Increasing the Accuracy of Detection of Audio-Visual Integration Differences in Autism Spectrum Disorders

Background:  

Individuals with autism often have difficulty integrating information across auditory and visual modalities (Bebko et al, 2006; Iarocci & McDonald, 2006). A commonly-used paradigm involves presenting 2 side-by-side visual displays, with an auditory track matched to only one display, while measuring the child’s preferential looking toward the synchronized screen. Using this technique, Bebko et al. (2006) identified a deficit in the processing of speech specific audio-visual intermodal information in children with autism. In many previous studies, preferences are represented by only 60% looking rates, even though chance looking patterns are 50% with two screens. We introduced a modification using a 4-screen display and presented pilot data previously at IMFAR (Hancock et al., 2008). Better discriminability of looking patterns was found, since this methodology lowered chance looking probabilities to 25%, resulting in larger magnitudes of differences from chance. The present study further investigates precision of the 4-screen technique (with eye-tracking), using a larger ASD sample size and including a TD comparison group. 

Objectives:  

 To assess the precision of the 4-screen versus 2-screen preferential looking paradigm (PLP). This is done in two stages. First, magnitude of effect (degree of difference from chance) on the 2- and 4-screen tasks for both ASD and TD groups is compared. Second, differences in looking patterns between ASD and TD groups are investigated using the 4-Screen PLP and eye-tracking data with linguistic and non-linguistic stimuli, and these results are compared with previous differences on 2-screen studies.

Methods:  

23 children with ASD and 23 TD children between the ages of 3 and 12 years were tested. The 4-Screen PLP involved displaying four identical videos, offset in time, with an auditory track synchronous to only one of the videos. Videos contained either linguistic (L, person telling a story) or non-linguistic (NL, hand playing a piano) stimuli. Preferential looking is assessed from differences in total time looking at each screen.

Results:  

Initial analyses indicate that the 4-screen variant produces more discriminable magnitudes of effect compared to chance: approximately 40% greater looking time at the preferred video over chance, vs. the 2-screen paradigm (approximately 10% greater than chance). Corrected for their relative chance levels (.25 vs .50), this is an 8-fold increase in precision in the ability to detect preference, which represents a major advance in sensitivity for the paradigm .  Two other results from previous two-screen studies were also preserved: First, for non-linguistic stimuli, both groups showed similar proportions of preferential looking to synchronous non-linguistic videos. In addition, we replicated the finding of an apparent speech-specific deficit in intermodal processing (IMP). Further analyses are ongoing.

Conclusions:  

These findings demonstrate the strongly increased sensitivity of using a 4-screen over a 2-screen paradigm in detecting differences in preferential looking in ASD and TD children with audio-visual information. Often subtle differences occur, and the ability to detect them will enable a clearer understanding of variables contributing to the language difficulties in ASDs. The corroboration of our previous findings of an apparent speech-specific IMP difficulty in ASDs lends additional confidence to the paradigm.