Similar to Romance and Balkan languages, Persian makes a formal distinction between indicative and subjunctive moods in its complement clauses. This paper proposes several generalizations about the distribution of the two moods, and evaluates the current theories of mood selection against Persian data. The proposal is that indicative appears when the complement is propositional, i.e., of type , and when one of the two other conditions holds: (1) the matrix predicate entails the content of the complement clause is in attitude holder’s context set, in which case the indicative mood makes no special semantic contribution, OR (2) the content of the complement clause or its negation is in speaker’s context set, in which case indicative signals the epistemic commitment of the speaker. Subjunctive, on the other hand, is selected when the conditions for the indicative are not met. As such, I argue that mood selection in complement clauses in Persian makes references to (1) the semantic type of the complement, (2) the semantics of the matrix predicate in terms of (non)veridicality, and (3) the speaker’s stand about the complement.