Articulatory–acoustic vowel space (AAVS) Parkinson disease Clear speech Vowel space Acoustics Perception

* Corresponding author. Tel.: +1 419 372 2518. E-mail address: [email protected] (J.A. Whitfield). http://dx.doi.org/10.1016/j.jcomdis.2014.06.005 0021-9924/ß 2014 Elsevier Inc. All rights reserved.

20

J.A. Whitfield, A.M. Goberman / Journal of Communication Disorders 51 (2014) 19–28

1. Introduction Parkinson disease (PD) is a progressive neurological disorder that results from degeneration of dopamine-producing cells (Damier, Hirsch, Agid, & Graybiel, 1999; Lang & Lozano, 1998). Movement deficits observed in PD are often attributed to basal ganglia dysfunction resulting from loss of dopaminergic input to the sensorimotor region of the striatum (Damier et al., 1999; Lang & Lozano, 1998; Redgrave et al., 2010). This results in decreased facilitation of cortical motor regions such as the supplemental motor cortex (Alexander, DeLong, & Strick, 1986; Mink, 1996; Redgrave et al., 2010). Hypokinesia is often marked in individuals with PD, and is suggested to result from increased difficulties initiating and properly executing automatic components of sequential behavior (Redgrave et al., 2010). The hypokinesia in PD is associated with a decrement in the velocity and amplitude of habitual movements, and has been shown to affect reaching (e.g., Torres, Heilman, & Poizner, 2011), walking (e.g., Morris, Iansek, McGinley, Matyas, & Huxham, 2005), and writing (e.g., Tucha et al., 2006). For example, individuals with PD exhibit abnormal handwriting kinematics characterized by reduced movement excursions, decreased maximum velocities, and longer movement durations compared to healthy control participants (Broderick, Van Gemmert, Shill, & Stelmach, 2009; Tucha et al., 2006). Similarly, the selfinitiated reaching movements of individuals with PD are characterized by decreased peak velocities and longer movement durations (Torres et al., 2011). Kinematic data collected from speech movements suggest articulatory movements are similarly affected, as individuals with PD exhibit decreased articulatory excursions and lower velocities of labial (e.g., Forrest, Weismer, & Turner, 1989) and lingual movements (e.g., Walsh & Smith, 2012) than control speakers. Because kinematic measurement of articulatory movements can be cumbersome, less feasible for patient populations, and costly to implement (e.g., Stone, 1997; Weismer, Yunusova, & Bunton, 2012), a number of studies have used acoustic measurements to infer articulatory function. Acoustic investigations of articulation often include measures derived from the first two formant frequencies of the vocal tract (F1 and F2, respectively), as relative changes in the formant pattern reflect changes in the degree and location of maximum vocal tract constriction, thereby relating acoustic events to the underlying physiology (Dromey, Jang, & Hollis, 2013; Huber & Chandrasekaran, 2006; Mefferd & Green, 2010; Tasko & Greilick, 2010; Weismer, Yunusova, & Westbury, 2003). A number of static and dynamic formant-derived measures have been used to examine differences in articulatory– acoustic behavior between individuals with and without PD. 1.1. Vowel space area Vowel space area (VSA) is an acoustic metric widely used to quantify articulatory function (e.g., Bradlow, Kraus, & Hayes, 2003; Ferguson & Kewley-Port, 2007; Lam, Tjaden, & Wilding, 2012). Traditional VSA is the two dimensional F1–F2 space formed by the first two formants of the corner vowels. For VSA, F1 and F2 are typically measured at the quasi-steady-state locations of corner vowel productions. The VSA measures have both triangular (e.g., Liu, Tsao, & Kuhl, 2005; Skodda, Visser, & Schlegel, 2011) and quadrilateral (e.g., Goberman & Elmer, 2005; Lam et al., 2012) forms that are reported in the literature. For both variants, the VSA is calculated as the area formed by connecting the corner vowels using the Euclidean distance between each coordinate in F1–F2 space. 1.1.1. VSA in PD and other motor speech disorders A number of investigations have used VSA to examine articulatory function in individuals with neurological motor speech disorders. Specifically, VSA has been used as a measure of vowel production in individuals with PD (e.g., Goberman & Elmer, 2005; McRae, Tjaden, & Schoonings, 2002; Tjaden, Lam, & Wilding, 2013; Tjaden & Wilding, 2004), amyotrophic lateral sclerosis (ALS; Turner, Tjaden, & Weismer, 1995; Weismer, Jeng, Laures, Kent, & Kent, 2001), multiple sclerosis (MS; Tjaden et al., 2013; Tjaden & Wilding, 2004), and cerebral palsy (CP; Liu et al., 2005). Some of these studies have found statistically significant differences between healthy controls and individuals with neurological disorders, while others have found no statistically significant differences. Relative to ALS, a number of studies suggest the speech of individuals with this disorder exhibited a restricted VSA (e.g., Lansford & Liss, 2014; Turner et al., 1995; Weismer et al., 2001). Specifically, a common finding reported in the literature is that the speech of individuals with ALS is characterized by smaller VSA relative to control speakers (e.g., Turner et al., 1995; Weismer et al., 2001). Turner and Tjaden (2000), however, failed to find a significant difference in VSA between neurologically healthy adults and individuals with ALS, although there was a trend for VSA to be reduced in speakers with ALS. Relative to PD, some studies have reported a significantly smaller VSA for individuals with PD when compared to control speakers (Tjaden et al., 2013; Tjaden & Wilding, 2004). Conversely, a number of other studies have reported no statistically significant differences in VSA between Parkinsonian and control speakers (Sapir, Ramig, Spielman, & Fox, 2010; Skodda et al., 2011; Skodda, Gro¨nheit, & Schlegel, 2012; Weismer et al., 2001). For example, Sapir et al. (2010) failed to observe pretreatment differences in vowel articulation as measured by traditional VSA between neurologically healthy controls and individuals with PD, though other acoustic measures differentiated the groups. Another study by Weismer et al. (2001) reported a non-significant trend for decreased VSA in the PD group as compared to the control group, though the PD group was perceived to be significantly less intelligible than control speakers. One possible explanation for the lack of consistent VSA findings is that traditional VSA may not be adequately sensitive to changes in the articulatory behavior, especially in individuals with motor speech disorders. This hypothesis is supported by

