Distribution of Visual Attention When Comparing Paired Faces in Typically Developing Infants and Infants Later Diagnosed with Autism

Background: Previous prospective research with infants later diagnosed with an autism spectrum disorder (ASD) has shown differences in how they scan a face (Jones & Klin, 2013). Many infant paradigms of face perception and memory require the infant to compare faces that vary on either novelty, or perceptual aspects such as emotion, gender, and attractiveness. No research exists on whether these infants also differ in how they compare or scan more than a single face.

Objectives: To study how 11-month old infants who were later diagnosed as being typically developing (TD), non-typically developing (NT) or having autism (ASD) differed in their eye tracking when presented with two faces to compare.

Methods: A prospective study design compared the distribution of visual attention of typically developing infants (n=49), infants assessed as non-typically developing (n=7) and infants diagnosed with ASD (n=7) at either 24, 36, or 48 months of age.  Eye tracking data were collected at 11 months of age while all infants completed a visual paired comparison task. Stimuli were face pairs of two different females who had similar (neutral/positive) facial expressions.  Infants received six different pairs of faces to compare.  The infants’ visual attention both within a face and across the pair of faces was determined by designating 14 areas of interest (AOIs) (see Figure 1).

Results: All three groups spent significantly (based on t-tests) more time looking at the bottom as opposed to the top half of the faces. Comparisons of the amount of looking to the left vs. right side of the face showed that the TD group looked longer to the right side of the face demonstrating a previously reported left visual field (LVF) bias (t(46)=1.98, p=.05). In contrast, this LVF bias was not seen in the other two groups. With respect to paired face comparisons, there were group differences in the proportion of congruent mouth-to-mouth saccades (F(2, 58)=6.71, p<.00). Infants with ASD made the largest number of mouth-to-mouth comparisons (M=.17, SD=.14), followed by the TD group (M=.06, SD=.09), and the NT group (M=.00, SD = .00). Group difference were also found in the proportion of scans that went from a non-internal facial feature of one pair member to a non-internal feature of the other pair member (F (2, 58) = 4.93, p = .01). Essentially, NT developing infants made more (M = .47, SD = .27) (t (52) = -2.950, p < .00) extraneous comparisons than did TD (M = .22, SD= .20) infants.

Conclusions: First, similar to previous studies of infant sibling populations (Dundas et al., 2012), only the typically developing infants displayed a LVF bias at 11 months of age. This suggests that the LVF bias observed during the viewing of singularly presented faces remains intact during the scanning of face pairs for typically developing infants. Second, differences found in visual attention distribution and saccade congruency were predictive of diagnostic outcome in early childhood. In particular, the mouth region seems primary in discriminating typically developing, non-typically developing, and autism outcome groups.