3 Data reduction

For the reduction of the spectra, we followed closely the method used in Paper I, which was itself similar to the process used in a series of papers on early-type galaxies (Simien & Prugniel 1997a,b,c, 1998: hereafter collectively referred to as SP). We thus limit ourselves, here, to a short abstract.

As in SP and Paper I, standard pre-processing was applied to the raw data, up to the rebinning in wavelength. The galaxy centers (r=0) were determined by a Gaussian fitting to a limited range ($\simeq 12''$) around the intensity peak. Compared to ellipticals, the centering was sometimes more difficult, because dust can cause strong asymmetries. In this case, and when the solution was obvious, we adopted the origin of the rotation symmetry, and we accordingly applied a correction to the r zero point. In the outer regions, cosmic-ray hits were removed with a median filter, and adjacent lines were combined with a variable weighting function (a Gaussian continuously wider faintward), by a process preserving the global fluxes. When present, emission lines were "cut'' at the level of the continuum. A Fourier-Fitting technique determined the central velocity dispersion $\sigma_0$ and, when possible, the radial profile $\sigma(r)$ of the dispersion, together with the projected rotation curve V(r) along the major axis. A two-pass mode (described in SP) allowed to remove cosmics on the inner lines, where the spatial resolution must be preserved.