Mark–recapture techniques were used to estimate persistence rates and detection probabilities of bird carcasses associated with the oil spill following the wreck of the M/V Selendang Ayu at Unalaska Island, Alaska. Only 14.6% of carcasses placed on beaches remained after 24 hours, and all carcasses that remained had been scavenged to some degree. Daily persistence rates for scavenged carcasses on subsequent days were substantially higher at 79.1%. Most carcasses (>98%) were removed by scavengers at night. When they made a single pass, observers searching beaches for carcasses that had washed ashore found only about 40% of carcasses known to be present. This detection probability did not vary between pairs of search teams or between beaches. Detection probability increased to about 70% when teams searched the same beach segment twice. Our data indicate that only a small fraction of beached carcasses would likely be found using standard beach survey protocols and search frequencies. These data emphasize the importance of measuring persistence and detection rates for each mortality event.