Abstract
The two hemispheres of the brain exhibit lateralization in processing visual stimuli, such as faces, Navon stimuli, and frequency gratings. Our "differential encoding" computational model accounts for these differences via a hypothesized asymmetry in the length of long-range lateral connections in the visual cortex. Here we show this model also matches human data in categorical and coordinate processing, and provide data that break a long-standing assumption in the field. Coordinate discriminations involve absolute positional judgments; categorical discriminations involve relative positional judgments. Kosslyn (1987) hypothesized that, due to relatively better low frequency processing, coordinate discriminations are right hemisphere dominant, whereas, due to relatively better high frequency processing, categorical discriminations are left hemisphere dominant. Slotnick et al. (2001) posited that task difficulty explains the inconclusive data in intervening years. Our model replicated the human data's lateralization patterns on Slotnick's stimuli--including the blob dot coordinate case where Slotnick contradicted Kosslyn's hypothesis. They resolved this anomalous result by stratifying along difficulty. Our model similarly increases lateralization when resources are limited, thereby making the task more difficult; however, task stratification results do not align with Slotnick's results. The post-hoc nature of the paper's analysis and inconsistencies across its figures raise questions on whether difficulty stratification adequately explains the anomalous result. Frequency analysis of our model associates task-relevant content of the blob dot coordinate experiment with high frequencies, not low frequencies, similar to categorical experiments. This suggests that Kosslyn's original hypothesis made a faulty assumption: coordinate and categorical stimuli do not perfectly correlate with low and high frequency information. The differential encoding model matches human results with reference to a specific neural computation, anatomical source, and a specific neurodevelopmental mechanism. We hope these results will motivate the vision community to explore intra-cortical connectivity in visual processing and how lateralization can inform us about human visual processing.
Meeting abstract presented at VSS 2017