Prediction of individual wave overtopping events is important in assessing danger to life and property, but data are sparse and hydrodynamic understanding is lacking. Laboratory-scale waves of three distinct types were generated at the Coastal Research Facility to model extreme waves overtopping a trapezoidal embankment. These comprised wave groups of compact form, wave groups embedded in a background wave field, and a solitary wave. The inshore wave propagation was measured and the time variation of overtopping rate estimated. The total volume overtopped was measured directly. The experiments provide well-defined data without uncertainty due to the effect of reflection on the incident wave train. The dependence of overtopping on a range of wave shapes is thus determined and the influence of wave-wave interactions on overtopping assessed. It was found that extreme overtopping may arise from focused waves with deep troughs rather than large crests. Furthermore, overtopping waves can be generated from small wave packets without affecting the applicability of results to cases in which there are surrounding waves. Finally, overtopping from a solitary wave is comparable with overtopping from focused wave groups of the same amplitude.