1 Introduction

The observed total radial velocity of a galaxy (found from the redshift) can be separated into two distinct parts: a cosmological part, from the expansion of the universe, and a peculiar motion associated with the galaxy proper motion. Measurements of galaxy peculiar velocities on large scales reveal the underlying mass density fluctuations, since galaxies will stream towards an overdense region and away from an underdense region. To determine peculiar motions, a distance-indicator relation has to be used to find redshift-independent distances to galaxies. The Tully-Fisher relation can be used for spiral galaxies. Elliptical galaxies have been found to populate a nearly planar region in the three-dimensional space defined by the central velocity dispersion, the effective (half-luminosity) radius, and the effective surface brightness; this region is called the Fundamental Plane (Lucey et al. 1991; J$\o$rgensen et al. 1993). The Fundamental Plane (FP) method for distance determination is an improvement on the $D_n - \sigma$ relation. It has a tighter correlation; therefore, a better precision in distances ($\sim$ 20%) can be achieved (J$\o$rgensen et al. 1996; Scodeggio et al. 1997).

Galaxy peculiar velocities are found from a comparison of the distances with the measured redshifts. There is strong observational evidence for the existence of large-scale flows in the local universe, induced by gravity (see Strauss & Willick 1995). The dipole anisotropy of the cosmic microwave background (CMB) radiation provides a natural velocity reference frame for the analysis of galaxy motions. The dipole anisotropy, determined from COBE, implies that the Local Group (LG) moves with respect to the CMB rest frame at 627 $\pm$ 22 km s^-1 towards $l = 276 \pm 3^{\circ}$, $b = +30 \pm 3^{\circ}$ (Kogut et al. 1993). If this has a kinematic origin then, sufficiently far away, galaxy peculiar velocities should converge to the CMB frame.

Until now, the only studies which have reported measurements of the velocity field as far out as 15 000 km s^-1 are those of Lauer & Postman (LP) (1994), using brightest cluster galaxies as distance indicators, and Riess et al. (1995), using Type Ia supernovae. LP checked the convergence of the LG dipole motion to the CMB dipole, with a surprising result: a strong signature of a very large-scale bulk flow was seen, with an amplitude of 689 $\pm$ 178 km s^-1 in the direction $l = 343^{\circ}$, $b = +52^{\circ}$. The LP study implies that the local rest frame fails to converge to the CMB frame, even in regions with radii $\sim$15 000 km s^-1. A bulk flow with the statistical significance of this result rules out a whole series of cosmological models at the >95% confidence level (Feldman & Watkins 1994; Strauss et al. 1995); the LP result is in disagreement with all viable models at present.

The LP sample extended to 15 000 km s^-1, with an effective depth of $\sim 8000$km s^-1. Therefore, the logical next step was to compare the LP result with peculiar velocities as found from applying the Tully-Fisher and FP methods to galaxies extending further out than any previous peculiar velocity studies. From Tully-Fisher studies of field and cluster spiral galaxies within 8000 km s^-1, Giovanelli et al. (1996, 1998a, 1998b) concluded that these galaxies do not show any evidence of such a bulk flow.

In order to investigate the reality of large-scale streaming motion on scales of up to 150 Mpc, we have studied the peculiar motions of 179 early-type galaxies in three directions of the South Equatorial Strip, at distances out to $\sim$ 20 000 km s^-1. We have obtained new and independent measurements of the peculiar velocity field of elliptical field galaxies at a depth similar to that of LP, using a combination of photometric and spectroscopic data. For further details of the project, see Müller (1997); the results for peculiar motions are analysed in Müller et al. (1998).

In this paper we present the spectroscopic data used in our study of the large-scale motions. From the spectra, galaxy redshifts were measured, and central velocity dispersions were obtained - the accurate determination of these is essential for the FP to be applied as a distance indicator. This paper is organised as follows. Sample selection and observations are described in Sect. 2, the basic reduction of the spectra is covered in Sect. 3, and the radial velocities and central velocity dispersions are derived in Sect. 4. The applied corrections are discussed in Sect. 5, and the results are given in data tables in Sect. 6. In Sect. 7 the results are compared internally and with results from the literature.