Abstract
Subjects are better at discriminating the direction of motion along the cardinal (horizontal or vertical) axes than the oblique axes; this is the oblique effect for motion (OEM). We used an equivalent noise paradigm to examine whether this effect results from under-sampling, or elevated additive internal noise on detectors tuned to oblique directions. Subjects judged whether the mean direction of a field of elements was clockwise or counter-clockwise of a static reference line. The directions of the elements were drawn from wrapped normal distributions with mean directions of 90 or 45 deg, and standard deviations ranging from 0.5 to 90 deg. Interestingly, we observed the OEM only at low levels of directional variance, but not at high. Efficiency was similar for oblique and cardinal conditions, so that the OEM is wholly due to a change in additive internal noise. However at high levels of directional variance, equivalent noise fails to capture the data. A channel-based averaging model (CBAM) that pools motion energy across a limited range of directions produces a better fit to our results. Within this scheme, the OEM arises from an increase in multiplicative internal noise on the obliquely tuned channels, rather than an increase in their directional bandwidth. A second experiment further tested the role of internal noise in the OEM and critically evaluated the CBAM. Subjects judged the direction of a set of elements (with directional s.d. = 8 deg) moving clockwise or anti-clockwise of a reference direction (90 or 45 deg) in the presence of mask elements whose mean direction was fixed at the reference direction (directional s.d. = 8 deg). By varying the number of elements in the target and mask we were able to show changes in directional sensitivity that could only be accounted for by elevated multiplicative internal noise at oblique directions. We conclude the OEM results from increased internal noise on the channels tuned to oblique directions of motion, and not changes in the bandwidth of those channels.