With computing power steadily rising, large eddy simulation (LES) is being used with increasing regularity, not only to explore fundamental physical issues associated with turbulence, but also to predict practical flows - the latter in preference to statistical modelling. However, a serious obstacle to the widespread use of LES is that resource requirements rise very steeply with Reynolds number (i.e. the flow speed and the dimensions of the flow) in flows that are bounded by walls. In many such flows, at least some of the gross characteristics are sensitive to frictional effects exerted by the wall on wall-remote regions, and in such circumstances the cost of LES can become comparable with to that of direct numerical simulation (DNS), and hence untenable. It is the current general view that this obstacle can only be overcome by adopting approximate modelling of the near-wall flow. The lecture will address this issue, introduce some approaches that combine simulation with approximate near-wall models, and illustrate the performance of such "hybrid schemes".