Phonatory and Articulatory Changes Associated With Increased Vocal Intensity in Parkinson Disease: A Case Study

Antonia B. Johnson Wilbur James Gould Voice Research Center The Denver Center for the Performing Arts This study examined changes in voice and speech production in a patient with Parkinson disease as he increased vocal intensity following 1 month of intensive voice treatment. Phonatory function and articulatory acoustic measures were made before and after treatment as well as 6 and 12 months later. Preto post-treatment increases were documented in sound pressure level in sustained phonation, syllable repetition, reading, and monologue. Consistent with mechanisms of intensity change reported in normal speakers, corresponding improvements were measured in estimated subglottal pressure, maximum flow declination rate, laryngeal airway resistance, open quotient, EGGW-25, harmonic-spectral slope, and maximum vowel duration. Measures of phonatory stability in sustained phonation and semitone standard deviation in reading and speaking showed changes accompanying increased vocal intensity. In addition, changes were measured in articulatory acoustic parameters (vowel and whole word duration, transition duration, extent and rate, and frication duration and rise time) in single-word productions. These findings indicate that this patient increased his vocal intensity using phonatory mechanisms that have been associated with the nondisordered larynx. In addition, the increased vocal intensity led to changes in articulation that were not targeted in treatment.

Increases in vocal intensity also can affect supraglottal articulator movements.Reorganization of articulatory movements and timing relationships of the lips and jaw and changes in acoustic segmental duration have been identified when normal subjects have produced loud speech (Schulman, 1989).Specifically, loud speech (90 dB SPL at 50 cm) elicited jaw openings greater than for normal speech, lip movement increases, and more complete lip closure.Movement changes were accompanied by shorter acoustic durations of intervocalic bilabial stops and longer vowels, resulting in a change in the relative timing of acoustic segments.The increased articulatory movement accompanying loud speech was not always a simple linear amplification of normal articulation, but reflected a more complex goaloriented reorganization of specific movements for the maintenance of phonetic features.For example, even as vowel duration increased, the phonological distinction between duration of long and short vowels and distinctions in vowel height were maintained.This goal-oriented reorganization of articulatory movements and timing relationships is consistent with descriptions in the speech production literature of functional units (e.g., coordinative structures or synergistic actions) governing the spatiotemporal interactions among glottal and supraglottal articulators (e.g., Browman & Goldstein, 1990;Fowler, 1980;Gracco, 1988Gracco, , 1994;;Kelso & Tuller, 1981;Kelso, Tuller, Bateson, & Fowler, 1984).Additionally, this cooperativity among articulators to meet phonetic task demands is consistent with clinical observations that treatment effects seem to generalize across the speech mechanism (Hardy, 1967;Netsell, 1986).Although segmental durations and timing relationships have been studied in a limited manner when normal subjects have produced comfortable and loud speech, they have not been studied following treatment designed to increase vocal intensity.
Although previous investigations have reported increases in the perceptual variable of loudness (Ramig et al., 1991;Ramig et al., 1994;Ramig, Fazoli, Scherer, & Bonitati, 1990), and its acoustic correlate of vocal intensity (Smith et al., in press) following the LSVT in patients with Parkinson disease, no study has evaluated these post-treatment changes relative to the mechanisms of intensity control that have been observed in normal speakers.An evaluation of the phonatory mechanisms associated with post-treatment changes in intensity would offer insights into the capacity of the Parkin-son disease patient to voluntarily regulate intensity as well as reveal details of compensatory vocal function.
Furthermore, given the displacement and velocity changes accompanying loud speech in normal speakers (Schulman, 1989), an articulatory acoustic analysis of the Parkinson patient's speech would offer insight into modifications in articulatory patterns concomitant with increased vocal intensity.Because of increased jaw displacement and the associated vocalic durational changes, second formant trajectory differences in transition duration and extent would be expected.The predicted increases in these parameters are anticipated because the articulators have more time to reach their target positions.Velocity increases in lip and jaw during loud speech (Schulman, 1989) are expected to be accompanied by increased velocity of tongue movement as reflected in transition rate.This is anticipated because of research by Munhall, Ostry, and Parush (1985), which indicated a uniform basis for the temporal coordination of speech articulators.Decreased frication duration and rise time and an increased ratio of vowel duration to whole-word duration would be expected to reflect improvements in laryngeal valving efficiency and accompanying oral/laryngeal interarticulator coordination for consonants and vowels (Abramson, 1977;Lindqvist, 1972;Lofqvist & Yoshioka, 1981).Knowledge of the phonatory and articulatory patterns associated with intensity change in Parkinson disease following treatment may contribute useful information to theories of voice and speech production as well as enhance the efficacy of treatment.

