Is the interstellar gas of starburst galaxies well mixed?

The extent to which the ISM in galaxies is well mixed is not yet settled. Measured metal abundances in the diffuse neutral gas of star--forming gas--rich dwarf galaxies are deficient with respect to that of the ionized gas. The reasons, if real, are not clear and need to be based on firm grounds. Far-UV spectroscopy of giant HII regions such as NGC604 in the spiral galaxy M33 using FUSE allows us to investigate possible systematic errors in the metallicity derivation. We still find underabundances of nitrogen, oxygen, argon, and iron in the neutral phase by a factor of~6. This could either be explained by the presence of less chemically evolved gas pockets in the sightlines or by dense clouds out of which HIIregions form. Those could be more metallic than the diffuse medium.