Long-lining operations for Dissostichus eleginoides off South Georgia were assessed for interactions with seabirds, and the effectiveness of measures employed to mitigate seabird mortality. In this study 61 lines were laid during April and May, all set at night using the Spanish double-line method with extra weights on the line and all deck-lights extinguished. Data were recorded during hauling using a randomised cluster sampling method, developed to allow representative data to be collected when 100% observer coverage could not be achieved. Thirteen seabird mortalities were recorded, all during setting operations and all but one during April, giving an average mortality rate of 0.018 birds/1000 hooks, and 0.099 birds/1000 observed hooks. Mortalities consisted of nine white-chinned petrels, two black-browed albatrosses and one unidentifiable bird caught on hooks; and one giant petrel killed flying into the side of the vessel. Few birds were generally seen following the vessel during setting operations. However, during April large numbers of white-chinned petrels were seen occasionally, and observed to dive; large numbers of black-browed albatrosses were seen when the moon was full, but were observed to dive on only one night. Live birds were observed to become caught on hooks during hauling on 23 occasions; no mortalities resulted, birds being released apparently unharmed. Black-browed albatrosses concentrated on taking returning bait off the line and accounted for eighteen birds caught; giant petrels concentrated on taking discarded offal, and accounted for five birds caught. An experiment to assess the effectiveness of streamer-lines in mitigating bird mortality found no significant difference due to a streamer-line when longlines were laid at night with no deck lights, and the line was properly weighted.