The metallicity of galaxies and its evolution with redshift is of paramount importance for understanding galaxy formation. Abundances in the interstellar medium (ISM) are typically determined using emission-line spectroscopy of H ii  regions. However, since H ii  regions are associated with recent SF they may not have abundances typical for the galaxy as a whole. This is true in particular for star-forming galaxies (SFGs), in which the bulk of the metals may be contained in the neutral gas. It is therefore important to directly probe the metal abundances in the neutral gas. This can be done using absorption lines in the Far UV. We have developed techniques to do this in SFGs, where the absorption is measured for sightlines toward bright SF regions within the galaxy itself. We have successfully applied this technique to a sample of galaxies observed with FUSE. The results have been very promising, suggesting in I Zw18 that abundances in the neutral gas may be up to 0.5 dex lower than in the ionized gas. However, the interpretation of the FUSE data is complicated by the very large FUSE aperture (30 arcsec), the modest S/N, and the limited selection of species available in the FUSE bandpass. The advent of COS on HST now allows a significant advance in all of these areas. We will therefore obtain absorption line spectroscopy with G130M (~1150–1450 Å) in the same sample for which we already have crude constraints from FUSE. The results will provide important new insights into the metallicities of galaxies, and into outstanding problems at high redshift such as the observed offset between the metallicities of Lyman Break Galaxies and Damped Lyman Alpha systems.