A new cosmological test is derived, based on the distribution of individual V/V-max in a complete redshift-limited sample of distant objects. The fundamental assumption is that, in any range of absolute luminosity, individual V/V-max are required to be uniformly spread over the [0: 1] range. Under the assumption of Pure Luminosity Evolution, this gives rise to a natural partition of the sample into high luminosity: redshift-limited and low luminosity magnitude-limited quasars. The behavior of V/V-max versus evolution and cosmology differs substantially in the two subsamples. This condition of uniformity is probed in any absolute magnitude bin, allowing a likelihood function to be computed from the Kolmogorov-Smirnov probabilities of each bin. Monte-Carlo simulations show that the test is mostly sensitive to the density parameter, but, under certain conditions, it also sets constraints on the space curvature and, to a lower extent, on the cosmological constant. Cross-tests between power law and exponential luminosity evolution laws are performed, showing that the functional form of luminosity evolution does not affect substantially the probabilities in the parameter space (Omega(0), Lambda). The efficiency of the test applied to two kinds of simulated quasar samples is examined: large number QSO sample, but limited to redshifts z < 2.2 or smaller in QSO number, but with higher a redshift limit. Two observational strategies are compared; aimed at the construction of such samples with the future instrumentation of the VLT. Finally, the test is applied to the UVX sample of Boyle et al. (1990). A low matter density, and a flat Universe without cosmological constant, are rejected: 0.2 < Omega < 0.8 within the 95% confidence level.
