Murray & Chiang (1997) developed a model wherein broad emission lines come from the optically thick base of a rotating, outwardly accelerating wind at the surface of an accretion disk. Photons preferentially escape radially in such a wind, explaining why broad emission lines are usually single-peaked. Less well understood are the observed shifts of emission-line peaks (from 1000 km s−1 redshifted to 2500 km s−1 blueshifted in C iv, with an average 800 km s−1 blueshift).