Observations and stellar synthesis models of early-type galaxies have shown that most of the visible and infrared light from these galaxies is supplied by a giant-dominated stellar population (O'Connell 1974, 1976b; Tinsley and Gunn 1976; Whitford 1977; Frogel et al. 1975a, b, 1977, hereafter referred to as Paper I). The various photometric indices used by these authors establish that the contribution of main sequence stars to the total flux at 2 µm is less than 10%. However, the detailed structure of the giant branch remains uncertain. As an example, M6 III stars contribute 37% of the flux at 2.2 µm in the model of O'Connell (1976b) that fits the infrared data best, but only 12% in the best-fitting model of Tinsley and Gunn (1976). Yet, the CO indices of these models differ by only 0.02 mag (Paper I).
A spectral feature which is sensitive to effective temperature in late-type stars, and can therefore set constraints on the luminosity function of the giant branch is the broad stellar absorption band due to H2O centered at 1.9 µm (Johnson and Mendez 1970; Frogel 1971; Baldwin, Frogel, and Persson 1973). In this paper we first discuss the dependence of the H2O absorption on stellar effective temperature and luminosity. We then present multiaperture H2O observations of the nuclei of 37 early-type galaxies and 5 globular clusters, and compare the data with those of Paper I and with the results of stellar synthesis models.