Seeing Things That Aren't There: Perception of Faces and Objects in Visual White Noise in Autism and Asperger's Syndrome

Background: Typical adults exposed to unstructured white noise report having perceived a smile on a face when they had been told that there might be a smiling face in the noisy image (Gosselin & Schyns, 2003). This induced illusory (“superstitious”) perception does not originate from the input signal (i.e. white noise), but from top-down influences on visual analyses of uninformative signals. Such tasks can help answer questions regarding the known but relatively unquantified atypical top-down processing mechanisms of persons with autistic spectrum diagnoses.

Objectives: Our goal was to investigate whether superstitious perception differed across groups of autistic and Asperger syndrome (AS) individuals and their age-and IQ-matched typical controls, as a means of discerning differences in top-down processing of visual stimuli in these populations.

Methods: Three groups of participants, autistic (N=13, Age=26 yrs, FS-IQ=104), AS (N=10, Age=24 yrs, FS-IQ=105), and typical (N=12, Age=24 yrs, FS-IQ=108) adults, were presented with face (half male, half female) and object (half animate, half inanimate)  images embedded in a fixed level of visual white noise, as well as with images of pure visual white noise. At the beginning of half of the trials, the textual cues of “FACE” and “OBJECT” were visually displayed prior to the presentation of a category-consistent stimulus or a visual noise stimulus. Participants were instructed to make a male/female judgment whenever a face appeared and an animate/inanimate judgment whenever an object appeared. Conversely, if they were uncertain of what they had perceived, they were to give an answer of “I don’t know”. Participants indicated their responses by pressing one of three buttons. Proportions of responses to individual categories of presented stimuli were analysed via separate mixed design analyses.

Results: When presented with objects in noise, AS participants showed a significant tendency to give “I don’t know” responses more often than the other two groups for both cued (p=0.006) or non-cued stimuli (p=0.03). When presented with white noise images, typical adults were significantly more likely than both autistic and AS participants to report face or object perception (by giving a response of “male”, “female”, “animal” or “object”). Autistic and AS groups were further distinguished from one another. Whereas autistic subjects gave comparable proportions of both face/object responses and “I don’t know” responses, nearly all of the responses to noise stimuli by AS participants were “I don’t know” (p<0.001 all cases).

Conclusions: Behaviourally, a gradient of susceptibility to superstitious perception can be inferred:  typical adults are highly susceptible; autistic adults somewhat susceptible; and AS adults appear unsusceptible. Perhaps more importantly, the present data suggest that while top-down processing drives typical visual perception, bottom-up processes mediate the perception of persons with autistic spectrum diagnoses. In the specific case of AS adults, perception may be driven exclusively by bottom-up processes.

Funding provided by an Autism Speaks Postdoctoral Fellowship Award to the first author and a NAAR operating grant to the third author.