This thesis investigates the evolution of galaxies as a function of look-back time, concentrating on early-type cluster galaxies. We demonstrate that selecting galaxies in the near-infrared produces samples that are representative of the galaxy population at intermediate and high redshift, and that such samples are likely to contain approximately the same proportion of all galaxy classes independently of z, avoiding the biases introduced by optical selection criteria. We have developed observing and data processing techniques using infrared arrays that yield high precision near-infrared photometry to very faint limits. Combining such data with optical photometry for the galaxy samples, we have been able to quantify the amount of colour evolution in early-type cluster galaxies up to redshifts z ~ 0.9.At z = 0.37 the mean colour-luminosity (c-L) relation for early-type cluster galaxies is compatible in slope and zero point with that of present-day ellipticals, indicating that the bulk of the stellar population at that epoch does not show any significant colour evolution. However, the scatter around the main c-L line is significantly larger than the observational errors, and distinctively non-Gaussian. We interpret this as evidence that a large fraction (~ 60%) of the early-type members have suffered a burst of star formation (involving ~ 10% of the galaxy mass) ~ 1 Gyr prior to the epoch of observation. In the redshift range 0.5 < z < 0.9, we detect systematic colour evolution with redshift in the red cluster galaxies. In particular, at z ~ 0.9 there are no galaxies as red as present day ellipticals at all wavelengths. The detected evolution is compatible with the passive ageing of stellar populations formed before z = 2. Superimposed on that, there is evidence for subsequent bursts of star formation happening on a substantial fraction of the cluster galaxies, but affecting only a small fraction of the total galaxy mass. We suggest this may be related to the Butcher-Oemler effect observed at lower redshifts. We did not detect significant amounts of galaxy luminosity evolution at 2µm, in contrast with the results for radio galaxies. We finally propose and test three new approaches to the study of normal field and cluster galaxies beyond z = 1: a study of gravitationally lensed galaxies, a search for galaxies producing clustered metallic absorption lines in QSOs and an infrared follow-up of ROSAT X-ray detected high redshift clusters.