Eye Tracking as a Measure of Responsiveness to Joint Attention In Infants at Risk for Autism

Background: Response to joint attention (RJA), as assessed by socially interactive measures like the Early Social Communication Scales (ESCS), is a key predictor of language skills in typically developing children and those with autism (Sigman & Ruskin, 1999). Reduced RJA in infancy is also predictive of an autism diagnosis (Rozga et al., 2010). Therefore, accurate measurement of RJA during infancy may facilitate early detection of autism. Because the ESCS yields a relatively blunt measure of RJA, eye-tracking assessments of RJA have recently been developed to assess the microstructure of eye movements associated with RJA (Senju & Csibra, 2008). Because most eye tracking assessments of RJA utilize pre-recorded video of a model turning towards objects, they are more consistent across administrations than the ESCS and thus potentially useful as standardized prognostic measures. However, opportunities for RJA that children typically experience are highly interactive. Do eye tracking assessments of RJA measure the same construct as RJA as assessed by the ESCS?

Objectives: The current study aimed to assess the ecological validity of an eye-tracking assessment of RJA by administering the ESCS and eye-tracking RJA to 39 eighteen-month infant siblings of autistic children, a sample therefore at heightened risk for diagnosis (Bailey et al., 1993). Relationships between both RJA measures and concurrent language scores were assessed. We predicted that both assessments of RJA would be correlated, and would be related to language skills.

Methods: During eye-tracking RJA, each infant watched a video of a model fixating 1 of 2 objects with direction of turn counterbalanced across trials. Eye movements were recorded with a Tobii 1750 eye tracker. Each trial consisted of a baseline phase, an infant-directed greeting and smile, and the model turning and fixating an object for 5 seconds. RJA was calculated by summing the duration of all looks from the model’s face to the correct object relative to the duration of looks from the face to either object. The ESCS yields two measures of RJA: a distal task wherein the examiner visually orients and points towards a poster and a proximal task wherein the experimenter points to pictures in a book directly in front of the infant. Language was assessed with the Mullen Scales of Early Learning. Total scores on the Autism Diagnostic Observation Schedule were used as an index of symptomatology

Results: The eye-tracking RJA was correlated with RJA during the distal ESCS task (p(39)=0.33, p<0.05) but not the proximal task. Only RJA during the proximal ESCS task was positively correlated with expressive (p(36)=0.51, p<0.01) and receptive language scores (p(36)=0.37, p<0.05).  When we controlled for autistic symptomatology, eye-tracking RJA and RJA–ESCS-distal remained correlated (r(31)=0.49, p<0.01); and RJA-ESCS-proximal was correlated with eye-tracking RJA (r(31)=0.45, p<0.01) and expressive language (r(31)=0.57, p<0.01).

Conclusions: While eye-tracking RJA and ESCS RJA appear to measure a similar construct, only proximal ESCS RJA related to language. The current study establishes the construct validity of eye-tracking RJA and suggests that it may provide a useful prognostic tool for assessing autism risk.