This paper extends the idea of classification schemes for static surface curvature into the temporal domain. We seek to identify regions in sequences of depth data that exhibit variations in shape change, and to characterise the nature of the deformation. From observing the change in principle curvatures we show how it is possible to decouple the type of change into one of fifteen classes, and also reveal the extent of alteration. Results are presented for synthetic and real data sequences, with additional alignment performed to accommodate global motion. This technique shows promise in analysing data from video-rate range sensors, with potential applications in biometric and psychological analysis of the face and other deformable objects.