Subject
The subject was selected from a larger group of patients with Parkinson disease who were participants in an investigation of voice treatment efficacy.He was selected for this case study because he was representative of early-stage Parkinson disease and his livelihood and daily living were dependent upon his oral communication.In addition, the validity, reliability, and scope of his multichannel experimental data were considered excellent.At the time the study began, he was 49 years old and employed as a family physician.He had been diagnosed with idiopathic Parkinson disease 2 years earlier and was in Stage II on the Hoehn and Yahr (1967) Scale.His Parkinson disease medications included Sinemet and Eldepryl.There was no change in medication during the course of the study, and neurological testing showed no progression of Parkinson symptoms.Neuropsychological testing revealed some possible mild attentional difficulties at the start of the investigation, with no progression of symptoms during the year of the study.
Initial clinical speech examination revealed a normal oral peripheral mechanism both in structure and function; hearing was within normals limits.At the initial speech evaluation, the patient reported that his voice had become softer during the previous year and was raspy at times.He had associated his raspy voice with frequent upper respiratory infections, but it was his impression that people could understand him most of the time.

Equipment
Multiple simultaneous signals were stored on a Sony PC-108M 8-channel digital audiotape (DAT) recorder.In addition, microphone and EGG signals were recorded on a Panasonic SV 3700 2-channel DAT recorder, which allowed higher bandwidth storage for subsequent acoustic analysis.Digitization was performed using a 16-bit Digital Sound Corporation A/D converter to a VAX 4000/200 computer and a Data Translation DT 2821 to a 486 PC.
A sound level meter (Bruel and Kjaer Type 2230) was positioned at 50 cm and a head-mounted microphone (AKG C410) at 8 cm from the subject's lips.A Synchrovoice Research Electroglottograph (EGG) was used to obtain the electroglottographic signal.A Rothenberg mask (Glottal Enterprises MS 100-A2) was held on the subject's face to collect the oral air flow signal.An intraoral pressure tube rested in the center of the oral cavity to allow the estimation of subglottal air pressure during /p/ closure.A pair of strain-gauge belt pneumographs (Respigraph-NIMS PN SY03) were used to measure rib cage and abdomen circumferential movement (Murdoch, Chenery, Bowler, & Ingram, 1989).A Collins wet spirometer (Model RS 2785) was used to measure forced vital capacity.Endoscopic examination was conducted with both an Olympus ENF-P3 fiberscope and Nagashima SFT-1 70 rigid telescope.

Procedure
Experimental data were collected twice within the week preceding treatment and twice during the week following treatment, and once 6 and 12 months later.The following procedure was carried out for each data collection session.After the respigraph bands were taped in position, the subject was seated in a medical examining chair in an IAC sound-treated booth.To limit extraneous movement, the subject's arms and legs were secured to arm and foot rests using 3-inch wide Velcro bands.After 2 minutes of tidal breathing, forced vital capacity (FVC) was measured.The subject was asked to take his deepest breath and blow out "as hard and fast and long as you can."This task was repeated three times at the beginning of the session and twice at the end.The best performance was taken as the FVC.The subject was asked to "inhale and exhale" using a 700 mL Inspirese bag.Care was taken to ensure that the subject's lips were tightly sealed on the mouthpiece for both the FVC and Inspirese tasks; to prevent any nasal air flow, nose clips were used for both procedures.
To determine the maximum duration of sustained vowel phonation, the subject was instructed to "take a deep breath and sustain /a/ for as long as you can."A timer with a second hand was within view of the subject, and he was encouraged to monitor his performance and sustain phonation maximally for each vowel.No instructions were given regarding loudness level.This task was repeated four times at the beginning of the recording session and twice at the end of the session.For analysis of respiratory excursions associated with vowel prolongations, the interval between vowels included a sigh and three to five tidal breaths.Because of potential instabilities in measures of maximum performance associated with lack of consistently high effort on the part of the subject (Kent, Kent, & Rosenbek, 1987), as well as the variability in the performance of Parkinson disease patients (Canter, 1965;King, Ramig, Lemke, & Horii, 1994), the experimenter (LR) was careful to elicit consistent high effort from the subject for each trial in each recording session as determined by her clinical judgment.
For the collection of air flow and intraoral air pressure data, the subject repeated three times a series of seven /pae/ syllables with the Rothenberg mask held in place by the experimenter (LR) and the air pressure tube in the middle of the oral cavity.Syllables were produced at normal pitch and loudness and flat intonation at a rate of 1.5 syllables per second, as modeled by the experimenter (LR).Samples of reading and spontaneous speech were obtained by asking the subject to read aloud the "Rainbow Passage" at a comfortable rate and loudness and to generate a 30-second monologue on a topic of his choice.
To gather data for the analysis of articulatory acoustics, the subject was asked to read individual words "as clearly as possible" from a series presented individually by videotape at approximately 5-second intervals.This style of presentation was used in order to control for changes in speaking rate.The 70 words in the test were all monosyllabic; data from 12 of the words will be described here.These words are wax, sigh, sip, ship, sew, hold, row, cash, hail, ate, shoot, and blend.The words were selected for analysis because they represent various patterns of F2 trajectories (e.g., gradually rising, sharply falling, etc. -Ansel & Kent, 1992;Kent et al., 1989;Weismer et al., 1992).
Laryngeal imaging and videolaryngostroboscopic examination were conducted using well-documented techniques (Bless, Hirano, & Feder, 1987).These data were collected within a week of the experimental voice recordings, but not on the same day.

