Measured abundances in interstellar gas will of course depend on how well, and on what timescale, the ISM is mixed and on what timescale fresh metal cools and becomes visible. Kunth and Sargent (1986) argue that H II regions are self-polluted within the ongoing burst providing the ejected oxygen can recombine fast enough to be observed in the H II zone while some can become neutral and be observed in the H I cloud. Pantelaki and Clayton (1987) dismiss this possibility from the fact that most of the ejecta should remain for a long time in the hot gas generated by SN events. Spiral galaxies display radial abundance variations, indicating that radial mixing is inefficient. On the other hand, barred spirals display smaller abundance gradients, since bar perturbations induce radial gas flows. Roy and Kunth (1995) discuss mixing processes in the ISM of gas rich galaxies, and conclude that dwarf galaxies are expected to show kpc scale abundance inhomogeneities. On the other hand, chemodynamical models (Hensler and Rieschick 1998), predict that the ISM will be well mixed and chemically homogeneous through cloud evaporation.
The observational situation is still not completely clear and rather few dwarfs have been subject to high quality studies of their chemical homogeneity. Most dwarf irregulars seem rather homogeneous (Kobulnicky and Skillman 1996, Kobulnicky 1998) with the exception of NGC 5253 where local N/H overabundances has been attributed to localised pollution from WR stars (Kobulnicky et al. 1997). There is also marginal evidence for a weak abundance gradient in the LMC (on the scale of several kpc, Kobulnicky 1998). The situation is less clear in BCGs: In IIZw40, Walsh and Roy (1993) found a factor two variation in the oxygen abundance. IZw18 appears to be rather homogeneous (e.g. Skillman and Kennicutt 1993, Vilchez and Iglesias-Páramo 1998, Legrand et al. 1999) while recent spectroscopy of SBS0335-052 (Izotov et al. 1999b) reveals small but significant variations in accordance (though to a much lesser extent) with previous results (Melnick et al. 1992).
The possibility that metallicities in the neutral gas phase are orders of magnitude below the H II region abundances would be an ultimate test of large scale inhomogeneities. Recent O/H abundance determination in the H I envelope of the very metal-poor compact dwarf IZw18 (Kunth et al. 1994) suggests the possibility of a striking discontinuity between the H I and H II gas phases: the measured O/H in the cold gas appears to be 30 times lower (i.e. ~ 1/1000 Z) than that of the associated H II region. Note however that Pettini and Lipman (1995) have strongly warned against the use of the O I interstellar lines in the deriving O/H for neutral gas, mainly because these lines are saturated and the velocity dispersion is unknown, (see also van Zee et al. 1998). A further HST observation of O and S lines in IZw18 has unfortunately not produced consistent results (unpublished). Nevertheless Thuan et al. (1997) circumvented this problem in the case of SBS0335-052 although their result awaits independent measurement of unsaturated lines such as the SII 1256 multiplet.
A crucial question would be how to interpret the presence of metals in the H I zone. If indeed IZw18 experiences its very first episode of star formation, the oxygen present in the ionised gas should originate from the ongoing burst and one can speculate that metals in the H I were produced at an earlier epoch from population III stars prior to the collapse of the proto galaxy. On the other hand the enrichment in the neutral gas could originate in a previous burst allowing for a time scale long enough to homogenise a cold cloud of 1 kpc diameter as discussed by Roy and Kunth (1995), but see Tenorio-Tagle (1996). To circumvent the problem, Legrand (1998) has conjectured that in between bursts, IZw18 had maintained a minimum continuous star formation rate of only 10-4 M/yr over the last 14 Gyrs. Such a SFR is comparable to the lowest SFR observed in low surface brightness galaxies. This scenario nicely explains the lack of galaxies with metallicities below IZw18, the absence of H I clouds without optical counterparts and the homogeneity of the metal abundances in dwarf galaxies. The question of the metallicity of the cold neutral gas is indeed very important for understanding how much enrichment has really occurred, since for many galaxies, a considerable fraction of the total baryonic mass is in the form of neutral hydrogen. If some dwarfs are not well mixed on large scales (e.g. LSBGs which have large H I-discs) they would appear more metal rich after converting a given fraction of gas into stars, because one would be observationally biased towards the star-forming, more metal rich regions. On the other hand, if some dwarfs can mix their whole ISM on not too long timescales, the fresh metals will be efficiently diluted and the galaxy appear more metal-poor. Galaxies with very turbulent ISM and those involved in mergers could possibly mix more easily on large scales.
Once ejected into the ISM, part of the metals may be locked up into dust grains. This is observed in the local ISM for some elements (see e.g. Pagel 1997) and may result in strange element ratios and apparent under-abundances in extragalactic H II regions, although the effect is believed to be small for H II regions, due to grain destruction. In fact Bautista and Pradhan (1995) find that in the Orion nebula, the depletion of iron into dust grains is probably a minor effect. In the colder gas associated with damped Ly systems (see Sect. 8.2) depletion onto dust grains may be important for some elements.