Comparison of Two Forms of Intensive Speech Treatment for Parkinson Disease

Yoshiyuki Horii Department of Communication Disorders and Speech Science University of Colorado-Boulder and The Wilbur James Gould Voice Research Center Denver Center For The Performing Arts This study investigated the effect of two forms of intensive speech treatment, (a) respiration (R) and (b) voice and respiration (Lee Silverman Voice Treatment [LSVT]), on the speech and voice deficits associated with Idiopathic Parkinson disease. Forty-five subjects with Idiopathic Parkinson disease completed extensive pretreatment neurological, otolaryngological, neuropsychological, and speech assessments. All subjects completed 16 sessions of intensive speech treatment, 4 times a week for 1 month. Preand post-treatment measures included intensity and maximum duration during sustained vowel phonation. Intensity, habitual fundamental frequency, fundamental frequency variability, and utterance and pause duration were measured during reading of the "Rainbow Passage" and conversational monologue as well. Family and subject self-ratings were completed preand post-treatment for the perceptual variables loudness, monotonicity, hoarseness, overall intelligibility, and initiation of conversation. Significant preto post-treatment improvements were observed for more variables and were of greater magnitude for the subjects who received the voice and respiration treatment (LSVT). Only subjects who received the LSVT rated a significant decrease post-treatment on the impact of Parkinson disease on their communication. Correlations between descriptive prognostic variables (i.e., stage of disease, speech/voice severity rating, depression, and time since diagnosis) and magnitude of treatment-related change indicated these factors did not significantly predict treatment effectiveness. These findings suggest that intensive voice and respiration (LSVT) treatment, focusing on increased vocal fold adduction and respiration, is more effective than respiration (R) treatment alone for improving vocal intensity and decreasing the impact of Parkinson disease on communication.

Recently, Ramig and colleagues (Countryman & Ramig, 1993;Countryman, Ramig, & Pawlas, 1994;Ramig, 1995;Ramig, Bonitati, Lemke, & Horii, 1994;Ramig, Horii, & Bonitati, 1991;Ramig et al., 1988) have reported positive findings following a voice treatment program for Parkinson disease (The Lee Silverman Voice Treatment [LSVT1) that focuses on increasing vocal loudness by increasing phonatory effort.Subjects have been trained to increase vocal fold adduction in maximum phonatory effort tasks and generalize the increased vocal effort to louder speech.Significant increases in maximum duration sustained vowel phonation, maximum fundamental frequency range, habitual fundamental frequency, and fundamental frequency variability in speech have also been documented following this intensive voice treatment (Ramig et al., 1994).Perceptual reports supported corresponding post-treatment increases in vocal loudness and improvements in speech intelligibility and functional communication (Ramig, 1992).The rationale for the Lee Silverman Voice Treatment is summarized in Table 1.
The interaction of the respiratory and laryngeal systems for control of intensity has been well described in nondisordered speakers (Stathopoulos & Sapienza, 1993a, 1993b).Increases in intensity have been associated with increased subglottal air pressure (Holmberg, Hillman, & Perkell, 1988;Isshiki, 1964) and vocal fold adduction (Colton, 1984;Titze & Sundberg, 1992).Given the significant role of the respiratory system in controlling subglottal air pressure (Hixon, Goldman, & Mead, 1973;Hixon, Mead, & Goldman, 1976;Stathopoulos & Sapienza, 1993a), speech treatment that is designed to improve inspiratory and expiratory respiratory muscle activity should contribute to increased subglottal air pressure and vocal intensity in patients with Parkinson disease.However, given the glottal incompetence frequently observed in patients with Parkinson disease, it is likely that treatment designed to increase vocal fold adduction also would be necessary to generate optimum gains in intensity.For example, Berke et al. (1990) concluded that increasing intensity (in an in vivo canine model) by medial adductory compression was more efficient than by increasing air flow.They reported that increasing air flow alone produced a significantly greater open quotient and vocal fold vibratory excursion.Clarification of the interactive roles of the respiratory and laryngeal systems in treatment-related change will contribute to enhancing the effectiveness of treatment for patients with Parkinson disease.Identification of patient suitability or prognostic factors for treatment success is another issue of importance in studies of speech treatment efficacy.Previous research has not considered characteristics of Parkinson disease patients such as age, sex, stage of disease, time since diagnosis, and severity of speech disorder in relation to the efficacy of intensive speech treatment.In addition, depression and dementia (cognitive ability or functioning), which are frequently observed in Parkinson disease (Duvoisin, 1984;Gauthier, Dalziel, & Gauthier, 1987;Mayeux, Williams, Stern, & Cote, 1986) and associated with performance limitations (Beatty et al. 1989;Cummings, 1986), should be considered for their potential impact on treatment-related change.Although preliminary work (Mead, Ramig, & Beck, 1989) suggested that moderate dementia is not a limiting factor for speech treatment success, neither cognitive status nor depression has been considered systematically for its impact on the efficacy of speech treatment in Parkinson disease.Clarification of the prognostic role of certain patient characteristics will contribute to maximizing treatment effectiveness.
This study was designed to evaluate the speech and voice characteristics of patients with Parkinson disease following one of two forms of intensive treatment: (a) treatment designed to increase respiratory support for speech (R); and (b) treatment designed to increase vocal fold adduction and respiratory support (LSVT).Pre-to post-treatment changes  (Bellaire, Yorkston, & Beukelman, 1986;Crystal, 1969).This study was also designed to document treatment-related changes in relation to subject characteristics including: age, sex, stage of disease, time since diagnosis, severity of speech/voice deficits, glottal incompetence, depression, and cognitive functioning.

