Aquatic hazing devices recently have been developed as a possible means of deterring waterbirds from oil spills, thereby reducing casualties. However, the effectiveness of these devices has not been examined with rigorous statistical tests. We conducted a study in the San Francisco Bay estuary to develop a design for testing the effectiveness of an aquatic hazing device on waterbirds in open water. Transects marked with poles at 100-m inter- vals up to 800 m from the hazing device were established at two sites separated by three km in the north bay. Alter- nating two-day test and control periods were conducted at each site. Observers in over-water blinds counted the number, species and behavior (swimming, diving, or preening) of birds on transects each day. Aerial surveys of birds within four km of the device were conducted at the beginning of each test. For both aerial and ground surveys, a three-way mixed model analysis of variance test was used to examine trial, distance from the device, and treatment (device on or off) fixed effects, and site as a random effect on numbers of Greater and Lesser scaup (Aythya affinis and A. marila), Surf Scoter (Melanitta perspicillata), and all other waterbirds. We could not detect a significant deter- rent effect of the hazing device in either aerial surveys of all ducks or scaup (all ducks, F,283 = 1.1; Scaup, F28,230 = 0.9, all n.s.; 3-factor ANOVA), or ground surveys for all ducks or scaup (all ducks, F28,23 = 1.0; scaup, F2s,230 = 0.9, all n.s.; 3-factor ANOVA). There was a significant trial-by-treatment interaction for Surf Scoters (F4,9 = 5.4, P = 0.02; 3-factor ANOVA), but Surf Scoter numbers fluctuated greatly among trials so the effect of the device on this species was not clear. Birds did not alter their behavior when the device was active. In general, although aquatic hazing devices have potential to reduce waterbird mortality in oil spills, the tested device was not effective as a deterrent for waterfowl in experimental trials on the estuary. Received 27 September 1999, accepted 3January 2000.