Visual Shape Discrimination in Autism: Linking Low- and Mid-Level Perception

Background: Studies investigating visual perception in autism have for the most part assessed lower (local) and higher (global) levels of processing in isolation. It therefore remains unknown whether altered perception at one level affects processing in the other. This relationship can be investigated through the use of Radial Frequency Patterns (RFPs; Wilkinson et al 1998; Grinter et al 2010), spatial stimuli that target mid-level visual analysis, a processing level that is involved in object perception. RFPs are closed-contour shapes that can be manipulated to create “bumps” or Radial Frequencies (RFs) along their quasi-circular contours. When RFPs have fewer bumps (i.e. 2, 3, 5 RFs), global processing is needed to discriminate them from a perfect circle, whereas local processing is advantageous when RFPs contain many bumps (i.e. 10 or more RFs). By further manipulating the local physical attributes (i.e. luminance vs texture) defining an RFP’s contour, one can assess whether a local attribute differentially affects global shape perception.

Objectives: To assess whether there is a functional relationship between low level and mid level visual processing in autism. 

Methods: Eighteen autistic and 17 non-autistic participants, matched for full-scale IQ and age (range 14-31 years), were asked to discriminate between perfect circles and RFPs, whose contours contained 2, 3, 5, and 10 bumps, and were either luminance- or texture-defined. The size, or amplitude, of the bumps was varied: the larger the amplitude, the easier it is to discriminate a RFP from a perfect circle. RFP discrimination thresholds were measured using a method of constant stimuli and a 2-ATFC procedure. Participants were asked which of two successively presented stimuli contained the RFP (target); the other was a perfect circle (amplitude = 0). All participants had to complete a total of 8 experimental conditions.

Results: Separate 2 (groups) X 4 (RFs) mixed factorial analyses of variance were conducted for luminance- and texture-defined RFPs. For both analyses, no significant interaction was identified. A main effect of group was found for both luminance and texture conditions (p < 0.01), indicating that autistics performed significantly worse across all RF conditions assessed. However, mean differences between the autistic and the control group were significantly greater across texture-defined RF conditions. As expected, a main effect of RFs (p < 0.01) was also demonstrated across groups, with decreased RFP discrimination found for conditions with fewer (i.e. 2 and 3) RFs compared to more (i.e. 5 and 10) RFs for both autistic and control groups.

Conclusions: Our findings demonstrate that autistics are less able to discriminate visual shapes defined by RFPs, particularly when the contours are defined by texture information. This suggests that the type of local information (luminance vs. texture) defining a shape’s contour differentially affects mid-level perception in autism. Such changes at low level, local visual processing, which are associated with altered mid-level perception, may in turn contribute to autistics’ atypical high level perception of objects and faces.