Expressivist analyses of evaluative discourse characterize unembedded moral claims as functioning primarily to express noncognitive attitudes. The most thorny problem for this project has been explaining the logical relations between such evaluative judgements and other judgements expressed using evaluative terms in unasserted contexts, such as when moral judgements are embedded in conditionals. One strategy for solving the problem derives logical relations among moral judgements from relations of ‘consistency’ and ‘inconsistency’ which hold between the attitudes they express. This approach has been accused of conflating inconsistency with mere pragmatic incoherence. In reaction to such criticisms several recent theorists have attempted to use alternative resources. The most sophisticated noncognitivists have postulated secondary descriptive meanings dependent on moral judgement's primary expressive meanings. Recent independent suggestions by Frank Jackson and Stephen Barker attempt to solve the embedding problem by utilizing such descriptive components of moral utterances. Unfortunately, this strategy fails to handle a certain sort of example using just the descriptive resources available to noncognitivists. It must rule valid arguments invalid in virtue of equivocation in the secondary descriptive meanings. The present paper explains the problem and suggests a moral for expressivist theories.