Fax +41 61 306 12 34E-Mail karger@karger.chwww.karger.com

Original Paper

Folia Phoniatr Logop 2011;63:216–220 DOI: 10.1159/000319971

Long-Term Phonatory Instability in Ataxic Dysarthria

Frank Boutsen   a Joseph R. Duffy   b Hani Dimassi   c Sarah S. Christman   a

a   University of Oklahoma Health Sciences Center, Oklahoma City, Okla. , and b   Mayo Clinic, Rochester, Minn. , USA; c   American University of Beirut, Beirut , Lebanon

Introduction

Phonation in humans is remarkably well controlled. Across glottal cycles, cycle-to-cycle variation is stable enough to yield a normal voice quality (not tight or breathy) that is virtually free of noise (or hoarseness). This control involves, among other things, normal mus-cle tonus and proprioceptive reflex activity, as well as complex online adjustments of pitch and airflow based on perception of one’s vocal output [1] . Maintenance of phonatory control for an extended period (e.g., during vowel prolongation) normally yields what is perceived to be a near-steady voice with nearly even loudness and pitch. Under conditions of nervousness and fatigue, but especially in neurologic motor speech disorders, particu-larly those affecting the cerebellum and its pathways, ab-normal long-term variations in loudness, pitch and/or quality can occur [2] . These variations can be (quasi-)cy-clical or noncyclical.

The distinction between these two forms of phonatory long-term instability, while perhaps resulting in a discon-tinuous perception of tremor and unsteadiness is neither simple nor clear. This is because to this date only phona-tory tremor has been defined, albeit mostly in terms of its characteristic frequencies. Quasi-cyclical variations in loudness and pitch at 1–2 Hz, 6–8 Hz and 10–12 Hz have been labeled as wow, tremor and flutter [3, 4] . Even so, the

Key Words

Phonation tremor � Ataxic dysarthria � Cerebellum

Abstract

Aims: Long-term phonatory instability can be quantified us-ing cyclical and noncyclical measures. The objective of this study is to evaluate phonation in ataxic dysarthria and a con-trol group of normal speakers to answer two main questions: (1) How common is elevated cyclical and noncyclical instabil-ity in ataxic dysarthria compared to that in a normal control group? (2) Is cyclical instability predictive of noncyclical in-stability? Methods: Vowel prolongations of ataxic-dysarthric and normal speakers were compared using the Motor Speech Profile module of the Computerized Speech Lab. Cyclical measures included tremor rate, amplitude and peri-odicity. Noncyclical measures included the coefficient of variation for loudness and frequency. Results: Noncyclical measures are elevated in a subset of speakers with ataxic dysarthria regardless of whether cyclical instability (vocal tremor) is present. Cyclical instability was detected in nearly half the patients. Interestingly, elevations in both types of measures also described phonation of a number of the par-ticipants in the control group. Conclusion: Combined use of cyclical and noncyclical measures can document aspects of phonation in ataxic dysarthria that have clinical implications.

Copyright © 2010 S. Karger AG, Basel

Published online: November 19, 2010

Frank Boutsen, PhD Department of Communication Sciences and Disorders, AHB 3082 1200 North Stonewall, The University of Oklahoma Health Sciences Center Oklahoma City, OK 73117-1215 (USA) E-Mail Frank-Boutsen   @   ouhsc.edu

© 2010 S. Karger AG, Basel1021–7762/11/0634–0216$38.00/0

Accessible online at:www.karger.com/fpl

Dow

nloa

ded

by:

Uni

vers

ity o

f Ala

bam

a

130.

160.

4.77

- 8

/26/

2013

6:4

2:11

AM

Long-Term Phonatory Instability Folia Phoniatr Logop 2011;63:216–220 217

extent and type of irregularity in tremulous voices re-mains underinvestigated leaving their classification (purely) in frequency categories less than balanced [5] . As for noncyclical long-term variations, they have not been uniformly described; they perhaps denote most closely what has been termed poorly modulated (abrupt or jerky) fluctuations in loudness and/or pitch that are irregular and are perceived as unsteadiness [6] .

