This chapter delves into robot dance as a prime example of directly mimicking human movements. In prior chapters, the objective of Learning from Observation (LfO) was not to replicate human movements, but to generate robot movements that achieve the same outcomes, such as identical face contacts, as those produced by the human movements. This approach addresses the hardware disparities between humans and robots. In contrast, this chapter introduces an LfO framework aimed at generating robot movements that are perceived by the audience as similar—though not necessarily identical—while addressing hardware discrepancies. This approach leverages Labanotation, a representation originating from the dance community.