New perspectives on the old problem of wall-generated turbulence I will be reporting on joint work with Tie Wei, Joe Klewicki, and Pat McMurtry. The scenario is steady highly turbulent flow through a channel or a pipe. The problem is to determine, at least approximately, the mean velocity profile. This is the turbulent analog of laminar Poiseuille flow, whose parabolic profile is a trivial matter. We deal with the averaged Navier-Stokes equations, which immediately reveal the well known fact that 2 length scales are needed. We show that many more in fact are appropriate, giving rise to many "layers". Experimental observations (born out by our analysis) show that four principal layers are each characterized by the dominance of two of the three terms in the averaged NS equation for streamwise momentum. A systematic (but necessarily incomplete, because the averaged equations form an underdetermined system) multiscale analysis together with reasonable structural assumptions lead to the various layers and scalings, and to many other new features of the profile as well.