New spectroscopic observations of a statistically complete sample of 50 very faint I-band selected galaxies with I(AB) < 22.5 show that, although they extend out to z approximately 1, the redshifts are generally low, [z]obs approximately 0.5 + 0. 1. This is the same as simple '' no-evolution '' predictions, despite a roughly threefold excess in the number-magnitude counts relative to a no-evolution q0 = 0.5 model. Within this cosmological framework, this supports the idea that the faint field galaxy population is dominated by a rapidly evolving population of numerous and presumably relatively low-mass galaxies with the spectral energy distributions of Irr galaxies and rest-frame V-band luminosities around M(V) approximately -20.0, i.e., about 1 mag below the present-day L*. On the other hand, a LAMBDA-dominated universe with q0 approximately -0.85 would require little evolution in the luminosity function at rest wavelengths lambda approximately 5500 angstrom back to z approximately 0.5, although substantial evolution of the luminosity function at shorter wavelengths is still required. The morphologies of half of the blue galaxies at z approximately 0.4 are peculiar in some way, possibly indicating interactions. The number of high-redshift luminous galaxies, with z > 0.7 and M(B) approximately -21.5, i.e., up to a magnitude above present-day L*, is roughly as expected from the present-day rest-frame B-band luminosity function without substantial luminosity evolution back to z approximately 1. However, these galaxies are much bluer than present-day galaxies of similar luminosities, and there are no galaxies in the sample with (V-I) colors as red or redder than those of present-day Sbc galaxies. This could simply indicate that normal massive spiral galaxies are forming stars more vigourously at z approximately 0.8 than at present. Alternatively, if the blue galaxies at high redshift are more luminous versions of the evolving population seen at z approximately 0.5, then the absence of redder objects at high redshifts would suport the idea that most massive galaxies have been built up by mergers over the last half-Hubble time.
