Data on the field-to-stream transport of sediment and chemicals from an agricultural watershed were collected in a three-year study to provide information for testing and evaluating mathematical models under development for predicting agricultural non-point source pollution. These models are prepared as tools to evaluate the effectiveness of different farm management practices in controlling losses of nutrients, pesticides, and sediment in field drainage to receiving waters. In the study, data were collected for small corn, soybean, and pasture fields; for two larger mixed-cover watersheds; and at three drainage-stream sites. During the study (1976-1978), annual rainfall (753 mm) and streamflow (124 mm) averaged a little below normal. Sediment losses were also low because of a lack of intense rains, averaging 2.6 t/ha from row-crops and 0.9 t/ha from the watershed as a whole. Soluble chemical losses (NH4-N, NO3-IM, PO<-P, Cl and TDS) in surface runoff were less than that deposited with rainfall. Because IMO3-N, Cl and TDS were concentrated in subsurface drainage, losses from the watershed as a whole due to streamflow (surface plus subsurface drainage) were significantly larger than losses from surface runoff alone. Average annual pesticide losses from the field were least for the shortest-lived herbicide (0.2% of that applied) and greatest for the most persistent (1.6%). No runoff event occurred within a week of application. On the basis of percentage applied, losses or export from the watershed as a whole were about 25% of the losses from the two individual fields studied. With the exception of the strongly adsorbed paraquat, at least 80% of the losses occurred in the water phase as opposed to that adsorbed on sediment.