In order to identify a candidate object as either a BL Lac, FSRQ or radio galaxy, an optical spectrum is required. To pinpoint the optical counterpart we used, where available, positions from either the NVSS or our ATCA survey (both of which have errors relative to the Digitized Sky Survey; note that the figure we use is significantly larger than typically quoted for NVSS and ATCA detections with F>50 mJy due to the non-planarity in the sky survey plates and the slight inconsistencies between the coordinate systems of each; see, e.g., Irwin, Maddox & McMahon 1994; Drinkwater et al. 1995) to obtain finders from the Digitized Sky Survey (using Skyview, McGlynn & Scollick 1996). For some sources which did not have NVSS positions before the time of optical observations, arcsecond positions from the Texas survey (Douglas et al. 1996) and NED (typically based on VLA data) were used. Magnitudes for all X-ray/radio sources with counterparts on the POSS and UKST plates which comprise the Digitized Sky Survey were obtained from the Cambridge APM and Edinburgh COSMOS projects (Irwin et al. 1994; Drinkwater et al. 1995), except where a blend of two or more sources were observed, in which case a magnitude was estimated by eye. All X-ray/radio sources without counterparts on the survey plates were imaged at either the KPNO 0.9m or the CTIO 0.9m telescopes. This allowed identification of all optical counterparts to R = 23.
Spectroscopic observations were conducted at the KPNO 2.1 m, MMT, Lick 3 m, ESO 2.2 m and 3.6 m, and CTIO 1.5 m telescopes. One object, WGAJ0449.4-4349, was observed by M. Ruiz at the CTIO 4 m telescope in January 1996; Dr. Ruiz has kindly allowed us to publish these data herein. In Table 1, we list all telescope runs (including, for completeness, the MMT and Lick runs, which were obtained for other projects, but during which a few DXRBS blazars were observed) and relevant details of the observing setup, such as approximate wavelength range and resolution. In Table 2, we list the details of the observations. Due to poor weather during the KPNO run (2 mostly cloudy nights out of 3) the data from this run are of lower quality than the ESO data. With the exception of the CTIO 4m and MMT observations, spectra were generally not taken at parallactic angle (mostly due to the difficulty of changing the position angle of the slit at the KPNO 2.1m, CTIO 1.5m and ESO 2.2m telescopes).
The spectra were reduced using standard IRAF routines. Data were overscan and bias-subtracted, and flatfielded using programs in the IRAF package noao.imred.ccdred, and spectra were extracted, wavelength-calibrated and flux-calibrated using programs in the package noao.twodspec. Cosmic rays were removed in the 1 and 2-dimensional data by hand.
A dereddening correction was applied to the data using the IRAF routine noao.onedspec.dered and assuming Galactic values of extinction derived from 21-cm measurements (Stark et al. 1992; Shafer et al., private communication).
We recorded the central wavelength, equivalent width, full-width at half-maximum, and flux in each spectral line. Those data will be given and analyzed in future papers. Except where noted, where only a single emission line was observed, it was assumed to be Mg II Å. Seven of 22 newly identified BL Lacs lack recognizable spectral features. These objects are not included in the redshift distributions discussed in § 5, and we have not computed luminosities for them. Higher signal-to-noise spectroscopic observations are required to obtain redshifts for these objects.