We recovered coded wire tags (CWTs) from a colony of Caspian terns Hydroprogne caspia on Brooks Island in San Francisco Bay, California, to evaluate predation on juvenile salmonids originating from the Sacramento and San Joaquin rivers. Subsamples of colony substrate representing 11.7% of the nesting habitat used by the terns yielded 2,079 salmonid CWTs from fish released and subsequently consumed by terns in 2008. The estimated number of CWTs deposited on the entire tern colony was 40,143 (ranging from 26,763 to 80,288), once adjustments were made to account for tag loss and the total amount of nesting habitat used by terns. Tags ingested by terns and then egested on the colony were undamaged, and the tags' complete numeric codes were still identifiable. The CWTs found on the tern colony indicated that hatchery Chinook salmon Oncorhynchus tshawytscha trucked to and released in San Pablo Bay were significantly more likely to be consumed by Caspian terns than Chinook salmon that migrated in-river to the bay; 99.7% of all tags recovered were from bay-released Chinook salmon. Of the CWTs recovered on the tern colony, 98.0% were from fall-run Chinook salmon, indicating a higher susceptibility to tern predation than for the spring run type. None of the approximately 518,000 wild Chinook salmon that were coded-wire-tagged and released in the basin were recovered on the tern colony, suggesting that the impacts on wild, U.S. Endangered Species Act-listed Chinook salmon populations were minimal in 2008. Overall, we estimate that 0.3% of the approximately 12.3 million coded-wire-tagged Chinook salmon released in the basin in 2008 were subsequently consumed by Caspian terns from the Brooks Island colony. These results indicate that CWTs implanted in juvenile salmon can be recovered from a piscivorous waterbird colony and used to evaluate smolt losses for runs that are tagged. Abstract We recovered coded wire tags (CWTs) from a colony of Caspian terns Hydroprogne caspia on Brooks Island in San Francisco Bay, California, to evaluate predation on juvenile salmonids originating from the Sacramento and San Joaquin rivers. Subsamples of colony substrate representing 11.7% of the nesting habitat used by the terns yielded 2,079 salmonid CWTs from fish released and subsequently consumed by terns in 2008. The estimated number of CWTs deposited on the entire tern colony was 40,143 (ranging from 26,763 to 80,288), once adjustments were made to account for tag loss and the total amount of nesting habitat used by terns. Tags ingested by terns and then egested on the colony were undamaged, and the tags' complete numeric codes were still identifiable. The CWTs found on the tern colony indicated that hatchery Chinook salmon Oncorhynchus tshawytscha trucked to and released in San Pablo Bay were significantly more likely to be consumed by Caspian terns than Chinook salmon that migrated in-river to the bay; 99.7% of all tags recovered were from bay-released Chinook salmon. Of the CWTs recovered on the tern colony, 98.0% were from fall-run Chinook salmon, indicating a higher susceptibility to tern predation than for the spring run type. None of the approximately 518,000 wild Chinook salmon that were coded-wire-tagged and released in the basin were recovered on the tern colony, suggesting that the impacts on wild, U.S. Endangered Species Act-listed Chinook salmon populations were minimal in 2008. Overall, we estimate that 0.3% of the approximately 12.3 million coded-wire-tagged Chinook salmon released in the basin in 2008 were subsequently consumed by Caspian terns from the Brooks Island colony. These results indicate that CWTs implanted in juvenile salmon can be recovered from a piscivorous waterbird colony and used to evaluate smolt losses for runs that are tagged ?? American Fisheries Society 2011.