6.2. Electron Density
The electron density can be estimated from the ratio of the two [S II] lines, at least for the portion of the narrow-line region (NLR) characterized by densities not greatly in excess of the critical density of [S II] (~ 3 x 103 cm-3), above which the lines become collisionally de-excited. As shown in Section 6.4, a range of densities, spanning nearly five orders of magnitude, exists in the NLRs of some LINERs and Seyferts. The [S II] densitometer strictly probes only the low-density regions. Figure 5b indicates that LINERs have smaller electron densities (median ne = 175 cm-3) than Seyferts (median ne = 290 cm-3), and the difference is highly significant according to the K-S test (PKS = 0.00078). Transition objects have smaller densities than LINERs, most notably in a considerable excess of low-density members, as seen in a large fraction of H II nuclei.
These density measurements are in substantial disagreement with previous
studies, which typically find densities on the order of 1000
cm-3
(Stauffer 1982b;
Keel 1983b;
Phillips et al. 1986).
The discrepancy can be traced
to a common culprit, which serves as an excellent lesson in the pitfalls of
measuring weak lines in galaxy nuclei. Careful inspection of
Figure 3
reveals that [S II] 
6716 is
affected by a stellar absorption line
due to Ca I 6718 [see also
Fig. 12 of
Filippenko & Sargent
(1985)].
Since the emission lines in the sources of interest generally have very low
equivalent widths (typically 2-3 Å), the absorption feature, though weak,
can significantly depress the ratio of [S II]
6716 to [S II]
6731;
this effect will artificially raise the derived electron density. The older
studies in question almost never performed starlight subtraction to the degree
of precision required to notice this, thereby systematically overestimating
the inferred densities.
It is interesting to point out that the electron densities among Seyfert nuclei appear to decrease with decreasing nuclear luminosity. In a sample of bright, mostly Markarian Seyfert 2 galaxies, Koski (1978) found that the average density, again as determined from [S II], is ~ 2000 cm-3, far greater than that encountered in the present sample of low-luminosity Seyferts. Although the systematic effect discussed above may also affect Koski's measurements to some degree, it probably cannot account for the large difference, especially in view of the much larger emission-line equivalent widths in his sample. The same trend of density variation is also clearly seen in the smaller sample of low-luminosity sources published by Phillips et al. (1983); for the 17 objects in which both [S II] lines were tabulated, I calculate < ne > = 340 cm-3.