Further Experiments

PAGE 11 OF 11

Experiment 2: Eye movement version of the subset experiment with slight modification

In experiment 1, the "world" changed from X's to the search image during the saccade. In experiment 2, we will follow up on the hypothesis that the system maintains indices automatically across a saccade as long as the world does not change in any radical way during the saccade.

Experiment 2 will be similar to experiment 1 with one difference: when subjects make a saccade in response to the late-onset cues, we will not change to the search image during the saccade. Instead, we will wait until the eye land and then at some delay into the fixation, we will change to the search image. (Click here to see an illustration)

Experiment 3: Eye movement version of the subset experiment with shift

Assuming that experiment 2's results replicate those of the original Burkell & Pylyshyn experiment, experiment 3, using the subset paradigm, we will begin to address the issue of how visual indices are maintained across a saccade. In other words, we want to begin to ask: what is the basis of the maintenance of visual indices across eye movements?

There are essentially two hypotheses concerning the basis of the maintenance of visual indices across eye movements: ones that appeal to a unique description of a individual object and ones that do not. The first type includes a proposal called the saccade-target theory (McConkie and Currie (e.g., Currie, McConkie, Carlson-Radvansky, & Irwin. "Role of the Saccade Target Object and the Perception of a Visually Stable World," to appear in Perception & Psychophysics). This theory postulates that unique properties of one object (the one that serves as the target of the saccade) are encoded and searched for in the second fixation in order to establish a cross-fixation correspondence.

The second option is exemplified by the visual index theory. In contrast to the saccade-target theory, visual index theory assumes that a small number of objects can be recovered from the second fixation as a side effect of their having been indexed and tracked across the saccade, without the benefit of an encoding of their properties. One way this could happen would be that the indices are adjusted by a signal that encodes the size and the direction of the saccade (i.e., an extra-retinal eye position signal). For example, by taking into account where the indexed locations in the world will be imaged on the retina after the saccade, the system could adjust the relationship between indices and retinal locations such that locations in the world indexed before the saccade are the same locations indexed after the saccade.

In order to test these two hypotheses, experiment 3 will be like experiment 2, except that during the saccade to the late-onset X's, the image will shift or displace so that the eyes will not land where they otherwise would have.

If visual indices employ something like an extra-retinal eye position signal, then (as long as the shift is large enough), any shift in any direction should be equally disruptive to the system. That is, after the saccade, the system should have difficulty processing information from objects that were indexed before the saccade because they have now changed their location in the world. The ability to search 3 or 4 indexed locations should degrade if the X's are displaced during a saccade.

On the other hand, the ability to search could depend more on where the eyes land relative to the late-onset items. For example, McConkie and Currie found that subjects in a natural viewing task had a distribution of landing positions around a saccade target object, i.e., sometimes overshooting and undershooting the center of the object. Thus, equal sized stimulus shifts in various directions caused the eyes to land at various distances from the saccade target object. They found evidence that detection of these equal sized stimulus shifts varied with how close the eyes landed relative to the saccade target object (a theory of detection based on an extra-retinal eye position signal would not have predicted an effect of this variable). If this variable affects search times and the overall pattern of search tasks in a similar fashion, then this would be evidence that the system uses something like a saccade target object to maintain visual indexes across saccades.

Please mail comments to:

Zenon Pylyshyn Center for Cognitive Science (Department of Psychology, Center for Cognitive Science), or
Christopher Currie Center for Cognitive Science (Department of Psychology, Center for Cognitive Science)