Abstract
Integrating multiple sensory estimates while weighting each according to its own reliability can minimise the uncertainty of the overall estimate. Whilst human adults are able to combine multiple sources of sensory information optimally, recent studies indicate that in typical development this ability does not develop until late childhood (Nardini, Bedford & Mareschal, PNAS 2010). Unusual sensory responses were described in Kanner's (1943) first reports of autism, and multisensory integration has been investigated in several studies (e.g. DeGelder, Vroomen, & Van der Heide, 1991; Keane, Rosenthal, Chun and Shams, 2010). The current study examined how adolescents with autism integrate two visual cues: texture and disparity cues to surface slant. Participants with and without autism (N = 13, N = 9), aged 12–16 years, were matched on gender and general cognitive ability (Wechsler Abbreviated Scales of Intelligence). We used a 2AFC procedure to measure participants' 75% discrimination thresholds for which of two planes was the more slanted based on disparity and texture cues, singly (D, disparity, slant signalled by disparities of random dots; T, texture, slant signalled by texture gradients of tiled surfaces viewed monocularly), or in combination (DT, slant signalled by disparities and texture gradients of tiled surfaces). We also measured changes in cue weighting in response to manipulations of the reliability of the texture cue. ANOVA showed an effect of condition corresponding to better discrimination given multiple than single cues, and an effect of group reflecting worse overall discrimination in the group with ASD. However, in both groups DT performance was consistent with optimal integration of D and T as predicted by a Bayesian ideal observer model. The ASD group also weighted texture less when it was less reliable, as predicted. This pattern of results indicates that adolescents with autism integrate visual cues to 3D structure in the same way as typically developing adolescents.
UK Economic and Social Research Council Grant RES-062-23-0819.