Speech Pitch More Variable in Autistic Children Than Typical Children

Daniel M. Keller, PhD

May 27, 2010

May 27, 2010 (Philadelphia, Pennsylvania) — Contrary to a common impression of monotonic speech in autism, children with autism spectrum disorder (ASD) were found to have a significantly greater variability in pitch compared with controls, according to a study presented here at the 9th Annual International Meeting for Autism Research. Thus, increased pitch variability may be a marker for ASD among children who can speak.

Israeli researchers were able to identify ASD with greater than 80% reliability by computing pitch across time and normalizing pitch histogram peaks as a measure of pitch variability.

Lead researcher Yoram Bonneh, PhD, senior research associate in the Department of Human Biology at the University of Haifa in Israel, said the original goal of the study was to quantify the abnormal voice quality and speech prosody (rhythm, stress, and intonation) often observed in autism "to get a quantitative measure that is not related to high-level things like social interactions and cognitive aspects."

The researchers studied 41 children with ASD who could speak (80% boys) and 41 controls (age range, 4-6 years; mean age, 5 years). They recorded the children's speech during 60-second periods while the children named a sequence of 36 pictures of simple, common objects that the experimenter pointed to. The speech spectrum and spectral variability were determined for each child by 32-millsecond time windows and using 31-Hz steps to give a spectral variability across time.

All the children could name all the objects without any apparent differences in speed between the groups.

"In the autistic children there was much more pitch variability...and drift of the pitch, while in the control child it's more stable," Dr. Bonneh said. Histograms of pitches between 0 and 400 Hz showed a much wider and shallower frequency distribution for children with ASD compared with controls, where the peaks were higher, sharper, and often showed 2 distinct peaks. In addition, there was more variability at certain frequencies for the ASD group.

By plotting the standard deviations in pitch (a measure of the pitch variability) and their numbers of occurrences for the 2 groups and by performing other transforms of the data, the researchers were able to accurately classify a subject as having ASD 85% of the time. Among all the subjects, they misclassified only 1 as autistic (1.2% false-positive rate). Ten autistic children were missed and classified as normal. No differences in results were seen for boys and girls in the control group.

Dr. Bonneh proposed 2 hypotheses to account for the study's findings. Each points to an underlying sensory motor deficit as opposed to high-level deficits in pragmatic or social interactions. (Pragmatic deficits refer to impairments in the semantic aspects of language and the appropriate use of language in social situations.)

First, Dr. Bonneh proposed that the control of pitch during speech may likely involve auditory feedback, suggesting an abnormal interaction between speech reception and production in autism.

"In typical speech models, you assume that there is a copy [in the brain] of what you're trying to produce compared with what you hear, and the error signal is used to correct your speech," he said. Problems with this mechanism would result in poor control over speech production, including pitch and pitch variability.

An alternative hypothesis is that there is more "neural noise" in the autistic brain, muddying the channels that transmit information from the speech model in the brain that one uses as a reference for the speech produced. "In order to compute the error signal now, the child generates a lot of pitch changes in order to overcome this noise," Dr. Bonneh explained.

Finally, a deficit in auditory feedback could also be the result of an inability of the child with ASD to "multitask," specifically, a problem of hearing and speaking at the same time.

D. Kimbrough Oller, PhD, professor of audiology and speech language pathology at the University of Memphis in Tennessee, said the results are not surprising to him in light of his own work on language development.

"It's quite clear that vocalization differences between autistic and nonautistic children exist. The problem is isolating them and characterizing them in a way that makes theoretical sense in terms of development and in terms of speech science," Dr. Oller said. "I'm not sure that either speech variability or spectral variability will be the best measures of the differences between autistic and nonautistic speech, but I'm not at all surprised to see them working."

He noted that many measures of speech differentiate autistic from nonautistic speech, and possibly a panel of tests could be helpful as a part of schemes for early diagnosis. "I think that they should be incorporated in the very near future, [and] we should see the possibility of them contributing to diagnosis and screening," he said.

Dr. Bonneh considers his study a first step toward the development of speech spectrum–based tools for early diagnosis of ASD. Some remaining questions he has are as follows:

  • Could speech spectrum analysis be used for early diagnosis (even of "baby babble")?;

  • Could speech spectrum differences reveal subclassifications of autism?;

  • Could speech spectral analysis help probe auditory perceptual deficits?; and

  • Are these findings specific to ASD or may they have wider application (eg, in dyspraxia, dyslexia, or ADHD)?

Dr. Bonneh and Dr. Oller have disclosed no relevant financial relationships. One coauthor (Yoram Levanon) owns a company that performs speech analysis. The work was initially funded by Cure Autism Now, now part of Autism Speaks.

9th Annual International Meeting for Autism Research (IMFAR): Poster 105.021. Presented May 20, 2010.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.