Most absolute seismic location methods are based on measurements of the arrival time of one or more phases at multiple stations. Geiger’s method seeks to minimize the residuals between observed and predicted arrival times using successive linear approximations to find the event parameters (location and origin time) resulting in the minimum total residual. The uncertainty is subsequently evaluated based on Gaussian statistics evaluated at the final location solution, and is typically parameterized in terms of a confidence or coverage ellipsoid. This traditional approach works well where the probability distribution described by the residuals has a single well-defined peak, and is approximately ellipsoidal in shape. As we show in this poster, the assumption of a single well-defined peak is reasonable for simple Earth models, but the assumption of an ellipsoidal shape can be unreasonable. Further, the addition of backazimuth measurements can result in uncertainty ellipses where the statistics are incompatible with basic physics con-straints. Geiger’s method was developed as an efficient way to solve the location problem at a time when computational constraints were limited. For many problems, it remains a suitable approach. However, modern computational resources make the full nonlinear location problem easily tractable. The results suggest a nonlin-ear approach should be used for many problems where there are limited observations at local distances, or where one is using azimuthal data.