Treatment
The subject, who had received no previous speech treatment, participated in 16 sessions of the Lee Silverman Voice Treatment within a 4-week period.The treatment was designed to increase vocal intensity by increasing phonatory effort, vocal fold adduction, and respiratory support.Treat-ment sessions included maximum phonatory effort tasks, such as drills on maximum duration /a/ phonation and maximum phonation frequency range for half of the session.The subject was encouraged to use this increased phonatory effort in louder speech production during the other half of the session.The details of this treatment approach have been summarized elsewhere (Ramig, 1995).It is important to point out that no effort in treatment was directed toward changing the subject's speech rate or improving his articulation.In the course of practicing loud voice, the subject may have spontaneously changed his rate or readjusted his articulation to preserve phonetic distinctions, but during treatment his attention was directed exclusively to increasing vocal loudness.
The clinician (AAP) reported that the subject was a cooperative and motivated patient once he became convinced that a louder voice would help people understand him.He attended all sessions and followed through on homework assignments.After the initial 16 sessions, the subject received no additional voice treatment.

Data Analvsis
The goal of this study was to evaluate the mechanisms associated with changes in vocal intensity in a patient with Parkinson disease.The choice of variables and measurements for this study was guided by reports of the mechanisms of intensity change in nondisordered speakers.Because of the physiological deficits that accompany Parkinson disease and the potential performance variability associated with the disorder (e.g., tremor, fatigue, cognitive limitations) as well as the response to neuropharmacological treatment (dyskinesia), valid and representative measurement of voice and speech production may be challenging.Application of any single measure that has previously been employed in nondisordered speakers may prove insufficient or not be feasible in this population.Therefore the simultaneous measurement of several apparently correlated variables as carried out here may be the most useful when documenting mechanisms of change.If parallel changes are observed in correlated variables, the interpretation of data may be conducted with greater confidence.

Laryngeal Measures
Acoustic variables.SPL.The signals from the sound level meter recorded during sustained phonation, reading, and monologue were digitized at 1 kHz and analyzed by custom software to derive the mean and standard deviation of intensity.In order to eliminate the contribution of any silent segments to the SPL analysis for running speech, a cursor system was employed to set a floor criterion corresponding to the lowest level of SPL during any phrase in the recording.The data points from the sound level meter that fell above this level were analyzed for central tendency and variation.Analyses were carried out on a VAX 4000/200 computer.
F o and STSD.Mean fundamental frequency and fundamental frequency variability during reading and monologue were obtained after digitizing the microphone recordings at 5 kHz and analyzing them with CSpeech on a 486 computer.
Phonatory stability and adduction.Measures representing the cycle-to-cycle stability of the voice were calculated after digitizing the sustained vowel phonation at 20 kHz.These digitized signals were analyzed with GLIMPES software (Titze, 1984) on a VAX 4000/200 computer to obtain the following measures: fundamental frequency, jitter and shimmer (with and without linear trend removed), coefficient of variation for frequency, coefficient of variation for amplitude, harmonic spectral slope, and harmonics-to-noise ratio.The electroglottographic (EGG) signal was analyzed with an in-house software program on a VAX 4000/200 computer to derive a measure called EGGW-25.This measure is based on the relative width of the EGG duty cycle at 25% of its height, and has been found to correlate with other measures of glottal adduction (Scherer & Vail, 1988).
Aerodynamic variables.Simultaneous recordings of air flow, intensity, and intraoral air pressure were digitized at 20 kHz and analyzed with custom software that interpolated between the pressure peaks during /p/ closure to allow the estimation of mid-vowel subglottal pressure.Flow and SPL values for their respective time-aligned channels were also obtained in this way, so that values for these measures represented the temporal midpoint of the vowel in each syllable.These analyses were performed on a VAX 4000/200 computer.
Laryngeal airway resistance was calculated by dividing the estimated mid-vowel subglottal pressure by the mean mid-vowel air flow.To measure maximum flow declination rate, the air flow signal was transferred to a 486 PC and was inverse-filtered with CSpeech 4.0.The maximum flow declination rate was measured as the magnitude of the downgoing peak from the derivative of the glottal flow signal for 10 successive cycles at the vowel midpoint.
Open quotient was measured using a custom software program (Stathopoulos & Sapienza, 1993a, 1993b) at a 20% AC flow criterion level.Between 30 and 50 consecutive cycles were measured from the glottal flow waveform at the vowel midpoint during the /pae/ task.The mean values from three vowels from each of three trials for each session were calculated.
Videostroboscopy.The videostroboscopic images were rated for the type of glottal closure (e.g., normal, bowed, posterior gap), as well as the degree of glottal incompetence and supraglottal hyperfunction.This latter rating was a judgment of the degree to which the true vocal folds were obscured by ventricular fold adduction or by anterior-posterior shortening of the glottis during phonation.The ratings were done as part of a larger study (Smith et al., in press) by four trained judges (three speech pathologists and an otolaryngologist).Randomized images were presented without an audio channel and were rated blindly by the judges.

