The flushing of Boston Harbor, a shallow, tidally dominated embayment with little fresh water input, is investigated using a depth-averaged model. The modeled tidal currents exhibit strong spatial variability and ebb/flood asymmetry due to complex topography and coastline geometry and were verified by shipboard acoustic Doppler current profiler measurements. At the inlets to the harbor, the asymmetry between flood and ebb gives rise to a net exchange of water, which acts over successive tidal cycles to flush the harbor. The flushing is examined by tracking water that starts out in Boston Harbor for 40 M2 tidal cycles. The tidal flushing is very efficient at mixing water in the vicinity of the inlets over several tidal cycles, but efficiency decreases with time as ``tidal mixing regions'' form on either side of the harbor inlets. When wind forcing is included, the wind-driven currents act to flush the tidal mixing regions, giving rise to more efficient flushing. The exception is when the wind is from the southwest, which confines the jet-like ebb flow from the harbor and therefore reduces the flushing efficiency. In general, flushing is shown to be a two-step process: (1) rapid exchange due to tides over a large region in the vicinity of the harbor inlets and (2) flushing of this region by wind-driven flow. The model also demonstrates that flushing is not uniform over the entire harbor but occurs rapidly in the deep tidal channels and slowly in the regions of weak tidal currents around the harbor periphery. Although the depth-averaged approach to flushing is appropriate over most of the harbor due to the harbor's shallow depth and broad depth distribution, the lack of bathymetric variability and the presence of locally important density driven currents in the Boston Inner Harbor indicates that flushing of this localized area must be approached with a three-dimensional model.