Much of the data from the ocean is Lagrangian in coming from subsurface floats or surface drifters moving with the flow. At the same time, many questions about the ocean involve transport and mixing, and therefore concern Lagrangian dynamics. These two facts have, however, remained separated as the large amount of Lagrangian data has not been used quantitatively. A new method will be described in this talk for assimilating such data into models. Our approach is based on dynamical systems ideas which also suggest new ways of placing floats and evluating models.