Though long-term phonatory instability may be a marker of cerebellar disease, it has received very limited attention in studies of ataxic dysarthria. Moreover, the quantification of long-term instability in these investiga-tions has been ambiguous. This is because nonspecific measures of long-term instability, such as standard de-viation (SD) or the coefficient of variation (CV), have been used to quantify overall instability that might or might not include tremor. In one study [7] SD was used to quantify long-term fluctuations in fundamental fre-quency (F 0 ) in patients with pure cerebellar atrophy and with olivopontocerebellar atrophy. SD F 0 was increased in about half of the patients in each group. Among the patients with increased SD F 0 , all but 2 had irregularly distributed shifts of F 0 that were perceived as pitch fluc-tuations. Phonation in the remaining 2 patients showed a low frequency (2.8 Hz) or intermittent rhythmic modula-tion (not quantified) that were perceived as tremor and quiver, respectively. Another study [8] used SD and CV indices of fundamental frequency (vF 0 ) and intensity (vAm) to quantify long-term phonatory instability in ataxic dysarthria. Findings again revealed elevated insta-bility and were felt to support the notion that long-term phonatory irregularities are a dominant feature of the voice in ataxic dysarthria. Perceptual-acoustic relation-ships were not addressed, however.

Voice tremor has been described as a rare manifesta-tion in ataxic dysarthria; however, its prevalence or pre-sentation has not been formally documented. Cerebellar voice tremor has only been described in small, mostly case studies, as a coarse 3-Hz tremor [2, 9] . Extending the review to include a study [3] of long-term instability in the speech of persons with multiple sclerosis, an impor-tant cause of ataxic dysarthria, brings into view the pos-sibility of coexisting types (and even subtypes) of long-term instability in ataxic dysarthria. In this study it is proposed that since vocal instability is complex, with dif-ferent aspects most likely being distributed rather than focally represented in cerebellar circuitry, multi-measure protocols of long-term instability are warranted so as to aid differential diagnosis. Accordingly, the use of two major types of measures is recommended, one that is cy-

clical (i.e., contains frequency information), that allows quantification of tremor rate and extent and a second, that is not cyclical, that would quantify total variation in phonation [2] .

The present study investigated long-term phonatory instability in ataxic dysarthria and a control group of normal speakers using cyclical and noncyclical measures. It sought to address two basic questions: (1) How com-mon is elevated frequency and non-cyclical instability in ataxic dysarthria compared to that in a normal control group? (2) Is cyclical instability predictive of noncyclical instability? In other words, can the two indices be disso-ciated? As to the latter question, one could, from a mea-surement perspective, argue that noncyclical measures like the SD or the CV would be elevated if tremor is pres-ent. A physiological perspective, on the other hand, could reasonably favor the opposite expectation that tremor and noncyclical variability (unsteadiness) are unrelated. For example, one could envision phonatory tremor and unsteadiness not unlike (voice) tremor and hypophonia (rigidity) in Parkinson’s disease as positive and negative symptoms with different underlying neurophysiology.

Materials and Methods

Participants Patients with a clinical diagnosis of ataxic dysarthria (n = 18)

and normal speakers (n = 19) participated in the study. The diag-nosis of ataxic dysarthria was made by 2 speech-language patholo-gists with expertise in neurogenic communication disorders. The diagnosis was based on perceptual analysis of the patients’ speech during connected speech, vowel prolongation, reading and diado-chokinetic speech tasks. Speech dimensions rated were those in-cluded in a study of dysarthria by Darley et al. [10] . For each of the patients, audio recordings of the speech tasks and reports of their neurologic and speech examinations were available. The 18 sub-jects were selected from among an original pool of 20 patients whose records indicated a diagnosis of ataxic dysarthria. Two pa-tients were excluded because a spastic component could not be ruled out by 2 of the authors (F.R.B., J.R.D.), primarily because mild strained voice quality was evident. For the remaining 18 pa-tients, review of the speech pathology written record and indepen-dent perceptual analysis of the audio recordings by 2 judges (F.R.B., J.R.D.) resulted in 100% agreement that the dysarthria was of the ataxic type. Among the 18 patients were 12 females and 6 males (mean age = 55; SD = 14.8). The etiology of the ataxic dysarthria was heterogeneous. Perceptual severity ratings of the dysarthria ranged from mild to marked (on a scale that included mild, mod-erate, marked and severe), with the modal rating being moderate.

