Perceptual learning of parametric face categories leads to the integration of high-level class-based information but not to high-level pop-out.

To date, the relative contribution of the different levels of the visual hierarchy to perception remains unclear. Typically, lower levels are seen as a sequence of simple feature-detectors, whereas the crucial part of visual processing is attributed to higher levels. This view is taken to an extreme by the reverse hierarchy theory (RHT), which predicts that lower-level information is processed only implicitly during the fist hierarchical sweep of information. Here, we investigate this issue based on two analyses. First, we test a vital prediction of RHT: the existence of high-level pop-out. For this, we combine perceptual learning of two classes of parametric faces with subsequent tests for pop-out. This controlled setting excludes low-level confounds and was recently shown to lead to distinct high-level representations - a prerequisite for high-level pop-out. Secondly, we explore the underlying form of category representation and implied contributions of the different levels during subsequent stages of perceptual training. This is accomplished by including class-external as well as class-internal target-distractor combinations. Our results show that, despite very early, near-perfect classification accuracy and extensive training, there is no sign of high-level pop-out. Moreover, whereas the subjects’ responses during the first sessions are best explained instance-based and dependent on low-level metric differences, later patterns exhibit the inclusion of high-level, class-based information that is independent of target-stimulus similarity. Finally, we show that the utilized forms of category representation are dependent on the task and the expressiveness of the different levels.

Video registration (1h27min)

(no podcast interview)