As one probes further into redshift space, it would be surprising if the Hubble sequence did not begin to break down as one approaches the initial epoch of galaxy formation. The interesting questions are where the system breaks down, and how. Does one class of galaxy within the system begin to gradually dominate over the overs, indicating that the sequence itself contains the ``galaxian ground state'' as one of its classification bins? Or do entirely new classes of galaxy emerge? With the advent of HST, we are now in a position to address such questions.
Recent work from deep HST imaging surveys [45, 43, 31, 2, 3, 42] coupled with ground-based spectroscopic work, [58, 20, 59, 36] has shown that much of the rapidly evolving faint blue galaxy population [16, 51, 58] is comprised of ``morphologically peculiar'' galaxies. This term has been rather liberally applied to encompass a vast range in observed galaxy forms, but in fairness more precise classifications have been difficult to apply, because at high redshifts galaxies are being observed in the rest-frame ultraviolet (a ``morphological K-correction''), where little is known about the appearance of the local galaxy population. However the conclusion that these systems are intrinsically peculiar seems secure, because the general effects of cosmological bandshifting on normal Hubble types has been determined from simulations . In general the observed faint peculiar systems do not resemble the appearance of bandshifted Hubble sequence galaxies. Furthermore the redshift range probed by most deep I-band HST imaging corresponds to z < 1 (with the exception of Lyman-limit selected systems discussed below), in which the effects of cosmological bandshifting on morphology are not yet extreme.
Figure 4. Morphologically segregated number counts from Brinchmann et al. (1998) , based on data from the CFRS/LDSS collaboration. The solid-line bins show counts as a function of redshift for irregular/peculiar/merger systems (top), spirals (middle), and ellipticals (bottom). Morphological classifications have been made from WF/PC2 images using an automated technique based on central concentration and asymmetry of galaxian light . The shaded region corresponds to the size of the ``morphological K-correction'' on the classification, accounting for the effects of observing the galaxies at bluer rest wavelengths as a function of redshift. Superposed on the observed histograms are the predictions of no-evolution (dashed) and 1 mag linear evolution to z = 1 (dot-dashed) models. At z ~ 1 approximately 40% of the galaxy population is morphologically peculiar.
Perhaps the clearest evidence for the increasing importance of morphologically peculiar systems as a function of redshift has been obtained by Brinchmann et al. . These authors applied an objective classification scheme, calibrated by simulations, to a set of ~ 300 HST I814-band images of galaxies with known redshifts taken from the CFRS  and LDSS  surveys. Because the statistical completeness of this sample is very well understood, reliable number-redshift histograms can be constructed for the various morphological types. The morphologically resolved n(z) result obtained by Brinchmann et al. is shown in Figure 4, and confirms that irregular/peculiar/merging systems are already greatly in excess of the predictions of no-evolution and mild-evolution models at redshifts z ~ 1. It is clear that by z ~ 1 approximately 1/3 of galaxies are morphologically peculiar.
What are these peculiar systems? The answer to this question is currently unknown. These galaxies are often referred to as ``irregulars'' in the literature, but it is probably a mistake to regard these systems as counterparts to local irregulars. As pointed out in the previous section, luminous irregulars are virtually unknown in the local Universe, while the high-redshift peculiar systems are generally both large and bright (3)
Let me conclude this section by pointing out that although (because of space limitations) my focus in these lectures is on peculiar galaxies at high redshifts, one should not lose sight of the importance of tracking systematic changes in the characteristics of morphologically normal systems. An interesting study has recently been completed by Lilly and collaborators , which seems to indicate little change in the space density of large spiral systems to redshift z = 1. The distribution of galaxian disk sizes is a sensitive probe of hierarchical formation scenarios (in which disk sizes are expected to strongly evolve with redshift). Attempts to understand the implications of this observation in the context of hierarchical models are underway .
(3) It is left as an exercise for
the reader to show that the selection function for a magnitude-limited
survey sampling a Schechter luminosity function results in a roughly
gaussian-shaped absolute magnitude distribution that peaks near
(3) It is left as an exercise for the reader to show that the selection function for a magnitude-limited survey sampling a Schechter luminosity function results in a roughly gaussian-shaped absolute magnitude distribution that peaks near L*.