Long-lining operations for Dissostichus eleginoides off South Georgia were assessed for incidental mortality and interactions with seabirds. Twenty-seven lines were observed and ninety-eight mortalities recorded over twenty sets of the line; no mortalities occurred during hauls. The 16 sets made at night contributed 15% of the overall mortality, all of white-chinned petrels: the four day time sets contributed 85% of overall mortality, with giant petrels, grey-headed albatrosses and black-browed albatrosses predominating. Our data suggest that grey-headed albatrosses, whose populations at South Georgia are in serious decline, were disproportionately affected in relation to their numbers in the vicinity of the fishing vessel; giant petrels also may be disproportionately affected, at least in relation to the size of their breeding population at South Georgia when compared to the albatrosses. Average mortality rate for the twenty sets was 0.48 birds/1000 hooks and maximum mortality 3.12 birds/1000 hooks. Mortality and interactions of birds with operations varied with site and time of day, and due to behavioural interactions between birds. Setting only at night would dramatically reduce albatross deaths, but would substantially increase white-chinned petrel mortalities. A streamer line made to CCAMLR specifications may also reduce mortalities but may be less effective during calm weather, intense feeding activity by seabirds, or when incorrectly constructed.