15866
Measurement of Crossmodal Integration of Expressive Affect Communication in Autism

Friday, May 16, 2014
Atrium Ballroom (Marriott Marquis Atlanta)
J. van Santen1, A. Kain1, A. P. Hill1, E. T. Prud'hommeaux2, R. Ludovise1, C. Conway3, G. Keepers1 and E. Fombonne4, (1)Center for Spoken Language Understanding, Oregon Health & Science University, Portland, OR, (2)University of Rochester, Rochester, NY, (3)Columbia University, New York, NY, (4)Psychiatry, Pediatrics & Behavioral Neuroscience, Oregon Health & Science University, Portland, OR
Background:  

There is evidence for atypical functional and anatomical connectivity in autism spectrum disorders (ASDs), as well as for reduced crossmodal integration in the receptive domain – plausibly a manifestation of the former.  The presence of an ADOS code called “Language Production and Linked Nonverbal Communication” suggests that reduced crossmodal integration may also exist in the expressive communication domain. However, there are few formal studies on the latter, due to a dearth of measurement methods.

Objectives:  

The objective is to create a method for measuring crossmodal integration of expressive communication (CIEC), specifically affect communication, and to demonstrate this method on a sample of children with ASDs, Typical Development (TD), and Specific Language Disorder (SLI).

Methods:

Participants: Twenty-five children ages 8-11 participated: TD (9), ASD (10), and SLI (4). Protocol: Children were trained to re-enact a brief story, pretending to be an actor/actress. Each produced between 19 and 46 videorecorded utterances. Modality-specific stimuli: Each utterance recording was processed by software to generate five stimulus types: three audio-only speech-related stimuli (speech; delexicalized speech, i.e., speech rendered unintelligible; text transcripts) and two video-only motor stimuli (facial expression; gesture). Ratings: Eleven undergraduate students rated arousal and valence of each stimulus on 5-point scales. These ratings were subsequently averaged over raters to form pooled ratings. Sequencing: stimuli were blocked by stimulus type and randomized by utterance and child, making it unlikely that raters knew when two stimuli related to the same utterance. Measurement of CIEC: For each individual child, separately for arousal and valence, correlations over utterances were computed between the pooled ratings of the five modality-specific versions of the utterances.  These cross-modal correlationsare proposed as a measure of CIEC.

Results:  

Reliability: The median inter-rater correlation was 0.52. We one-hundred times randomly formed two rater sub-groups and computed corresponding sub-group averages for each stimulus.  The correlations between these averages ranged from 0.825 to 0.875, indicating that ratings pooled over all eleven raters had adequate reliability.   Subsequent analyses used these pooled ratings. Validity:  First, correlations among the motor modalities and among the speech-related modalities were higher than those between motor and speech-related modalities. Second, using multiple regression, speech ratings were well-predicted from delexicalized speech ratings and text ratings. Both facts confirm that the ratings behaved as expected. Presence of CIEC: Combining data over participants, each of the twenty pairwise cross-modal correlations was positive for both valence (mean of 0.31; p<0.0001) and arousal ratings (mean of 0.42; p<0.0001).  Because of the way the stimuli were sequenced and their modality-specificity, these correlations strongly suggest the presence of CIEC. CIEC differences between groups: Cross-modal correlations were significantly larger in the TD group than in the ASD group for valence, but not for arousal.

Conclusions:  

A new paradigm was proposed for measuring crossmodal integration of expressive communication. Affect of individual utterances is rated independently in five modalities, followed by per-child crossmodal correlational analysis. Results provide evidence for the reliability and validity of the rating paradigm, and suggest that, as predicted, crossmodal integration is reduced in children with ASD.