Respiratory Measures
Lung volume.Measures of lung volume were derived from the sum of the signals from the abdominal and rib cage bands, which was digitized at 300 Hz.Calibration values were derived from the voltage of these signals during the 700 cc Inspirese exchange task.Initiation and termination volumes were measured for the phonation tasks relative to resting expiratory level as established during tidal breathing before each task.These analyses were performed using CSpeech.

Articulatory Acoustic Measures
Measures of articulatory acoustic variables were made after digitizing the 12 selected test words at 20 kHz on a 486 computer.Wide-band spectrographic (300 Hz) and waveform displays of the digitized signals were generated by CSpeech.
Temporal parameters.Words were segmented and measured according to conventional criteria (Klatt, 1975;Peterson & Lehiste, 1960).Since one of the goals of this study was to investigate evidence of articulatory effects associated with treatment designed to increase vocal intensity, the measurements of frication duration, rise time and vowel duration to whole-word duration ratio were chosen as preliminary indices of laryngeal/oral interarticulatory coordination.These were selected because visual inspection of PD spectrograms in a previous study indicated increased preand post-word aspiration and lengthened frication durations (Johnson, 1993).Pre-and post-word aspiration was included in whole-word duration measures because relative timing variations in laryngeal and oral opening and closing gestures for consonants have been shown to produce contrasts in voicing and aspiration (Lofqvist & Yoshioka, 1981).Although it might have been desirable to quantify the segmental characteristics of voice onset time (VOT) and preand post-word aspiration duration, as well as the interval between /s/ and the following vowel in an /sV/ context, it was reasoned that the selected measures, along with the temporal and trajectory measures already obtained, would contribute to a useful overall description of laryngeal-oral articulatory coordination.
Measures of pre-and post-treatment frication duration and rate of onset of frication noise (rise time) were made using the operational definitions of Howell and Rosen (1983).The rate of onset of frication (rise time) was used as an indicator of changes in the articulatory valving action of the larynx, which is critical for generating early peak glottal opening in voiceless fricatives (Abramson, 1977;Kingston, 1990).Whole-word duration and the ratio of vowel duration to whole word duration (VD/WWD) were measured to provide a basic index of relative timing relations between the larynx and the supraglottal articulators.
Trajectory patterns.Second formant trajectory measures were extracted for each word using the conventional criteria of Weismer et al. (1988).In cases where more than one segment met the criteria of a transition segment within a single trajectory, the steepest slope corresponding to that obtained for normal geriatrics (Weismer et al., 1992) was used.Formant trajectory measures included (a) transition extent (TE), operationally defined as the frequency change along the transitional segment; (b) transition duration (TD), operationally defined as the duration of the transitional segment; and (c) transition rate (TR) or slope, defined as TE/TD.The transition rate can be either positive or negative depending on the direction of F 2 movement.
Temporal data were analyzed for individual words and for groups of words classified as monophthong or diphthong.This classification was chosen because research by Forrest, Weismer, and Turner (1989) has shown that when PD dysarthric speakers are compared to normal geriatric speakers, they tend to have reduced second formant durations and transition extents for complex vowels (which correspond to diphthongs), but not for monophthongal vowels.In this study, the categories of monophthong and diphthong were defined according to Shriberg and Kent (1982).Temporal and trajectory data are presented for these phonemic categories to allow a determination of commonalities in temporal and trajectory characteristics of vowellike sounds produced with a gradually changing vocal tract configuration (diphthong) compared to relatively static vocal tract configuration (monophthong).Note that one of the diphthongs /al/ (sigh) is phonemic and cannot be reduced to a monophthong.The other diphthongs, /el/ (ate) and /oU/ (sew) are nonphonemic and can be reduced to monophthongs.According to Shriberg and Kent (1982), the diphthongal forms /el/ and /oU/ occur most commonly in heavily stressed syllables, whereas the monophthongal forms /e/ and /o/ usually are found in weakly stressed syllables and occur much less frequently than the diphthongal forms.Since all words in this study with nonphonemic diphthongs had one syllable, they have been classified as diphthongs.

