Abstract
The perceived brightness of a uniform region can be profoundly altered by the presence of a light-dark luminance gradient along its border, inducing the Craik-Cornsweet-O'Brien effect (CCOB). Dakin & Bex (2003; “Natural image statistics mediate brightness “filling-in”. Proc Roy Soc, 270) proposed that such illusory changes in brightness arise as a result of the visual system reconstructing the image most likely to have elicited a particular series of responses from a bank of isotropic band-pass filters. The model effectively boosts the low spatial frequency (SF) structure of the image and suggests an early, possibly pre-cortical, mechanism for ‘filling-in’. Here we sought to assess the role of low SFs in driving the CCOB and to identify the neural site of brightness filling-in. In a block design fMRI study, healthy human volunteers viewed (a) the CCOB illusion (b) CCOB with low SFs scrambled (c) CCOB with high SFs scrambled or (d) an identical pattern containing physical luminance change (psychophysically) matched to the perceived contrast of the CCOB. Stimuli in conditions (a–c) had identical power spectra. All conditions phase reversed at 1Hz and blocks were interleaved with a fixation baseline. Using high-resolution fMRI (1.5mm3 at 3T) we imaged the lateral geniculate nucleus (LGN) and primary retinotopic visual areas. The LGN preferentially responded to the CCOB illusion. Scrambling low SFs destroyed the illusion and significantly reduced LGN activity. V1 showed a qualitatively similar pattern of responses. To test whether these observations might reflect monocular signals we examined whether perception of the illusion changed under dichoptic viewing conditions. When only the light or dark portion of the CCOB is visible to each eye the illusory percept disappears; the CCOB effect is strictly monocular. Taken together these findings suggest that brightness filling-in may be a sub-cortical phenomenon, contingent on the presence of low SF structure.
Supported by The Wellcome Trust