At its last meeting, WG-FSA noted a declining trend in the mean lengths of toothfish caught around South Georgia and Shag Rocks over the period 1995 – 1999. We re-analysed the mean lengths of fish taken on individual longline sets from which length frequencies were sampled by observers and found the same declines in mean lengths for the period 1997 – 1999 as previously noted by WG-FSA, but also that this trend was halted in 2000. A GLM analysis of these data shows that both area and depth are highly significant factors in explaining part of this variation over time, but not all of it. Taking depth and area into account there is less of an apparent decline in mean length over the period 1997-1999. The overall picture that emerges is one of variable mean lengths, both within and between seasons but with no consistent overall trend. The analysis also suggested that the exploitable population around South Georgia and Shag Rocks is quite spatially and temporally heterogeneous in terms of its length distribution. Separately, it is known that the effort distribution is also highly heterogeneous. The combined effects of these findings were examined in a set of preliminary analyses of length densities by area and depth zone. The results indicate that the annual distribution of effort by area and depth zone does have an important effect on the overall length-specific selectivity and that it is also likely that historical changes over years in the depth distribution of effort have led to different length-specific selectivity curves applying in different years. Combined length-specific selectivity curves were estimated for each of the years 1997-2000. Notable differences were found amongst these, but there was a consistent tendency for fish over 90 -100 cm to have a lower relative selectivity than smaller fish.