Studied were four primary sources of uncertainty in krill density estimates from acoustic surveys. The variance in system calibration with a standard sphere was evaluated in relation to sphere material and diameter, water temperature, and pulse length (Demer and Hewitt, 1992). Calibration bias was investigated by comparing the theoretical and actual target strength (TS) values of four different standard spheres with the results of a calibration by self-reciprocity (Demer and Hewitt, submitted). Combining the results of these experiments, the accuracy and precision of a system calibration with a standard sphere were estimated as -1.2 dB and ±0.3 dB, respectively. Uncertainty in estimating krill TS was then investigated through in situ measurements (Hewitt and Demer, 1991). The TS data provided corroboration to an empirical model developed from a linear regression. However, TS values have been observed to vary as much as 8 dB, depending on the time of day (unpublished data). Also, Monte Carlo simulations have demonstrated the potential errors in developing empirical models from linear regressions of zooplankton scattering data (Demer and Martin, 1995). Therefore, the accuracy and precision krill TS estimates were conservatively estimated to be 0 dB and ±4.0 dB, respectively. The uncertainty in species delineation was also investigated, through the development of a statistical technique for remote species identification (Demer et al., submitted). The technique was used to apportion the integrated echo energy between two predominant scatterers, Euphausia superba and Salpa thompsoni. These studies indicated that scattering from S. thompsoni contributed to a positive bias in the krill density estimates of 0.6 dB. Finally, uncertainty due to the diel vertical migration (DVM) of Antarctic krill above a down-looking transducer was quantified through time-depth-density analyses (Demer and Hewitt, 1995). A method was developed for compensating acoustic biomass estimates for the effects of DVM. Applying the compensation function to survey data, the resulting biomass estimates were an average of 1.8 dB higher than those calculated disregarding biases due to DVM, with a standard deviation of 0.6 dB. Combining these four sources of uncertainty, the overall bias in krill density estimates was estimated to be 2.4 dB. From a boot-strap simulation the total variance was estimated as ±0.9 dB. Compensating for these components of uncertainty resulted an increase in krill density estimates of74%, with confidence limits of ±55%.