Abstract
Previous temporal masking studies have found that the visual system possesses independent temporal frequency (TF) channels: one low-pass and one or two higher bandpass channels peaking ∼ 10Hz. Using iso-oriented targets and masks, we replicated this finding. However, when targets and masks were cross-oriented we found no evidence for the low-pass channel, but still found a high bandpass channel. The orientation-dependence of the low-pass channel suggests a cortical mechanism, whereas the orientation-invariant bandpass performance supports recent evidence that temporal masking may be mediated sub-cortically (Freeman, Durand, Kiper & Carandini, 2002). Follow-up masking experiments using (unoriented) spatio-temporally filtered noise revealed evidence for two bandpass channels (peaking ∼ 1.5 and 10Hz). Interestingly, these channels appear to interact asymmetrically: high TFs mask low TFs, but not vice versa. We also varied spatial frequency using spatio-temporally filtered noise. These data indicate that while the temporal channel asymmetry is seen at several spatial frequencies, the high-pass channel shifts to higher temporal frequencies as spatial frequency is reduced, and a third intermediate temporal channel emerges. The implications of these spatio-temporal interactions for figure-ground, form and motion processing are discussed. We propose that the asymmetric temporal channel interactions may serve to equalize the 1/TF power spectrum present in natural image sequences, and may also contribute to motion deblurring.