Acoustic markers of dysarthria in children with cerebral palsy: comparison of speech tasks

Frits van Brenk, Anja Kuschmann

Research output: Contribution to conferencePosterpeer-review


Dysarthria is the most frequent communication impairment in children with cerebral palsy (CP), affecting functional communication in over 50% of children (Parkes et al., 2010). Speech characteristics include shallow, irregular breathing, harsh and/or breathy voice, hypernasality and imprecise articulation (e.g. Ansel & Kent, 1992; Hodge & Wellman, 1999; Nordberg et al., 2014; Workinger & Kent, 1991). Although the presentation of dysarthria in children with CP can vary considerably, it is generally assumed that at least one - but often all - speech subsystems, i.e. respiration, phonation, resonance and articulation, are affected by the motor control issues. Treatment approaches for children with dysarthria due to CP focus on improving intelligibility, which is why considerable research efforts have been made to identify those features that impact most on intelligibility. Perceptual evaluations of speech produced by children with dysarthria and CP point to difficulties with articulation, voice quality and speech rate as the primary contributors to reduced speech intelligibility (e.g. DuHadway & Hustad, 2012 ; Nordberg, et al., 2014; Workinger & Kent, 1991). In order to quantify these perceived deviations, recent research has turned to acoustic measures, and identified deviations in articulation rate and F2 range as some of the main speech features in children with CP to differ from typically-developing children (e.g. Allison & Hustad, 2018; Lee et al., 2014). Measuring acoustic correlates offers the advantage of objectively capturing those changes to the acoustic signal that lead to the perception of impaired speech characteristics in children with dysarthria and CP. In addition, the measures allow the quantification of differences in speech features produced by children with CP and their typically-developing peers. To date, acoustic characteristics of speech in children with dysarthria and CP were analysed using sets of single words or short sentences. Research on adult dysarthria has shown the potential of measuring acoustic features in connected speech to capture and identify changes in speech production, e.g., vowel space area (Rusz et al., 2013), articulatory working space (Yunusova et al., 2005), and long term average spectra (Tjaden et al., 2010). In addition, studies have shown that speaking tasks have an influence on articulation with more complex tasks such as monologues being more likely to elicit articulatory difficulties (Rusz et al., 2013). This raises the question whether analysing acoustic features in speech produced by children with dysarthria could yield similar information to draw conclusions as to their segmental and suprasegmental speech features. The current study aims to answer this question by identifying acoustic markers that may aid in the characterization of speech in children with dysarthria due to CP, and by evaluating the suitability of different functional speech tasks side by side. Methodology Speakers: Audio recordings were taken from eight children with dysarthria due to CP and eight age-, gender- and dialect-matched typically-developing children. (CP: six boys and two girls, mean age = 12.0 years, range = 7-18 years; CON: six boys and two girls, mean age = 11.8 years, range = 7 – 20 years). The data were collected as part of a project on prosodic abilities in children with CP (Kuschmann & Lowit, 2018). Three children had been diagnosed with dyskinetic CP, two with spastic CP, and two with ataxic CP. The motor speech difficulties experienced by the children with CP ranged from mild to severe. Speech tasks: Acoustic measures were obtained from the following eight speech tasks: two monologue tasks differing in topic, a picture description task, a story retelling task, a set of single words from the Children's Speech Intelligibility Measure (CSIM), sets of single words and two-word phrases from a picture naming task, and a set of single sentence productions. Acoustic analysis: Across the speech tasks, suitable voiced fragments for acoustic analyses were identified and marked. Non-lexical fillers (e.g., uh or um) were excluded. Acoustic measures were quasi-automatically obtained by means of custom scripts executed in the speech-analysis software PRAAT. The following measures were obtained to investigate aspects of vocal intensity, prosody, and articulatory working space: Sound Pressure Level (mean, sd, 90th-10th percentile range), Fundamental Frequency (mean, sd, 90th-10th percentile range), and Second Formant Interquartile Range (3rd quartile – 1st quartile). Statistical analysis: A series of one-way ANOVAs were performed to compare the two groups. In the first step, groups were compared by pooling the outcome of acoustic parameters over the different speech tasks. Subsequently, group differences were evaluated for each task separately. Results and Discussion Preliminary results are reported for two of the children with CP (CP01 with a diagnosis of spastic dysarthria, and CP02 with a diagnosis of ataxic dysarthria) and their matched controls. Across all speech tasks and when taking the speakers with CP together, higher values were found for four outcome measures: F0 SD: F (1, 30) = 10.06, p = .003; F0 Range: F (1, 30) = 7.31, p = .011; SPL SD: F (1, 30) = 42.1, p < .001; and SPL Range: F (1, 30) = 37.3, p < .001. These measures were consistently found to be higher in the children with CP. However, differences emerged when comparing the speaker pairs individually: speaker CP02 showed higher values of F0 SD and F0 Range compared to his control peer, whereas speaker CP01 showed lower measures. These findings appear reflective of the type of dysarthria: relatively lower intensity and fundamental frequency excursions may be founded in the cluster of prosodic insufficiency in the case of the speaker with spastic dysarthria (CP01). Speech production findings of excessive and variable patterns of stress in the speaker with ataxic dysarthria (CP02) may have contributed to the findings of higher intensity and pitch variation, showing the potential of these acoustic measures for the quantification of speech characteristics associated with dysarthria. Furthermore, comparison of speech tasks showed that the monologue tasks, picture description task and the story retelling task accounted for the largest group differences. This finding indicates that these tasks were most effective in distinguishing speaker groups, highlighting the potential of connected speech tasks for the evaluation of speech difficulties in speakers with dysarthria due to CP.
Original languageEnglish
Publication statusPublished - 2018
EventAnnual ASHA convention 2018 - Boston, US
Duration: 15 Nov 201817 Nov 2018


ConferenceAnnual ASHA convention 2018


  • acoustic markers
  • dysarthria
  • cerebral palsy
  • speech tasks
  • comparative study


Dive into the research topics of 'Acoustic markers of dysarthria in children with cerebral palsy: comparison of speech tasks'. Together they form a unique fingerprint.

Cite this