J.A. Whitfield, A.M. Goberman / Journal of Communication Disorders 51 (2014) 19–28

21

studies that have used both acoustic VSA and perceptual data to examine articulatory deficits of individuals with motor speech disorders. A number of investigations have shown VSA to be significantly related to speech intelligibility (Ferguson & Kewley-Port, 2007; Turner et al., 1995; Weismer et al., 2001) as VSA has been shown to account for 45% of the variance in intelligibility scores of individuals with ALS (Turner et al., 1995). Other studies, however, suggest VSA may not robustly predict perceptual clarity in individuals with PD and MS, accounting for less than 15% of the variance in scaled intelligibility scores (McRae et al., 2002; Tjaden & Wilding, 2004). Overall, traditional VSA has been associated with mixed findings when attempting to differentiate individuals with neurological disorders from control speakers, and traditional VSA also fails to consistently relate to intelligibility. This disconnect may again arise from inadequate sensitivity of the traditional VSA measures to capture a representative distribution of articulatory movement in specific populations of patients with motor speech disorders. 1.1.2. Speaking condition-related VSA changes in Parkinson disease In addition to examining between group differences in articulation, a number of investigations have used traditional VSA to examined relative changes in articulatory function between speaking conditions in individuals with PD (Goberman & Elmer, 2005; McRae et al., 2002; Tjaden et al., 2013; Tjaden & Wilding, 2004; Sapir et al., 2010). For example, a recent study by Tjaden et al. (2013) examined habitual, clear, loud, and slow, speech in individuals with PD and MS. Across all participants, traditional VSA was reported to be largest in the clear speech condition. Though some studies have found significant differences between speaking conditions (e.g., Tjaden et al., 2013), a number of studies have reported no significant VSA differences between habitual and experimental speaking conditions for individuals with PD (Goberman & Elmer, 2005; McRae et al., 2002; Tjaden & Wilding, 2004). For example, McRae et al. (2002) found trends, but failed to find a statistically significant increase in traditional VSA for individuals with PD between habitual and slow speaking conditions. Similarly, Tjaden and Wilding (2004) found no significant differences in traditional VSA between slow, loud, and habitual speech for individuals with PD, although scaled intelligibility in the loud condition was rated higher than the habitual condition. Goberman and Elmer (2005) also failed to find a significant difference in VSA elicited from clear versus habitual speaking samples. Traditional VSA has not been entirely successful in tracking changes in articulatory movement, although kinematic data suggest that individuals with PD can modulate articulatory movement in cued conditions that elicit clearer speech behavior (Darling & Huber, 2011; Dromey, 2000; Kleinow, Smith, & Ramig, 2001). Therefore, traditional VSA may not possess adequate sensitivity to capture subtle differences or changes in articulatory behavior between and within clinical populations (e.g., Sandoval, Berisha, Utianski, Liss, & Spanias, 2013; Sapir et al., 2010; Skodda et al., 2011). 1.1.3. Alternative vowel space measures A number of studies have developed alternative measures to address potential weaknesses of the traditional VSA measures. For example, the Vowel Articulation Index (VAI; Skodda et al., 2011) and its reciprocal, the Formant Centralization Ratio (FRC; Sapir et al., 2010) both use formant values in an effort to more sensitively examine vowel space. Also, automated connected-speech VSA assessment has been used by Sandoval et al. (2013) in another attempt to improve accuracy of vowel space measurement. Their method calculates the peripheral vowel space area of formant frequency data using a convex-hull algorithm in Matlab (Sandoval et al., 2013). Aside from Sandoval et al. (2013), these efforts at increasing sensitivity of vowel space measurement have continued the practice of examining a limited number of points (Sapir et al., 2010; Skodda et al., 2011). Both of these new vowel measures, the VAI and FRC, along with traditional VSA measures are calculated from the midpoint of the three to four most peripheral vowels, providing a snapshot of one ‘‘steady-state’’ moment in the speech signal. This may explain the failure of VSA to consistently track small changes in vowel articulation. 1.2. Current investigation The aim of the current investigation was to introduce the novel articulatory–acoustic vowel space (AAVS), and examine its ability to characterize the articulatory function in individuals with PD. First, the ability of the AAVS to detect differences in articulatory–acoustic function between individuals with PD and neurologically healthy individuals was tested. Second, the ability of the AAVS to track clarity-related changes in articulatory function of individuals with PD was tested using a clear speech paradigm. 2. Methods 2.1. Participants A total of 22 individuals completed an informed consent process and served as speakers for the current study. The present study examined 12 individuals with PD, including 6 males (mean age = 74.3; SD = 10.7; range = 55–84 years) and 6 females (mean age = 70.8; SD = 7.1; range = 58–77 years). All participants in the PD group were originally recorded by Goberman and Elmer (2005), and the previous and present studies were both approved by the Human Subjects Review Board at Bowling Green State University. All participants in the PD group were diagnosed with idiopathic PD and were on anti-Parkinsonian medication (e.g., Sinemet CR, Artane, Requip, Mirapex, Amantadine). Type and severity of dysarthria were determined by

