24 / 56
130
JCPSLP
Volume 14, Number 3 2012
Journal of Clinical Practice in Speech-Language Pathology
www.fon.hum.uva.nl/praat). PRAAT was first released
in 1995 and is regularly maintained by its developers (P.
Boersma and D. Weeninck, University of Amsterdam). It
has been used extensively for analysis of both healthy and
impaired speakers. Comprehensive manual and tutorials
on the website provide guidelines for checking for errors in
measurement that can occur more frequently with the more
variable speech of dysarthria.
Aims
The aim of this study was to demonstrate the use of a small
number of easy-to-collect acoustic measures using a free
software program, PRAAT (Boersma & Weenink, 2010), for
three prototypical dysarthria cases: one spastic, one ataxic,
and one flaccid dysarthria case. The list of measures
presented here is by no means comprehensive, but rather
provides an introduction to using the PRAAT software and
perhaps an incentive to explore it more fully. We report the
results of these acoustic analyses, compare them with
available normative data, and how they relate to perceptual
judgements.
We predicted that the individuals with spastic or flaccid
dysarthria would demonstrate abnormal vocal quality
measures (e.g., jitter, shimmer, HNR), associated with
perceived abnormal vocal quality. The individual with ataxic
dysarthria and notable pitch breaks and vocal tremor was
expected to show high variability of
f0
during sustained
ah
production. We expected that all would demonstrate
reduced speech rate in diadochokinetic and connected
speech tasks. Further, the individuals with spastic and
ataxic dysarthria would deviate from normal on objective
measures of prosody (i.e., relative duration,
f0
and/or
intensity across syllables in connected speech as measured
by the PVI), reflecting the perception of equal stress or
scanning speech, respectively. Perception of monopitch or
monoloudness should be reflected as lower PVI values for
f0
and dB (PVI_
f0
, PVI_dB), respectively.
Yumoto & Gould, 1982). Of note, software programs have
different algorithms for calculating these measures which
may yield differing results (Maryn, Corthals, De Bodt, Van
Cauwenberge, & Deliyski, 2009). It is best to use norms
generated by the selected software and standardise data
collection methods to achieve highly reliable measurement
over time. Further, the software may generate some
erroneous
f0
measurements (e.g., excessively high values
at the edges of vowels) that distort maximum and average
measures. Care is taken to omit these from the selection
used for calculations (see Figure 1).
Analysis of prosody also involves measuring frequency
and intensity, as well as segment or syllable durations, but
at word or connected speech level. English is a stress-
timed language that generally alternates stressed and
unstressed syllables in a word or sentence. One measure
proving useful for capturing this pattern is the pairwise
variability index (PVI), which is a normalised measure of
relative duration,
f0
, or intensity over a word or speech
sample (Ballard, Robin, McCabe, & McDonald, 2010;
Courson, Ballard, Canault, & Gentil, 2012; Low, Grabe,
& Nolan, 2000; Vergis & Ballard, 2012). Specifically, one
calculates the difference in duration (or
f0
or intensity)
over two consecutive vowels and divides the difference
by their average. This calculation is done pairwise for
the whole sample and the average PVI value used as an
index of stress variability. Low et al. (2000) reported that in
British-English average PVI for vowel duration (PVI_Dur) in
sentences containing all stressed words (100% stressed)
is ~30 and rises to ~78 for sentences with alternating
stressed and unstressed words (50% stressed). The
Grandfather passage (Darley, Aronson, & Brown, 1975)
contains about 60% stressed words so PVI values below
30 indicate equal and excess stress.
Most of the recommended acoustic measures of
speech can be made using free downloadable speech
acquisition and analysis programs, such as PRAAT (http://
Audrey McCarry
(top), and Kirrie
J. Ballard
Table 1. Demographic and injury data for the three participants with dysarthria and three age- and gender-matched control participants
Participant
Age Sex PTA
(months)
CT results
TPO
Injury
Dysarthria ASSIDS
Participant 1
39 M 3.5
Large left SAH and SDH and 10 mm
midline shift, craniotomy and evacuation of
haemorrhage
3
Fall
Mild-
moderate
Spastic
84% (single
words)
94%
(sentences)
Control 1
41 M
Participant 2 27 F
1
Left occipital penetrating wound with bullet
fragmentation and swelling of bilateral
cerebellar hemispheres, SAH and SDH
surrounding occipital lobes and cerebellar
hemispheres, left parietal craniectomy and
debridement of foreign body
18
Focal
open
head
injury
Moderate
Ataxic
86% (single
words)
95%
(sentences)
Control 2
30 F
Participant 3 26 M 6
EDH, left SDH, base of skull, temporal
and sphenoid fracture, left cerebellar
haematoma, bilateral craniotomy,
hydrocephalus and meningitis, CSF
drainage and ventriculoperitoneal shunt
15 Motor
vehicle
accident
Severe
Flaccid
26% (single
words)
Sentences not
attempted
Control 3
25 M
Note: PTA: post-traumatic amnesia; TPO: time post-onset; ASSIDS: Assessment of Intelligibility for Dysarthric Speech (Yorkston, Beukelman &
Traynor, 1984); SAH: subarachnoid haemorrhage; SDH: subdural haemorrhage; EDH: extradural haemorrhage; CSF: cerebrospinal fluid.