Abstract
It is well documented that odd elements are easily detected when embedded in a field of distractors that are similar to each other and differ categorically from the odd element in one feature, such as color, orientation, motion direction or shape. We have claimed that this “pop-out” effect is due to high-level rather than low-level mechanisms (Hochstein & Ahissar, 2002). We now ask whether the relevant shape needs to be defined by real edges or do illusory contours (Kanizsa, 1979) also induce a pop-out effect, though their perception is slower, and to what extent does such pop-out depend on illusory-contour training history. Previous studies investigated the possibility of a pop-out effect when subjects search for a set of illusory-contour inducers among a field of non-inducers (Grabowecky & Treisman, 1989; Davis & Driver, 1994), while we tested detection of an odd illusory shape among illusory distractor shapes (see Gurnsey et al., 1992). We tested subjects with different levels of training on a Kanizsa illusory contour task, detecting an illusory parallelogram among illusory triangles. Set-size effects and mean response time were analyzed to determine if the classical tests for pop-out (relative set-size independence for odd element present trials) is relevant also for slowly induced percepts. A significant difference in performance, as well as in learning rate, was found between subjects with different levels of training. In contrast, a control experiment with real (rather than illusory) figures showed significantly less dependence on training level. These results suggest that with training detection of Kanizsa figures is parallel - i.e. these figures do pop-out!
Supported by “Center of Excellence” grant #8009 from the Israel Science Foundation and the US-Israel Bi-National Science Foundation.