We forecast the sensitivity with which the Murchison Widefield Array (MWA) can measure the 21 cm power spectrum of cosmic hydrogen. The MWA is sensitive to roughly a decade in scale (wavenumbers of k $ 0:1Y1 h Mpc À1 ). This amounts primarily to constraints on two numbers: the amplitude and the slope of the 21 cm power spectrum on the scales probed. We find, however, that the redshift evolution in these quantities can yield important information about reionization. We examine a range of theoretical models, spanning uncertainties in the nature of the ionizing sources and the abundance of minihalos during reionization. Although the power spectrum differs substantially among these models, a generic prediction is that the amplitude of the 21 cm power spectrum on MWA scales (k $ 0:4 h Mpc À1 ) peaks near the epoch when the intergalactic medium (IGM) is %50% ionized. Moreover, the slope of the 21 cm power spectrum flattens as the ionization fraction increases and the sizes of the H ii regions grow. With regards to detection sensitivity, we show that the optimal MWA antenna configuration for power spectrum measurements would pack all 500 antenna tiles as closely as possible in a compact core. Detecting the characteristic redshift evolution of our models will help to confirm that observed 21 cm fluctuations originate from the IGM, and not from foregrounds, and will provide an indirect constraint on the evolution of the volume-filling factor of H ii regions during reionization. After two years of observations, the MWA can constrain the filling factor at an epoch when x i h i$ 0:5 to within roughly AE x i h i$ 0:1 at 2 confidence. Subject headingg s: cosmology: theory -intergalactic medium -large-scale structure of universe