Abstract
When attention is spread diffusely across multiple objects, observers can create statistical summaries of object features, such as average size (Ariely, 2001; Chong & Treisman, 2003) or the average location of a group (Alvarez & Oliva, 2008). We tested whether such summary representations are available for another primary visual feature, orientation. Participants judged the average orientation of either a set of six lines or six Gabor gratings, with either homogenous or heterogeneous orientations, presented for 800ms. Participants then chose between the correct average and a foil alternative that was rotated by 7–42 degrees. In all conditions, accuracy became significantly higher than chance at various foil distances, suggesting that, at minimum, coarse averaging was possible in each condition. However, accuracy in the heterogeneous condition was lower than that in the homogeneous condition, suggesting that orientation averaging is not as precise as orientation perception in a display containing redundant orientation information. We also found that participants were more accurate at averaging lines than gratings, despite a control study showing that 800ms was ample time to extract equally precise orientation from a single line or grating. Furthermore, using heterogeneous instead of homogenous displays caused a larger averaging impairment for gratings, compared to lines. This suggests that average orientation is more accessible when boundary features convey the orientation information rather than when only the surface features carry the orientation information, despite the fact that early visual areas often favor gratings over lines and that gratings contain redundant orientation information. Our study suggests that the estimation of the average orientation is possible, though imprecise, and that the boundary information may play a critical role in the processing of average orientation.