Abstract
Are statistical summary representations (SSRs) for multiple dimensions processed independently or are they part of a bound scene gist representation? Previous research suggests that multiple SSRs, the means for two sets of circles, can be computed with no cost (Chong & Treisman, 2005). However, there appears to be a cost for encoding averages from multiple dimensions. Emmanouil and Treisman (2008) presented participants with arrays of objects that varied on two dimensions and then either pre-cued or post-cued which average should be reported. Post-cue performance was lower than pre-cue performance, however the cost may have occurred during reporting rather than encoding. If a bound scene gist representation is stored, performance may be impaired when only one dimension is reported. In the current study, participants viewed arrays of 16 lines that varied in length, orientation, or both. Across four blocks of trials, participants were asked to adjust a test line to the (1) average orientation of a group of lines that varied only in orientation, (2) the average length of a group of lines that varied in only length, (3) the average length and orientation of a group of lines that varied in both, or (4) the average length or orientation (determined via a post cue) for a group of lines that varied in both dimensions. Orientation performance supported the bound gist representation hypothesis: Although there was a cost for reporting only orientation when both orientation and location were encoded (post-cue condition), encoding and reporting both was just as accurate as encoding and reporting only orientation. However, for length performance, encoding and reporting both resulted in greater errors than encoding and reporting length alone. These results suggest that all dimensions are not encoded equally; orientation and length may be encoded as a bound unit, while length may also be encoded independently.
Meeting abstract presented at VSS 2015