Long-lining operations from a single vessel fishing for Dissostichus eleginoides off South Georgia (Sub-area 48.3) were assessed over 48 days during April and May. All lines were set at night using the Spanish double-line method with extra weights on the line and all deck-lights extinguished to mitigate against bird mortality. Data were recorded during hauling using a randomised cluster sampling method; the known number of coils laid during a night's setting operations provided the sampling frame, and the sampling units consisted of lengths of line between marked connecting lines. Previous catch rates were used to set daily sampling effort, and accurately achieved a CCAMLR sampling target of 60 fish/day. The method was further developed to integrate different, and varying, levels of effort at two observation stations into one daily sampling schedule. Data for total length, maturity and sex were taken. Mean total length of females caught was higher than males, and length at first maturity was also larger. The proportion of mature fish in the male population sampled was 52.0%, but was only 24% in the female population. The mean daily loss rate of toothfish observed at the hauling point was only 0.8% of the mean daily catch rate; total by-catch by numbers was less than 5%. Consistent CPUE by numbers between catches in 1994 and 1997 despite reduction in CPUE by weight indicated that exclusion effects may be important in determining the catch taken by longlines. Continuing decreases in size of fish caught relative to the length at first maturity of females indicate that reproductive over-exploitation may become a threat.