The paper develops an experimental test of a baseline model of strategic communication by a reporter who wants to convince an evaluator of being well informed. We anatomically dissect strategic behavior through a number of treatments that control for the beliefs of, as well as learning by, the evaluator. The evidence we find is broadly consistent with reporters best replying to fixed beliefs by evaluators. When we control for learning by evaluators, the amount of misreporting is again broadly consistent with theoretical predictions. However, human evaluators find it difficult to assess the informativeness of reports and to learn the strategies played by reporters. In turn, when interacting with human evaluators, reporters end up misreporting more than predicted by baseline equilibrium theory.