Subjects
Forty-five patients with idiopathic Parkinson disease volunteered to participate in this study.All subjects were residents of the Denver, Colorado, area.All subjects participated in baseline pretreatment neurological, neuropsychological, otolaryngological, and speech assessments.Subjects were stratified on the variables age, sex, and stage of Parkinson disease (Hoehn & Yahr, 1967), score on the motor examination section (section III) of the Unified Parkinson's Disease Rating Scale (UPDRS) (Fahn, Elton, & members of the UPDRS Development Committee, 1987), time since diagnosis, magnitude of glottal incompetence, and severity of speech disorder.These stratified subjects were randomly assigned to one of two treatment groups: respiration (R) or voice and respiration (LSVT).
An analysis of variance revealed no significant differences between the two groups of subjects at baseline for the variables age, F(1,43) = .4368,p = .51,stage of disease, F(1,43) = 2.6113, p = .11,score on the motor examination section of the UPDRS, F(1,40) = .0309,p = .86,time since diagnosis, F(1,43) = .6609,p = .42,glottal incompetence, F(1,19) = 1.8274, p = .19,and severity of speech disorder, F(1,43) = .6704,p = .41.Lack of subject compliance with pre-and post-treatment assessments and experimental and treatment protocols resulted in unequal group sizes and gender distribution.Difficulty recruiting female patients resulted in a limited female sample (n = 12).Descriptive characteristics of the 45 subjects obtained during baseline assessments are summarized in Table 2 and details of the baseline assessments are described below.Subjects initiated no other forms of treatment (e.g., physical therapy, occupational therapy, psychological counseling) while they participated in this study.
Patients were excluded from participation in this study if they had a form of Parkinson disease other than idiopathic or a laryngeal pathology that would contraindicate phonatory effort treatment (e.g., vocal nodules, gastric reflux).Voice treatment efficacy data from patients with non-idio-pathic Parkinsonism such as Shy-Drager syndrome, multiple system atrophy, progressive supranuclear palsy and bilateral thalamotomy are summarized elsewhere (Countryman & Ramig, 1993;Countryman et al., 1994).

Baseline Neurology, Neuropsychology, Otolaryngology, and Speech Assessments
Neurological assessment.Routine neurological assessment was carried out on all subjects within 1 month of the initiation of treatment.Stage of Parkinson disease, score on the motor examination section (section III) of the UPDRS, time since diagnosis, and anti-Parkinson medications are summarized in Table 2. Subjects were considered neuropharmacologically stable before and during treatment and did not change medications or dosage during the treatment period.
Neuropsychological assessment.A battery of neuropsychological tests was administered within 1 month of the initiation of treatment to determine status of cognitive functioning in these Parkinson disease subjects.These measures focused on attention and concentration, learning and memory, vocabulary and auditory verbal comprehension, and visual spatial skills.Four subtests from the Wechsler Adult Intelligence Scale-Revised were administered (Vocabulary, Block Design, Object Assembly, Digit Symbol;Wechsler, 1981).In addition, parts A and B of the Trail Making Test (Reitan & Wolfson, 1985), the Digit Vigilance Test (Lewis & Kupke, 1977), Story and Figure Memory Test (Heaton, Chelune, & Lehman, 1978), and the Complex Ideational Material auditory comprehension subtest from the Boston Diagnostic Aphasia Exam (Goodglass & Kaplan, 1972) were included.These measures produced 13 scores, all of which were converted to T scores (mean of 50, standard deviation of 10) that correct for demographic variables of age, education, and sex (Heaton, Grant, & Matthews, 1991).After completion of the cognitive evaluation, subjects were interviewed by the clinical neuropsychologist, who then rated depression using the Montgomery Asberg Depression Rating Scale (MADRS; Montgomery & Asberg, 1979).At this time subjects also completed a self-report depression scale, the Beck Depression Inventory (BDI; Beck, Ward, Mendelson, Mock, & Erbaugh, 1961) as well as the Sickness Impact Profile (SIP; Bergner, Bobbit, Carter, & Gilson, 1981).The SIP instrument assesses the extent to which patients perceive their illness to affect 12 different areas of their lives.For the purposes of this paper, the individual scores for the impact of illness on communication and social interaction were included.Both the BDI and the SIP were completed pretreatment and post-treatment after completion of the last speech treatment session.Pretreatment neuropsychological data for each individual subject as well as group means are summarized in Table 3. Results of analysis of variance revealed no significant differences between treatment groups in cognitive functioning, level of depression, or sickness impact on communication or social interaction at baseline.However, a main effect for gender was revealed for the self-rating of depression, F(1,41) = 7.368, p = .010,as well as clinical rating of depression, F(1,41) = 9.971, p =    .003,with the females (n = 12) being more depressed than the males (n = 33).
Otolaryngological assessment.An otolaryngological history and videolaryngostroboscopy examination were obtained on all subjects prior to beginning treatment.Laryngeal imaging and videolaryngostroboscopy examination were conducted utilizing well-described techniques (Bless, Hirano, & Feder, 1987).The nasal passage was topically anesthetized with 4% lidocaine spray.Endoscopic examination was conducted with both an Olympus ENF-P3 fiberscope and Nagashima SFT-1 70-degree rigid telescope.Light sources included both an Olympus halogen constant light source and a Nagashima LS-3A laryngoscope.The subjects were instructed to say /i/ at a comfortable pitch and loudness.Images were recorded with a CCD camera, using a 35-mm lens for the fiberscope and 60-mm lens for the rigid telescope, and a VHS or SVHS tape recorder.The data for pretreatment glottal incompetence ratings were prepared by creating a "master" study videotape that included segments from the laryngostroboscopic videotapes of each subject.The order of the subject videotapes was determined by a computer random number generator.Six subjects' recordings were repeated to assess intrajudge reliability.The audio signal from each subject tape was removed to eliminate auditory perceptual cues.Instead, a "prompter" was overdubbed onto the audio track information that identified the tasks performed during examination of each patient to allow raters to identify them.Glottal incompetence was rated during comfortable pitch, comfortable loudness /i/.Three speech-language pathologists and one otolaryngologist with experience in voice research assessed the recordings.
They independently and as a group viewed a practice tape, and met several times to develop standard rating criteria.Ninety-five percent intrarater agreement was achieved on the practice tape.Following the practice procedures, they independently viewed and rated the master tape.Twentyfour percent of the glottal incompetence ratings were reanalyzed to determine reliability.Interexaminer reliability ranged between .78 and .94whereas intraexaminer reliability ranged between .87 and 1.0.The magnitude of glottal incompetence pretreatment is summarized in Table 2 for 25 of the 45 subjects.Videostroboscopy ratings could not be obtained on 20 subjects due to their inability to tolerate the procedure, excessive involuntary movement, or technical problems.Results of analysis of variance revealed no significant difference between the two treatment groups in glottal incompetence at baseline.Additional details of pretreatment (Perez, Ramig, Smith, & Dromey, in press) and preto post-treatment laryngological characteristics (Smith et al., in press) have been previously reported.
Speech assessment Standard speech and voice assessments were completed at the time of the first pretreatment data collection session.None of the subjects exhibited oral motor and/or speech and voice characteristics uncommon to Parkinson disease.The primary speech and/or voice characteristics and severity of speech disorder rating for each subject determined by clinical observations are presented in Table 2.
In order to provide additional descriptive information on the subjects and to evaluate the relationship between magnitude of speech and voice disorder and severity of Parkinson disease, correlation coefficients were completed  on these baseline data for all 45 subjects.Severity of speech and voice disorder rated on a 1-5 point scale at the time of clinical intake (pre-treatment) were correlated (Pearson product moment) with the variables time since diagnosis (r = .43,p = .003),rating on the motor examination section of the UPDRS (r = .63,p = .001),Parkinson disease stage (r = .62,p = .001),and depression rating (r = .45,p = .004).These findings suggest that subjects who had Parkinson disease longer and were more motorically impaired also had more severe speech and voice deficits.

