How does the retina encode natural inputs?

Biological sensory systems provide precise and reliable information about the world. They do so despite the considerable challenges posed by limited hardware and the difficulty of the task that they face. As a field, we have made a good deal of progress in understanding how such systems work using simple artificial stimuli, such as spots, spatial gratings or noise in the case of vision. Unfortunately, this understanding fails to capture sensory responses to more complex stimuli, such as those encountered under natural conditions. At least two issues contribute to this failure. First, natural stimuli recruit multiple nonlinear circuit mechanisms, and how such mechanisms work in concert is poorly understood. Second, natural stimuli have a rich and high-dimensional structure, which is impossible to probe fully in experiments and which confounds attempts to identify the specific stimulus features that elicit responses. Our approach to this challenge is to reduce the full problem of sensing naturalistic stimuli to manageable parts and directly interrogate and manipulate the underlying circuits and mechanisms to reveal their roles. I will describe recent experiments applying this approach to understand the building blocks of computation in the vertebrate retina.