We present a new catalogue of faint southern galaxy clusters identified from examination of high-contrast film derivatives of a set of 55 prime focus AAT photographic plates taken in one or both of two pass-bands (J = b(J) and F = r(F). Visual scans of these films have been compared with machine measurements of the original plates and the results demonstrate that the films offer ready access to the positions (but not photometry) of large numbers of faint objects to limiting magnitudes of b(J) congruent-to 24.4 +/- 0.3 and r(F) congruent-to 22.9 +/- 0.2. Our cluster candidates are chosen according to their contrast, sigma(cl), above the fluctuations in the field counts as determined locally on the same film. A total of 112 candidate clusters are tabulated down to a contrast level of sigma(cl) = 2.0 (at which point marginal effects become apparent). Cross-correlations with published catalogues (and spectroscopic measurements) confirm that the bulk of these clusters are more distant than those found by Abell and co-workers. By applying our search criteria to simulated fields containing clusters of known richnesses, and with fluctuations in the field counts matching those seen in the data, we determine our cluster list should be complete above a limit sigma(cl) = 3.25. Given recent estimates of the volume density of clusters of various richnesses, our simulations predict the observed number of F clusters seen above the sigma(cl) = 3.25 limit, but predict too few J clusters. To test for the origin of this excess of blue clusters we have measured redshifts for 95 faint galaxies in 29 of the 44 clusters with sigma-cl > 4.0 using a variety of telescopes and instruments. A spectroscopically complete sub-sample in 16 fields allows us to determine the absolute space density. The spectroscopy confirms the reality of the clusters with sigma-cl > 4.0 and shows a redshift range for the F sample similar to that expected without luminosity or cluster evolution; the J sample extends to higher redshifts than expected. A detailed study of the highest redshift clusters suggests the abundance of high-redshift J clusters arises either from projections from foreground spiral-rich groups or via recent star formation within the distant cluster environments. In support of the latter hypothesis, we demonstrate how bursts of star formation in a subset of cluster galaxies would transform the J contrast leaving the F contrast largely unaffected - in agreement with the basic trend observed. We discuss the implications of our results for further searches for high-z clusters. Only with the advent of wide-field infrared imaging, or X-ray selected samples, might galaxy-dependent selection effects associated with variations in stellar populations at a given epoch be overcome.
