Given a knowledge base encoded as a set of first-order sentences T, one would naturally like to be able to update T as new information arrives. Unfortunately, there is no clear-cut application-independent choice of semantics for updating T, and a variety of candidate semantics have appeared in the database, AI, and philosophical literature [Abiteboul 85, Alchourrón 85, Appelt 88, Borgida 85, Dalai 88, de Kleer 87, Fagin 83, 86, Foo 89, Forbus 89, Gärdenfors 88ab, Ginsberg 86, 88ab, Goodman 83, Harman 86, Hegner 87, Jackson 89, Katsuno 89, Lewis 73, Liu 88, Oddie 78, Pollack 76, Reiter 87, Rescher 64, Satoh 88, Weber 86, Winslett 86abc, 88bc]. These works vary widely in motivation, terminology, and goals; the applications discussed in these papers range from logical theory revision, counterfactuals, and philosophy of language (all active areas of research among modern philosophers, as well as the likes of Aristotle, Hume, and Mill) to diagnosis, database updates, reasoning about the effects of actions in the physical world, and discourse understanding.