I first discuss Chihara's claim that the presence of Liar-paradoxical sentences presents no problem for our understanding of natural languages, and argue that this cannot be held as easily as he suggests. I then consider the theories advanced by Martin, van Fraassen, Kripke and Burge which attempt to meet some of the problems involved. I argue that the claim in the first two theories that Liar sentences are ill-formed cannot be maintained, and that Burge's theory is methodologically unsound and seriously (...) incomplete. Kripke's theory, though it gives implausible truth-values to some sentences, is more satisfactory. Finally, I discuss Gupta's revision theory, and provide my own alternative which has the following advantages. First, it facilitates our understanding of Kripke's and Gupta's theories, and of their relationship. Second, it provides a third theory in which we can understand how the presence of the Liar does not radically infect a language. (shrink)