Abstract
Two competing low-level origins of the Poggendorff illusion have been proposed (1) The angle of the pointer is biased by cross-orientation inhibition (2) the position of the intersection of pointer and inducing lines is mislocated by large-scale neural blurring. We tested the latter by getting 8 observers to make saccades to the acute-angle intersection point in a Poggendorff figure from a remote fixation position. On half the trials the position of the target intersection was shifted during the saccade, either in the direction of the Poggendorff bias or in the opposite direction, to prevent the observers from recalibrating their response. These shifts were not detected in a forced-choice test during the experiment. Some observers showed a bias of the first saccade endpoint into the acute angle, but most did not. All observers showed a large Poggendorff bias when making a saccade from a pointer to an imagined extrapolation point on a landing line. These findings suggests that mislocalization is not the origin of the Poggendorff bias. Previous psychophysics using a matching method have been used to argue that the intersections are mislocated (Morgan, Vision Research ,1999). However, this conjecture was based upon a method of adjustment which may be susceptible to decisional biases. Using a more rigorous 2AFC method with roving pedestal no consistent mislocation bias was found, although it was present in some observers. We conclude that the Poggendorff bias arises from a bias in the direction in which the pointers appear to point, although (paradoxically) not in the perception of their orientation. The Poggendorff bias is computationally interesting, since it raises the question: why does pointing use only information from the pointer termination, where it is least reliable?
Meeting abstract presented at VSS 2015