Abstract
When recognizing an object, low spatial frequencies (LSFs) are processed before higher spatial frequencies (HSFs), presumably through the faster magnocellular pathway. People suffering from autism spectrum disorders (ASD) however may not benefit from such a precedence of LSF information, several studies indicating a deficit in processing related to the magnocellular pathway (e.g., Sutherland & Crewther, 2010) and a preference toward HSFs rather than LSFs (e.g., Deruelle, Rondan, Gepner, & Tardif, 2004). Our study compared the time course of spatial frequency (SF) use in object recognition in neurotypical and ASD subjects. Forty-five neurotypical and 18 ASD subjects participated to the study. On each trial, a short video (333 ms) was presented to subjects. This video was created by selecting one of 86 object images, all equalized in SF content, and by sampling randomly its SFs across time. An object name immediately followed and subjects had to indicate if it matched the object (it did on half the trials) as quickly as possible without making too many errors. We then performed multiple linear regressions on SF x time sampling planes and accuracy. Most SFs between 0.08 and 4.42 cycles per degree (cpd) during almost all stimulus presentation (0 to 325 ms) led to more accurate responses for both groups (p< 0.05; Zmax=7.10). Interestingly, SFs between 3.75 and 4.33 cpd in the 58-100 ms time window (p< 0.05; Zmax=4.03) and SFs between 5.00 and 5.58 cpd in the 67-100 ms time window (p< 0.05; Zmax=3.99) led to more accurate responses for autistic subjects than for neurotypical subjects. These results indicate that while both groups use LSFs and intermediate SFs throughout object recognition, autistic subjects use more HSFs at a shorter latency. This suggests a different time course of SF extraction for autistic subjects.
Meeting abstract presented at VSS 2015