22

J.A. Whitfield, A.M. Goberman / Journal of Communication Disorders 51 (2014) 19–28

two ASHA-certified Speech–Language Pathologists with prior experience working with motor speech disorders. All participants exhibited hypokinetic dysarthria with 9 of the 12 participants exhibiting mild-to-moderate dysarthria and 3 participants characterized as having severe dysarthria. Participants with PD self-reported no other neurological diagnoses and had not received any speech–language therapy. Ten neurologically healthy older adults (OA), 5 males (mean age = 65.8; SD = 6.1; range = 57–73 years) and 5 females (mean age = 71.8; SD = 8.8; range = 58–81 years), served as a control group for the current investigation. An independent samples t-test confirmed there was no difference in mean age between the PD and control groups [t(21) = 1.054; p > 0.05]. Control participants were Caucasian, non-smokers, free of neurological diagnoses, with no self-reported or observed deficits in speech, language, or hearing. In addition, all participants spoke Standard American English and demonstrated functional paragraph-length reading abilities. In addition to the speakers, a total of 4 female participants (mean age = 22.5; SD = 2.38; range = 21–26 years) completed the informed consent process and participated as listeners for an auditory perceptual rating task. Listeners were free of neurological diagnoses, and self-reported no current or prior deficits in speech or language. Each of the listeners passed a pure-tone hearing screening in both ears for the 250, 500, 1000, 2000, and 4000 Hz frequencies at 25 dB. 2.2. Recording procedures Participants with PD were asked to read a set of stimuli that included the first paragraph of the Rainbow Passage (Fairbanks, 1960). The participants with PD were then asked to repeat the readings, but with the added instruction: ‘‘Produce the items as clearly as possible, as if I am having trouble hearing or understanding you.’’ Speech samples from a larger data set served as control data for the current study. The OA control participants also read a set of stimuli that included the first paragraph of the Rainbow Passage (Fairbanks, 1960). Control participants, however, were not given the clarity instruction and, therefore, only habitual samples were collected. All speech samples were recorded in a quiet room onto a portable digital audio recorder (PD Group: Sony PCM-M1; OA Group: Marantz PMD661; sampling rate = 44.1 kHz) using a table-top microphone (Shure SM-58) at a constant mouth to microphone distance of 15 cm. For the current investigation, the first sentence of the Rainbow Passage was analyzed to examine articulatory–acoustic differences between the habitual speech of individuals with PD and OA controls. In addition, this same reading sample was used to examine the ability of individuals with PD to modulate their speech given a simple clear speech instruction. 2.3. Formant trajectory trace and articulatory–acoustic vowel space The formant trajectory trace (FTT) is a visual history of the formant trajectories of an entire utterance in F1–F2 space (Fig. 1). To obtain the FTT display, Linear Predictive Coding (LPC) values for F1 and F2 were calculated every millisecond, providing a means of displaying a continuous formant trajectory in F1–F2 space across all voiced segments of an entire utterance. For the current study, the rendering of the FTT involved extraction of formants using PRAAT-based LPC analysis (Burg method; window length = 0.05 s; time step = 0.001 s), following concatenation of the voiced portions of the signal (i.e., removal of all aperiodic segments and silent portions of the signal). Based on comparison of PRAAT-based LPC analysis to an overlaid wideband spectrogram, it was noted that the PRAAT-based LPC analysis default settings yielded a number of inaccurate formant values. To ensure that the current analysis was based on accurate first and second formant values, adjustments were made to the formant frequency range and number of formants for each individual sample. After adjustments, any remaining outliers were removed manually, resulting in an accurate F1–F2 trajectory for each sample. The F1 and F2 LPC values for a given utterance were then read into a custom MATLAB script that was created to calculate the density distribution of the FTT data in F1–F2 space (Fig. 1). This custom script calculates density distribution based on an overlaid 625 block grid in the F1–F2 plane. The F1–F2 data were then plotted (black circles), with the scaled density represented in color (blue = low/no density; red = high density). From these data a number of measurements are made to quantify various aspects of the articulatory–acoustic trajectories taken from connected speech. Fig. 1 shows a model FTT

Fig. 1. A model formant trajectory trace (FTT) for the first sentence of the Rainbow Passage produced using habitual speech by an older adult control female participant. Black circles represent F1–F2 trajectories, and background color represents density distribution (blue = low/no density; red = high density). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

J.A. Whitfield, A.M. Goberman / Journal of Communication Disorders 51 (2014) 19–28

