Topographic maps of the core-mantle boundary (CMB), obtained by inverting different subsets of International Seismological Centre arrival time data involving either waves refracted through the boundary (PKP) or waves reflected from the topside (PcP), do not agree. We report on the implications for models of CMB structure when data from bottomside reflections (notably PKKP) are included in the inversion. Travel time residuals with respect to a reference Earth model (PREM) of all core phases are positively biased, but the mean residual is different for different phases and must be accounted for prior to inversion. The new data illustrate that models of CMB topography can explain only a small fraction of the data variance. The new CMB model involves both topography and a laterally varying boundary layer. The CMB itself is smooth compared to previous results. ¿ American Geophysical Union 1989