It is important to remind ourselves of the definition of LINERs. Although rigorous boundaries have little physical meaning and are, to some extent, arbitrary, classification is operationally necessary. Heckman (1980b) originally defined LINERs strictly using the optical forbidden lines of oxygen: [O II] 3727 > [O III] 5007 and [O I] 6300 > 0.33 [O III] 5007. Compared with the spectra of Seyfert nuclei or H II regions, the low-ionization states of oxygen in the spectra of LINERs are unusually strong relative to its high-ionization states. Recognizing the arbitrariness of this definition, Heckman drew attention to a group of ``transition objects'' whose spectra were intermediate between those of ``pure'' LINERs (as defined above) and classical Seyfert nuclei.
As a consequence of the near coincidence between the ionization potentials of hydrogen and neutral oxygen, the collisionally-excited [O I] line in an ionization-bounded nebula arises predominantly from the ``partially-ionized zone,'' wherein both neutral oxygen and free electrons coexist. In addition to O0, the conditions of the partially-ionized zone are also favorable for S+ and N+, whose ionization potentials are 23.3 eV and 29.6 eV, respectively. Hence, in the absence of abundance anomalies, [N II] 6548, 6583 and [S II] 6716, 6731 are strong (relative to, say, H) whenever [O I] 6300, 6363 are strong, and vice versa. This theoretical expectation and the empirical evidence that extragalactic H II regions rarely exhibit [N II] 6583 / H 0.6 (e.g., Searle 1971) have led some subsequent investigators to short-cut Heckman's original definition of LINERs. For instance, it has become customary to classify emission-line objects solely on the basis of the [N II] / H ratio (e.g., Keel 1983b; Keel et al. 1985; Phillips et al. 1986; Véron-Cetty & Véron 1986). While this convention does permit a convenient first-order separation between nuclei photoionized by stars (small [N II] / H) and those photoionized by a harder, AGN-like spectrum (large [N II] / H), it provides no information on the excitation level of the AGN-like objects - in other words, one cannot distinguish LINERs from Seyfert nuclei. There are two additional complications. A classification scheme that relies on [N II] / H alone obviously is sensitive to variations in the abundance of N, which appears to be enhanced in some galactic nuclei (Storchi-Bergmann & Pastoriza 1989, 1990; Ho, Filippenko, & Sargent 1996d). The net effect would be to falsely designate star-forming nuclei having enhanced N abundance as AGNs. Moreover, the reliability of the [N II] / H ratio depends critically on the accuracy of the separation between the emission and absorption components of the H line. Although the ability to model and remove the stellar contribution to the integrated spectra is an inherent limitation to any method of classification (see Section 3 and Section 4), it is preferable to use as many line ratios as possible to strengthen confidence in the classification assignment.
In the work to be discussed below, I will be using the classification criteria advocated by Veilleux & Osterbrock (1987), which are motivated in part by the principles of Baldwin, Phillips, & Terlevich (1981). Based on the dereddened line-intensity ratios [O III] 5007 / H, [O I] 6300 / H, [N II] 6583 / H, and [S II] 6716, 6731 / H (H and H refer only to the narrow component of the line), the Veilleux-Osterbrock system is not only relatively insensitive to extinction corrections, but also conveniently falls within the spectral range of the optical survey to be described in Section 4. For concreteness, the following definitions will be adopted: H II nuclei ([O I] < 0.08 H, [N II] < 0.6 H, [S II] < 0.4 H), Seyferts ([O I] 0.08 H, [N II] 0.6 H, [S II] 0.4 H, [O III] / H 3), and LINERs ([O I] 0.17 H, [N II] 0.6 H, [S II] 0.4 H, [O III] / H < 3). Although the adopted definition of LINERs differs from that of Heckman, inspection of the full optical spectra of Ho, Filippenko, & Sargent (1993) reveals that emission-line nuclei classified as LINERs based on the Veilleux & Osterbrock diagrams almost invariably also satisfy Heckman's criteria. This is a consequence of the inverse correlation between [O III] / H and [O II] / [O III] in photoionized gas with fairly low excitation ([O III] / H 3; see Fig. 2 in Baldwin et al. 1981).
In addition to these three categories of nuclei, Ho et al. (1993) identified a class of ``transition objects'' (in retrospect, a poor choice of terminology) whose [O I] strengths are intermediate between those of H II nuclei and LINERs. Although O-star models with an appropriate choice of parameters can account for their line-intensity ratios of these objects (Filippenko & Terlevich 1992), an alternative explanation is that these objects are composite systems having both an H II region and a LINER component (Ho et al. 1993). We will define transition objects using the same criteria as for LINERs, except that 0.08 H [O I] < 0.17 H.
It should be emphasized that the classification process is not always straightforward, since the three conditions involving the low-ionization lines do not hold simultaneously in all cases. In view of potential selective N enhancement in galactic nuclei, less weight is given to the [N II] / H ratio than to either [O I] / H or [S II] / H. [O I] / H, if reliably determined, deserves the most weight, since it is most sensitive to the shape of the ionizing spectrum. Figure 1 shows sample spectra of the various classes of objects outlined above.
Figure 1. Sample optical spectra of the various classes of emission-line nuclei.