23

derived from the first sentence of a Rainbow Passage produced by an older adult female control participant using habitual speech. Notice for this utterance the formant trajectory traverses into the peripheral corner vowel regions, though the highest concentration of formant data is located in a relatively central location in F1–F2 space. 2.3.1. Articulatory–acoustic vowel space The first measure developed from the FTT is the articulatory–acoustic vowel space (AAVS). This space is calculated as the square root of the generalized variance of the F1–F2 data, resulting in an elliptical space representing the average bivariate variability in F1–F2 space. The generalized variance is equivalent to the product of the variance of each dimension and the proportion of the unexplained variance between each variable. The square root of the generalized variance is then taken to provide a measure of variability that corresponds to the bivariate standard deviation of the formant trajectory in the F1–F2 plane. Because the peripheralization of the F1 and F2 values reflects greater deviation of the vocal tract area function from a hypothetical neutral vowel, the AAVS represents a global articulatory–acoustic range of motion or working space used for an entire utterance. 2.4. Perceptual ratings: Preparation of audio samples The first sentence of the Rainbow Passage was used during the perceptual rating task. First, listeners rated paired samples of the clear and habitual stimuli produced by each speaker with PD. Following the rating of the PD sample pairs, listeners rated the speech clarity of the habitual samples produced by the control participants. To eliminate differences in mean intensity, samples were normalized using the root-mean-square method, so all audio files were the same mean amplitude during the listening experiment. To minimize potential order effects, both the speaker presentation order and speech condition order were randomized for the PD samples. Participants did not receive any information on the speakers prior to the perceptual rating task and were not made aware of disordered versus control groups. 2.5. Perceptual rating protocol During the perceptual rating task, individual listeners were seated in a quiet laboratory room in front of a computer and loudspeaker. The volume of the loudspeaker was between 67 dB A and 70 dB A for each listener. For each paired sample, the listeners completed ratings using two 100 mm visual analog scales to rate the clarity of each sample (left anchor = unclear; right anchor = very clear). In addition to the two perceptual ratings, the listeners completed a forced-choice option to indicate which sample was clearest. This was completed for each of the 36 randomly-ordered paired samples. Listeners were permitted to listen to the samples multiple times. To ensure the rating were accurate and valid as a rating of articulatory precision/clarity, listeners were provided with the following verbal and written directions ‘‘. . .Listen to how clearly the vowels and consonants are articulated. Your task is to (1) rate the relative clarity of each sample, and (2) indicate which sample is clearer.’’ Similar directions have been used in other studies examining perception of speech clarity (Searl & Evitts, 2012; Tasko & Greilick, 2010). Perceptual listener reliability was completed to isolate the best listeners to be used in subsequent analyses. A total of four listeners were recruited for the study, and each listener rated a total of 40 samples, including 6 samples repeated for reliability. To be included in the analysis, listeners needed both a forced choice accuracy of 85% and a correlation greater than 0.7 between the original and repeated visual-analog scale rating. Three participants met this criterion and were included in the study. The average Pearson Product Moment Correlation (PPMC) for all original and repeated ratings given by these 3 listeners was 0.79 (range = 0.70–0.86). All subsequent perceptual analyses used the mean rating of these three most reliable listeners. 2.6. Measurement reliability Intra-judge acoustic measurement reliability was completed through a re-analysis of 6 samples (17%) by the original researcher. Inter-judge acoustic measurement reliability was completed by an analysis of the same 6 samples by a second researcher not affiliated with the current study. Based on this PPMC analysis, the inter-judge reliability was 0.95 (p < 0.05) across all measures, and the intra-judge reliability was 0.98 (p < 0.05). 2.7. Statistics To examine between-group differences in articulatory–acoustic behavior between individuals with and without PD a 2  2 Multivariate Analysis of Variance (MANOVA) was completed using both the AAVS measure and the mean listener rating of speech clarity. For this analysis, group and sex were the independent variables. To examine clarity-related changes in articulatory–acoustic behavior produced by individuals with PD, a 2  2 repeated-measures MANOVA was completed using both the AAVS measure and listener rating of speech clarity with speaking condition and sex as the independent variables. Because both listener ratings of speech clarity and the AAVS were measured for all groups and conditions, the degree of correspondence between the perceptual and acoustic findings can be used to confirm the ability of the AAVS to track perceptually relevant change in articulatory clarity.

J.A. Whitfield, A.M. Goberman / Journal of Communication Disorders 51 (2014) 19–28

