The thesis that the visual system represents objects has garnered empirical support from a variety of sources in recent decades. But what kinds of things qualify as “objects” in the relevant sense? Are they ordinary three-dimensional bodies? Are they the facing surfaces of three-dimensional bodies? I argue that there is no fact of the matter: what we have are equally acceptable ways of assigning extensions to the relevant visual states. The view I defend bears obvious similarities to Quine’s thesis that linguistic reference is inscrutable. Importantly, though, I argue that even if Quine was wrong about inscrutability as a thesis about language and thought, the case for the inscrutability of visual reference remains strong.