Automatic Detection of Emotional Prosody in Children and Adults with Autism Spectrum Disorders

Thursday, May 11, 2017: 5:30 PM-7:00 PM
Golden Gate Ballroom (Marriott Marquis Hotel)
M. Gomot1, J. Charpentier2, J. Malvy3, F. Bonnet-Brilhault1, E. Houy-Durand1 and M. Latinus2, (1)UMR930 Inserm, Université François-Rabelais de Tours, Tours, France, (2)UMR930, INSERM, Université François –Rabelais de Tours, Tours, France, (3)CHRU Tours, Tours, France

Although intolerance of change is a main feature of Autistic Spectrum Disorder (ASD), the brain processes underlying this aspect of the disorder remain poorly understood. This oversensitivity to changes may lead to an inability to adapt to new sensory inputs, and especially to the ever changing social environment. In this respect the processes involved in pre-attentional detection of changes in stimulus features have been investigated in ASD using the mismatch negativity (MMN), an event-related potential that reflects error detection caused by a deviation from a learned regularity. However, although ASD is also characterized by an inability to detect and adapt to changes in emotional states, only few MMN studies in this population have investigated automatic deviancy detection in an emotional context. These studies used different deviant emotions that where always compared to neutral standards, and showed that eMMN was smaller or less lateralized in ASD than in controls. Objectives:

In order to determine if particularities observed in previous studies are related to emotion processing or to change detection abnormalities, the present study addresses direct comparison of automatic change detection of neutral and emotional deviants with strictly controlled acoustic parameters, in children and adults with ASD.


Thirty-two adults (16 typical; 16 ASD) and thirty 7-12 years old children (15 typical; 15 ASD) were presented with two voice deviants, neutral (p=.085) and angry (p=.085) (prosodic variation of the vowel ‘aaa’) embedded in a repetitive neutral stimuli sequence. Brain electrical activity was recorded using a 64-channels Biosemi EEG system, while participants were watching a silent movie. A genuine MMN was measured for each condition as the subtraction of the stimuli presented in an equiprobable sequence from the same deviant stimuli in the oddball sequence.


Comparisons of age groups and conditions during typical development indicate a late maturation of the brain processes associated with emotional prosody discrimination. Findings in both children and adults with ASD revealed an atypical processing of both emotional and neutral deviancy. In children with ASD, the amplitude of the responses to voice stimuli from the equiprobable sequence were significantly smaller than in controls for both the neutral and the emotional prosody, as where the responses to deviancy (nMMN and eMMN). In adults with ASD responses to voice stimuli were found normal, for both the neutral and the emotional prosody. However the MMN to neutral and emotional deviancy presented atypical brain distributions.


The detection of prosody deviancy was found altered in children with ASD, together with atypical responses to voice, and this regardless of expression. In adults despite a normalization in voice processing, the detection of vocal deviancy remains affected.

Altogether these findings suggest that voice and prosody processing can improve with age in ASD, but that the lack of expertise in vocal encoding during childhood would lead to a long lasting inability to discriminate between different vocal emotions. This study points to specific difficulty in the online processing of emotional changes in ASD which potentially plays a crucial role in social interaction deficits regardless of age.