Seminar in Computational Linguistics

In the project Functional consequences of misarticulation in children’s connected speech we examine associations between different types of misarticulation (or ”speech errors”) in children’s speech, and their consequences when occurring in spoken language. Simply put, the central question is: ”Which speech errors are the worst?”. On the one hand, we have explored this question through cross-linguistic corpus investigation, where we have simulated speech errors frequently occurring in children’s speech, and studied the linguistic effects of these simulations. On the other hand, we have explored the question by examining listeners’ reactions when listening to recordings of children with speech sound disorders. We have employed an audience response system-based method to examine if specific speech errors evoke more listener reactions, either in terms of not understanding the spoken message (i.e. intelligibility), or in terms of perceiving speech as odd (i.e. acceptability). As these aspects are not only dependent on the nature of the speech, but also on factors tied to the listener, we have included listeners from different groups: speech-language pathologists, other adults, and 6-10-year-old children. I will present some of our findings thus far, and discuss what clinical and theoretical implications these may have.