This paper uses actual krill length-frequency data in an attempt to address the problem of the determination of adequate sample size to obtain representative krill length-frequency distributions from commercial catches. The possible effect of two other factors (within-trawl variability and sample decomposition) which may influence length-frequency data quality is also considered. Attention is drawn to the following:

• all measurements of length from commercial krill; catches should be made, as far as possible, by a single observer/vessel;

• consideration still has to be given to the minimum length differences which should be detectable between catches. Account must be also taken of the desired biological characteristics which are to be discerned (e.g., length-with-age). At present, and in the interests of both statistical rigour and efficiency of measurement, it is proposed that length measurements made on commercial krill catches should be grouped into 2 mm size classes;

• biological implications associated with the detection of specific differences in length should be considered in conjunction with the need to collect information on maturity stages;

• for most purposes a minimum sample size of at least 100 animals/trawl is necessary to obtain statistically meaningful differences between samples; and

• the problem of how frequently commercial krill catches should be sampled still requires consideration.