Instrumentation
In order to measure intensity, frequency, and duration data, sound pressure level (B & K 2230 sound level meter) and microphone (AKG C410) signals were transduced and recorded onto a Sony PC-108M 8-channel digital audiotape recorder (DAT).Forced vital capacity was measured with a Collins wet spirometer (Model RS 2785).

Tasks
Experimental tasks were designed to sample a broad continuum of speech production to address both underlying mechanisms of treatment-related change as well as functional aspects of speech production.Data were collected while subjects performed the following tasks: forced vital capacity, maximum duration sustained vowel phonation, reading of the phonetically balanced "Rainbow Passage" (Fairbanks, 1960) and 25-30 sec of conversational monologue.Because the sound level meter and microphone offer noninvasive signal transduction, measures of intensity, frequency, and duration could be collected from all subjects.
In order to evaluate the subjects' impressions of treatment effectiveness and the impact of treatment on functional communication, perceptual rating scales were completed by all 45 subjects at the time of data collection.Family members from 20 of the 45 subjects completed the perceptual rating scales as well.

Procedures
Pretreatment experimental data were collected within the week before treatment was initiated.Post-treatment data were collected within the week following treatment.For assessment of intrasubject variability and reliability, 58% of the subjects repeated their pretreatment data collection and 36% of the subjects repeated their post-treatment data collection.For experimental data collection, subjects were seated in a medical examining chair located in an IAC sound-treated booth and transducers were positioned and calibrated using standard procedures.The sound level meter was positioned 50 cm from the lips and the microphone was positioned on a head mount 8 cm from the lips.A calibration tone of known intensity was generated at the subjects' lips at the beginning and end of each recording session.
After 2 minutes of tidal breathing, forced vital capacity was measured.Subjects were asked to take their deepest breath and blow out "as hard and fast and long as you can."This task was repeated two or three times at the beginning and end of the recording session and the best performance was taken as maximum forced vital capacity.Because of motor and balance problems of some subjects, all subjects performed the task in a seated position.One subject could not perform the task preor post-treatment.
To obtain maximum duration sustained vowel phonation, subjects were instructed to "take a deep breath and sustain 'ah' for as long as you can."A timer with a second hand was within view of the subjects and they were encouraged to monitor their performance and sustain phonation for a longer duration each time.No instructions were given for loudness level.This task was repeated four times at the beginning of the recording session and two times at the end of each recording.For a few subjects, only three of the initial four consecutive maximum duration sustained phonations were possible, due to fatigue.Because of potential instabilities in measures of maximum performance associated with subjects' inconsistent generation of maximum effort (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 was careful to elicit consistent maximum effort performance from each subject at each recording trial as determined by her clinical decision.The same experimenter collected all the data.
Samples of reading and conversational speech were obtained by asking subjects to read aloud the "Rainbow Passage" at a comfortable rate and loudness and generate a 25-30 sec monologue on a topic of their choice.
A visual analog scale (Kempster, 1984;Schiffman, Reynolds, & Young, 1981) was used to obtain subject self-ratings as well as family member ratings of loudness, monotonicity, hoarseness, overall intelligibility, and the extent to which the subject initiated conversation.

