Abstract
The visual system contains mechanisms that are tuned for orientation. One technique for measuring the bandwidth of tuning is to measure detection thresholds for a target presented simultaneously with a high-contrast mask. For example, Campbell & Kulikowski (1966) measured detection thresholds for a sinewave grating masked by a second grating that varied in orientation. Thresholds were highest when the target and mask had the same orientation, and declined significantly when the target and mask orientations differed by as little as 10 deg. This result has been interpreted as showing that the grating target is detected by a quasi-linear filter that is narrowly tuned to orientation. Here we show that the estimate of orientation bandwidth depends critically on the nature of the mask.
We first replicated Campbell & Kulikowski (1966). Detection thresholds were measured for a horizontal Gabor (3 cpd) masked by a grating of a similar spatial frequency and a fixed, suprathreshold contrast. Mask orientation was offset from the target's orientation by 0, 15, 30, 45, 60, 75, or 85 deg. Thresholds declined with increasing orientation offset; estimated bandwidths (full-width at half-amplitude) were 20 deg. Next, we measured thresholds for targets masked by two suprathreshold gratings that were symmetrically offset from the target orientation by 0, 15, 30, 45, 60, 75, or 85 degrees. Surprisingly, thresholds were nearly constant for orientation offsets ranging from 0 to 75 deg; the resulting estimates of channel bandwidth were approximately 150 deg. Control experiments ruled out the possibility that the different results were due to differences in the contrast energy in the mask. Similar results were obtained with younger (∼22 years) and older (∼65 years) observers.
We currently are conducting experiments examining the nature of the nonlinearity that produces the very broad orientation bandwidths found with two-component masks.
This work was supported by NSERC, CIHR, and the Canada Research Chairs program