The computational dynamics of shape orientation perception

How does the brain transform retinal information into representations of oriented objects? The most comprehensive computational explanation to date - the coordinate-system hypothesis of orientation representation - proposes that this transformation relies on the computation of four parameters that jointly define the relationship between a shape and its environment: axis correspondence, polarity correspondence, tilt direction, and tilt magnitude. The goal of this research was to investigate whether these parameters are computed in parallel or serially and, if so, in which order. To do so, we conducted three same/different experiments in which targets and probes could differ by either one of two parameters (A and B) or both (A + B). Under the assumption that response times in such tasks reflect the rate at which evidence for a difference is accumulated, the conjunction condition (A + B) should result in faster response times if the two parameters (A and B) are processed in parallel. In contrast, if the two parameters are processed serially, response times for A + B should be equivalent to those for the first parameter (e.g., A) and faster than those for the second parameter (B). In this framework, the results of the three experiments suggest that axis correspondence is computed first, followed by all the other parameters, computed in parallel.