To obtain a representative sample of the catch taken by toothfish longliners, CCAMLR observers are recommended to sample a quota of the first 30 fish that are caught during a sampling session. Since toothfish are often aggregated on a line, this means that fish within aggregations are sampled with less catch effort than those outside; or more generally, sampling effort is applied in inverse proportion to toothfish density on the line. The present recommendation therefore indicates sampling with unequal probabilities which, if unaccounted for, can lead to serious biases in estimates of population vital rates. To resolve this, either the probabilities need to be incorporated in the analysis of sampling data, or sampling effort should be proportional to catch effort. As vessels generally record the number of baskets taken while hauling a line, observers can use this to establish a point at which to start a session. Because there is little evidence of a consistent trend in fish sampled over a line, it does not particularly matter if an observer starts sampling before or after the correct basket arrives; but, preferably, the observer should start as close as feasible, and not be consistently late or early.