Abstract
We assayed the contributions of image Fourier amplitude spectra (AS) and color in two experiments focusing on rapid categorization of affective versus neutral natural scenes. Focusing on the initial feed-forward sweep of activation through the visual hierarchy, we used briefly flashed (~33 ms) scenes that were immediately backward masked with visual textures. Previous studies hint that low-level AS information might guide rapid detection of some image categories (e.g., human faces). In Experiment 1, we used a method developed by Gaspar and Rousselet (2009) to determine whether AS information is used in image categorization, running 3 separate groups: (i) original images, (ii) images with AS information swapped within category and (iii) images with AS swapping between category. A linear support vector machine (SVM) using AS information only was able to discriminate aversive vs. neutral images with ~70% accuracy. Findings from human observers indicate that AS information contributes to affective image categorization only insofar as it destroys image amplitude-phase interactions. In Experiment 2, we focused on the role of color for rapid affective image categorization. Trichromacy provides putative advantages in food detection, detection of social cues in red-skinned conspecifics, enhanced edge and object parsing ability, and enhanced memory encoding and retrieval for some (color-diagnostic) objects. Participants viewed affective and neutral natural scenes either in (i) true color, (ii) red-green (R-G) inverted false color, (iii) blue-yellow (B-Y) inverted false color or (iv) monochromatic viewing conditions. Accuracy findings suggest that false color (particularly R-G inversion) and monochromatic images impaired performance for emotional but not for neutral content, suggesting that chromatic information may help guide affective image categorization.
Meeting abstract presented at VSS 2018