I have reinterpreted Rucinski's line profiles (broadening functions) for the hot A-type W Ursae Majoris binary AW UMa in terms of various contact models, finding that the detached/semidetached geometry he postulates does not fit the system's light variation or its line profiles. The highly overcontact geometry that actually fits the system's ellipsoidal light variation reproduce the total breadth of the lines but give profiles too broad to fit their shapes. The best solution to this dilemma is differential rotation of both stars in which the mid-latitudes are rotating 20-30 per cent slower than synchronously. This approach preserves the traditional light-curve solution and explains the drift of ripples seen in the line profiles. In addition, there are obviously other flows superimposed on the rotation which make the profiles asymmetrical. Fitting the line profiles is complicated by uncertainties in how the equivalent widths of the lines change with effective temperature and how much they are limb darkened. The contribution of any point on the stellar surface to the equivalent width depends only very weakly on local temperature throughout the range of W UMa binaries. The limb darkening of the lines is probably small, although the effect remains unexplored. I also speculate about the nature of hypothetical star-spots in this system.