Large numbers of sea otters (Enhydra lutris) dies following the 1989 E
xxon Valdez oil spill. Two previous studies estimated spill-related mo
rtality, one from the difference between counts of otters before and a
fter the spill, and the other from the recovery rate of tagged carcass
es that had been released at sea. I used a derivative of the second ap
proach, but revised it to account for (1) moribund offers that had hau
led out on shore; (2) carcasses that were collect at sea; and (3) diff
erences in search effort in different areas of the spill-consideration
s that were not addressed previously. The mortality estimation procedu
re presented here, and applied to Prince William Sound, Alaska, had si
x input parameters, two with fixed values (the number of spill-related
carcasses collected [391] and the number of otters taken alive that d
ied in captivity [84]) and four that were assigned a range of values (
the recovery rate for carcasses on the beach [60-90%], the proportion
of recovered carcasses that were onshore [70-95%]), the proportion of
offshore deaths recovered onshore (20-50%), and the proportion of mori
bund otters that hauled out (20-60%). Empirical estimates for these pa
rameters were not available after the spill, but a range of plausible
values for each was delimited using other available data. Randomly-sel
ected combinations of values within these ranges produce mortality est
imates for the sound ((x) over bar = 750, 5-95% quantiles approximate
to 600-1000) that were lower than indicated by previous studies. These
lower estimates do not diminish the tragic nature of the mortality, b
ut instead highlight the significance of both the process and informat
ional basis in producing estimates of catastrophic loss. Estimates of
loss based on faulty or inadequate data may mislead investigations of
population recovery and may jeopardize public trust in scientific asse
ssments of future catastrophes.