Conventional velocity analysis of seismic data is based on normal moveout of common-depth-point (CDP) traveltime curves. Analysis is done in a hyperbolic framework and, therefore, is limited to using the small-angle reflections only (muted data). Hence, it can estimate the interval velocities of compressional waves only, since mode conversion is negligible when small-angle arrivals are concerned.We propose a new method which can estimate the interval velocities of compressional and mode-converted waves separately. The method is based on slant stacking or plane-wave decomposition (PWD) of the observed data (seismogram), which transforms the data from the conventional T-X domain into the intercept time-ray parameter domain. Since PWD places most of the compressional energy into the precritical region of the slant-stacked seismogram, the compressional-wave interval velocities can be estimated using the 'best ellipse' approximation on the assumption that the elliptic array velocity (stacking velocity) is approximately equal to the root-mean-square (rms) velocity.Similarly, shear-wave interval velocities can be estimated by inverting the traveltime curves in the region of the PWD seismogram, where compressional waves decay exponentially (postcritical region).The method is illustrated by examples using synthetic and real data.