Abstract
The Just Noticeable Difference (JND) is a common measure used to investigate models for 3D cue combination. It is typically assumed that the JND reflects the variance in the 3D shape estimate from depth cues when viewing an object. Although it is straightforward to measure the JND for a cue consistent standard by using a cue consistent comparison, it is less obvious for a cue inconsistent standard. Using a cue consistent comparison, as some experimenters have done, will lead to a measured JND that reflects the two different variance structures of the standard and comparison instead of the desired JND that reflects only the estimation variance associated with the standard. In this study, we seek to show that the comparison stimulus must have the same cue conflicts as the standard when measuring the JND. To accomplish this, we employ two crucial conditions when measuring the JND of a cue inconsistent standard: one where the comparison is cue consistent and the second where the comparison is cue inconsistent with the same cue conflicts as the standard. Our results show that the measured JND from these conditions are significantly different yet consistent with the predictions of the Maximum Likelihood Estimation (MLE) model for 3D cue combination. This study demonstrates that it is important to strategically choose the type of comparison depending on what variance structure the JND is expected to reflect.