I critically analyse two causal analyses of seeing, by Frank Jackson and Michael Tye. I show that both are unacceptable. I argue that Jackson's analysis fails because it does not rule out cases of non-seeing. Tye's analysis seems to be superior to Jackson's in this respect, but I show that it too lets in cases of non-seeing. I also show that Tye's proposed solution to a problem for his theory -- which involves a robot that mimics another (unseen) robot -- (...) fails. Finally I show that his 'variability' requirement is not necessary, because there are cases where someone can see an object even though the variability that Tye requires does not exist. (shrink)