Expressivism faces four distinct problems when evaluative sentences are embedded in unassertive contexts like: (1) If lying is wrong, getting someone to lie is wrong, (2) Lying is wrong, so (3) Getting someone to lie is wrong. The initial problem is to show that expressivism is compatible with (1)-(3) being valid. The basic problem is for expressivists to explain why evaluative instances of modus ponens are valid. The deeper problem is to explain why a particular argument like (1)-(3) is valid. The deepest problem is to explain the meanings of evaluative conditionals like (1). Expressivists can solve the initial and basic problems simply by acknowledging that evaluative sentences have minimal truth aptness, but the deeper and deepest problems require more. The deepest problem cannot be solved even with the semantics of Gibbard and Blackburn, as is shown by an extension of Dreier's hiyo argument.