Acoustic surveys to estimate krill biomass require that sound backscattered by krill be identified and distinguished from all other types of backscatter. Traditionally this has been achieved by sampling characteristic types of acoustic target with nets. More recently partitioning of backscatter into species groups has been attempted using differences in echo strength at two acoustic frequencies (ΔMVBS). Here we compare net and acoustic data from 2 cruises around South Georgia in 1996 in order to assess the functionality of acoustic target identification techniques. Mean volume backscattering strength (MVBS) data at 120 and 38 kHz were collected with a Simrad EK500 echo sounder, and net samples were collected with an RMT8. Around 80 % of the targets thought to be krill on the basis of their appearance on echo-charts were also identified as krill on the basis of their difference in backscatter at 38 and 120 kHz (ΔMVBS = MVBSI20 - MVBS3S; where ΔMVBS between 2 & 12 dB). Biomass values estimated from krill identified using echo-chart appearance or dB difference were broadly similar (regression: dB classification = 0.94 visual classification, r2 = 0.99). Krill size was predicted from scattering models using the 2 frequency data and compared with that obtained in net hauls. This comparison revealed that a simplified bent cylinder model was a better predictor of krill length (predicted length = 8.79 + 0.685 observed length, r2 = 0.77) than a fluid-filled sphere model.