Measurement Reliability
Thirty percent of all the phonatory function data on each measure were reanalyzed to assess measurement reliability.A paired t-test revealed no significant differences between original and repeated analyses.Pearson correlation coefficients ranged from 0.998 to 1.000 between the original and the reanalyzed data.
Twenty percent of the segment duration and formant trajectory measurements were repeated to assess judgment reliability.Reanalysis of the vowel segment durations yielded an average measurement difference of 3.7 msec (range 0-10 msec).Remeasurement for whole word durations yielded a measurement difference of 20 msec or less.The difference values do not exceed the differences of interest reported here.
Visual inspection of the data was performed to examine the main trends over time.The application of inferential statistical techniques to single-case studies is debatable (Barlow & Hersen, 1984;Kratochwill & Levin, 1992;McReynolds & Kearns, 1983).The present approach, which incorporated an irreversible treatment, makes the majority of these tests inappropriate and a visual inspection of the data more suitable, particularly since our purpose is to examine and describe changes rather than to generalize treatment results to a population of patients.

Layngeal Measures
Acoustic variables.SPL.Sound pressure level (SPL) data for sustained vowel phonation, vowel midpoint for the /pae/ syllable series, reading, and monologue are presented in Table 1 and plotted in Figure 1.Sound pressure level for vowels made the greatest pre-to post-treatment increase (20 dB), followed by reading (12 dB) and then monologue (5-6 dB).
F o and STSD.Mean fundamental frequency and fundamental frequency variability (STSD) for reading and monologue are presented in Table 2.The mean F o for these tasks increased by approximately 15-20 Hz from pre-to posttreatment, whereas the STSD rose by a little more than half a semitone for reading; the increases for the monologue task were more variable (up to about one semitone).Aside from the 6-month monologue sample, the values for these variables remained above pre-treatment levels through the 12month follow-up.
Phonatory stability and adduction.Acoustic measures from maximum sustained vowel phonation are presented in Table 3. Increases were measured in the duration (15-18 seconds) and sound pressure level (20 dB) of sustained phonation from the pre-to post-treatment condition, which were maintained through the 12-month follow-up.Mean fundamental frequency increased 50 Hz pre-to post-treatment, and at 12 months follow-up was 20-30 Hz above pre-treatment levels.Perturbation measures of amplitude (coefficient of variation of amplitude, and shimmer with and without linear trend) and frequency (coefficient of variation of frequency, and jitter with and without linear trend) showed increased stability pre-to post-treatment, which was maintained throughout the 12-month follow-up.Spectral measures (harmonic spectral slope and harmonics-to-noise ratio) revealed increased high frequency spectral energy and reduced spectral noise pre-to post-treatment that were maintained through the 12-month follow-up.The data on EGGW25 are also presented in Table 3.The values for this variable increased from approximately 0.51 to 0.66 (reflecting more adduction) from pre-to post-treatment and remained above pre-treatment levels (0.59) through the 12month recording.

Aerodynamic variables
The simultaneous aerodynamic and sound pressure level data for the /pae/ syllable repetition task are presented in Table 4, and four of the aerodynamic variables are plotted in Figure 2.Estimated subglottal pressure increased 2-3 cm H 2 0 from the pre-to post-treatment condition and remained about 1 cm H 2 0 above pre-treatment levels through the 12-month follow-up recording.Maximum flow declination rate increased over 300 Us/s from the pre-to the posttreatment condition and remained 150 Us/s above pretreatment levels through the 12-month recording.
Laryngeal airway resistance increased following treatment by 0.015 cm H 2 0/cc/sec and remained above pre-treatment levels through the 12-month follow-up recording.The mean midvowel air flow decreased somewhat following treatment, returning almost to the level of the first pre-treatment recording by the time 12 months had passed.Open quotient, as measured at the 20% AC flow criterion, decreased by approximately 0.1 following the intensive voice treatment and remained below pre-treatment levels through the 12month follow-up.Forced vital capacity data are presented in Table 4 and show no pattern of difference throughout the course of the study.
Videostroboscopy.Pre-treatment, the subject's vocal folds were judged to be bowed by all four judges; posttreatment they were judged to have a posterior glottal gap-a less severe deficit in adduction.On a 5-point scale (1 = normal, 5 = most severe), the degree of glottal incompetence was rated 1.33 pre-treatment and 1.00 posttreatment, for both soft and loud phonation.For follow-up recordings, this measure was 1.00 at 6 months and 1.25 at 12 months.Supraglottal hyperfunction was also reduced in the post-treatment condition.Both the degree of ventricular fold adduction and anterior-posterior shortening of the glottis decreased from 1.50 to 1.00 pre-to post-treatment.At 6 months, both anterior-posterior shortening and false fold hyperadduction were given ratings of 1.50, whereas at 12 months, the measures were 1.00 and 1.75 respectively.

Respiratory Measures
Lung volume.Respiratory kinematic data are presented in Table 5 and show that during maximum duration sustained vowel phonation, the subject consistently exhaled further below resting expiratory level immediately post- treatment when compared to pre-treatment.Six months after treatment, the subject used both higher initiation lung volumes and lower terminations during this task.The large respiratory range used for the maximum phonation task at 6 months suggests that the subject might not have reached his full respiratory range during the FVC task on this occasion.

