Purchase this article with an account.
Karla K. Evans, Anne Treisman; Perception of natural scenes; is it really attention-free?. Journal of Vision 2004;4(8):129. doi: 10.1167/4.8.129.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
How much is possible outside the focus of attention? Recent studies have argued that perception of natural scenes contradicts common findings of limited capacity of visual attention. They have shown that participants detect a natural category target in real scenes, just as well in a dual task situation as when the scene receives full attention (Li et al. 2002). Furthermore they are able to search two scenes in parallel in a go-no-go natural category detection task (Rousselet et al. 2002). We believe that this ability may not be due to high-level object representation in the absence of attention. Participants may be able to detect unbound features of category targets in parallel and then use these to discriminate between scenes that do or do not contain the target. In the first study in which the task was to detect the presence of an animal in a rapid sequence of scenes, we obtained some evidence supporting this idea. Detection of an animal was considerably worse in scenes that also contained humans than in scenes without humans, and participants did poorly registering the target's location. The second study showed some limits to this surprising attention capacity in an attentional blink paradigm. Participants observed a rapid serial presentation of real scenes while searching simultaneously for two targets, either from the same or from different target categories (animals and vehicles). Results show that category detection in natural scenes is not exempt from the temporary refractoriness caused by detection of a first target. It is more severe when the targets are from different categories, suggesting that search for two sets of features at once is difficult or impossible. We conclude that the ability observed in previous studies may depend on detection of disjunctive feature sets rather than high-level identification of category targets, and that there are limits to the number or the variety of features that can be simultaneously processed
This PDF is available to Subscribers Only