Voice Patterns of Turn-Taking Interactions in Adults English Speakers with Autism Spectrum Disorder

Background: Individuals with Autism Spectrum Disorder (ASD) reportedly display atypical modulation of speech described as awkward, monotone, or sing-songy (Shriberg et al., 2001). These patterns are robust indicators of social communication deficit (Paul et al., 2005) and contribute to reaching a diagnosis of ASD. Fusaroli et al. (2013 ; IMFAR 2014 ; IMFAR 2015) showed that Recurrence Quantification analysis of acoustic features could be used to successfully identify voice patterns characteristic of adults with ASD and train machine learning algorithms to accurately (80-86%) discriminate autistic from non-autistic speakers in directed speech recordings.

Objectives: Our aim was to replicate and extend the results obtained by Fusaroli et al. (2013, 2014, 2015) in a turn-taking interaction, i.e. (1) characterise the speech patterns of adults with ASD in a Q&A setting, (2) characterise the corresponding changes in the interlocutor’s speech patterns, and (3) employ the results in a supervised machine-learning process to determine whether acoustic features predict diagnosis and severity of the symptoms. We were also interested to evaluate how valid the model built based on directed speech data would be on turn-taking data

Methods: The context of a previously published study of memory in ASD (Maras et al., 2013) provided audio recordings of 17 ASD and 17 matched Typically Developing (TD) adults recalling details of a standardised event they had participated in. Part of the recording consisted in a Q&A between experimenter and participant, i.e. a turn-taking interaction. Transcripts were time-coded, and pitch (F0), speech-pause sequences and speech rate were automatically extracted. We conducted traditional statistical analysis on each prosodic feature. We then extracted non-linear measure of recurrence: treating voice as a dynamical system, we reconstructed its phase space and measured the number, duration and structure of repeated trajectories in that space (Marwan et al., 2007). The results were injected to train a linear discriminant function algorithm to classify the descriptions as belonging either to the ASD or TD group. The model was developed and tested using 1000 iterations of 10-fold cross-validation (to test the generalizability of the accuracy) and variational Bayesian mixed-effects inferences (to compensate for biases in sample sizes).

Results: Preliminary analysis of a subset of recordings suggest similar results to those obtained with directed speech: ASD individuals ASD produce highly regular speech patterns organized in frequently repeated short sequences (200-400 ms), supporting clinical reports of monotony. While features are similar across modes of communication, the coefficients discriminating individuals with ASD and controls need to be re-trained in the context of a turn-taking interaction. Interestingly, the interlocutor’s speech patterns are as informative about diagnosis as the participants’.

Conclusions: The current data suggest than ASD adults produce highly regular patterns of speech in turn-taking interaction. Importantly this measurement captures some aspects of the clinical reports, which contribute to reaching a diagnosis of autism. Further analysis will establish whether voice patterns in turn-taking interaction are particularly distinct in ASD compared to directed speech, indicating whether interactive conversation exacerbates voice pattern atypicalities, reflecting the reported “awkwardness” in the interaction with ASD individuals ASD.