Articulatory Acoustic Measures
Temporal parameters.Temporal data for word production are presented in Table 6.The data represent the mean segment durations and standard deviations for all words taken together.The mean whole word duration for all words increased by about 100 msec following treatment, but decreased 75-180 msec below pre-treatment levels at the 6-and 12-month follow-up recordings.The average vowel duration for all words also increased by about 100 msec following treatment and, though decreasing from posttreatment levels, remained 20-80 msec above pre-treatment levels through the 6-and 12-month follow-up recordings.Thus, at the 6-and 12-month follow-up recordings, average vowel duration for all words was above pre-treatment levels, whereas whole word duration was below pretreatment levels.
The ratio of vowel duration to whole word duration across the 12 words may provide a rough ordinal index of relative timing relations among speech movement gestures.These ratios increased from .31 and .32before treatment to .39 after treatment and .45 and .47 in the 6-and 12-month follow-up recordings.Increases were primarily due to a decrease in frication duration and a decrease in postvocalic aspiration (which was included in the whole word duration measurement).
Frication duration decreased post-treatment when measurements were averaged across words and when individual words were compared.Mean frication duration decreased from 133 msec to 82 msec for/h/, and from 223 msec to 204 msec for /s,S/.Mean duration of fricatives was longer than the 50-100 msec reported for normal speakers, but similar to mean frication duration reported for cerebral palsied speakers of 218 msec (Ansel & Kent, 1992).
Rise time decreased after treatment from a mean of 83 msec to 57 msec for /h/.Mean rise time change was inconsistent for /s, f/ after treatment, with decreases measured for mean rise time for three out of five words with initial /s, / (sip, ship, shoot).Before treatment, mean rise time for all /s, f/ fricatives was 125 msec (range 109-184 msec); after treatment, mean rise time was 125 msec (range 102-172 msec).For nondisordered speakers, the mean rise time of isolated real-word fricatives in the initial position is reported to be 123 msec (Howell & Rosen, 1983).
Trajectory patterns.Measures associated with second formant trajectories are presented in Table 7.The mean transition duration (TD) for all words together increased over 30 msec from the pre-to the post-treatment condition and    Note.MPT = maximum phonation time (seconds).SPL = mean sound pressure level (dB SPL).F = mean fundamental frequency (Hz).CVA = coefficient of variation for amplitude.Shimmer 1 = amplitude perturbation with linear trend.Shimmer 2 = amplitude perturbation-linear trend removed.CVF = coefficient of variation for frequency.Jitter 1 = frequency perturbation with linear trend.Jitter 2 = frequency perturbation-linear trend removed.HSS = harmonic spectral slope (dB/octave).HNR = harmonics-to-noise ratio (dB).EGGW-25 = EGG pulse width adduction measure using 25% height criterion.maintained a comparable difference through the 12-month recording.Increases in transition duration immediately following treatment were greater for diphthongs (approximately 60 msec increase) than for monophthongs (with no overall increase).In the 12-month follow-up recording, although this difference between diphthongal and monophthongal change became smaller, the mean TD for diphthongs remained above pre-treatment levels.The greatest increase in TD post-treatment and the strongest maintenance of that increase were found for words with vocalic nuclei or off-glides in the high front position (e.g., ship, sigh) and words reflecting lip rounding features (row and sew).
The mean transition extent (TE) of the second formant trajectory for all words taken together increased by 100 Hz following treatment and made further increases in the follow-up sessions.Mean TE values for diphthongs were practically identical to monophthongs before treatment, and both showed similar increases post-treatment.However, in the follow-up conditions, monophthongs increased in mean TE by 60-140 Hz in contrast to diphthongs, which only maintained their previous gains.In general, the greatest gains in TE for individual words were in words containing high front vowels and diphthongs with high front off-glides (e.g., sip, ship, sigh).
In general there was little overall TR change.This was primarily due to increases in TD's being accompanied by increases in TE.However, comparison of individual words across conditions indicates a general trend for increases in TR primarily for words with front monophthongs and for diphthongs with high front off-glides (e.g., sip, ship, blend, wax, sigh, ate).This was due primarily to increases in TE's being considerably greater than increases in TD.

