Fig 1.
Two possible accounts for the attentional demand of statistical properties.
Panel A illustrates the same-cost account of the encoding of statistical properties: Attention can be focused on a single object or distributed over a set, and it allows observers to perceive the attended information in essentially the same way, no matter whether this information is an object feature or statistical property. Panel B illustrates the cost-free account of the encoding of statistical properties: Attending to a single object allows observers to perceive its features, whereas the statistical properties of a set that is out of attentional focus can be simultaneously perceived for free without any additional cost.
Fig 2.
A. Stimulus display. This stimulus display consisted of a rectangle (or a 4×4 array of rectangles) in the top-left or bottom-right quadrant and a circle (or an array of circles) in the top-right or bottom-left quadrant. The circle (or the array of circles) could be reddish or greenish. The rectangle (or the array of rectangles) could be vertical or horizontal. When an array was reddish (or vertical), this just meant that 12 of the 16 items in the array were reddish (or vertical). The participants were asked to report one of these features, as specified by the post-cue. B. Sequence of display. The stimulus display was preceded by a pre-cue and followed by a post-cue. The post-cue always specified the target, whereas the pre-cue specified the target in half of the trials (pre-cue condition) but not in the other half (post-cue condition).
Fig 3.
Panel A shows the observers’ accuracies in all 16 conditions. There were significant pre-cueing advantages for both the color and orientation tasks regardless of whether each task was about object features or statistical properties. Panel B shows the average pre-cueing advantages for the color and orientation tasks. The predictions are given to illustrate the predicted directions of effects and are not meant to be numerically precise. It is clear that the results of the experiment are consistent with the same-cost account of the encoding of statistical properties and inconsistent with the cost-free account of the encoding of statistical properties. Error bars show within-subject 95% confidence intervals [22].