Participants in the control group included 12 females and 7 males. As a group, they were slightly younger than the dysarthric group (mean age = 48; SD = 8.9). Each of the normal-speaking participants passed a hearing screening and denied a history of neurologic, speech and/or voice disorders.

Dow

nloa

ded

by:

Uni

vers

ity o

f Ala

bam

a

130.

160.

4.77

- 8

/26/

2013

6:4

2:11

AM

Boutsen   /Duffy   /Dimassi   /Christman   Folia Phoniatr Logop 2011;63:216–220 218

Tasks Participants were asked to produce the vowel /ah/ for as long

and as steadily as possible, until they ran out of air as part of a comprehensive speech evaluation. They were asked to use their natural voice on a single trial, and on occasion twice or three times. Repeat trials were required either because of failure to fol-low the task requirements (e.g., a participant phonated for a very short period or conducted the task with inappropriate loudness or pitch or in a singing voice) or to check the validity of previously made auditory perceptual observations. The same clinical proce-dure was also followed with the control subjects. The trial that was accepted by the clinician as most valid was used for analysis in both groups.

Data Analysis Motor Speech Profile (Kay Elemetrics Motor Speech Profile,

CSL Model 4300B, Lincoln Park, N.J., USA) was employed to quantify long-term instability. A 4-second window was selected from the temporal midpoint of each digitized sustained phona-tion, where phonation was continuous and devoid of onset and termination effects. Noncyclical, total variation parameters em-ployed included vF 0 and vAm. The vocal tremor analysis module in MSP was used to detect and quantify amplitude and frequency modulation, henceforth labeled as frequency and amplitude trem-or, in terms of rate, magnitude, and periodicity. The computation-al methods that are employed by MSP for these measures are per-turbation analysis and autocorrelation. A smoothed (55th order) perturbation function is used to estimate the magnitude of the tremor. Rate and periodicity are derived from autocorrelation analyses of the energy and F 0 contours of the signal using a 30-ms window with 5-ms frame advance. Tremor rate is determined as the rate corresponding to the strongest component when a 1 to20 Hz component range is considered in the autocorrelation anal-ysis [Deliyski D: personal commun., 2005]. Periodicity is comput-

ed as the ratio between the energy of the strongest autocorrelation component and the energy after direct current removal. For ex-ample, in a pure tone the energy of the component identified by the autocorrelation equals that of the total energy and periodicity and is, therefore, 100% [Deliyski D: personal commun., 2005]. It should be noted that MSP provides values for magnitude that arise from nonperiodic modulations even when tremor (rate) is not de-tected. For purposes of consistency and in line with the intent to use the tremor analysis module as a cyclical measure, the decision was made to only report magnitude in association with tremor and as nondetected along with periodicity if tremor is not detected.

Since the acoustic measures of long-term instability were based on automatic processing of the 4-second phonation seg-ments without user intervention, identical values were found for repeated analyses of the same segments, and no further measures of agreement were calculated.

Because the data set was relatively small and positively skewed in its distribution, the Wilcoxon Exact test with a Monte Carlo resampling procedure was employed to analyze the results. The Wilcoxon test is a nonparametric statistic that tests the central tendency of the data by allocating ranks to the observations. The Monte Carlo resampling procedure is a technique that corrects the p value for small samples. The descriptive analysis was carried out on SAS 8.1, and the Wilcoxon test was carried out on StatXact Software.

Results

Results summarized in table  1 show that frequency tremor was detected in 50% of the ataxic speakers and in 26% of normal speakers. This difference in prevalence, while seemingly large, failed to reach statistical signifi-cance, most likely because of the combined influence of large variances and small sample sizes. Table  1 further makes apparent that amplitude tremor detection rates were less than those for frequency tremor; they were ob-served in the vocal performance of approximately 1/3 of the members in each group. When analyzed in terms of rate, periodicity and magnitude ( table 2 ) amplitude tremor in the ataxic speakers was characterized by lower frequen-cy (mean = 2.64 Hz; SD = 0.78), lower periodicity and higher magnitude than was observed in the normal speak-ers (mean = 4.23; SD = 2.13). As such, these data are con-sistent with previous accounts of cerebellar vocal tremor as being slow and coarse [2, 9] . Of the aforementioned de-scriptive differences only tremor rate differentiated the vocal tremor of the groups to an extent that was statisti-cally significant. As for the relationship between CV pa-rameters and vocal tremor in both groups, table 3 shows that among the ataxic dysarthric speakers vAm was statis-tically elevated compared to that of the normal speakers regardless of whether amplitude tremor was present or not. A similar trend was also observed for vF 0 . That is, the