Discussion
The purpose of this study was to document the phonatory mechanism associated with intensity change in a patient with Parkinson disease and to investigate the concomitant effects on articulation.Although the findings reported here are from a single subject, they are consistent with larger group data on several of the variables, including sound pressure level, fundamental frequency variability, estimated subglottal pressure, maximum flow declination rate, maximum duration of sustained vowel phonation, and articulatory acoustic variables (Johnson, 1993;Ramig, Bonitati, Lemke, & Horii, 1994;Ramig, Dromey, & Johnson, 1994).
The sound pressure level data in the present study document increases in intensity in sustained vowels, reading, repetition of the /pae/ series, and monologue that occurred after treatment and which were maintained up to 12 months without additional treatment.These data indicate that the patient was able to voluntarily increase his vocal intensity and to maintain this increase in spite of the physiologic impairment resulting from his neurological condition.Although intensity increases were measured in all speech tasks, the greater improvements for sustained phonation than for the other speech tasks are likely due to the fact that maximum vowel duration and high effort level were drilled extensively during treatment.Furthermore, maximum phonatory effort and increased SPL are more easily achieved in a task that requires constant vocal fold adduction.The reading and monologue tasks, on the other hand, require more dynamic adductory gestures, as well as more attention  Note.VD = vowel duration.WWD = whole word duration.
and concentration, and might therefore have been more difficult to maintain at a high effort level.
To examine the respiratory/phonatory mechanisms responsible for these SPL changes, we considered variables previously associated with vocal intensity: subglottal air pressure and vocal fold adduction.Pre-to post-treatment increases were observed in estimated subglottal air pressure that were maintained through the 12-month follow-up.Sound pressure levels measured during the /pae/ task are consistent with simultaneous estimates of subglottal air pressure.Two factors can be associated with the subject's increased subglottal air pressure.The first would be the increased effort expended by the subject in speech production.Higher lung volumes and greater expiratory force would allow generation of more subglottal air pressure.The respiratory excursion data are consistent with increased respiratory effort.A second factor involved in the maintenance of higher subglottal air pressure during prolonged phonation would be the improved valving efficiency of the larynx.If glottal adduction had not improved, the increased respiratory drive would have been partially wasted because of inefficiency at the glottis.However, the longer and louder sustained /a/ vowels produced by this subject after treatment suggest that this was not the case.Aerodynamic, videolaryngostroboscopic, and electroglottographic data are indicative of post-treatment increases in vocal fold adduction that were maintained through the 12-month follow-up, suggesting that this was an important factor in the increased SPL.
Maximum flow declination rate (MFDR) increased substantially following treatment and remained at higher than pre-treatment levels through the 12-month follow-up.This variable has been found to correlate highly with sound pressure level in nondisordered speakers (Gauffin & Sundberg, 1989) and indicates that the subject was able to approximate the vocal folds with greater speed and shut off the flow more quickly at higher intensities.Although a more rapid approximation of the vocal folds might raise a concern that the treatment has the potential for leading to damaging vocal behavior, the videostroboscopic findings of improved vocal fold adduction and decreased false fold adduction after treatment indicate that this was not the case.The subject was able to increase the excitation of the vocal tract through a more rapid flow shut-off, with improved true vocal fold approximation but without hyperfunctional behavior.These findings suggest that the patient was able to approach a more normal glottal configuration by overcoming the hypoadduction that had been a major contributor to his reduced intensity before treatment.
The open quotient, laryngeal resistance and harmonic-TABLE 7. Mean and standard deviation temporo-spectral measurements for selected second formant trajectories for two repetitions of all 12 words; also for selected categories of words including monophthongs and diphthongs (see Table 6).
spectral slope data are consistent with post-treatment increases in vocal intensity and vocal fold adduction.Smaller open quotients are associated with longer closed phases and increased glottal efficiency (Colton, 1984).The main variable identified in lowering the open quotient is increased glottal adduction (Scherer, 1991).Laryngeal airway resistance was higher following treatment, suggesting that the subject was able to adduct the vocal folds more fully and use the air supply more efficiently.The shallower harmonic spectral slope after treatment is consistent with louder phonation.This observation documents the presence of increased high frequency energy and, when linked with greater power in the fundamental, led to stronger and more resonant phonation.
The acoustic perturbation data indicate that this subject's phonatory stability (in both amplitude and frequency domains) improved with increased sound pressure level.It is reasonable to speculate that increased vocal intensity and adduction contributed to more stable vocal fold oscillation.Since the treatment tasks included 10 to 15 high-effort maximum duration sustained vowel phonations daily, it could be hypothesized that as a result of treatment, the muscles of phonation became stronger and steadier in their function.This is consistent with the suggestion of Orlikoff and Kahane (1991) that when subjects phonate more loudly, there is recruitment of additional motor neurons and a higher frequency of firing in the active motor units, which could result in smoother muscle contraction and lower perturbation measures.Improved perturbation values with increased vocal fold adduction have been reported previously in Parkinson disease patients following treatment (Ramig et al., 1990) and appear to be additional benefits of a simple focus on louder phonation.
