Shadow processing: Top-down or bottom-up?

The existence of an automatic shadow discounting mechanism in the visual system (Rensink andCavanagh, 2004 Perception 33 1339 ^ 1358) is still under debate (eg Elder et al, 2004 Perception33 1319 ^ 1338). Revisiting Rensink and Cavanagh's search task, we found that, independently ofthe exact task requirements (target-present/target-absent or target-present only), search wasslowed for upright shadow stimuli relative to inverted, but not for a control condition in whichneither stimulus orientation could be interpreted as shadows. Differences emerged consistentlyin absolute reaction times but not search slopes. Similarly, only the number of eye movementsbut not fixation durations increased in the shadow-like condition. Taken together, our data arein line with Rensink and Cavanagh's observation that shadow-like images are difficult to process.However, rather than an automatic bottom ^ up driven mechanism to discount shadows, our datasuggest the existence of a top ^ down mechanism that leads to delayed processing of shadow-like stimuli.[Supported by the BRACE Charity.]