A new decomposition has been developed in which turbulent processes in shear flows may be represented as a combination of organized and more random turbulent motions. Each component is modeled as a summation of its characteristic eddies, of strength that varies in time and space as a function of the entire process. The contribution of all turbulent eddies of the more random component are estimated with an adaptive turbulence filter, which recognizes this component as the orthogonal partner to organized motion, with a power density spectrum of appropriate shape. The decomposition recovers organized motion from time and space series of data in a physically meaningful way, and can be used to characterize interaction between coherent and more random motions. It also provides an estimate for the turbulence in shear flows that are too complex for a meaningful average motion to be identified.