Treatment
Both forms of treatment were designed to be intensive (16 sessions in one month) and high effort (subject was pushed to maximum performance levels throughout the entire duration of each session).Each type of treatment stimulated high effort tasks in repeated drills for half of each 50-minute treatment session and then stimulated carryover of the drilled behavior to speech tasks during the final half of each session.The implementation of high effort, intensive treatment is based on neurology and physical therapy practices (England & Schwab, 1959;Hallet & Khosbin, 1980;McDowell, Lee, & Sweet, 1986) that suggest when pushed to higher effort levels, patients with Parkinson disease can compensate for bradykinesia and improve task performance.The daily training of high effort to increase magnitude of speech motor output is based upon the potentially similar problems in scaling the magnitude of output (e.g., stride length in walking, letter stroke in writing) observed in patients with Parkinson disease (Brooks, 1986;Grill, Demirchi, Mc Shane, & Hallett, personal communication;Muller & Stelmach, 1991;Stemach, 1991).We speculate that intensive high effort speech treatment teaches patients to rescale the magnitude of speech motor output (Ramig, 1995;Ramig, Bonitati, Lemke, & Horii, 1994) and that daily treatment with intensive practice (Schmidt, 1975(Schmidt, , 1988) ) and feedback (knowledge of results) (Adams, 1971(Adams, , 1986) facilitates this training.
The respiratory treatment (R) was designed to train increased respiratory muscle activity for inspiration and expiration and sustained expiration to enable subjects to generate increased volumes and subglottal air pressure for speech (Leinonen & Laakso, 1990;McKenzie & Gandevia, 1986;Netsell & Daniel, 1979;Panzarella, 1995;Yorkston, Beukelman, & Bell, 1988).The R treatment trained subjects to increase respiratory effort through tasks such as maximum inhalation and exhalation (Hardy, 1983;Netsell & Rosenbek, 1986), maximum duration of the voiceless continuants /s/ and /f/, and sustained intraoral air pressure (with and without leak tube) (Iowa Oral Performance Instrument [IOPI]; Leith & Bradley, 1976;Robin, Goel, Somodi, & Luschei, 1992).Subjects were stimulated to take deep, frequent breaths during reading and speaking and were given visual feedback (NIMS Respigraph System PN SY03) of the excursion of their rib cage and abdomen during both sustaining of /s/ and /f/ and while reading and speaking.In the respiratory treatment group, no attention was directed toward increasing vocal fold adduction or increasing vocal loudness.
The voice and respiratory treatment (LSVT) focused on increased vocal loudness through increasing vocal fold adduction.Tasks such as pushing the hands together or pushing down or lifting up on the arms of a chair or table while phonating (Aronson, 1990;Froeschels, Kastein, & Weiss, 1955) were used to stimulate increased vocal fold adduction.Drills were carried out on tasks such as maximum duration "ah" phonation and maximum fundamental frequency range.Subjects were encouraged to "think loud" during these tasks and use the same loud, good quality voice generated during sustained phonation during reading and speaking.Attention was directed toward the respiratory system only in the form of encouraging these subjects to take deep breaths frequently and speak "on top of the breath."However, because maximum effort phonatory tasks such as maximum duration of sustained vowel phonation also stimulate respiratory function, the treatment was considered to address both voice and respiration.Feedback for adequate loudness was given using a voice light (lights up at target intensity level) and tape recorder (for self-monitoring of intensity).Training materials for the LSVT approach are summarized elsewhere (Countryman, Ramig, & Pawlas, 1994;Ramig, 1995;Ramig, Bonitati, & Winholtz, 1994;Ramig, Pawlas, & Countryman, 1995).
It is important to point out that the treatment goal of increasing vocal fold adduction in Parkinson disease is designed to maximize the efficiency of the phonatory source.Because of the glottal incompetence frequently observed in patients with Parkinson disease, increased vocal fold adduction may allow patients the ability to generate adequate intensity.It is never the goal of treatment to increase vocal fold adduction so the voice becomes pressed or hyperadducted.The goal is a voice with sufficient loudness, generated with maximum phonatory efficiency (Ramig, 1995).
The treatment intensiveness, daily homework, daily quantification of treatment variables and carryover were all stimulated equally in both treatment groups.Subjects in both treatment groups made progress during treatment as reflected in their daily treatment variables (e.g., maximum durations and ranges).No attention was directed toward improvement of intonation, articulation, or rate in either treatment group.Two clinicians delivered the treatment to all the subjects; both clinicians administered both forms of treatment and were randomly assigned to individual patients.The clinicians worked closely together to ensure consistency and equivalent high effort and motivation across both forms of treatment.A typical treatment session consisted of 25 min of repeated and intensive maximum effort drills and 25 min of high effort speech production tasks moving through a hierarchy of single words, phrases, sentences, reading, and conversation during the 4-week period.Performance feedback (knowledge of results; Adams, 1971Adams, , 1986) ) was provided to subjects using a variety of instruments: tape recorder, stop watch, voice light, respigraph, IOPI, visipitch, and vocal demodulator (Winholtz & Ramig, 1992) depending upon the treatment group.

