Noncognitive analyses of evaluative discourse characterize moral discourse as primarily functioning to express attitudes that are not, strictly speaking, representational in the way that ordinary beliefs are representational. But, since expressivists must explain our practices of making evaluative judgments as we do, they owe us an explanation of the logical relations between these evaluative judgments and other judgments. For it is part of our ordinary evaluative practices to make inferences based upon and leading to evaluative judgments. The most thorny problem for this project has been to explain the logical relations between evaluative judgments and other judgments best expressed using evaluative terms in unasserted contexts, such as clauses embedded in conditionals. Because one may use evaluative terms in such unasserted contexts without expressing the attitudes usually associated with them in asserted contexts, it becomes hard to explain why there should be logical relations between the judgments expressed. The noncognitivist who has given this problem the most sustained attention over the years is Simon Blackburn,' and recently he has been joined by Allan Gibbard.2