Abstract
Purpose:
Models predict performance for detection of change in multi-element patterns in terms of the number of elements (set-size), irrespective of pattern shape and element arrangement. We set to investigate this assumption.
Methods:
Stimuli were composed of D6 patches (set-sizes of 1, 8 and 35) positioned as if on the contour of specific shapes (a circle or an eight-lobed sinusoidally deformed circle (RF8)). Mean eccentricity was 2.7° around fixation. The orientation of individual patches was arranged either to be tangential to the contour or random (same absolute orientations but random location). Subjects (n = 3) indicated in which of the sequentially presented stimuli one patch had changed its orientation.
Results:
A decrease in performance with increasing number of elements in the random condition showed the expected set-size effect. Interestingly, no set-size effect was observed when the elements were tangential to a circle. When they were tangential to the RF8 contour, performance decreased from 1 to 8 elements (as in the random condition) but, paradoxically, improved on average when increasing from 8 to 35 elements.
Conclusions:
Our results show two new factors determining change detection. First, the difference in performance between random vs. tangential conditions suggests that change detection does not only depend on set size but also on the arrangement of elements. Second, the difference between circular vs. RF8 shapes indicates that the shape of the pattern plays an important role in change detection.