Proper orthogonal decomposition (POD) has proven to be an excellent tool forextract- ing the energy-containing temporal and/or spatial structures of a given flow field. These structures are portrayed in modes that can be used to describe a flow and significantly reduce the size of time-varying data sets by only retaining those most energetic modes. Unfortunately, many flow fields, such as wakes shedding from cylinders and jet flows, have a dominant mode that is associated with the convection of structures. POD techniques in an Eulerian frame are successful at extracting the convectionof the energy-containing structures, but are unable to identify the next-order phenomena such as vortex growth, bursting, and merging. Herein, a new technique called Lagrangian POD is proposed, in which POD techniques are applied in windows that move so as to track the features; the tracking is done by cross-correlating the various temporal snapshots. Shown here are results that have been obtained with Lagrangian POD on experimental cylinder wake data, where one can clearly identify several physical phenomena of an individual vortex in the wake such as convection, dissipation, and spreading.