Aims: To quantify and derive statistical relationships with which to predict the delivery of faecal bacteria ( Escherichia coli) to a pastoral stream, by overland flow. Methods and Results: A large- scale ( 1050 m 2) rainfall simulator, located upon a steep ( 18 degrees) grazed hillside in New Zealand, was used to simulate 11 heavy rainfall events. Overland flow was generated and sampled throughout each event, before discharging to a headwater stream. The samples were subsequently analysed to determine the concentration of E. coli. Statistical analysis showed that the time elapsed since the last period of grazing was a statistically significant predictor of both the total number ( load) and concentrations of E. coli in overland flow. Between 10(5) and 10(8) E. coli per m 2 of hillside were delivered to the stream within overland flow during each event, and peak concentrations ranged between 10(3) and 10(7) most probable number per 100 ml. Conclusions: Under heavy rainfall on steep pastoral land, overland flow can transport substantial levels of faecal bacteria to streams. Under such conditions, it is unlikely that vegetated buffer strips will be particularly effective at attenuating bacteria within overland flow. Significance and Impact of the Study: This work has improved understanding of the importance of overland flow as a process contributing to the contamination of pastoral streams by faecal bacteria. In addition, the predictive relationships derived can be incorporated within catchment models.els.