JCPSLP vol 14 no 3 2012

Figure 1. Waveform and spectrogram for the word “vegetables”, as displayed in PRAAT, with the first vowel highlighted in the waveform (upper panel). In the spectrogram (lower panel), the top overlaid dotted line represents the fundamental frequency as it changes over the word (displayed in blue within PRAAT), the bottom overlaid line represents the vocal intensity (displayed in yellow within PRAAT). Note the erroneous pitch values just prior to the ‘b’ and at the onset of the final schwa.

Acoustic measurements Vocal quality. Vocal quality was assessed during sustained phonation, which represents stable vocal performance with minimal demands for vocal tract adjustments. First, the average duration (msec) was measured over three successive attempts at sustained phonation. Second, a 3-second stretch of the sustained vowel was selected for measurement from the middle of the sustained phonation, not including the first 25 msec or the terminal part of the phonation (Kent et al., 2000). The PRAAT Voice Report function was used to calculate average f0 , standard deviation of f0 , jitter (local), shimmer (local), and HNR (http://www.fon.hum.uva.nl/praat/manual/Voice.html; e.g., Kent et al., 2000). Speech rate and prosody. Speech rate was measured as syllables spoken per second for AMR, SMR, and reading. In connected speech, stress variability was measured with the Pairwise Variability Index (PVI). Instructions for calculating PVI_Dur, PVI_f0 and PVI_dB for the first 20 syllables/vowels in the Grandfather reading are given in Table 2 and Figure 1. Higher PVI values represent greater variation; PVI values close to zero indicate equal stress, monopitch, or measurements using intra-class correlation coefficients (inter-rater reliability: ICC 2, 1, absolute agreement, single measures). Inter-rater reliability was excellent ( ≥ 0.75; Cicchetti, 1994) for vowel duration (ICC: 0.78, 95% CI 0.51 –0.89), vowel peak f0 (ICC: 0.78, 95% CI 0.62 –0.86) and vowel peak dB (ICC: 0.75, 95% CI 0.02 –0.92). Absolute agreement resulted in the wide CI band for vowel peak dB however, the average difference in dB measures was not clinically significant at 2.07 dB (SD = 1.12). Intra-rater reliability was also high for the three measures (ICC: 0.85, 95% CI 0.72 –0.93; ICC: 0.92, 95% CI 0.84 –0.96; and ICC: 1.0, 95% CI 0.99 –0.1, respectively). Results Each participant presented with prototypical perceptual features consistent with spastic dysarthria (P1), ataxic monoloudness (i.e., dysprosody). Reliability of measurement Inter-rater reliability was calculated on all manual

dysarthria (P2), and flaccid dysarthria (P3) (Duffy, 2005). Acoustic measures are presented in Table 3 along with comparative data from healthy age- and sex-matched adults. The nonparametric Wilcoxin Matched Pairs Signed Ranks test was used to compare the PVI for each syllable pair in a patient’s sample with those for the matched control. The relationships between perceptual and acoustic measures are reported below. Participant 1 Vocal quality Duration of sustained ah was reduced, consistent with the reduced respiratory-phonatory control and short phrase length. P1’s average f0 was higher than normal, although low pitch has been more often associated with the increased laryngeal tone of spasticity (Duffy, 2005). While variability of f0 , jitter, and shimmer for the sustained ah production were within normal limits, the harmonics-to- noise ratio (HNR) was slightly below the recommended threshold, indicative of mild vocal hoarseness. This likely relates to the perception of P1 having a strained-strangled voice quality. Speech rate and prosody P1’s speech rate was perceived as mildly slow, consistent with AMR and SMR rates being about 1 syllable/sec below the normal range and 1.7 syllables/sec slower than normal for reading. Prosodic variation in the reading task was measured with the Pairwise Variability Index. PVI_Dur was significantly reduced compared to the control sample, consistent with the perception of mild equalisation of stress. PVI_ f0 and PVI_dB were not significantly different to the control, despite the perception of reduced pitch and loudness variation in the reading sample. Participant 2 Vocal quality Duration of sustained ah was well below the average expected for healthy speakers, suggestive of poor respiratory-phonatory control. P2 displayed irregular pitch breaks and vocal tremor. Average f0 was within the normal range but standard deviation of f0 was very high, possibly influenced by brief pitch breaks. Jitter was below the

132

JCPSLP Volume 14, Number 3 2012

Journal of Clinical Practice in Speech-Language Pathology