Data analysis
Sound pressure level (SPL) measures for sustained phonation, reading and the monologue were analyzed using a custom-built software program.The sound level meter signal (Bruel and Kjaer 2230) was preamplified and then digitized at 5000 samples per second into a VAX 4000/200 system computer through a 16-bit resolution DSC-200 A/D converter.The software program displayed the sound level meter signal in decibels (dB).To avoid the inclusion of nonvoiced or pausal segments in the analysis, data points below the noise floor were eliminated from the signal.This was achieved by cursor-marking the first and last negativegoing peaks of the signal and then identifying the percentage below which nonvoiced or pausal data points occurred.The percentage was input into the analysis program and the signal was displayed with the percentage cutoff line.This enabled the analyzer to determine if the correct percentage was chosen.After the correct percentage was determined, the program displayed the mean SPL of the signal.
The SPL was measured from five to six maximum duration sustained "ah" vowel phonations and one reading of the "Rainbow Passage" per recording session for all 45 subjects.Due to limited and variable utterance duration, the SPL for conversational monologue was obtained from 29 of the 45 subjects pre-to post-treatment.The average duration of these monologues ranged from 25-30 sec.
Duration of sustained vowel phonation was also calculated by the SPL program using the markings between the first and last negative going peaks.Duration and SPL data were obtained from the same five to six sustained vowel phonations for each subject.After obtaining the mean SPL for each individual /a/, data were averaged across all max-Downloaded From: https://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929260/ on 10/24/2018 Terms of Use: https://pubs.asha.org/ss/rights_and_permissions.aspximum duration sustained vowel phonations to obtain an overall average mean SPL for sustained phonation.
To obtain measures of mean fundamental frequency and fundamental frequency variability during reading and monologue, the microphone signal was digitized at 5000 samples per second into the same VAX system described above.The files were downloaded onto a 486 computer and then analyzed using CSpeech software (Milenkovic, 1987).The program calculated mean fundamental frequency (Fo) and standard deviation in Hertz.Using standard procedures, the Hertz standard deviation was then converted to express frequency variability in semitones (STSD) (Horii, 1994).Two subjects in the R group had unanalyzable habitual fundamental frequency and fundamental frequency variability data during reading due to severe dysphonia.
Utterance and pause duration data were obtained from both the "Rainbow Passage" and the 25-30 sec monologue.The data of 4 subjects during reading of the "Rainbow Passage" were unanalyzable; thus, 41 passages and 29 monologues were analyzed.The recorded voices were played back, fullwave rectified and smoothed by an RC lowpass filter (Dobkin, 1969).The rectifying and smoothing extracted the intensity envelope in real time.This intensity envelope was digitized by a 12-bit analog-to-digital converter at a rate of 1000 times per second and stored on a 386 microcomputer disk using CSRE (Canadian Speech Research Environment) software.
The digital intensity envelope was subsequently analyzed for utterance and pause durations by software developed by Horii (1983).Given a specification of the maximum amplitude threshold for pause, the minimum duration of pause, and the minimum duration of utterance, the program identified utterances and pauses.The program then printed out means and standard deviations of utterance and pause durations reported here.Standard procedures for analysis of visual analog scales (Boeckstyns & Backer, 1989) were used to obtain perceptual data from the subjects and families.

Reliability
For an assessment of intra-subject reliability, 58% of the subjects completed a second pretreatment voice recording and 36% of the subjects completed a second post-treatment voice recording.One hundred percent of the subjects completed a second set of perceptual self-rating scales preand post-treatment.Correlation coefficients reflecting intrasubject reliability for all acoustic and perceptual variables are summarized in Table 4. T-tests for related measures were carried out on the acoustic and perceptual variables obtained in the two preand the two post-recordings.There were no significant differences among any of these repeated measures.Reliability of perceptual ratings completed by the families are also included in Table 4. Sixty percent of the 20 families completed a second pretreatment visual analog scale and 20% completed a second post-treatment visual analog scale.
Twenty percent of the data were reanalyzed to determine Note.Correlation coefficients are based upon the following sample sizes: 58% of the 45 subjects repeated the pre-treatment and 36% of the 45 subjects repeated the post-treatment acoustic measures; 100% of the 45 subjects repeated the preand post-treatment perceptual self-ratings; 60% of the 20 families repeated the pretreatment and 20% of the 20 families repeated the post-treatment perceptual ratings.
measurement reliability.Correlation coefficients were 0.99 for all measures of SPL, STSD, and F of the "Rainbow Passage," F during conversation, and maximum duration of sustained vowel phonations, and 0.92 for STSD in conversational monologue.Twenty percent of the perceptual data from the subjects' self-ratings as well as the families' perceptual ratings were remeasured to determine measurement reliability.Correlation coefficients were 0.99 for all categories of the visual analog scale.

Acoustic Variables
An analysis of variance carried out to determine pretreatment comparability between the two groups on acoustic and perceptual variables revealed no significant differences.In addition, there were no statistically significant differences on these acoustic and perceptual variables between the two groups of males or the two groups of females at baseline.
A three factor (treatment group x gender x time) repeated measures analysis of variance (ANOVA) was used to evaluate the statistical significance of pre-to post-treatment changes (Campbell & Stanley, 1966).F and p values for the acoustic variables are reported in Table 5.All means and standard deviations of each acoustic variable for each group and gender pre-to post-treatment are summarized in Table 6.

SPL-Vowels
Statistical analysis of SPL during sustained vowel phonation revealed a significant time (pre-to post-treatment) by treatment group (R or LSVT) by gender interaction [F(1,41) = 7.37, p = .01](Table 5).Both the males and females who received the LSVT increased SPL post-treatment (mean increases of 13.96dB [SD 5.01] and 9.89dB [SD 5.66], respectively) (Table 6 and Figure la).Whereas the females who received the R treatment increased on average 1.99 dB (SD 6.09), the males who received the R treatment decreased on average 3.23 dB (3.83) pre-to post-treatment.

SPL-Reading and Conversation
Statistical analysis for SPL during reading revealed a significant time by treatment group by gender interaction, F(1,41) = 4.85, p = .033.Both the males and females who received the LSVT increased in SPL pre-to post-treatment (mean increases of 9.13dB [6.05] and 3.39dB [1.91], respectively) (Table 6; Figure 1 b).The females who received the R treatment also increased SPL (mean 3.45dB [4.77]); the males who received the R increased on average 1.92dB (2.30).
Statistical analysis for SPL during conversational monologue revealed a significant main effect for time, F(1,25) = 14.02, p = .001.There were no significant interactions between treatment group or gender (Table 5; Figure 1c).Subjects in both treatment groups significantly improved SPL in conversation from pre-to post-treatment.

