Evaluating models of collinearity judgment for reliability and scale

In prior work from this laboratory, we have examined how accurately subjects: can judge collinearity, ie alignment, as a function of the angular position of stimulus elements. In those experiments, subjects were presented with a line segment (or a pair of dots) which varied across the 360 degrees range of angular positions, and were required to select and mark a point which was perceived as being collinear. We found that the models of error for each subject consisted of a complex set of peaks and valleys which are herein described as S excursions. The error tendencies appeared to be idiosyncratic, in that each subject manifested a different profile of these excursions. Here we report the results of three experiments. In the first, we tested subjects across five sessions, and found that the five models of error were fairly comparable for a given subject. In the second and third experiments, we tested at close intervals of angular position, and found evidence for localized sources of error. We continue to find that the error model for each subject is idiosyncratic, and argue that the excursions may be due to defects in a system which combines responses from position-encoding fields that vary in size.