Visual attention is often described as operating at multiple levels. Attention can be directed to a location in space (“where”), to a visual feature such as color or motion direction (“what”), or to an object or surface as a coherent perceptual entity. While spatial and feature-based attention are relatively easy to reconcile with the known organization of visual cortex, the mechanisms that allow selective processing of whole objects remain less well understood.

Transparent-motion stimuli provide a powerful way to study this problem. In these displays, two moving dot patterns are superimposed in the same region of space, yet observers perceive two distinct moving surfaces, one moving atop the other. Because the surfaces occupy the same location, spatial selection could not seemingly allow for the selective processing of one surface over the other. Yet Valdes-Sosa, Cobo, and Pinilla (1998) introduced a transparent-motion paradigm that demonstrated that observers can selectively process one of the two dot fields. In their design, observers report the translation direction of either a cued or an uncued dot field. It was found that observers report the translation direction of the cued dot field more reliably than the translation direction of the uncued dot field. Because translation direction is unpredictable and uncorrelated with the rotation direction, these results rule out feature-based attention to motion direction. Later studies also ruled out color-based selective processing (Mitchell, Stoner, Fallah, & Reynolds, 2003; see also Stoner and Blanc, 2010).

There have been numerous follow-up studies that have used variants of that design to investigate the behavioral and neural correlates of object-based attention. In our current set of studies we use a delayed-onset variant of this design. One of the two surfaces appears slightly later than the other; one of the dot fields then briefly translates, and observers report its direction. Stoner and Blanc (2010) and Çatak et al. (2022) discuss how this paradigm constitutes a simplified version of the original Valdes-Sosa et al. paradigm. They also offer a brief review of the various other studies that have built on the original work of Valdes-Sosa and colleagues.

The demonstrations below illustrate the basic cued and uncued conditions used with this design. For clarity, they are played at half speed with extended translation durations; only half of the dots within each field translate coherently. The left stimulus shows translations of the delayed surface; the right shows translations of the non-delayed surface. The delayed onset serves as an exogenous attentional cue (Reynolds, Alborzian, & Stoner, 2003) such that attention is briefly attached to the delayed dot field: Most observers find the translations easier to judge on the delayed surface — a result that has been interpreted as evidence for object- or surface-based selection.

Stoner and Blanc (2010) asked whether these effects were based on the global direction of motion and color configurations (as might be expected by demonstrations like that above) or rather were specific to dots of the two dot fields. Specifically, they swapped the color and motion directions of the dot fields at the time of translation. They found that the performance advantage was specific to the individual dots not the global feature configuration. This result is surprising given previous studies that have shown that the perception of these moving surfaces is relatively insensitive to the individual dots that happen to compose these dot fields (as observed in the demo above; refs to be added). These findings suggest that the selective process underlying these performance effects is spatially fine-grained (see Çatak et al., 2022). Stoner devised a model that can account for these results based on feedback from area MT and/or area MST onto area V1 neurons (Stoner, 2018, SFN poster). This model predicts that increasing the density of these dot fields would reduce the performance asymmetry as spatial selection, even if operating at a fine spatial scale, should break down as dots from the two fields crowd one another. Investigating the impact of spatially crowding these dots is one of the manipulations used in ongoing experiments.

Another area of investigation is the role of depth-order in these stimuli. As observed above, perceptual depth-ordering of these dot fields happens despite the absence of explicit depth cues. This suggests the possibility that depth may be an organizing principle by which the visual system selects one set of dots over the other. By explicitly introducing depth-from-disparity cues into these stimuli, we are investigating this hypothesis in another set of experiments.