24

3. Results 3.1. AAVS and perceptual ratings of speech clarity: PD versus control differences Table 1 shows the means and standard deviations of the AAVS measures and perceptual clarity ratings for the male and female speakers in both groups. A MANOVA was completed to determine if there were any AAVS and/or perceptual clarity rating differences between individuals in the PD group and individuals in the neurological healthy control group (habitual productions only). The analysis revealed a significant multivariate main effect for group [F(2,17) = 10.336; p = 0.001; h2 = 0.549] and sex [F(2,17) = 9.482; p = 0.002; h2 = 0.527]. The group by sex interaction failed to reach significance [F(2,17) = 0.491; p = 0.621; h2 = 0.055]. Univariate analysis revealed a significant effect of group for both perceptual rating of speech clarity [F(1,18) = 17.896; p = 0.001; h2 = 0.499] and the novel AAVS [F(1,18) = 5.921; p = 0.026; h2 = 0.248]. In terms of habitual productions, clarity ratings were significantly higher and the AAVS area was significantly larger in the control group compared to the PD group. As expected based on vocal tract size differences, there was a significant main effect of sex for AAVS [F(1,18) = 19.427; p < 0.001; h2 = 0.519]. Specifically, females had significantly larger AAVS values than male speakers. No sex-related differences in perceptual rating of speech clarity were observed [F(1,18) = 1.610; p > 0.05; h2 = 0.082]. In addition, no significant univariate interactions were observed for either dependent variable (p > 0.05). The univariate main effect of group for perceptual rating of clarity and the AAVS measure are shown in Fig. 2. 3.2. AAVS and perceptual ratings: Clarity-related changes in PD To examine the ability of individuals with PD to modulate speech clarity and the utility of AAVS to measure these potential changes, a 2  2 repeated-measures MANOVA was competed. The AAVS and the mean perceptual speech clarity of the first sentence of the Rainbow Passage produced by individuals with PD in habitual and clear conditions were used as the dependent variables. Table 1 shows the means and standard deviations of the AAVS measure and perceptual clarity ratings for individuals with PD in both the habitual and clear speaking conditions. Multivariate analysis revealed a significant main effect for speaking condition [F(2,9) = 21.285; p < 0.001; h2 = 0.825] and sex [F(2,9) = 5.603; p = 0.026; h2 = 0.555]. The condition by sex interaction failed to reach significance [F(2,9) = 1.894; p = 0.206; h2 = 0.296]. Univariate analysis revealed a significant main effect of condition for both perceptual rating of speech clarity [F(1,10) = 19.543; p < 0.001; h2 = 0.662] and the AAVS [F(1,10) = 24.718; p < 0.001; h2 = 0.712]. As a group, the clear speech condition for individuals with PD was rated to be significantly clearer and exhibited a significantly larger AAVS area compared to the habitual speech condition. Indeed all participants exhibited an increase in AAVS between the habitual and

Table 1 Means and standard deviations (SD) of the articulatory–acoustic vowel space (AAVS) and perceptual rating of speech clarity (clarity rating) for males and females in the control group (OA) and in individuals with Parkinson disease (PD). AAVS (kHz2) Female

Male

Female

Mean = 38.45, SD = 5.20

Mean = 64.59, SD = 9.77

Mean = 63.00, SD = 9.92

Mean = 64.60, SD = 17.96

Mean = 26.88, SD = 5.73 Mean = 31.97, SD = 7.81

Mean = 49.33, SD = 21.58 Mean = 61.43, SD = 21.30

Mean = 39.22, SD = 8.13 Mean = 56.83, SD = 14.50

Mean = 49.33, SD = 3.90 Mean = 66.11, SD = 11.55

60

*

*

AAVS (kHz2)

50

Clairty Rang (mm)

OA Group Habitual PD Group Habitual Clear

Clarity rating (mm)

Male

40 30 20 10 0

Habitual OA

Habitual

Clear PD

100 90 80 70 60 50 40 30 20 10 0

*

Habitual OA

*

Habitual

Clear PD

Fig. 2. Estimated Means and standard error for the articulatory–acoustic vowel space (AAVS; left) and perceptual rating of speech clarity (clarity rating; right) for participants in the older adult (OA) control group and Parkinson disease (PD) group. Clarity ratings range from 0 (‘‘unclear’’) to 100 (‘‘very clear’’).

J.A. Whitfield, A.M. Goberman / Journal of Communication Disorders 51 (2014) 19–28

25

Fig. 3. The formant trajectory trace (FTT) for a male speaker (upper panes) and a female speaker (lower panes) from the Parkinson disease (PD) group showing the first sentence of the Rainbow Passage produced using habitual (left panes) and clear (right panes) speaking conditions. Black circles represent F1–F2 trajectories, and background color represents density distribution (blue = low/no density; red = high density). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

clear conditions that ranged from 3 percent to 43 percent (mean increase = 22.45%; SD = 11.28%). Fig. 3 shows both a habitual and clear speaking condition FTT produced by two individuals from the PD group. In addition, a significant main effect of sex was observed for perceptual rating of speech clarity [F(1,10) = 5.831; p = 0.036; h2 = 0.368] and the AAVS [F(1,10) = 8.272; p = 0.016; h2 = 0.453]. Female speakers with PD were rated on average to be clearer and exhibited a larger AAVS values than the male speakers with PD. For both measures, the univariate group by sex interaction was not significant (p > 0.05).

4. Discussion The articulatory–acoustic vowel space (AAVS) was created as an acoustic measure of articulatory range of motion across a continuous speech sample. The current investigation examines the utility of the AAVS in documenting differences in articulatory behavior between individuals with PD and neurologically healthy controls. Additionally, within-group clarityrelated changes were examined in individuals with PD. The AAVS and perceptual measures were coupled to examine both the sensitivity of the AAVS in tracking perceptually relevant changes in articulation, and to evaluate the acoustics of articulatory deficits in PD. 4.1. PD versus controls Acoustic and perceptual data were used to examine articulatory function of the habitual speech of individuals with PD compared to neurologically healthy age-matched controls. The novel AAVS measure detected differences in articulation between the habitual speech of individuals with PD and control participants, and these differences were confirmed by perceptual clarity ratings. Specifically, individuals with PD were perceived by listeners to be significantly less clear than control speakers, and they also exhibited smaller AAVS than healthy controls. These data confirm that the AAVS is sensitive to differences in articulatory behavior between healthy and disordered populations. Since formant vowel area is seen to relate to tongue positioning and globally reflect tongue range of motion (e.g., Maeda & Honda, 1994; Weismer et al., 2001), the AAVS findings show that the habitual speech of individuals with PD is characterized by decreased articulatory range of motion compared to controls. These results are in agreement with a number of studies that have reported acoustic differences in vowel space measures between individuals with PD and neurologically health control speakers (e.g., McRae et al., 2002; Tjaden & Wilding, 2004).

