This talk describes joint work with Horst-Moritz Maus, Shai Revzen, Christian Ludwig, Johann Reger and Andre Seyfarth. We utilize motion capture movies of individuals to construct low dimensional models of treadmill running. Our starting point was to assume that the data could be represented as a noisy limit cycle in a hybrid (piecewise smooth) dynamical system. Springmass (SLIP) models are concrete physical systems that have long been used as caricatures of running. Feedback that determines foot placement is an essential ingredient of the SLIP models. Our primary goal was to find still low dimensional extensions of SLIP that predict center of mass location for one and two steps ahead almost as well as is possible using all of the motion capture data.