Abnormal speech spectrum and increased pitch variability in young autistic children

Abstract

Children with autism spectrum disorder (ASD) who can speak often exhibit abnormal voice quality and speech prosody, but the exact nature and underlying mechanisms of these abnormalities, as well as their diagnostic power are currently unknown. Here we quantified speech abnormalities in terms of the properties of the long-term average spectrum (LTAS) and pitch variability in speech samples of 83 children (41 with ASD, 42 controls) ages 4-6.5 years, recorded while they named a sequence of daily life pictures for 60 s. We found a significant difference in the group's average spectra, with ASD spectra being shallower and exhibiting less harmonic structure. Contrary to the common impression of monotonic speech in autism, the ASD children had a significantly larger pitch range and variability across time. A measure of this variability, optimally tuned for the sample, yielded 86% success (90% specificity, 80% sensitivity) in classifying ASD in the sample. These results indicate that speech abnormalities in ASD are reflected in its spectral content and pitch variability. This variability could imply abnormal processing of auditory feedback or elevated noise and instability in the mechanisms that control pitch. The current results are a first step toward developing speech spectrum-based bio-markers for early diagnosis of ASD.

Keywords: ASD; autism; pitch variability; speech.

Figure 1

Long-term average spectrum (LTAS) analysis of 1-min speech (naming) of ASD and control children. (A) Examples of normalized LTAS of five controls (C1–C5) and five ASD children (A1–A5). Note that the controls have higher and more numerous spectral peaks, whereas the ASD curves are shallower. (B) Group average of normalized LTAS for 42 controls and 41 ASD children. Error bars denote one SE of the mean. (C) Group average of the stability of the spectra across time (SD divided by the power) for each frequency. Note in (A) that the controls exhibit sharper peaks and a more periodic (harmonic) structure of the spectra. Note in (C) that the ASD spectra are in general more variable across time.

Figure 2

Pitch analysis of 1-min speech (naming) of ASD and control children. (A) Examples of 15-s pitch time courses for one control (top) and one autistic (bottom) children, demonstrating the difference in variability. (B) Group averages of pitch range and SD. (C) Examples of pitch occurrence histograms (across time in 10-ms windows) in five ASD children (A1–A5) and five controls (C1–C5); the same 10 children shown in Figure 1. The x-axis denotes pitch frequency (Hz) and the y-axis denotes a normalized occurrence histogram across time. (D) Group average of the pitch histograms for 41 ASD children and 42 controls. Error bars denote one SE of the mean, with points around 220 Hz showing a highly significant difference (p < 0.002). (E) A scatter plot for the pitch histogram height (log units, x-axis) and pitch SD (y-axis), with each point corresponding to one child. (F) Occurrence histogram for the data in (E) along the x-axis (pitch histogram maxima) with a Gaussian fit for each group, presented in SD units (average across groups) around the average of the means. The difference between the groups corresponds to a d′ of 1.76 and a threshold criterion allows 86% success in classification, with a sensitivity of 80%, a specificity of 90%, and a positive predictive value of 89% (four controls classified as ASD and eight ASD as controls).