Stimulus Control Procedure for Reducing Vocal Stereotypies in an Autistic Child

Stereotyped vocal behavior exhibited by a seven-year-old child diagnosed with autism spectrum disorder and maintained by automatic reinforcement was placed under stimulus control through discrimination training. The training consisted of matching a green card (SD) with free access to vocal stereotypy and a red card (SD-absent) with interruption of stereotypy and vocal redirection. At the same time, appropriate behaviors were reinforced. After discrimination training, the child rarely engaged in vocal stereotypy in the red card condition and, to a greater extent, in the green card condition, demonstrating the ability to discriminate between the two different situations. After the training, the intervention began. Once they reached the latency criterion in the red stimulus condition, the child could have free access to vocal stereotypy (green card condition). The latency criterion for engaging in stereotypy was gradually increased during the red card condition and progressively decreased during the green card condition. The intervention follows a changing criterion design. This study indicates that stimulus discrimination training is a useful intervention to reduce vocal stereotypy in an autistic child.

Artificial Intelligence for the Measurement of Vocal Stereotypy

Both researchers and practitioners often rely on direct observation to measure and monitor behavior. When these behaviors are too complex or numerous to be measured in vivo, relying on direct observations using human observers increases the costs of conducting research and monitoring interventions in practice. To address this issue, we conducted a proof of concept examining whether artificial intelligence could measure vocal stereotypy in individuals with autism. More specifically, we used an artificial neural network with over 1,500 minutes of audio data from eight different individuals to train and test models to measure vocal stereotypy. Our results showed that our artificial neural network performed adequately (i.e., session-by-session correlation near or above .80 with a human observer) in measuring engagement in vocal stereotypy for six of eight participants. That said, researchers need to conduct additional research to further improve the generalizability of the approach.