Diadochokinetic speech in individuals at clinical high risk for schizophrenia

Abstract

Past work has shown that individuals at clinical high risk (CHR) for developing psychosis have difficulty with motor coordination (e.g., Mittal etal., 2013). Here we ask whether these motor disruptions also affect their ability to produce speech, resulting in more variable speech productions. CHR participants and matched controls completed a diadochokinetic speech task, in which they first repeated syllables (i.e., papap…, tatat…, kakak…) and then alternated between different syllables (i.e., pataka....) as quickly and accurately as possible. We detect and localize the acoustic syllables /pa/, /ta/ and /ka/ automatically using a deep learning algorithm that was specially designed for the detection of `acoustic objects', discrete events in the acoustic signal (Segal et al., 2019). We predict that, due to less stable control of their articulators, CHR individuals will show increased acoustic variability and more overlap between different sound categories relative to healthy controls.