Holistic Processing of Unfamiliar Faces and Novel Objects in Autism

Background:

Numerous studies have suggested that individuals with autism may have difficulties perceiving items (e.g. faces) as single, unified percepts (holistic processing), instead defaulting to perceiving the item’s components as discrete parts (feature-based processing). There is some debate as to whether this perceptual difficulty in autism is specific to faces, which are typically processed holistically, or if this difficulty extends to visual processing in general.

Objectives:

1)      Used a novel eye-tracking paradigm to assess the degree to which typically developing (TD) individuals holistically process unfamiliar faces and novel objects.

2)      Investigate holistic processing difficulties in autism, and determine whether they are specific to faces or generalizeable to visual object stimuli.

Methods:

Twenty-eight individuals to date (data collection ongoing; ages 8-21 years) with and without autism have completed an eye-tracking paradigm in which participants were presented with pairs of stimuli from different viewpoints and responded as to whether the two stimuli were the same or different. Four types of stimulus-pairs were presented: low-ambiguity faces (50% morphs), high-ambiguity faces (80% morphs), low-ambiguity novel objects (highly distinct), and high-ambiguity novel objects (highly similar). 72 trials of each condition were presented, and pilot studies matched accuracies across stimulus types.

This eye-tracking paradigm was developed specifically to compare holistic and feature-based perceptual strategies. By comparing the number of within-item saccades (holistic strategy) to between-item saccades (feature-based strategy), one can assess the extent to which an individual relies on a particular visual processing strategy (see Figure 1). Thus, the ratio of within-item saccades to between-items saccades was calculated for each individual with each stimulus type (W/B-ratio), and used as an index of holistic processing.  Higher ratios were indicative of greater use of holistic processing strategies.

Results:

W/B-ratio scores (Figure 2) were compared across stimulus types and diagnostic groups in a three-way ANOVA (diagnostic group x stimulus type x ambiguity level). A main effect of ambiguity was observed (F=9.28, p=0.005), as was a trend towards a main effect of diagnosis (F=2.70, p=0.11). No main effect of stimulus type was observed (F=0.10, p=0.75). Two-way interactions between stimulus type and ambiguity (F=7.95, p=0.01) as well as diagnosis and ambiguity (F=4.94, p=0.04) were observed, but no interaction between diagnosis and stimulus type (F=0.28, p=0.60). Finally, a significant three-way interaction between diagnosis, stimulus type, and ambiguity was observed (F=6.47, p=0.02).

Conclusions:

We confirmed that TD participants processed faces holistically, regardless of ambiguity level. In contrast, for novel objects, they applied a holistic processing strategy when the objects were high-ambiguity, but used feature-based processing when the objects were low-ambiguity. This faces versus objects difference was not observed in individuals with autism. While individuals with autism exhibited a numeric decrease in their holistic processing from high- to low-ambiguity stimuli, this change did not vary according to stimulus domain, suggesting that while TD individuals process faces differently than objects, individuals with ASD did not switch between processing styles based on ambiguity or stimulus type. Instead, their viewing patterns for all stimulus classes reflected an attempt at a holistic strategy.