Vision science has demonstrated considerable success in characterizing early- and mid-level visual processes through reductionism and tightly controlled experiments. However, an understanding of high-level visual processing will benefit from new technologies and approaches that bring vision science closer to the real world in all its rich complexity. We propose strategies to address problems inherent in conventional approaches. These include more natural, rich, and complex stimuli and tasks; technologies such as virtual reality that represent visual space more veridically; deeper recognition of the importance of the participant as an active agent rather than a passive observer; deeper consideration of the ecological goals of vision in the context of evolution and development, including for artificial neural network models; and new analytic and theoretical approaches that treat complexities of the natural world as data, not confounds. We provide diverse examples of how such approaches have advanced our understanding of visual processing in everyday life.
Culham et al. (Thu,) studied this question.