This paper addresses the notion of accuracy in reported content in direct and indirect speech by focusing on the use of masculine and feminine forms in Japanese. By analyzing naturally-occurring examples of direct reports, the paper suggests that direct speech is similar to indirect speech in that the reported content is transformed and thus “inaccurate” in similar ways to indirect speech. The analysis also shows that reporters use contextual clues to signal to the hearer that the direct reports are not to be taken literally. These clues include incongruous indexical expressions used in the reporting and reported contexts as well as mitigation expressions that approximate the accuracy of the reported content used with reporting verbs. The study provides supportive evidence that distinctions between direct and indirect speech are less clearcut than traditionally believed and that these distinctions may be based on the functions and contexts in which each form is used rather than verbatim report for direct report and inclusion of the reporter’s voice in indirect report.