Are mirror-symmetric objects of special importance for 3D shape perception? A reply to Sawada and Pizlo (2022)

Abstract

Yu, Todd, and Petrov (2021) and Yu, Petrov, and Todd (2021) investigated failures of shape constancy that occur when objects are viewed stereoscopically at different distances. Although this result has been reported previously with simple objects such as pyramids or cylinders, we examined more complex objects with bilateral symmetry to test the claim by Li, Sawada, Shi, Kwon, and Pizlo (2011) that the perception of those objects is veridical. Sawada and Pizlo (2022) offer several criticisms of our experiments, but they seem to suggest that the concept of shape is defined by what is computable by their model. If stimuli are used that cannot be discriminated by their model, they are dismissed as degenerate, and tasks that cannot be performed by their model are assumed to be based on something other than shape. This allows them to disregard empirical evidence that is inconsistent with their model. We argue, in contrast, that all reliable aspects of shape perception are deserving of explanation. We also argue that there are many different attributes of shape and many different sources of information about shape that may be relevant in different contexts. It is unlikely that all of them can be explained by a single model.

Effects of violating the assumptions of equal variance and independent and identically distributed random variables: A case using {B}ayesian statistical modeling

“…I found that Yu et al (2021, see also Sawada & Pizlo, 2022Petrov et al, 2022) provide a good example of how a violation of assumptions can distort the results of an analysis. These authors proposed a Bayesian statistical model with some assumptions about their data but the data clearly violated these assumptions.…”

“…Consider Experiment 1 of Yu et al (2021). They tested the effect of an object's distance on the perceived depthinterval of the object's shape based on binocular disparity under two conditions of the object's size (the fixedphysical-size condition and the fixed-projected-size condition, see also Sawada & Pizlo, 2022;Petrov et al, 2022). There were three different viewing distances (0.7 m, 1.5 m, and 2.3 m) in the experiment.…”