Abstract
Recent research has shown that participants are better able to categorize briefly presented natural scene images that have been rated as “good” exemplars of their category, and that this is reflected in the distributed patterns of neural activation obtained through fMRI (Torralbo, et al., 2009). The effect of typicality on categorization/decision processes is well documented (see Rosch, 1978), but it is possible that such effects may also reflect differences in perception. Here we asked whether subjects might actually ‘see’ good exemplars of a category better than bad exemplars. We asked subjects to simply report whether a very briefly presented (19 ms – 60 ms) image was intact or scrambled. Images drawn from six natural scene categories (beaches, city streets, forests, highways, mountains and offices) were rated as either “good” or “bad” exemplars of their categories. These images were presented in either their original intact state, or 100% phase scrambled (Sadr & Sinha, 2004), and then followed by a perceptual mask. Note that the subjects were never instructed to categorize the scenes nor were they explicitly notified that the image set contained good and bad category exemplars. We measured participants' d′ separately for good and bad images, and found that participants were better able to discriminate intact from scrambled images when the images were good category exemplars than bad category exemplars. These results suggest that knowledge about scene category actually allows observers to ‘see’ natural scene images better, regardless of whether scene category is relevant to task.
This work is funded by the NIH (LFF, DB, DW).