Maximum Duration of Sustained Vowel Phonation
Statistical analysis of maximum duration of sustained vowel phonation revealed a statistically significant main effect for time, F(1,41) = 9.29, p = .004.There were no significant interactions between treatment group or gender (Table 5).Males and females from both treatment groups significantly increased maximum duration of sustained vowel phonation from pre-to post-treatment (Table 6).

Fundamental Frequency (Mean FJ and Variability (STSD)-Reading and Conversation
Statistical analysis of mean habitual fundamental frequency while reading revealed a significant main effect for time, F(1,39) = 35.58,p = .001.There were no significant treatment group or gender interactions (Table 5).Males and groups significantly increased STSD from pre-to posttreatment.
Statistical analysis of fundamental frequency during the conversational monologue revealed no significant changes pre-to post-treatment for either gender or treatment group.Statistical analysis of corresponding fundamental frequency variability (STSD) during the monologue revealed a significant time by treatment group interaction, F(1,25) = 8.48, p = .007,(Table 5).Subjects who received the LSVT significantly increased their STSD during monologue from pre-to posttreatment whereas the subjects who received the R treatment decreased their STSD during monologue from pre-to post-treatment.Despite the statistically significant finding, individual variability was apparent in the magnitude of change across treatment groups and in the intrasubject reliability.

Rate
Statistical analysis for utterance duration during reading revealed a significant main effect for time, F(1,37) = 19.30,p = .001.Both males and females in both treatment groups statistically significantly reduced their utterance duration during reading from pre-to post-treatment (Table 6).Statistical analysis for utterance duration during monologue revealed no significant main effect or interactions pre-to post-treatment (Table 5).Statistical analysis for pause duration during reading of the "Rainbow Passage" revealed a significant time by treatment group interaction, F(1,37) = 6.61, p = .014.Both groups increased on this measure; however, only subjects who received the R treatment demonstrated a statistically significant pre-to post-treatment increase (Table 5).Statistical analysis for pause duration during the monologue revealed no significant main effect or interactions pre-to post-treatment (Table 5).

Forced Vital Capacity (FVC)
FVC data are summarized in Table 6.Statistical analysis revealed no significant main effect for time or interactions (Table 5).females from both treatment groups significantly increased their fundamental frequency from pre-to post-treatment (Table 6).Statistical analysis of corresponding fundamental frequency variability (STSD) while reading the "Rainbow Passage" revealed a significant main effect for time, F(1,39) = 43.29,p = .001.Again, there were no significant interactions (Table 5).Male and female subjects from both treatment F and p values for the analysis of variance carried out on the perceptual variables are presented in Table 7. Means and standard deviations for the perceptual variables for subject self-ratings are presented in Table 8.
Family ratings (n = 20) of preto post-treatment perceptual variables are presented in Table 8.Family ratings of loudness revealed a significant main effect for time, F(1,16) = 5.73, p = .029(Table 7).As seen in Table 8, families of subjects from both treatment groups rated statistically significant preto post-treatment improvement in loudness.Family ratings of overall intelligibility revealed a significant time by treatment group interaction, F(1,16) = 6.74, p = .019.Although families of subjects from both treatment groups rated improvement on overall intelligibility preto post-treatment, only the families of the subjects who received the LSVT rated statistically signifi- cant improvement.Family ratings of monotone, hoarseness, and initiating conversation were not statistically significantly different pre-to post-treatment.Families rated that female subjects initiated conversation significantly more often than male subjects, F(1,16) = 8.66, p = .010.

Neuropsychological Variables of Self-Rated Depression and Sickness Impact
The Beck Depression Inventory (BDI: subject self-rating of depression) and the Sickness Impact Profile on communication and social interaction were completed pre-and post-treatment.The F and p values for the analysis of variance are presented in Table 9. Pre-and post-treatment means and SDs are presented in Table 10.Statistical analysis revealed no significant main effect or interactions for the BDI (n = 35) pre-to post-treatment.Statistical analysis of the SIP on communication (n = 35) revealed a significant time by treatment group interaction, F(1,31) = 6.86, p = .014.Only subjects who received the LSVT rated a statistically significant reduction in the impact of their sickness on their communication from pre-to post-treatment.Statistical analysis of the SIP on social interaction (n = 35) revealed no significant main effect or interactions pre-to post-treatment.

Correlations to Treatment-Related Change
The relationships between magnitude of treatment-related change and patient suitability or prognostic factors for treatment success were examined.Subject pre-and posttreatment change in SPL during reading the "Rainbow Passage" and conversational monologue were correlated (Pearson product moment) with pretreatment age, stage of disease, rating on the motor examination section of the UPDRS, time since diagnosis, severity of speech disorder, glottal incompetence, dementia (cognitive ability), and depression.

All Subjects
There were no significant correlations for the group as a whole between pre-to post-treatment change in SPL and age, stage of disease, rating on the motor examination section of the UPDRS, time since diagnosis, severity of speech disorder, glottal incompetence, cognitive ability, and depression variables or subject ratings of the impact of their disease on communication and social interaction.
All males.No significant correlations were observed for male subjects between prognostic variables and magnitude of treatment-related change in SPL.
All females.No significant correlations were observed for female subjects between prognostic variables and magnitude of treatment-related change in SPL.
LSVr subjects.There was a significant correlation (r = .56,p = .023)between the magnitude of pre-to posttreatment change in SPL in monologue (n = 16) and mean T score pretreatment.Subjects who had greater cognitive ability pretreatment had greater post-treatment increases in SPL in monologue than subjects with lower cognitive abilities.For the males only in the LSVT group there were no significant correlations between magnitude of treatmentrelated change in SPL and prognostic variables.For the females only (n = 5) there was a significant negative correlation (r = -.88, p = .05)between magnitude of pre-to post-treatment change in SPL on the "Rainbow Passage" and chronological age.Younger female subjects had greater increases in SPL than did older female subjects.
R subjects.No significant correlations were observed for this group as a whole or male or female subjects in this group between prognostic variables and magnitude of treatment-related change in SPL.

