Abstract
The visual system achieves efficient recognition of complex scenes by grouping multiple features and objects into ensembles. It has been suggested that ensemble perception does not require and in some cases even impairs conscious access to the property of individual elements. So, is ensemble perception possible even if individual elements are phenomenally invisible? To address this question, we examined the effect of invisible elements on the perception of average texture orientation. In Experiment 1, a line-based texture pattern with a 3.7 deg diameter was briefly flashed (17 ms) to one eye and was followed by a circular Mondrian-based mask with a 2.0 deg diameter (17 ms) to the other eye with SOA of ~33 ms. The orientation of texture elements followed a Gaussian distribution with a particular mean and s.d. (4.0 deg), but the mean orientation within the central region overlapping with the mask was tilted by +/-30 deg relative to the surrounding region. Nine observers indicated whether texture elements were visible inside the central region and whether the mean orientation of the whole texture was tilted clockwise or anti-clockwise. We found that perceived mean orientation matched the mean orientation of visible elements in the surrounding region, thereby indicating that the observers did not use invisible elements in the central region to estimate the mean orientation. In Experiment 2, the mask was replaced by an annular contour so that the mask could not interfere with texture elements in the central region. Remarkably, perceived mean orientation matched the mean orientation of all elements including invisible elements within the central region. Together, results suggest that human vision does not require phenomenal awareness of individual elements in order to spatially average ensemble orientations.