The textbook regression discontinuity design assumes that people do not have precise control over their own scores on the assignment variable. Without this assumption, people may manipulate their assignment scores above or below a cutoff threshold to enter a desired treatment condition. Applied researchers often try to test the manipulation assumption by posing a no-manipulation Null hypothesis under which the density function of the assignment score is smooth, and then forming test statistics that can reject the Null. McCrary (2008) provides a formal version of the test that relies on local linear regressions to estimate discontinuities in the density of the assignment variable. Frandsen (2013) presents a modified version of McCrary’s density test that accounts for naturally occurring discreteness in the assignment variable. Further afield, the statistics literature offers a family of tests for the presence of multiple modes or peaks in a density function, such as Sliverman’s Critical Bandwidth Test and Hardigan’s Dip Test. In this paper, we examine how modality tests can be applied to the RDD manipulation problem, as well as present simulation evidence on the performance of the modality tests, Frandsen’s test, and McCrary’s test under alternative models of assignment scores and manipulation.