Four models of speech dynamics are compared for robustness and accuracy: Adaptive Flux Interpolation, Flow-Based Prediction, Delta and Delta-Delta Coefficients. These models are applied to a number of speech representations. The interpolative approach is shown to be superior to prediction in all cases.