Table 1. Frequency of detection of amplitude tremor and frequen-cy tremor by group (n = 37)

Variables Group

normal a taxia

n % n %

Frequency tremorDetected 5 26.30 9 50.00Nondetected 14 73.69 9 50.00Totals 19 100.00 18 100.00

p value = 0.184Amplitude tremor

Detected 7 36.84 6 33.33Nondetected 12 63.16 12 66.68Totals 19 100.00 18 100.00

p value = 1.000

p v alues are derived from Fisher’s exact test used due to small cell frequencies.

Dow

nloa

ded

by:

Uni

vers

ity o

f Ala

bam

a

130.

160.

4.77

- 8

/26/

2013

6:4

2:11

AM

Long-Term Phonatory Instability Folia Phoniatr Logop 2011;63:216–220 219

between-group difference in vF 0 approached significance when frequency tremor was detected and it reached sig-nificance when frequency tremor was not detected.

Discussion

The results of this study generally agree with those of previous acoustic studies [6–9] of long-term phonatory instability in ataxic dysarthria. Specifically, long-term noncyclical instability of vowel prolongation, particular-

ly that involving frequency, is elevated in ataxic dysar-thria and occurs with a prevalence of about 50%, similar to that shown in previous studies [7, 8] . With regard to cyclical phonatory instability, the data reveal that fre-quency tremor was detected in nearly half of the patients. Like amplitude tremor, which was less common, frequen-cy tremor was most often in the 3-Hz range. This range is consistent with that reported in the literature [2, 9] . The data further indicate that elevated noncyclical instability in the ataxic speakers is not predicated on the presence of vocal tremor either for the frequency or the amplitude

Table 2. Mean and medians of tremor rates (Hz), periodicity (%), and magnitude (%) by group for amplitude and frequency tremor in participants with vocal tremor

Variables Amplitude F requency

normal (n = 7) ataxia (n = 6) norm al (n = 5) ataxia (n = 9)

Tremor ratesMedian 4.60 2.50 5.10 3.00Mean 8 SD 4.2382.13 2.6480.78 5.0280.75 3.6582.13

p value = 0.192 p value = 0.016Periodicity

Median 46.00 28.95 30.30 33.10Mean 8 SD 42.14814.54 36.22816.22 30.87810.11 31.99811.99

p value = 0.843 p value = 0.797Magnitudes

Median 3.80 7.45 1.20 0.90Mean 8 SD 5.6084.00 8.6884.65 3.7285.71 0.9680.36

p value = 0.169 p value = 0.202

p v alues are derived from Wilcoxon’s exact test with Monte Carlo resampling.

Table 3. Mean and median of vAm and vF0 for group (normal-ataxic dysarthria) and tremor status (detected/not detected)

Amplitude CV (vAm) F requency CV (vF0)

normal ataxia no rmal ataxia

Tremor detectedMedian 9.94 17.82 1.18 1.48Mean (SD) 11.0 (10.33) 21.39 (12.15) 1.1 (0.13) 1.8 (0.98)n 7 6 5 9

p value = 0.035 p value = 0.078Tremor not detected

Median 12.53 21.71 0.65 1.38Mean (SD) 13.15 (8.23) 23.67 (10.19) 0.82 (0.44) 1.61 (0.10)n 12 12 14 9

p value = 0.004 p value = 0.007

p v alues are derived from Wilcoxon’s exact test with Monte Carlo resampling.

Dow

nloa

ded

by:

Uni

vers

ity o

f Ala

bam

a

130.

160.

