The results of a deep imaging and spectroscopic optical survey are presented. Three small areas of sky at 13 h, 17 h and 22 h have been imaged to faint levels in four passbands (U', B, V, and I) in the optical window. These reach approximately 27th magnitude for galaxy images, at which point the surface density of galaxies is roughly 2.5 x 10(5) deg-2. A complete catalog of objects to I(AB) approximately 24.5 is provided as the basis for future studies of the faint extragalactic population. Spectroscopic observations of a small but essentially complete sample of galaxies in and around the 22 h field have yielded redshifts for a high fraction of galaxies with B(AB) < 24.1. Detailed comparison of our deep images with similar data from other observers shows some broad similarities but also some significant differences. In particular, our faintest galaxies are in general considerably redder, and the fraction of galaxies with colors that are unrepresentable by normal galaxies is correspondingly much smaller. We also see evidence for a decline in the steep slope of the B-band number counts for magnitudes fainter than B approximately 24. The spectroscopic and imaging data indicate that the median redshift for B less-than-or-equal-to 24 galaxies is still low, <z> approximately 0.4, and close to that predicted from nonevolving models of the local galaxy population, despite the substantial enhancement that is seen in the galaxy number counts relative to these same models. A nonparametric analysis confirms that this excess relative to the nonevolution case persists even if the faint galaxy population is used to define the "nonevolving" population, emphasizing that the effect is not caused by a gross misrepresentation of the true local population. The general relationship between the surface number density of a sample of galaxies and their modal redshift is discussed for various parametric forms of the evolution of the luminosity function. The low redshifts encountered at B approximately 24 extend the trend noted by other workers to considerably fainter levels, and this indicates that the evolution for galaxies to z approximately 0.5 is best characterized as an increase in the apparent phi* rather than in L*. Some mechanisms for producing an apparent increase in phi* are reviewed, including luminosity-dependent luminosity evolution, an incorrect cosmological volume element, and the possibility of seeing galaxies that have subsequently disappeared in the sense of failing to meet the detection criteria for local galaxies. One of the faint and very blue "flat-spectrum" galaxies in this field has the spectrum of an extragalactic H II region at z = 0.13, but the other two remain stubbornly unidentifiable, including SSA 22-24, for which we had earlier suggested a redshift of 3.3 which we are not able to confirm. The nature of these flat-spectrum objects is discussed, and it is concluded that a high-redshift population remains the most likely, but unproved, explanation. Finally, we explore the implications of the very high number surface density of galaxies seen at 26th magnitude and highlight the difficulties of accounting for these galaxies with models that conserve the numbers of galaxies.