The increases in fundamental frequency variability for reading and monologue following treatment may be interpreted in at least two ways.Repeated maximum fundamental frequency range exercises were included as part of the daily phonatory effort treatment.It can be speculated that the cricothyroid muscle activity that was stimulated in these tasks generalized to improved intonation in connected speech.Another explanation is that as the subject's voice became stronger-and thus the effectiveness of his oral communication increased-his attitude and emotional outlook also improved and were reflected in the intonation of his speech.Both of these factors may have played a role in the greater fundamental frequency variability observed in this subject's speech following treatment.The improved intonation in connected speech is indicative of the patient's capacity to compensate for the laryngeal and respiratory muscle rigidity associated with his condition.
The acoustic data on vowel characteristics and second formant trajectories provide additional evidence of a change in the articulatory valving efficiency of the larynx, as well as for modifications to the relative timing of speech segments.Specifically, post-treatment and follow-up data in this study are consistent with evidence of articulatory reorganization found by Schulman (1989) in normal subjects producing loud speech.These included increased average vowel duration for all words accompanied by maintenance of phonological distinctions between durations of long and short vowels.The post-treatment and follow-up increases in the ratio of vowel duration to whole word duration across the 12 words also suggest changes in relative timing relations among speech movement gestures.These increases were primarily due to a decrease in frication duration and a decrease in post-vocalic aspiration.
Another index of increased oral/laryngeal articulatory coordination is found in the post-treatment decrease in mean frication duration for /h,s,J/ and the decrease in /h/ rise time.As described by Lofqvist and Yoshioka (1980), during voiceless fricative production, rapid glottal adjustments allow the high flows that are required for the turbulent noise source.These laryngeal movements are also precisely coordinated with supralaryngeal movements (particularly tongue gestures) to produce specific fricative sounds (Benguerel, Hirose, Sawashima, & Ushijima, 1978).The decreases in post-treatment mean frication duration for all words and mean rise time for 5 of 7 words provided acoustic evidence of more rapid glottal adjustments and may suggest modifications in coordination of the glottal valving gesture with the oral constriction.
Evidence of articulatory modifications accompanying louder phonation is apparent in second formant trajectory changes.As described previously, increases across all words for second formant transition duration, extent, and rate had been predicted because the expected increases in jaw displacement and vowel duration accompanying louder speech would allow more time for supraglottal articulator movement, including the tongue, to reach the target position.As predicted, there were increases in transition extent (TE).However, there was considerably more gain in TE for words containing high front vowels and diphthongs with high front off-glides.Contrary to prediction, there was no overall mean transition duration (TD) increase for monophthongs following treatment.However, mean TD increased by approximately 60 msec for diphthongs.Thus, even though there was more time "available" for supraglottal movement because vowel duration increased for all vowel types, in general the transition duration stayed the same for vocalic nuclei produced with a relatively static vocal tract configuration (monophthong), but increased for vowel production with a gradually changing vocal tract configuration (diphthong).
These acoustic results are a preliminary indication that changes in segmental durations and timing relationships in a subject following treatment to increase vocal intensity may represent reorganization of articulatory movements and timing relationships that deserve further investigation with a larger sample of PD subjects.The post-treatment acoustic data on maintenance of phonetic distinctions between long and short vowels may be indicative of one or both of the following factors.First, the data may reflect goal-oriented reorganization to maintain phonetic distinctions.Second, the data may be indicative of the biomechanical properties accompanying increased jaw displacement combined with the need for a gradually changing or static vocal tract configuration for diphthongs or monophthongs respectively.
Loud speech might be viewed as a naturally occurring scaling transformation, which modifies the activity of all muscles in the articulatory linkage, or coordinative structure, yet also preserves some consistent temporal relationships.Other naturally occurring scaling transformations include changes in speaking rate and degree of prosodic stress (Kelso, Saltzman, & Tuller, 1986).Change in speaking rate has been referred to as one of the major contributors to phonetic modification (Kelso et al., 1986).The present data suggest that scaling changes in intensity could be another cause of articulatory reorganization and resultant phonetic modification.
The findings of this study contribute to our understanding of the mechanism of intensity change in one patient with Parkinson disease.The respiratory and laryngeal changes accompanying treatment designed to increase vocal intensity included increases in vocal fold adduction and subglottal air pressure.These mechanisms, which have previously been observed in normal speakers as they increase intensity, have been documented here in a patient with Parkinson disease.These findings suggest that compensation by behavioral means may modify, in part, the phonatory and articulatory acoustic effects of the physical pathology of this disease.The effect of this treatment on articulation, which was not targeted during treatment, suggests that a focus on improved phonation can have benefits beyond the larynx.

FIGURE 2 .
FIGURE 2. Means and standard deviations for estimated subglottal pressure, maximum flow declination rate, laryngeal airway resistance, and open quotient from three sets of 7 /pae/ syllables per session.

TABLE 1 . Means and standard deviations for sound pressure level (dB SPL at 50 cm) for vowels, syllables, reading, and monologue.
a"n for mean" indicates how many tokens contribute to the mean and standard deviation values reported here.FIGURE 1.

TABLE 2 . Mean fundamental frequency (Hz) and semitone standard deviation for one reading passage and one monologue for each of the six recording sessions.
M (Hz) STSD M (Hz) STSD M (Hz) STSD M (Hz) STSD M (Hz) STSD M (Hz) STSD

TABLE 5 . Respiratory mean initiation and termination values for six sustained vowels during each recording session (liters relative to REL).
Note.Initiations = lung volume relative to REL at which vowel was initiated.Terminations = lung volume relative to REL at which vowel ended.Twelve month respiratory data are not presented because of technical difficulties.