A Computerized Interactive Game for Remediation of Prosody in Children with Autism

Background: People communicate both by what they say (utterance content) and how they say it (prosody). Prosody refers to the rhythm, loudness, timing, stress, and melody of speech. Prosody allows people to emphasize what is important to them. Prosody conveys feelings, intentions, attitudes (e.g. sincerity, contempt); the role someone has for the speaker (e.g., one speaks differently to a son vs. a policeman). Prosody conveys a host of meanings that go beyond utterance content. There is evidence that individuals with autism spectrum disorders (ASD) have prosodic deficits. They may have impaired expressive prosody (e.g., speak in a flat, or sing-song voice) and/or impaired receptive prosody (e.g., unable to grasp feelings conveyed by others through prosody). Such prosodic difficulties may severely affect social communication in ASD.

Objectives: Creation and evaluation of a computer-assisted system for remediation of expressive and receptive prosody in children with ASD.

Methods: (1) Participants. Five children with High Functioning ASD, ages 6-10, participated. (2) System. A computer program was developed that displays -- and allows control by a therapist of -- interactive “drama books” containing videotaped scenarios, each consisting of a series of interrelated scenes enacted by child and adult actors. An identified, on-screen targeted child is shown in each scenario for whom the ASD child will speak. Each drama opens with one scene, and the next scene – out of a possible three or four alternatives – occurs based on a spoken response by the ASD child to the other on-screen characters. Importantly, it is both what and how something is said that drives the selection of events, highlighting prosody’s role in affecting others and changing the course of events. In total, 23 drama books were developed. (3) Protocol. Each child participated in five half-hour remedial sessions, administered by an experienced therapist. Each session involved six drama books. The order of the drama books was randomized between children.

Results: (1) The correlation between session number and appropriateness rating (averaged over subjects) was 0.41 (t(29)=2.369, p<0.0125, one-tailed). A linear model that took into account the effects of differences in difficulty between drama books also yielded a significant result (t(164)=1.847, p<0.035. one-tailed). A permutation test for the same hypothesis (where we permuted the session numbers randomly 10,000 times) yielded p=0.0337, one-tailed. (2) All children thoroughly enjoyed the process. Their behavior also strongly suggested that they experienced the filmed scenarios as having a high level of realism.

Conclusions: (1) The study showed the feasibility of building a system of this type (including a substantial amount of content development that required coordination of a large movie “crew”), and of creating a remediation protocol that children enjoy and that does not put undue strains on the therapist (key here was that the user interface for the therapist was extremely simple). (2) The children improved during the sessions. While it remains to be seen whether this improvement generalizes to behaviors in natural contexts, this is certainly a promising result given that only five sessions were given spaced one week apart.