Orthogonal neural representations support perceptual judgements of natural stimuli.

We studied how the structure of the mid-level neural representation of multiple visual features supports robust perceptual decisions. We combined array recording with parametrically controlled naturalistic images to demonstrate that the representation of a central object's position in monkey visual area V4 is orthogonal to that of several background features. In addition, we used human psychophysics with the same stimulus set to show that observers' ability to judge a central object's position is largely unaffected by variation in the same background features. This result supports the hypothesis that orthogonal neural representations can enable stable and robust perception in naturalistic visual environments and advances our understanding of how visual processing operates in the real world.