Abstract
Relatively little is known regarding the cortical representation of variations in texture appearance in humans. We used a carry-over fMRI design (Aguirre 2007 Neuroimage) to test if 1) the perceptual similarity of textures is reflected in the similarity of neural population codes; and 2) if these representations co-localize with neural populations sensitive to spatial frequency.
A texture space was created using two gray-scale photographs of natural textures (stucco and wood grain). One perceptual dimension was defined by parameter adjustment of steerable pyramids between the textures, and the other dimension by variation of image scale (essentially a change in spatial frequency distribution). Multiple exemplars of each point in the texture space were created by steerable pyramid synthesis. Iterative behavioral testing (RT to distinguish same and different pairs) established the orthogonality of the two dimensions and the equal, regular, and linear effect of perceptual steps in the space.
During scanning, participants observed a continuous stream of texture exemplars (1500 msecs / texture), ordered to counterbalance position in the texture space. Participants monitored for the infrequent appearance of a texture target not from the space. Exemplars for a given point in the texture space was randomized amongst three alternatives, minimizing pixel-to-pixel correspondence of textures. Covariates modeled a continuous modulation of neural adaptation proportional to changes in either texture or scale.
Across subjects, voxels sensitive to texture variation and scale variation were observed within ventral visual areas. Texture variation generally produced a proportional recovery from adaptation within ventral LOC, while scale variation produced recovery in earlier visual areas. The degree of overlap in the cortical response to the two dimensions of texture space varied across subjects. We are currently investigating if the degree of overlap may be related across subjects to performance on behavioral tests of integral or separable perception of the dimensions.