Recent work suggests that feature constraints in the training datasets of deep neu- 1 ral networks (DNNs) drive robustness to adversarial noise (Ilyas et al., 2019). 2 The representations learned by such adversarially robust networks have also been 3 shown to be more human perceptually-aligned than non-robust networks via image 4 manipulations (Santurkar et al., 2019; Engstrom et al., 2019). Despite appearing 5 closer to human visual perception, it is unclear if the constraints in robust DNN 6 representations match biological constraints found in human vision. Human vision 7 seems to rely on texture-based/summary statistic representations in the periphery, 8 which have been shown to explain phenomena such as crowding (Balas et al., 2009) 9 and performance on visual search tasks (Rosenholtz et al., 2012). To understand 10 how adversarially robust optimizations/representations compare to human vision, 11 we performed a psychophysics experiment using a metamer task similar to Freeman 12 & Simoncelli (2011); Wallis et al. (2019); Deza et al. (2017) where we evaluated 13 how well human observers could distinguish between images synthesized to match 14 adversarially robust representations compared to non-robust representations and a 15 texture synthesis model of peripheral vision (Texforms (Long et al., 2018)). We 16 found that the discriminability of robust representation and texture model images 17 decreased to near chance performance as stimuli were presented farther in the 18 periphery. Moreover, performance on robust and texture-model images showed 19 similar trends within participants, while performance on non-robust representa- 20 tions changed minimally across the visual field. These results together suggest 21 that (1) adversarially robust representations capture peripheral computation better 22 than non-robust representations and (2) robust representations capture peripheral 23 computation similar to current state-of-the-art texture peripheral vision models. 24 More broadly, our findings support the idea that localized texture summary statis- 25 tic representations may drive human invariance to adversarial