Objectives: To develop a quantification method for the abnormal speech in ASD and conduct a preliminary assessment of the method on a large sample.
Methods: we recorded 82 children (41 autistic, 41 controls) ages 4 to 6 years (mean 5) while naming a sequence of daily life pictures pointed by the experimenter for 60 sec in a quite room in their preschools. We computed pitch across time and normalized pitch histogram peaks as a measure for pitch variability.
Results: Contrary to the common impression of monotonic speech in autism, the ASD children had significantly larger pitch variability across time. A measure of this variability yielded more than 80% success in classifying ASD in the sample.
Conclusions: Speech abnormalities in ASD are reflected in increased pitch variability during speech. Controlling pitch is likely to involve auditory feedback, and the current findings could imply abnormal interaction between speech reception and production in autism. It could also reflect elevated neural noise in the mechanisms that control speech, providing support to a neural-noise theory of autism, and possibly an early measure of such noise. The current results are a first step towards the development of speech-spectrum based tools for early diagnosis.