It was apparent from some of the earliest redshift surveys that the central regions of galaxies often show evidence of strong emission lines (e.g., Humason, Mayall, & Sandage 1956). A number of studies also indicated that in many instances the spectra revealed abnormal line-intensity ratios, most notably the unusually great strength of [N II] relative to H (Burbidge & Burbidge 1962, 1965; Rubin & Ford 1971). That the optical emission-line spectra of some nuclei show patterns of low ionization was recognized from time to time, primarily by Osterbrock and his colleagues (e.g., Osterbrock & Dufour 1973; Osterbrock & Miller 1975; Koski & Osterbrock 1976; Costero & Osterbrock 1977; Grandi & Osterbrock 1978; Phillips 1979), but also by others (e.g., Disney & Cromwell 1971; Fosbury et al. 1977, 1978; Danziger, Fosbury, & Penston 1977; Penston & Fosbury 1978; Stauffer & Spinrad 1979).
Most of the activity in this field culminated in the 1980s, beginning with the recognition (Heckman, Balick, & Crane 1980; Heckman 1980b) of LINERs as a major constituent of the extragalactic population, and followed by further systematic studies of larger samples of galaxies (Stauffer 1982a, b; Keel 1983a, b; Filippenko & Sargent 1985; Phillips et al. 1986; Véron & Véron-Cetty 1986; Véron-Cetty & Véron 1986; see summary in Table 1). At optical wavelengths the nuclear component in a ``normal'' galaxy is generally much weaker than the stellar background of its bulge. In addition to having very small equivalent widths, many of the emission lines are severely blended and diluted by stellar absorption lines. Thus, adequate removal of the stellar contribution to the integrated spectrum is an absolute prerequisite to any quantitative analysis of the emission-line component for the sources in question. In this regard, with the exception of the survey by Filippenko & Sargent (which will be considered later), the rest suffer from several major drawbacks (Table 1). Although most of the surveys attempted some form of starlight subtraction, the accuracy of the methods used tended to be fairly limited (see discussion in Ho, Filippenko, & Sargent 1996c), the procedure was sometimes inconsistently applied, and in two of the surveys subtraction was largely neglected. The problem is exacerbated by the fact that the effective aperture used for the observations was quite large, thereby admitting an unnecessarily large amount of starlight. Furthermore, most of the data were collected with rather poor spectral resolution ( 10 Å). Besides losing useful kinematic information, severe blending between the emission and absorption components further compromises the ability to separate the two.