Locomotion through the environment creates characteristic patterns of image motion, referred to as retinal flow. This flow can be purely radial, if observer motion is on a straight path and no eye movements are made. In this case, the computation of heading direction (among others) is straightforward. When pursuit eye movements are also present (creating a parallel flow), the overall retinal flow pattern becomes more complex by adding a lateral component to each radial motion vector. In order to extract the radial component of such a combined retinal flow, the parallel flow needs to be discounted in some way. One obvious solution would be to use signals that created the eye movements in the first place (efference copy) or feedback from the eyes. There is evidence that such extraretinal or non-visual decomposition takes place and can be effective. The other solution is to use visual decomposition, for which there is also some evidence when certain limitations (low speeds of eye rotation, environments rich in depth cues) are respected. We report evidence for a mechanism that could underlie such a visual decomposition.

In our research we employed flow patterns in which the radial and the parallel flow were transparently superimposed. This stimulus is similar to the one used by Duffy & Wurtz (1993) to demonstrate the illusory shift of the focus of expansion. We assume that the two components of the pattern will have to be separated at some level. Our goal was then to examine in detail the complementary masking between parallel and radial motion flow and to relate it to parameters known to influence the illusion. It was thus of interest whether one motion is prioritized over the other, whether the parallel/radial motion interaction is speed dependent and whether it is influenced by the presence of acceleration in the radial flow. We also explored the spatial extent of this interaction by limiting dot lifetime and by confining the component motions to non-overlapping segments of the display.

Our results indicated that there exists a strong and persistent asymmetrical masking effect, such that parallel motion was easily interfered with by radial motion, while radial flow was fairly resistant to the presence of parallel flow. In other words, as compared to a radial jitter noise, a coherent radial motion significantly decreased sensitivity to parallel flow. In contrast, sensitivity to radial motion was little affected by the presence of coherent parallel flow compared to parallel jitter. Changing dot lifetime did not affect this result. Removing the velocity gradient from the radial flow preserved the asymmetry, but eliminated the advantage of expansion over contraction. Presenting the flow fields in non-overlapping segments preserved the masking effect of radial on parallel all the way up to 180 deg segments, while the expansion advantage disappeared for segments larger than 90 deg.

We believe that these results provide further evidence for a purely visual compensation for the image motion produced by oculomotor pursuit: there is an asymmetrical interaction between the radial and parallel components of the retinal motion flow which preserves the integrity of the former and minimizes the influence of the latter. Although we can only speculate about the nature of this interaction, it seems to operate on a large spatial scale and takes into account the depth information carried by the motion flow. In addition, the suppression of the parallel flow is optimal under the same conditions, which support heading accuracy in the presence of simulated eye movements. Therefore it is likely that the visual processes we describe contribute to the maintenance of stable and accurate behavioral direction.

References:

Duffy CJ and Wurtz RH (1993). An illusory transformation of optic flow fields. Vision Res. 33(11): 1481-1490.