Eye-Tracking of Social Information Processing As an Outcome Measure for Clinical Trials

Thursday, May 12, 2016: 11:30 AM-1:30 PM
Hall A (Baltimore Convention Center)
D. Yurovsky1, G. W. Gengoux1, A. Y. Hardan1, T. W. Frazier2 and M. C. Frank1, (1)Stanford University, Stanford, CA, (2)Cleveland Clinic Center for Autism, Cleveland, OH
Background: Social information processing is both a central impairment in autism spectrum disorder (ASD) and also a key target for clinical intervention. Measures of social information processing typically rely either on parent reports or in-person interactions in an unfamiliar laboratory context, neither of which is ideal for providing objective and generalizable measurements. Parent report may be biased (and will likely not be blind to treatment condition for many treatments); clinical interactions, especially with an unfamiliar experimenter, can underestimate social information processing abilities because of the challenging nature of the context. Eye-tracking measures provide one promising alternative approach for measurement: Eye-trackers are inexpensive, but yield substantial amounts of high-reliability data in even a short measurement session. Thus, eye-tracking may be an important tool for assessing the effectiveness of social information processing interventions.  

Objectives: Our goal was to develop a fast, reliable, low-demand eye-tracking measure for estimating children’s social information processing and word learning abilities, and to use this measure as an exploratory endpoint for an ongoing clinical trial.

Methods: We created a 5-minute video of actors engaging in monologues and dialogues, in which the labels for two novel toys were introduced. While children watched these videos, their eye-gaze was tracked using a 120Hz SMI corneal reflection eye-tracker. To quantify children’s ability to follow speakers’ social cues, we measured their proportion of time looking to the toy that the actors were talking about. Subsequently, we measured whether children learned the correct label for each toy via their gaze during a two-alternative word recognition test.

Results: We first used our paradigm to measure social information processing in a group of 236 typically developing children aged 1-5. We found that individual differences in following social cues during the monologues and dialogues were highly correlated with individual differences in word learning (r = .56). We then tested a group of 40 1-8 year-old children diagnosed with ASD. As expected, these children were less skilled at both following social cues and learning the words for toys than their typically-developing counterparts (Figure), but their social information processing was again highly correlated with their ultimate word learning (r = .6). Finally, in on-going work we have used this measure as part of a randomized controlled trial of a Pivotal Response Treatment program (PRT-P). So far, 18 children have been tested 3 times at 6-month intervals. Preliminary analyses show that children in the intervention group performed better on both social information processing and word learning relative to the control group at the second and third time-points.

Conclusions: Individual differences in social information processing can be quantified reliably with short free-viewing eye-tracking measures. These individual differences predict learning in both typically-developing children and those with ASD. Further, preliminary results from our on-going randomized controlled trial show that these measure are sensitive enough to pick up effects of a PRT intervention. Our measure---and measures like it---are thus a promising direction for the rapid assessment of changes in social information processing targeted by clinical interventions.