Dissociating Attention and Eye Movements in a Quantitative Analysis of Attention Allocation

In a recent paper, we introduced a method and equation for inferring the allocation of attention on a continuous scale. The size of the stimuli, the estimated size of the fovea, and the pattern of results implied that the subjects' responses reflected shifts in covert attention rather than shifts in eye movements. This report describes an experiment that tests this implication. We measured eye movements. The monitor briefly displayed (e.g., 130 ms) two small stimuli (≈1.0° × 1.2°), situated one atop another. When the stimuli were close together, as in the previous study, fixations that supported correct responses at one stimulus also supported correct responses at the other stimulus, as measured over the entire session. Yet, on any particular trial, correct responses were limited to just one stimulus. This pattern suggests that the constraints on responding within a trial were due to limits on cognitive processing, whereas the ability to respond correctly to either stimulus on different trials must have entailed shifts in attention (that were not accompanied by eye movements). In contrast, when the stimuli were far apart, fixations that had a high probability of supporting correct responses at one stimulus had a low probability of supporting correct responses at the other stimulus. Thus, conditions could be arranged so that correct responses depended on eye movements, whereas in the "standard" procedure, correct responses were independent of eye movements. The results dissociate covert and overt attention and support the claim that our procedure measures covert attention.