26

J.A. Whitfield, A.M. Goberman / Journal of Communication Disorders 51 (2014) 19–28

Some studies using traditional VSA measures have failed to detect a significant difference between neurologically health controls and individuals with motor speech disorders (e.g., McRae et al., 2002; Skodda et al., 2011; Turner & Tjaden, 2000; Weismer et al., 2001). For example, Weismer et al. (2001) observed no statistically significant differences in traditional VSA between individuals with PD and neurologically health controls, although significant differences in scaled intelligibility ratings were observed. Overall, individuals with PD may be exhibiting subtle acoustic differences in articulation characterized by a more centralized F1–F2 space used rather than a more restricted peripheral vowel space area. This explanation may account for the smaller AAVS values observed in the current study. Traditional measures of VSA may not capture these subtle differences. 4.2. Clarity-related changes in PD In addition to the PD versus control speaker comparison, the current investigation also examined clarity-related changes in articulatory behavior in individuals with PD using both perceptual and acoustic AAVS data. A significant increase in both the perceptual rating of clarity and the novel AAVS measure was observed between the habitual and clear speaking conditions for individuals with PD. The current findings suggest that the AAVS can sensitively track clarity-related changes in articulation for individuals with PD. A number of previous investigations have used traditional VSA to examine clarity-related changes in articulatory function of individuals with motor speech disorders (e.g., Goberman & Elmer, 2005; McRae et al., 2002; Tjaden et al., 2013; Tjaden & Wilding, 2004; Turner et al., 1995; Turner & Tjaden, 2000). Some of these studies have failed to find differences in traditional VSA between speaking conditions for individuals with PD (e.g., Goberman & Elmer, 2005; McRae et al., 2002; Tjaden & Wilding, 2004). Indeed, speech samples used in the current study were originally collected for a previous investigation examining clear speech in PD (Goberman & Elmer, 2005), which failed to find a significant difference in clear and habitual traditional VSA. Overall, the AAVS studied here is shown to be sensitive to within-speaker articulatory clarity changes, while the traditional VSA measure has not been shown to be as sensitive to these changes. 4.3. Articulatory–acoustic deficits in PD Relative to speech motor control in individuals with PD, the current results are in line with kinematic and acoustic data, which suggest individuals with PD exhibit hypokinesia in the articulatory system (e.g., Forrest et al., 1989; Tjaden et al., 2013; Tjaden & Wilding, 2004; Walsh & Smith, 2012). In the current study, the AAVS revealed differences in articulatory behavior between the PD and control groups, suggesting that individuals with PD exhibit smaller articulatory–acoustic range of motion when compared to control speakers. The current data further show that the hypokinetic articulatory behavior of individuals with PD is perceptually salient, and is reflected in the formant trajectories of the acoustic signal. Additionally, these data confirm that individuals with PD can modulate the speech motor system to increase articulatory range of motion and speech clarity when given a simple prompt. Similar findings have been reported in both the speech and limb motor control literature (Bryant, Rintala, Lai, & Protas, 2009; Darling & Huber, 2011; Sadagopan & Huber, 2007; Ringenbach, Van Gemmert, Shill, & Stelmach, 2011). Accordingly, cue-based interventions have been shown to be quite effective in the rehabilitation of gait (e.g., Bryant et al., 2009), manual motor (e.g., Ringenbach et al., 2011), and speech motor impairments (e.g., Fox, Ebersbach, Ramig, & Sapir, 2012) in individuals with PD. Neurological data suggest associative and limbic regions of the basal ganglia are less affected in individuals with PD than the sensorimotor regions, providing a possible neurological explanation for behavioral improvements observed when individuals with PD operate under volitional or cued rather than an habitual mode of control (Redgrave et al., 2010). 5. Limitations The derivation of the AAVS is more labor-intensive than the traditional VSA; however, the current results suggest that the AAVS more robustly tracks articulatory behavior. Although LPC analysis was individually adjusted for each sample, both intra-judge and inter-judge reliability were strong. Additionally, the current study was completed based on only one stimulus phrase, and future research should examine the effect of phonetic makeup of multiple stimuli on AAVS results. Finally, voiceless segments were required to be removed from each sample prior to AAVS analysis. Although a voiced-only stimulus could have decreased the need for editing, semi-automatic extraction of voiced elements would allow that the AAVS to be more feasible for all stimuli. 6. Conclusions The ability of this novel articulatory–acoustic vowel space (AAVS) to detect PD-related differences in articulatory– acoustic behavior was confirmed and supported by perceptual data. More importantly, the AAVS was sensitive to withinperson changes in articulatory–acoustic behavior, as it tracked perceptually salient differences in articulatory clarity in a group of individuals with PD. The AAVS may, therefore, prove useful in tracking changes in speech behavior in other populations of individuals with motor speech impairments. Future investigation should examine the AAVS in other

J.A. Whitfield, A.M. Goberman / Journal of Communication Disorders 51 (2014) 19–28

27

populations and in other contexts such as tracking treatment-related change in articulatory function or disease progression in progressive neurological disorders. Financial and Non-financial Disclosures The authors have no financial or nonfinancial relationships to disclose. Appendix A. Continuing education CE questions (1) (a) (b) (c) (d) (e) (2) (a) (b) (c) (d) (e) (3) (a) (b) (c) (d) (4) (a) (b) (c) (d) (e) (5) (a) (b) (c) (d) (e)

