Tentative Conclusions

PAGE 10 OF _ 11

The main prediction was that our results should replicate the original Burkell & Pylyshyn study. For this to happen, the indices would, effectively*, have to maintain contact with the world locations of the 3 and 4 late-onset X's. This is because while the late-onset X's and the subset search items shared the same location in the world they occupied different retinal locations. Note that after the saccade (i.e., the first fixation in the search image), our experiment is like the Burkell & Pylyshyn's study in terms of type of search items and the visual angle of stimulus. We can draw the following tentative conclusions from the data so far.

    1. The system does maintain the indices across the saccade, but there is a cost of accessing the indices after an eye movement. (This seems unlikely because one would expect a similar cost of moving the eyes in the 0 late-onset feature search condition, i.e., where the search set is all the items on the screen. Further, aside from the location of the first fixation in the search image, after the eye movement, the experiment, visually, was like the original Burkell & Pylyshyn study).
    2. The indices cannot, effectively*, maintain contact with the world locations of the late-onset X's, at least under the current conditions of the experiment. Specifically, it is possible that the system maintains indices automatically across a saccade as long as the world does not change in any radical way during the saccade.

If the world does change during a saccade, then what one might predict is exactly what we see: some sort of memory based task for the world locations of the 3 or 4 late-onset X's. This trans-saccadic memory representation may allow the locations of the late onset X's to be recovered after the saccade. Once recovered, the system can re-attached the indices to those locations, and only the objects at those locations are processed.

The example scanpaths in the 3 & 4 late-onset conditions show the eyes serially visiting at least some of the items in the late-onset cue locations. The eye movement pattern may reflect the spatial-temporal properties of a process of using a trans-saccadic memory of locations to re-attach visual indices. This process may be responsible for the longer response times in the 3 & 4 late-onset feature search conditions.

Next

*There are, presumably, many ways this could be done, e.g., an extra-retinal eye position signal could be used to adjust the location of the indices with respect to new, incoming visual information after each saccade.