In 2001, WG-FSA asked the Intersessional Subgroup on Sampling Catches from Longlines to develop recommendations on 1) subsampling methods using frames and sampling units based on time and gear; 2) the allocation of observer effort within longline haul and between hauls; and 3) the allocation of observer effort directed toward fishery target species versus ecological interactions. Both subsampling methods use the hauling completed within a day as a frame, and theoretically cover all fishing activity by sampling every day of hauling. Both methods essentially follow a multi-stage cluster sampling design, which could be implemented more rigorously if the CCAMLR objective was changed from sampling 60 fish/day, to sampling a set length of each line or number of hours of each day. The principal advantage of the time-based method is that it gives a frame that can be monitored easily using a watch, but the frame is based on the mean time that is taken for hauling which is itself a random variable. The gear-based approach uses a fixed frame, but monitoring the amount of line hauled can be difficult when sampling the target species. Sensitivity analyses used to examine present guidelines on sample sizes, indicated that increasing the number of fish sampled for length would not substantially increase the precision of the estimates for mean length. From the limited evidence available, seabird mortality rates are estimated sufficiently precisely at 25% coverage to be reasonably sure of detecting increases to unacceptable levels. Sampling toothfish for biometric data in the factory and observing the incoming line may usefully be allocated to different days, if done so randomly.