The face is an interesting object to animate for several reasons: it is an important channel of communication and therefore important to any human body animation, and it is a complex object in that it is composed of many nonrigid interacting nonarticulated regions. In this paper, we examine the face, and present it as as a hierarchically structured regionally defined object. Based on this regional decomposition, and a set of primitive actions, we describe an encoding of a large set of high level facial action descriptors. We also present an application which studies the interaction between intonation and facial expressions for a given emotion. It offers a higher level of representation of the action units by grouping them into specialized functions (lips shape for phonemes, eyebrow movements). An animation system linked to facial motion property is also presented.