Spatial scale processing of natural stimuli (e.g. faces, scenes) can be flexible depending on type of categorization task and prior experience (Gosselin & Schyns, 2001, V. Res., 41, 2261-2271). We argue that attentional modulation of spatial frequency (SF) channels in early vision can account for such flexibility (Sowden, Özgen & Schyns, 2002, J. Vis., 2, 283a). Last year we reported evidence (Özgen, Sowden & Schyns, 2002, J. Vis., 2, 202a) suggesting that such early mechanisms may indeed be involved: we sensitised observers bottom-up to scenes with different SF content at separate retinal locations. Perception of hybrid images (those that contain both a low-pass and a high-pass scene of different categories) was orthogonal at different retinal locations.

Here we explored whether a top-down attentional process could also drive flexible scale use. In a sensitisation stage observers were required to categorise (highway or city) briefly presented low and high-pass filtered scenes that were combined with high and low-pass noise respectively. Prior to each stimulus a symbolic auditory cue (the words “coarse” and “fine”), expected to act top-down, signalled the SF bandwidth that would contain the scene on the next trial. In a subsequent test stage, unknown to observers, we interleaved hybrid images and continued cueing observers to expect a scene at one or the other scale. Orthogonal perceptions of hybrid images depending on the cues suggest that a top-down attentional process can also result in the flexible processing of spatial scale. Such evidence indicates that in contrast to the stimulus-determined attention to a single narrow-band SF channel reported for some tasks (cf. Majaj, Pelli, Kurshan & Palomares, 2002, Vis. Res., 42, 1165-1184) when explicit top-down cues are provided endogenous switching between channels is possible.