In this paper I study the interaction of cheap talk and verifiable disclosure. I use a canonical model of a sender-receiver game of verifiable disclosure a la Milgrom (1981), and augment it by allowing the use of cheap talk in addition. Specifically, there is a single dimensional state privately observed by the sender. The sender's preferences are state-independent. In particular, they depend only on the receiver's posterior mean state. The sender can verifiably reveal the state exactly, and/or use cheap talk to communicate with the receiver.

Individuals are evaluated on their factual knowledge in a myriad of settings such as job interviews and exams. In this paper I ask: How to optimally design such an evaluation scheme?