Erin Goddard¹, Kateryna Marchenko¹, Colin Clifford¹; ¹UNSW, Sydney

The human ventral visual cortex contains a mosaic of visual areas, including areas with higher level responses to objects, places and faces. It also contains several regions that are highly responsive to color. However, characterising how these areas vary in their response specialisations for basic visual features is complicated by the multidimensional nature of their responses. Recent work (Goddard & Mullen, Neuroimage, 2020), using Representational Similarity Analysis (RSA) of fMRI data showed two independent changes in response properties along the ventral visual pathway (V1, V2, V3, hV4, VO1, VO2): a decreasing dependence on spatial frequency, and an increasing dependence on the presence of color contrast. This previous work used simple stimuli (radial gratings) designed to preferentially stimulate different subcortical inputs to cortex. Here, we collected fMRI data (n=20) while participants viewed complex stimuli (filtered natural images) to target areas with higher-level responses. Across 12 stimulus conditions, all images included a range of spatial frequencies (either low, high or very high), were either colored or greyscale, and their content was either intact or phase randomised. Intact color images were either naturally colored (original) or unnaturally colored (color contrast reversed). We combined pairwise classification analyses with RSA to compare functional response preferences across the entire ventral occipital-temporal cortex, using the cortical parcellation of Glasser et al. (2016, Nature). Our data replicate the trends found previously for simple stimuli: along the ventral visual pathway, responses to filtered images showed decreasing dependence on spatial frequency, and increasing dependence on color. Across all areas, responses were modulated by whether the images were intact or phase-scrambled. However, there was an interaction between color responses and image structure, with higher-level visual areas showing greater evidence of color responses for intact than for phase-scrambled images. No regions showed clear evidence of responses that varied with color naturalness.

Acknowledgements: This work was supported by the Australian Research Council (DP220100747 to EG)