PAGE 4 OF _ 11

II. Experimental Paradigm: Subset Selection
(Adapted with permission from Pylyshyn's
VISUAL INDEXES, PRECONCEPTUAL OBJECTS, AND SITUATED VISION, to appear in special issue of Cognition on Objects and Attention)

This paradigm was first reported in Searching through subsets: A test of the visual indexing hypothesis ("Spatial Vision", 1997, 11(2), 225-258) by Burkell & Pylyshyn. It provided evidence for the idea that the visual system has a mechanism for picking out and accessing individuals in the way assumed by the visual index theory. They showed that sudden-onset location cues (assumed to cause the assignment of indexes) were used to control search so that only the precued locations were visited during the search. They concluded that indexes could be used to determine where to direct focal attention.

In these studies (see Figure 2 below) a number of placeholders (11), consisting of black X's, appeared on the screen and remained there for one second. Then an additional 3-5 placeholders ("late-onset cues") were displayed. After 100 ms one of the segments of each X disappeared and the remaining segment changed color, producing a display of right-oblique and left-oblique lines in either green or red.

In this experiment, the overall displays contained elements with all 4 combinations of color and orientation features, so a search of the entire display would always constitute a conjunction search. Critically, however, the subset that was precued by late onset cues could be either a feature or a conjunction subset. (If it was feature subset, the target differed from all other items in the search set by no more than one feature. If it was a conjunction subset, only a combination of two features could identify the target because some of the nontargets differed from it in one feature and others differed from it in another feature).

Figure 2: Sequence of events in the Burkell & Pylyshyn (1997) study. In the first frame the observer sees a set of placeholder Xs for 1000 ms. In the second frame, "late onset" placeholders appear for 100 ms, signaling the items that will constitute the search subset. In the third frame, all placeholders change to search items and the subject must try to find the specified target in one of two conditions. In the top display the target differs from all the nontargets by one feature (color) whereas in the bottom display, a combination of two features is required to distinguish the target. In the experiment the bars were either red or green and the faint circles did not appear - they are only for expository purposes.

The critical question was: Will the observed search behavior be determined by the property of the entire display or only the property of the subset. They found clear evidence that only the property of the subset (whether it constituted a simple-search or a conjunction-search task) determined the relation between number of search items and reaction time. This provided strong evidence that only the cued subset is being selected as the search set. Notice that the distinction between a single-feature and a conjunction-feature search is a distinction that depends on the entire search set, so it must be the case that the entire precued subset is being treated as the search set: the subset effect could not be the result of the items in the subset being visited or otherwise processed one by one.