Summary
The following pre-to post-treatment changes were statistically significant for all subjects regardless of treatment group: SPL during conversation, mean habitual fundamental frequency and fundamental frequency variability during reading, maximum duration sustained vowel phonation and utterance duration during reading, and perceptual self-ratings of monotonicity, intelligibility, and initiation of conversation, and family ratings of loudness.Only for subjects who received the LSVT were the following pre-to post-treatment changes statistically significant: SPL during vowels, SPL during reading, fundamental frequency variability in monologue, perceptual self-ratings of loudness (males), family ratings of overall intelligibility, and the Sickness Impact Profile for the impact of Parkinson disease on their communication.Only for subjects who received the R treatment were the following pre-to posttreatment changes statistically significant: SPL during read-  The data of this study document that intensive speech treatment, particularly the LSVT, made significant short-term improvements in speech and voice characteristics of the subjects studied.Although the R treatment improved some aspects of speech and voice, the subjects who received the LSVT made greater and more consistent post-treatment changes in intensity and fundamental frequency variation.Furthermore, only subjects who received the LSVT reported significant reductions in the impact of Parkinson disease on oral communication following treatment.Given the low vocal loudness reported in patients with Parkinson disease (Aronson, 1985;Blonder, Gur, & Gur, 1989;Boshes, 1966;Critchley, 1981;Darley, Aronson, & Brown, 1969a, 1969b) and its association with reduced speech intelligibility (Ramig, 1992), these findings distinguish these two forms of speech treatment on an important variable.
These findings are logical outcomes when the focus of treatment for each group is considered.The changes in maximum duration sustained phonation and utterance and pause duration are consistent with the goals of the R treatment.It is likely that R treatment contributed to significant post-treatment increases in duration, magnitude, and frequency of inspiration and expiration and the respiratory system control of volume, pressure, and flow (Dromey et al., 1995).However, only following the LSVT were consistent and significant post-treatment increases in intensity observed.It is likely that the improved respiratory support together with the increased vocal fold adduction observed following the LSVT (Brosovic, 1994;Smith et al., in press) resulted in these significant post-treatment increases in intensity.These findings suggest that treatment that trained both increased vocal fold adduction and respiratory support (LSVT) was necessary to generate significant increases in intensity post-treatment in these subjects with idiopathic Parkinson disease.Increasing respiratory support alone was not sufficient to generate significant and consistent increases in intensity in these glottally incompetent subjects.
The intensity changes pre-to post-treatment for the LSVT group were of a perceptually significant magnitude (Pichney, Durlach, & Braida, 1986;Pickett, 1956) for sustained phonation and reading.The pre-to post-treatment increase in intensity for conversational monologue was statistically significant; however, the magnitude was somewhat less than observed in sustained phonation and reading.Given the cognitive disorders and difficulty with complex tasks observed in patients with Parkinson disease (Berardelli, Dick, Rothwell, Day, & Marsden, 1986;Brooks, 1986), this was not surprising.The finding of a statistically significant correlation between cognitive levels and magnitude of pre-to posttreatment changes in intensity in monologue suggests that it is difficult for the more cognitively impaired subjects with IPD to generalize treatment effects.In addition, increased intensity may be more easily achieved and maintained in a task that requires constant vocal fold adduction (e.g., sustained phonation).The reading and monologue tasks require more dynamic abductory and adductory gestures, as well as more attention and concentration, and might therefore have been more difficult for subjects to maintain at a high phonatory effort level.
The changes observed in habitual fundamental frequency and fundamental frequency variability in these subjects are notable given the fact that these variables were not directly trained in treatment.While both treatment groups improved post-treatment, the magnitude and consistency of improvements of the LSVT subjects were greater.Training of adduction and maximum frequency ranges may be the basis for these post-treatment increases in fundamental frequency and fundamental frequency variability.Another explanation may relate to the perception of the LSVT subjects that their communication skills were much less affected by their disease post-treatment.Only the LSVT subjects rated a significant reduction in the impact of their sickness on their communication skills post-treatment.It may be speculated that as the LSVT subjects improved their ability to communicate, their confidence, attitude, and affect improved.This attitude could be reflected in increased pitch and intonation during speech.Higher pitch and increased pitch variability have been associated with descriptive characteristics of happiness, joyfulness, and confidence (Fairbanks & Pronovost, 1939;Scherer, London, & Wolf, 1973).Improvements in confidence, attitude, and affect are also consistent with our subjective observations of the subjects who received the LSVT immediately post-treatment.This finding suggests that improved communication can play an important role in the psychosocial well-being of individuals with Parkinson disease.
Although activities of respiration were the focus of the R treatment and activities of vocal fold adduction and respiration were the focus of the LSVT, it was apparent by clinical observation that post-treatment changes were observed in magnitude and precision of articulatory gestures as well (Ramig, Dromey, Johnson, & Scherer, 1994).These observations were most apparent in subjects who received the LSVT and are consistent with reports of Schulman (1989) who studied articulatory changes following loud speech in nondisordered speakers.These post-treatment changes in untreated articulation have been quantified in a patient with IPD following a course of the LSVT (Dromey et al., 1995).Such findings are in agreement with those of Scott and Caird (1983) who reported that phonation in particular was a prerequisite for intelligible, effective speech in Parkinson disease and as voice production improved other aspects of speech also improved.In addition, they are consistent with those of Rubow and Swift (1985) who reported a positive relationship between improved loudness and articulatory skills in one patient with Parkinson disease.These observations suggest that multiple levels of speech production in patients with Parkinson disease can benefit from a single treatment focus of increased phonatory effort (Ramig et al., 1994).This is useful information since a treatment that is simple is optimal for IPD patients given their known cognitive disorders and difficulty with complex tasks (Beneckel, Rothwell, Dick, Day, & Marsden, 1986;Yanagisawa, Fujimato, & Tamaru, 1989).
The perceptual data presented here were considered as an initial probe of the impact of treatment-related changes on functional communication.Although the ratings of subjects and family members may be biased and less reliable, they represent one important level of assessment of the perceived impact of treatment on functional communication.These findings suggest that both forms of treatment influenced subject and family perceptions of overall communication skills and are consistent with general positive treatment effects across both groups.However, they also suggest that, consistent with acoustic data, subjects who received the LSVT perceived the greatest impact of the treatment on these perceptual variables.
These findings of significant treatment-related change are in contrast to previous reports from Allan (1970), Aronson (1985), Greene (1980), Sarno (1968), and Weiner and Singer (1989) who concluded that speech treatment was not effective for patients with Parkinson disease.The most reasonable explanation for the positive findings reported here is the focus and style of treatment.Previous approaches to treatment for Parkinson disease have focused on articulation and rate and were administered in the classic speech treatment style of once or twice a week.Both treatments studied here were administered 4 days a week for 16 high effort sessions and focused on respiration (R) or voice and respiration (LSVT).Based upon our findings here, we suggest that intensive treatment with the focus on voice and respiration (LSVT) is effective in increasing vocal intensity in patients with Parkinson disease.These findings are consistent with Scott & Caird (1983) and Robertson & Thompson (1984) who also concluded that intensive voice treatment could increase loudness in patients with Parkinson disease.We suggest that increasing vocal loudness through high phonatory effort (LSVT), teaches patients to rescale the magnitude of their speech motor output.Such interpretation is consistent with amplitude resealing changes discussed in other motor systems of patients with Parkinson disease (Teulings & Stelmach, 1991).High effort respiratory (R) treatment generated fewer significant changes in the speech of these patients.Therefore, it appears that high effort with the focus on phonation is a critical combination in the successful treatment of patients with Parkinson disease.
Of the 45 subjects in this investigation, prognostic variables of stage of disease, time since diagnosis, severity of speech disorder, glottal incompetence, and depression did not significantly influence post-treatment effects.Age and cognitive status may influence magnitude of treatmentrelated change (with younger and more cognitively intact subjects improving more on certain measures).This observation must be studied more closely before conclusive statements are made.It should be noted, however, that most of the subjects studied here were in the mild-tomoderate range of their disease.Further research should be carried out in order to examine more severely involved as well as very recently diagnosed subjects to assess the generalizability of treatment across all stages of IPD.Given the heterogeneous nature of Parkinson disease, it is reasonable to assume that all subjects would not respond to treatment in the same way.This variability is clearly reflected in individual subject responses to treatment.It is important to keep in mind that subject heterogeneity may reduce statistically significant group effects despite positive findings for individual subjects.This is an important consideration in evaluating the efficacy of treatment.
The gender differences observed in certain of the pre-to post-treatment data deserve further consideration.The smaller number of females in both treatment groups, the heterogeneity of the females in the LSVT group and the overall higher depression scores for females limit conclusive statements at this time.However, the findings indicate that further study of the effects of intensive speech treatment for female subjects with idiopathic Parkinson disease is essential.
Databased efficacious treatment programs for patients with progressive neurological disorders are virtually nonexistent (Yorkston, Beukelman, & Bell, 1988).Given the degenerative nature of Parkinson disease, the accompanying cognitive impairment, generalized involvement at multiple levels of the speech mechanism, depression, and comprehensive physical impairment, documentation of statistically significant pre-to post-treatment changes for group data of the magnitude reported here is an important step in enhancing the efficiency of treatment for this population.The maintenance of these changes is being addressed in ongoing work (Ramig, Countryman, O'Brien, & Smith, 1995).
Although the degenerative course of some neurological disorders cannot be altered at this time, improved oral communication may be an important component in developing the highest levels of functioning and independence for each individual.

