Acoustic Variability in Dysarthria: Effects of Articulation Rate and its Relationship with Intelligibility

Abstract

Introduction The characteristics of normal and disordered speech motor control processes have been a well-covered and popular topic of research in the past decades and have been studied by an array of experimental and empirical approaches. One way to assess motor control in these areas is to estimate the stability of movement patterns, e.g., by means of the spatiotemporal index (STI) (Smith, Goffman, Zelaznik, Ying, & McGillem, 1995), and functional data analysis (FDA) to obtain separately measured spatial and temporal variability (Lucero & Koenig, 2000). In research to date, speech variability measures have largely been applied to articulatory movement signals captured using direct measurement techniques such as electromagnetic articulography. One of the largest drawbacks of these methods is the invasive and complex nature of the technology to collect data, involving intra- and extra-oral attachment of sensors to lips, tongue, palate, velum, and jaw. These methodologies require expensive, large and specialist equipment, making them not easily acceptable for use with participants with motor and sensory impairments. Recent developments in speech processing have made it possible to assess and quantify variability of continuous or quasi-continuous acoustic properties as an indirect measure of speech movement stability. Measures of acoustic variability have been shown to differentiate between dysarthria types and correlate with established clinical assessments (Anderson, Lowit, & Howell, 2008; van Brenk & Lowit, 2012). This study further investigates acoustic variability characteristics in speakers with dysarthria, by assessing the effects of rate reduction and rate increase on acoustic stability. Reducing rate has long been used as a management strategy in the behavioural treatment of hypokinetic dysarthria (Yorkston, Hakel, Beukelman, & Fager, 2007). Rate reduction is thought to enhance intelligibility for speakers with dysarthria because a slower rate of speech production gives talkers increased time to achieve contrastive vocal tract configurations and may increase coordination within and among the speech subsystems. However, not all speakers with dysarthria exhibit improved speech intelligibility when using a slower-than-typical rate (Tjaden, Sussman, & Wilding, 2014; Tjaden & Wilding, 2004). In addition, research on the effects of rate changes on speech motor control has found that slowing down rate might decrease the stability of speech movements. Kleinow, Smith, & Ramig (2001) reported higher STI values of lower lip movements in sentences repeated at slow rate compared to habitual and fast rate conditions in speakers with hypokinetic dysarthria, possibly because they moved away from their ‘stable’ speech mode. In contrast, McHenry (2003) reported lower STIs in a stretched-words slow speaking condition, compared to habitual and fast rate conditions, and STIs in the fast condition were higher compared to the habitual condition. From these results, the authors concluded that individuals with dysarthria benefit from a reduced speaking rate. Whilst these discrepancies across studies may partly be explained by differences in rate reduction modelling procedures, it is clear that variability of speech movements across different rate conditions needs further evaluation. Furthermore, little is known about how variability measures relate to functional communication metrics such as intelligibility. The current study aimed to investigate whether individuals with hypokinetic dysarthria, ataxic dysarthria, and healthy control speakers differ in spatial and temporal variability of a range of acoustic parameters including sound pressure level, fundamental frequency, and first formant frequency as a factor of speaking rate, and how these findings relate to measures of intelligibility. Methodology Audio recordings were taken from 59 participants: 23 speakers with Parkinson’s disease and mild to moderate hypokinetic dysarthria (HD), 9 speakers with various neurological diseases and mild to severe ataxic dysarthria (AD), and 27 age-matched healthy control speakers (CON). Speakers were asked to repeat the sentence “Tony knew you were lying in bed” approximately 20 times, during Habitual, Slow, and Fast rate. The contours of Sound Pressure Level (SPL), Fundamental Frequency (F0), and First Formant Frequency (F1 were extracted, and subjected to Functional Data Analysis to obtain spatial variability and temporal variability. A series of linear mixed model analyses were carried out to compare variability measures across groups and speech tasks. The statistical models were applied separately to the two variability measures and three speech parameters, with Group and Task as fixed factors and Subject as random factor. Fifteen undergraduate SLT students participated in the intelligibility judgments. For each speaker with dysarthria, ordinal ratings of intelligibility were obtained from a monologue fragment by means of a 1 (completely unintelligible) to 9 (completely intelligible without effort) point Likert scale incorporating intelligibility and listener effort (Dobinson, 2007). Pearson two-tailed correlations were calculated to evaluate the strength of association between acoustic variability and intelligibility. Results & Discussion Comparisons between groups showed that spatial and temporal variability values of SPL and F0 contours were higher in the AD group compared to the HD and CON group, possibly reflecting the higher severity in the AD group. Differences between the HD and AMC groups were absent. There were no effects of group for F1 spatial variability, but F1 temporal variability was higher in both dysarthric groups compared to the control speakers. The presence and direction of differences in variability across rate conditions were largely dependent on speech parameter. Overall findings pooled across groups showed higher spatial variability for SPL and F0 in the slow rate condition compared to the habitual and fast conditions, whereas temporal variability of F1 was higher in the fast condition compared to the other two conditions. The results of the correlational analyses indicate robust negative associations of intelligibility with SPL temporal variability for the HD speakers, and with SPL spatial variability for the AD speakers, respectively. For both groups, correlations were strongest in the fast and slow conditions. In relation to F0 spatial and temporal variability, the HD group showed significant correlations with intelligibility in the fast rate condition. The AD group showed negative correlations of intelligibility with F1 temporal variability in the slow condition. The overall results of this study suggest (1) that deviating from habitual rate generally results in decreased stability of different aspects of speech production and (2) a convoluted relationship between acoustic variability and intelligibility, with associations largely dependent on dysarthria type and speech parameter.

Frits van Brenk

External consultant in the Motor Speech Disorders Laboratory

My research and teaching interests include motor speech disorders, clinical linguistics, experimental psycholinguistics, speech science, and research methods.