1 Introduction

Recent work has identified the GHz Peaked Spectrum (GPS) radio sources as the most likely candidates for the progenitors of the large scale powerful radio sources (e.g., Fanti et al. 1995; Readhead et al. 1996a,b; O'Dea & Baum 1997). The GPS sources are powerful but compact radio sources whose spectra are generally simple and convex with peaks near 1 GHz (O'Dea et al. 1991; De Vries et al. 1997; for a review see O'Dea 1998) and are entirely contained within the extent of the narrow line region ($\le 1$ kpc). O'Dea (1998) estimates up to 10% of the sources listed in high frequency selected samples are GPS, underlining their relative large abundance. Still their relation to the large scale sources is not clear. It is fairly well established now that the extended radio galaxies and quasars should be unified with the compact, core dominated quasars and BL Lac objects through the combined effects of radio jet orientation and anisotropic obscuration (e.g., Urry & Padovani 1995). These objects are considered mature, well-developed radio sources. GPS objects, however, are proposed to be young radio sources which will evolve into the 10 -- 100 kpc scale radio sources. Constraints on numbers vs. linear size are consistent with simple models in which these sources propagate from the $\sim 100$ pc to Mpc scales at roughly constant velocity through an ambient medium with a density profile $\rho(R) \propto R^{-2}$, while the sources decline in radio luminosity as $L_{{\rm rad}} \propto R^{-0.5}$ (Fanti et al. 1995; O'Dea & Baum 1997; Readhead et al. 1996b; De Young 1997; Begelman 1998). In addition, our multi-color optical (O'Dea et al. 1996) as well as near-IR imaging (De Vries et al. 1998a, 1998b; De Vries 1999) has shown that host galaxy colors of nearby GPS objects are indeed consistent with non- or passively evolving ellipticals, with absolute magnitudes somewhat fainter than brightest cluster members, similar to the hosts of intermediate sized and large radio source classes. Similar conclusions have been reached by Snellen et al. (1996). Determination of the rest-frame broad-band colors (which requires redshift information) in connection with stellar synthesis modeling has proven essential for these investigations.

GPS sources have been identified with both galaxies and quasars. Whereas almost all of the GPS quasars have been identified and spectrally investigated, the GPS galaxy spectral coverage is far less complete. The optically faint end of our GPS sample (O'Dea et al. 1991), combined with the unidentified part of this list, form a considerable fraction ($\sim$20%) of the total GPS galaxy population. In a previous paper (De Vries et al. 1995, hereafter Paper I), we reported on identification and spectroscopic work using 2 m class telescopes. Here we present our subsequent efforts on 4 m class instruments, to which we add a small amount of near-IR imaging data from the 10 m Keck-I telescope. While the imaging part of this program was fairly successful in identifying optical counterparts of the radio sources, spectroscopy of these objects proved often to be severely photon limited. Therefore, in order to complete the optical and spectroscopic content of the optically faint part of our sample, we need to use 8 m class telescopes. In a further paper, we will report on our upcoming Very Large Telescope (VLT) results. This push towards higher redshifts and consequent enlargement of the GPS redshift baseline will improve our understanding of both the radio source and cosmological evolution of powerful radio sources (e.g. De Vries et al. 1998b; De Vries 1999).