We present new data on the redshifts and Hα profiles of 18 medium to high redshift QSOs observed by us using the FIGS spectrometer at the Anglo-Australian Telescope facility and compare the data to that of the strong UV lines obtained in spectra taken at nearly the same time. We find that the Hα line is redshifted by an average of 1100 km s−1 with respect to CIVλ1549. Low ionization lines, such as OIλ1305 and MgIIλ2798 are shifted by similar amounts. These results are difficult to reconcile with any simple models currently available, including those where dust obscuration is solely responsible for the observed velocity shifts. The similarity between the velocities of Hα and MgII, OI provides some support for models where the Balmer lines are produced predominantly in a warm HI region, perhaps by X-ray heating, while the Lyman lines arise mainly in a population of optically thin clouds. A velocity separation between the two cloud populations, along with some obscuration, could explain the main features. However, detailed differences between lines of similar ionization suggest that the true situation is more complex.

A search for correlations between the velocity shift and other quantities revealed only one, namely with the ratio of the CIII]λ1909 and CIVλ1549 line fluxes. The meaning of this is obscure, but may relate to density or projection effects. A number of other correlations proposed by previous workers as tests for their models were sought but not found.