The role of local separation in spatial frequency discrimination

Spatial frequency discrimination could simply reflect the ability to discriminate local differences in width when sinusoidal gratings are used. We introduced random fluctuations in the half-cycle widths of both extended 1c/° gratings and those restricted to 1.5 cycles (i.e. two bars of identical phase), as we hypothesised that discrimination for large gratings would be more robust to this distortion in the presence of a genuine spatial frequency sensitive mechanism that averaged information across a wide area. To exclude the possibility of a local mechanism that averaged through scanning eye movements, experiments were repeated with a short presentation time. We also repeated the experiment with a fixation point and the central 4° of the grating masked. Discrimination thresholds were more robust to spatial distortion for large gratings vs. two bars, and were not significantly altered by either a short presentation time or masking of the central part of the grating. We therefore find that spatial frequency discrimination performance is undistinguishable from width discrimination performance when regular sinusoidal stimuli are used. When spatial distortion is introduced to make measures of spatial width unreliable, spatial pooling of information is both possible and results in clear performance benefits.