I argue for a version of the causal analysis of seeing which I call the 'potential information' analysis. I proceed initially by considering some standard causal analyses, those of Tye and Jackson. I show that these analyses are too weak, for they allow cases of hallucination to count as seeing. I argue that what is central to seeing is that our visual experiences provide a means of gaining true beliefs about objects. This, however, does not mean that we must actually (...) gain true beliefs about objects in any particular case. Rather, what must be the case is that a perceiver of our sort could gain true beliefs about objects on the basis of experiences like ours. I defend this analysis against various objections, making important qualifications to it as I do so. (shrink)