What type of movement deficits are typically observed in Parkinson disease? bradykinesia hypokinesia rigidity tremor all of the above Speech characteristics observed in individuals with dysarthria secondary to Parkinson disease typically include which of the following? monopitch deceased amplitude and velocity of articulatory movements monoloudness accelerated rate all of the above The current study examined which novel measure of vowel space measured from connected speech? the habitual speech of individuals with PD was perceived to be less clear than healthy controls individuals with PD could did not increase speech in the clear condition the habitual speech of control speakers was perceived to be less clear than the speech of individuals with PD no difference in the perception of the habitual speech of individuals with PD when compared to control speakers Relative to the speech of individuals with Parkinson disease, the current study found: individuals with PD an articulatory deficit characterized by restricted range of articulatory motion individuals with PD can increase speech clarity when given a simple prompt the novel AAVS can track changes in articulatory behavior all of the Above none of the above The current study supports which of the following claims: individuals with PD an articulatory deficit characterized by restricted range of articulatory motion individuals with PD can increase speech clarity when given a simple prompt the novel AAVS can track changes in articulatory behavior all of the Above none of the above.

References Alexander, G. E., DeLong, M. R., & Strick, P. L. (1986). Parallel organization of functionally segregated circuits linking basal ganglia and cortex. Annual Review of Neuroscience, 9(1), 357–381. Bradlow, A. R., Kraus, N., & Hayes, E. (2003). Speaking clearly for children with learning disabilities: Sentence perception in noise. Journal of Speech, Language and Hearing Research, 46(1), 80–97. Broderick, M. P., Van Gemmert, A. W., Shill, H. A., & Stelmach, G. E. (2009). Hypometria and bradykinesia during drawing movements in individuals with Parkinson’s disease. Experimental Brain Research, 197(3), 223–233. Bryant, M. S., Rintala, D. H., Lai, E. C., & Protas, E. J. (2009). An evaluation of self-administration of auditory cueing to improve gait in people with Parkinson’s disease. Clinical Rehabilitation, 23(12), 1078–1085. Damier, P., Hirsch, E. C., Agid, Y., & Graybiel, A. M. (1999). The substantia nigra of the human brain II. Patterns of loss of dopamine-containing neurons in Parkinson’s disease. Brain, 122(8), 1437–1448. Darling, M., & Huber, J. E. (2011). Changes to articulatory kinematics in response to loudness cues in individuals with Parkinson’s disease. Journal of Speech, Language and Hearing Research, 54(5), 1247–1259. Dromey, C. (2000). Articulatory kinematics in patients with Parkinson disease using different speech treatment approaches. Journal of Medical Speech Language Pathology, 8(3), 155–162. Dromey, C., Jang, G. O., & Hollis, K. (2013). Assessing correlations between lingual movements and formants. Speech Communication, 55, 315–328. Fairbanks, G. (1960). The rainbow passage. Voice and articulation drillbook (pp. 124–139). New York, NY: Harper & Row. Ferguson, S. H., & Kewley-Port, D. (2007). Talker differences in clear and conversational speech: Acoustic characteristics of vowels. Journal of Speech, Language and Hearing Research, 50(5), 1241–1255. Forrest, K., Weismer, G., & Turner, G. S. (1989). Kinematic, acoustic, and perceptual analyses of connected speech produced by Parkinsonian and normal geriatric adults. The Journal of the Acoustical Society of America, 85(6), 2608–2622.

28

J.A. Whitfield, A.M. Goberman / Journal of Communication Disorders 51 (2014) 19–28

