Abstract
Humans are able to process spatial scale flexibly depending on task demands and prior knowledge (e.g. Schyns & Oliva (1999) Cognition, 69, 243–265). Could such ability be due to attentional modulation of early visual processing mechanisms such as spatial frequency (SF) channels (Sowden, Özgen & Schyns (2001) Perception, 30, s91)? Consistent with this, a series of experiments in our lab suggested that sensitization to spatial scale is retinal location specific. In a ‘sensitization’ stage observers were initially shown low or high-passed scenes, combined with noise at complementary SF's (e.g. Low SF (LSF) scene + High SF (HSF) noise), and asked to categorize them (beaches vs. cities, or highways vs. cities). During this stage, one type of sensitization image was consistently presented to each visual field (e.g. LSF scene + HSF noise on the left and HSF scene + LSF noise on the right; or v.v.). In a subsequent test stage, ‘hybrid’ images were interleaved among the sensitization stimuli. Hybrids combine a low-passed scene from one category with a high-passed scene from the alternative category. Observers showed a bias to report the category of the hybrid component that corresponded to the SF's that they were sensitized to in a particular retinal location. None of our observers reported seeing both components of a hybrid. Such retinotopic differentiation of scale use is consistent with observers having conscious access to spatial frequency channels. We show that our findings can emerge despite an overall low or high bias, and that this varies systematically as a function of the selected HSF and LSF filter cutoffvalues. This work was supported by BBSRC Grant No. 90/S131186 awarded to Paul Sowden & Philippe Schyns.