Field measurements of wave overtopping over a rubble mound breakwater have been carried out at Zeebrugge, Belgium, during the period 1999–2003. The overtopping volumes of water are collected in an overtopping tank located on the breakwater crest behind the access road. A short side of the tank is made of a weir, allowing the collected water to flow out continuously, and allowing measurements to be made for long duration. Calculation of the inflow discharge is based on the outflow discharge over the weir and the volume in the tank. This calculation requires an accurate measurement of the water depth inside the tank. The water depth is measured indirectly using a pressure sensor near the bottom. The measurement setup for wave overtopping at the breakwater is described in detail. Eight-wave records with wave propagation normal to the breakwater have been used in the analysis of the field data. Wave heights range between 2.6 and 3.9 m (i.e., up to 63% of the design wave height) for wave records with a typical duration of 1–2 h. The measured average overtopping rate q has been determined for all records using three methods ( q<1 l/s m). Average overtopping rates from the field data have been compared with predicted values from prediction formulae by van der Meer et al. [van der Meer, J.W., Tönjes, P., de Waal, H., 1998. A code for dike height design and examination. In: Proceedings International Conference on Coastlines, Structures and Breakwaters. Institution of Civil Engineers, London, Thomas Telford, London, pp. 5–19], Owen [Owen, M.W., 1980. Design of seawalls allowing for wave overtopping. Hydraulics Research, Wallingford, Report No. EX 924, UK], and Besley [Besley, P., 1999. Wave overtopping of Seawalls. Design and Assessment Manual. Hydraulics Research Wallingford. R&D Technical Report W178, ISBN 1 85705 069 X]. Three crest freeboard values have been used in the analysis. The calculated reduction factor for surface roughness used in the formulae is 0.51. Best agreement is observed for van der Meer et al.'s and Owen modified by Besley's prediction formulae using the actual average crest level. In this case, measured average overtopping rates are within the 95% confidence intervals of these prediction formulae.