Fox, C., Ebersbach, G., Ramig, L., & Sapir, S. (2012). LSVT LOUD and LSVT BIG: Behavioral treatment programs for speech and body movement in Parkinson disease. Parkinson’s Disease, 2012. Goberman, A. M., & Elmer, L. W. (2005). Acoustic analysis of clear versus conversational speech in individuals with Parkinson disease. Journal of Communication Disorders, 38(3), 215–230. Huber, J. E., & Chandrasekaran, B. (2006). Effects of increasing sound pressure level on lip and jaw movement parameters and consistency in young adults. Journal of Speech, Language and Hearing Research, 49(6), 1368–1379. Kleinow, J., Smith, A., & Ramig, L. O. (2001). Speech motor stability in IPD: Effects of rate and loudness manipulations. Journal of Speech, Language and Hearing Research, 44(5), 1041–1051. Lam, J., Tjaden, K., & Wilding, G. (2012). Acoustics of clear speech: Effect of instruction. Journal of Speech, Language and Hearing Research, 55(6), 1807–1821. Lang, A. E., & Lozano, A. M. (1998). Parkinson’s disease. New England Journal of Medicine, 339(15), 1044–1053. Lansford, K. L., & Liss, J. M. (2014). Vowel acoustics in dysarthria: Speech disorder diagnosis and classification. Journal of Speech, Language and Hearing Research, 57(1), 57–67. Liu, H. M., Tsao, F. M., & Kuhl, P. K. (2005). The effect of reduced vowel working space on speech intelligibility in Mandarin-speaking young adults with cerebral palsy. The Journal of the Acoustical Society of America, 117, 3879–3889. Maeda, S., & Honda, K. (1994). From EMG to formant patterns of vowels: The implication of vowel spaces. Phonetica, 51(1–3), 17–29. McRae, P. A., Tjaden, K., & Schoonings, B. (2002). Acoustic and perceptual consequences of articulatory rate change in Parkinson disease. Journal of Speech, Language and Hearing Research, 45(1), 35–50. Mefferd, A. S., & Green, J. R. (2010). Articulatory-to-acoustic relations in response to speaking rate and loudness manipulations. Journal of Speech, Language and Hearing Research, 53(5), 1206–1219. Mink, J. W. (1996). The basal ganglia: Focused selection and inhibition of competing motor programs. Progress in Neurobiology, 50(4), 381–425. Morris, M., Iansek, R., McGinley, J., Matyas, T., & Huxham, F. (2005). Three-dimensional gait biomechanics in Parkinson’s disease: Evidence for a centrally mediated amplitude regulation disorder. Movement Disorders, 20(1), 40–50. Redgrave, P., Rodriguez, M., Smith, Y., Rodriguez-Oroz, M. C., Lehericy, S., Bergman, H., & Obeso, J. A. (2010). Goal-directed and habitual control in the basal ganglia: Implications for Parkinson’s disease. Nature Reviews Neuroscience, 11(11), 760–772. Ringenbach, S. D., Van Gemmert, A. W., Shill, H. A., & Stelmach, G. E. (2011). Auditory instructional cues benefit unimanual and bimanual drawing in Parkinson’s disease patients. Human Movement Science, 30(4), 770–782. Sadagopan, N., & Huber, J. E. (2007). Effects of loudness cues on respiration in individuals with Parkinson’s disease. Movement Disorders, 22, 651–659. Sandoval, S., Berisha, V., Utianski, R. L., Liss, J. M., & Spanias, A. (2013). Automatic assessment of vowel space area. The Journal of the Acoustical Society of America, 134(5), EL477–EL483. Sapir, S., Ramig, L. O., Spielman, J. L., & Fox, C. (2010). Formant centralization ratio: A proposal for a new acoustic measure of dysarthric speech. Journal of Speech, Language and Hearing Research, 53(1), 114–125. Searl, J., & Evitts, P. M. (2012). Tongue-palate contact pressure, oral air pressure and acoustics of clear speech. Journal of Speech, Language and Hearing Research, 56(3), 826–839. Skodda, S., Gro¨nheit, W., & Schlegel, U. (2012). Impairment of vowel articulation as a possible marker of disease progression in Parkinson’s disease. PLoS One, 7(2), e32132. Skodda, S., Visser, W., & Schlegel, U. (2011). Vowel articulation in Parkinson’s disease. Journal of Voice, 25(4), 467–472. Stone, M. (1997). Laboratory techniques for investigating speech articulation. In W. J. Hardcastle, J. Laver, & F. E. Gibbon (Eds.), The handbook of phonetic sciences (pp. 11–32). Oxford, UK: Blackwell Publishing Ltd. Tasko, S. M., & Greilick, K. (2010). Acoustic and articulatory features of diphthong production: A speech clarity study. Journal of Speech, Language and Hearing Research, 53(1), 84–99. Tjaden, K., & Wilding, G. E. (2004). Rate and loudness manipulations in dysarthria: Acoustic and perceptual findings. Journal of Speech, Language and Hearing Research, 47(4), 766–783. Tjaden, K., Lam, J., & Wilding, G. (2013). Vowel acoustics in Parkinson’s disease and multiple sclerosis: Comparison of clear, loud and slow speaking conditions. Journal of Speech, Language and Hearing Research. http://dx.doi.org/10.1044/1092-4388 Torres, E. B., Heilman, K. M., & Poizner, H. (2011). Impaired endogenously evoked automated reaching in Parkinson’s disease. The Journal of Neuroscience, 31(49), 17848–17863. Tucha, O., Mecklinger, L., Thome, J., Reiter, A., Alders, G. L., Sartor, H., & Lange, K. W. (2006). Kinematic analysis of dopaminergic effects on skilled handwriting movements in Parkinson’s disease. Journal of Neural Transmission, 113(5), 609–623. Turner, G. S., & Tjaden, K. (2000). Acoustic differences between content and function words in amyotrophic lateral sclerosis. Journal of Speech, Language and Hearing Research, 43(3), 769–781. Turner, G. S., Tjaden, K., & Weismer, G. (1995). The influence of speaking rate on vowel space and speech intelligibility for individuals with amyotrophic lateral sclerosis. Journal of Speech, Language and Hearing Research, 38(5), 1001–1013. Walsh, B., & Smith, A. (2012). Basic parameters of articulatory movements and acoustics in individuals with Parkinson’s disease. Movement Disorders, 27(7), 843–850. Weismer, G., Jeng, J. Y., Laures, J. S., Kent, R. D., & Kent, J. F. (2001). Acoustic and intelligibility characteristics of sentence production in neurogenic speech disorders. Folia Phoniatrica et Logopaedica, 53(1), 1–18. Weismer, G., Yunusova, Y., & Bunton, K. (2012). Measures to evaluate the effects of DBS on speech production. Journal of Neurolinguistics, 25(2), 74–94. Weismer, G., Yunusova, Y., & Westbury, J. R. (2003). Interarticulator coordination in dysarthria: An X-ray microbeam study. Journal of Speech, Language and Hearing Research, 46(5), 1247–1261.