We study the impact of the assumed velocity distribution of galactic dark matter particles on the interpretation of results from nuclear recoil detectors. By converting experimental data to variables that make the astrophysical unknowns explicit, different experiments can be compared without implicit assumptions concerning the dark matter halo. We extend this framework to include the annual modulation signal, as well as multiple target elements. Recent results from DAMA, CoGeNT and CRESST-II can be brought into agreement if the velocity distribution is very anisotropic and thus allows a large modulation fraction. However constraints from CDMS and XENON cannot be evaded by appealing to such astrophysical uncertainties alone.