Nelson Goodman claims to have given us a criterion for likeness of meaning that is more stringent than simple coextensiveness and yet that avoids the familiar extentionalist objections. The notion of a nominal compound plays a key role in his account. I show that Goodman's comments concerning this notion are inadequate, that his comments concerning expressions like unicorn-picture are subject to two serious objections: they don't support his claims about likeness of meaning and they make English an unlearnable language.