The ESO Slice Project (ESP) galaxy redshift survey IV. A discussion of systematic biases in galaxy redshift determinations

We present a detailed discussion of the redshift errors associated to the ESO Slice Project measurements. For a subsample of 742 galaxies with redshifts determined both from the absorption lines (V_{abs}) and from the emission lines (V_{emi}), we find an average difference <V_{abs} - V_{emi}> ~ +100$ km/s. We find that a similar effect is present in another, deeper redshift survey, the Durham/Anglo-Australian Telescope faint galaxy redshift survey, while is absent in surveys at brighter magnitude limits. We have investigated in detail many possible sources of such a discrepancy, and we can exclude possible zero-point shifts or calibration problems. We have detected and measured systematic velocity differences produced by the different templates used in the cross-correlation. We conclude that such differences can in principle explain the effect, but in this case the non-trivial implication would be that the best-fitting template does not necessarily give the best velocity estimate. As we do not have any a priori reason to select a template different from the best-fitting one, we did not apply any correction to the ESO Slice Project velocities. However, as for a small number of galaxies the effect is so large that it is likely to have a physical explanation, we have also taken into account the possibility that the discrepancy can be partly real: in this case, it might help to understand the role of gas outflows in the process of galaxy evolution. In view of the future large spectroscopic surveys, we stress the importance of using different templates and making them publicly available, in order to assess the amplitude of systematic effects, and to allow a direct comparison of different catalogues.