Attention to pattern depth depends on pattern dimensionality

Document Type

Conference Document



Embargo Period



Stereoscopic depth, attention, binocular disparity




Cognition and Perception | Cognitive Neuroscience


The perceived stereo depth separating two stimuli usually varies with the horizontal disparity difference between the stimuli. This, however, is not the case when one or both stimuli are one-dimensional. Instead, perceived depth depends on the difference between the disparity vectors of the two stimuli; relative disparity magnitude and direction both matter, interactively (Farell, Chai, Fernandez, Vis. Res., 2009). Here, we compare judgments of the depth of 1-D and 2-D stimuli, asking how attention affects this interaction. Our displays contained a central stimulus whose disparity varied across trials. This stimulus was either 1-D (a grating) or 2-D (a plaid). The stimuli surrounding the center were oblique-disparity plaids, the location of one being designated as relevant. The task was to judge the relative depth of the central stimulus and the relevant plaid; the remaining plaids were irrelevant throughout the block of trials and were to be ignored. Psychometric functions for depth judgments of the grating and the relevant plaid shifted laterally in response to the plaid's disparity direction (parallel or orthogonal to the grating's disparity). Interestingly, the disparities of irrelevant plaids produced exactly the same effect. Thus, attention failed to distinguish relevant and irrelevant stimuli when observers judged a grating-plaid pair. By contrast, when the stimuli being judged were both plaids, psychometric functions were affected neither by the disparities of irrelevant stimuli nor by the disparity direction of relevant stimuli. Attentional filtering of disparity signals thus succeeded only when observers judged the depths of 2-D stimuli. The judged depth of a 1-D stimulus varied with all the disparities in the display, whether relevant or irrelevant, revealing a disparity field that could be useful in transforming ambiguous 1-D component disparities into coherent object depths.


Journal of Vision