FIGURE 1 .
FIGURE 1. Means and standard deviations for sound pressure level (SPL) before and after treatment for the LSVT and R treatment groups, showing male, female, and overall group values.Tasks included: (a) sustained phonation, (b) reading of the Rainbow Passage, and (c) conversational monologue.All measures of SPL were made at a distance of 50cm.

Discussion-
This study was designed to compare the effects of two forms of intensive speech treatment (respiration [R] and voice and respiration [Lee Silverman Voice Treatment; LSVTJ) on the speech and voice deficits of subjects with Idiopathic Parkinson disease.The study also examined the magnitude of treatment-related change relative to subject characteristics including: age, stage of disease, time since diagnosis, severity of speech/voice disorder, glottal incompetence, depression, and cognitive ability.

TABLE 3 . Neuropsychological characteristics of the Lee Silverman Voice Treatment (voice and respiration) subjects and the respiration-only subjects.
Note.T scores below 40 are considered impaired.Scores above 9 on self-rating of depression suggest some depression.Scores of 6 or greater on clinician ratings of depression are suggestive of depression.Scores that are larger on the SIP indicate more impact of the sickness in that area.Scores of 0.00 on the SIP indicate no difficulty in that area.

TABLE 8 . Perceptual variables for the subjects' (n = 45) self-ratings by group and gender as well as the families' ratings for both groups. Standard deviations are presented in parentheses.
Perceptual characteristics of loudness, hoarseness, intelligibility, and initiation of conversation are on a scale of 0-100% with 0% the most severe and 100% the least severe.