Computerized binocular infrared pupillography was used to measure the transient pupillary light reflex (PLR) in both children with autism spectrum disorders (ASDs) and children with typical development. We found that participants with ASDs showed significantly longer PLR latency, smaller constriction amplitude and lower constriction velocity than children with typical development. The PLR latency alone can be used to discriminate the ASD group from the control group with a cross-validated success rate of 89.6%. By adding the constriction amplitude, the percentage of correct classification can be further improved to 92.5%. In addition, the right-lateralization of contraction anisocoria that was observed in participants with typical development was not observed in those with ASDs. Further studies are necessary to understand the origin and implications of these observations. It is anticipated that as potential biomarkers, these pupillary light reflex measurements will advance our understanding of neurodevelopmental differences in the autism brain.