Butterworth et al. (1992) put forward an extension of the harvesting model of Beddington and Cooke (1983) to relate potential krill yield to a pre-exploitation survey estimate of krill biomass. In this paper, the approach is extended further so as to incorporate most of the amendments specified by the Third and Fourth Meetings of the Working Group on Krill (WG-Krill). The most important of these extensions is integration over the ranges of uncertainty for a number of model parameters. Results are provided for the probability of spawning biomass falling below various fractions of its median pre-exploitation level (K^sp), as a function of the fraction of the biomass estimate which is set as the catch for a 20-year period. Three possible fishing seasons are considered. The model extensions requested by the Third Meeting make little difference to the results of Butterworth et al. (1992). Winter fishing is marginally preferable to a summer harvest. However, the imposition of an upper bound of 1.5 yr^-1 on the effective annual fishing mortality, as specified by the Fourth Meeting, results in marked reductions in the probabilities of krill spawning biomass falling below specified fractions of (K^sp).