Abstract
We wish to study observers' ability to divide attention more and more finely in a search task, i.e., to study the spatial modulation transfer function of attention. Prior to viewing a black-and-white test stimulus, subjects are shown a map of the search regions (defined by a red-green grating) where attended stripes—within which a target (T) can occur—are green, and nontarget areas are red (or vice versa for other observers). Observers tend to ignore the fine structure required by the map unless false targets (FTs) are placed in unattended areas to force confinement of attention to the to-be-attended stripes. As the stripes become thinner [spatial frequency (SF) of the attend/nonattend grating increases] search performance deteriorates (ARVO, 2001; VSS, 2002). Unfortunately, there is a confounding effect of FT-crowding: at high SFs, FTs are closer to Ts than at lower SFs. We discriminate the effects of attention and FT-crowding by, unbeknownst to the observers, embedding occasional one-FT trials (stimuli with only one FT) in a series of stimuli with many FTs.
Results: For identical one-FT stimuli (obviously with the same FT-T distance), search performance still consistently decreases with an increase in the to-be-attended SF. Attempting to divide attention finely has a cost (which is why observers don't do it unless forced to by FTs). However, crowding also has a big effect—more and/or closer FTs impair T detection.
Discriminating the effects of FT-crowding from spatial frequency fall-off enables a more accurate model of the maleability of visual spatial attention, per se.