We investigate the science goals achievable with the upcoming generation of groundbased CosmicMicrowave Background polarization experiments, focusing on one particular experiment, QUaD, a proposed bolometric polarimeter operating from the South Pole. We calculate the optimal sky coverage for this experiment including the effects of foregrounds and gravitational lensing. We find that an E-mode measurement will be sample-limited, while a B-mode measurement will be detector-noise-limited. We conclude that a 300 deg2 survey is an optimal compromise for a two-year experiment to measure both E and B-modes, and that a ground-based polarization experiment can make an important contribution to B-mode surveys. QUaD can make a high significance measurement of the acoustic peaks in the E-mode spectrum, over a multipole range of 25 < ℓ < 2500, and will be able to detect the gravitational lensing signal in the B-mode spectrum. Such an experiment could also directly detect the gravitational wave component of the B-mode spectrum if the amplitude of the signal is close to current upper limits. We also investigate how QUaD can improve constraints on the cosmological parameters. We estimate that combining two years of QUaD data with the four-year WMAP data can improve constraints on bh2, mh2, h, r and ns by a factor of two. If the foreground contamination can be reduced, the measurement of r can be improved by up to a factor of six over that obtainable from WMAP alone. These improved accuracies will place strong constraints on the potential of the inflaton field.