4.77

- 8

/26/

2013

6:4

2:11

AM

Boutsen   /Duffy   /Dimassi   /Christman   Folia Phoniatr Logop 2011;63:216–220 220

parameter. That is, among the ataxic dysarthric speakers vAm was statistically elevated compared to that of the normal speakers regardless of whether amplitude tremor was present or not. A similar trend was also observed for vF 0 .

The acoustic analyses also revealed elevated levels of cyclical and noncyclical variation in the normal controls. Thus, neither type of measure of vocal instability yielded quantitative indices that have sufficient specificity to dis-tinguish the vocal performance of ataxic and normal speakers. The results of this study do suggest that, among the cyclical indices, tremor rate is a useful diagnostic in-dicator (i.e., tremor rate is slower in ataxic speakers than in normal speakers) whose specificity is likely enhanced when used in combination with magnitude and periodic-ity.

Finally, as for the implications of this investigation for future lines of inquiry, there are several observations that need further exploration. While either type of vocal in-stability can be manifest in amplitude or frequency pa-rameters, or both, our data, as well as previous research, suggest that noncyclical variation is more frequently manifested in the frequency parameter. In contrast, cycli-cal variation was expressed in the frequency and ampli-tude parameters with no trend for one or the other, or

their combination, to occur more frequently. In order to meaningfully interpret these data, more needs to be known about the mechanisms that control parameter ex-pression in vocal instability. Specifically, it would be of interest to know whether variation in the frequency pa-rameter, regardless of the measurement type more direct-ly implicates a ‘local’ impairment limited to vocal fold function as opposed to more widespread system involve-ment, such as respiration, that might influence fluctua-tions in the amplitude parameter. This is because pos-tural tremor involving hypotonic muscles of the larynx that may have abnormal reflex properties can be a mech-anism for isolated frequency tremor of the larynx [9] . However, it is quite possible that instead of local laryn-geal oscillatory behavior, respiratory muscle drive as-sumes abnormal agonist or antagonist (reactive) roles in disordered laryngeal-respiratory coupling. Our observa-tion that in some patients amplitude and frequency mod-ulation were in phase, while in others they were out of phase, suggests that ‘coupling’ may indeed be a process that deserves evaluation in explaining heterogeneity in vocal tremor in these patients. These observations, as well as our general finding that types of vocal instability are descriptive of only a subset of ataxic dysarthric patients, warrant further research.

References

1 Schulz GM, Varga M, Jeffires K, Ludlow CL, Braun AR: Functional neuroanatomy of hu-man vocalization: an H215O PET study. Cereb Cortex 2005; 15: 1835–1847.

2 Aronson AE: Clinical Voice Disorders: An Interdisciplinary Approach. New York, Thieme, 1990.

3 Hartelius L, Buder EH, Strand EA: Long-term phonatory instability in individuals with multiple sclerosis. J Speech Lang Hear Res 1997; 40: 1056–1072.

4 Buder EH, Strand E: Quantitative and graph-ic acoustic analysis of phonatory modula-tions: the modulogram. J Speech Lang Hear Res 2003; 46: 475–490.

5 Kreiman J, Gabelman B, Gerratt B: Percep-tion of vocal tremor. J Speech Lang Hear Res 2003; 46: 203–214.

6 Duffy JR: Motor Speech Disorders: Sub-strates, Differential Diagnosis and Manage-ment. St Louis, Mosby-Yearbook, 1995.

7 Ackermann H, Ziegler W: Acoustic analysis of vocal instability in cerebellar dysfunc-tions. Ann Otol Rhinol Laryngol 1994; 103: 98–104.

8 Kent RD, Kent JF, Duffy JR, Thomas JE, Weismer G, Stuntebeck S: Ataxic dysarthria. J Speech Lang Hear Res 2000; 43: 1275–1289.

9 Ackermann H, Ziegler W: Cerebellar voice tremor: an acoustic analysis. J Neurol Neu-rosurg Psychiatry 1991; 54: 74–76.

10 Darley FL, Aronson A, Brown J: Motor Speech Disorders. Philadelphia, Saunders, 1975.

Dow

nloa

ded

by:

Uni

vers

ity o

f Ala

bam

a

130.

160.

4.77

- 8

/26/

2013

6:4

2:11

AM