By checking all records as well as the banders, 116 discrepancies in each 
1000 records were eliminated; an equal number of minor nature were found 
to be connected with dates of recovery. As used in this study, each 1000 
banding records are thought to contain .2 questionable or spurious reports; 
undetectable errors in a mallard file were estimated at 0.5 per thousand. 
Variations in the manufacture of bands and in usage by the 
birds make for band loss or illegibility that will potentially bias 
survival data. This bias is probably greatest in data for large birds 
in a marine environment. 
26 
