high or higher than at any other epoch observed to date (many are forming stars rapidly enough to be consistent with the classical picture of galaxy formation). Beyond that, there is a great deal of promise in early attempts to describe the star formation history of the universe since z≥4, but there are a number of caveats that must be heeded before the census is much more than broadly indicative. Finally, it is now entirely feasible to map out the large scale structures delineated by these early galaxies with presently available observational facilities. Initial attempts in these directions will be described below.
A strong motivation for pushing studies of galaxy evolution beyond z~1 is the relatively coherent picture of the z<1 universe that has resulted from a number of recently completed redshift surveys (15–18). Broadly speaking, there are two principle conclusions emerging from the studies, which at first glance may appear contradictory: first, the “luminosity density,” particularly at blue and UV rest wavelengths, is a strongly increasing function of redshift from the present time to z~1; second, it appears that “big” galaxies, or those that populate the bright end of the luminosity function and would generally fall into the morphological categories of early type spiral, S0, or elliptical, have evolved very little since z~1. Evidence for these statements appears to be very strong, as redshift surveys selected in many different ways yield the same qualitative results, and the results are also supported by Hubble Space Telescope (HST) morphological studies (19, 20) and by kinematic studies of individual galaxies (21). A simple (qualitative) way of reconciling the two general inferences is that the change in the luminosity density is dominated by relatively small galaxies undergoing enhanced star formation, whereas larger (more massive) systems evolve relatively quiescently. There is some evidence that near z~1 the enhanced star formation activity is beginning to “migrate” to more massive systems with increasing redshift (16). In any case, it is fairly clear that the “formation epoch” of most massive galaxies must be prior to the epoch corresponding to z~1. If the “bottom-up”, hierarchical picture of galaxy formation (often described using catch-phrases like “gradual merging of subgalactic fragments”) is correct, then apparently most of the activity relevant to the formation of massive galaxies must have occurred at much higher redshifts. It is unclear whether this challenges the prevailing theoretical views of galaxy formation, or not.
Prior to a couple of years ago, QSOs and radio galaxies represented our sole window into the high redshift (postrecombination) universe, with many very successful surveys accumulating an impressive number of objects (22–24). Still, it was not completely clear what these relatively rare, hyperluminous active galactic nuclei (AGN) were telling us about the state of the galaxy formation/evolution process in general, and their surface densities were too low to permit a great deal of information on their clustering properties on small and intermediate scales. There is reason to believe that the formation of luminous AGN and the formation of massive galaxies ought to go hand in hand, but in the end it would be very reassuring to see pure, unadulterated star formation at high redshift (should such a thing exist).
Our own attempts to understand the nature of “normal” galaxies beyond z~1 originally grew out of the perspective on the high redshift universe afforded by working in the area of QSO absorption lines. Here, while spectroscopic surveys for field galaxies were still turning up only a few galaxies beyond z~0.7 (circa 1990), metal line QSO absorption systems were known, as were their statistics and some details concerning their chemical and other physical properties, to well beyond z~3. The original motivation for a targeted search for galaxies associated with known QSO absorption systems at z>3 (3, 4) was that it would be a means of testing whether finding objects using the Lyman discontinuity would be viable, having seen that the Lyman α emission line searches were producing mostly null results. The reasoning was that, if an object whose redshift was known exactly, with an approximately known position (i.e., near the QSO line of sight), could be found using a specially designed set of broad-band filters optimized for the detection of Lyman continuum breaks near z~3, then one would simultaneously demonstrate that the technique works, and obtain a believable redshift (or, at least the basis for a plausibility argument) for something that would always remain impossible to confirm directly using conventional spectroscopy. Happily, it turns out that we were being overly pessimistic about the prospects for spectroscopy (7).
It is well known that the nature of the spectra of galaxies is such that there is very little in the way of spectroscopic features to facilitate secure redshifts for z≥1.3, when the [OII] λ3727 line (or the 4,000-Å break region for earlier type systems) begins to leave the useful spectroscopic window. It is also the case that observing samples of galaxies selected by apparent magnitude, although very useful for some purposes, will end up being a very inefficient means of accumulating large samples of the most distant galaxies, as the median redshift of even the faintest spectroscopic samples is still considerably smaller than z~1 (25). As it has turned out, obtaining spectra of galaxies at higher redshifts, say z≥2.7, is (practically speaking) considerably easier than in the “spectroscopic no man’s land” of the z~1–2.5 regime. The reason is that there are many strong spectral features (including the infamous Lyman α emission line in many cases, but also including very strong resonance lines in the rest-frame far-UV that are both interstellar and stellar in their origins) that appear in the wavelength range over which optical spectrographs have by far the best sensitivity and the lowest background. Because of this, it turns out to be relatively straightforward (given the Keck telescopes!) to obtain spectra of large samples of galaxies that