This paper reviews the data used to estimate the abundance of juvenile Dissostichus eleginoides in Division 58.5.2 and examines the sensitivity of the assessment of yield to these results. Primarily, it is proposed to revise the use of historical survey data by reassigning them to the strata established in 2003. In addition, the sensitivity of the estimates of abundance to the latest length-at-age model is examined. The best fits using CMIX to estimate cohort densities were obtained using the updated assignment of survey data to the 2003 stratification along with the latest estimates of length at age. The quality of model fit and estimates of abundance of specific cohorts were sensitive to the number of cohorts assumed to be represented in the length density data. Assuming survey design produces unbiased estimates of length density, we suggest that best estimates of cohort abundance will be achieved when mixture analysis is informed by accurate estimates of length at age, an independent estimate of relative abundance of cohorts present in any survey year such as could be derived through an age length key, and vulnerability of the stock to the surveys. The consequences to the 2005 assessment of yield of toothfish of changing the time series of abundance of juvenile fish and the length-at-age model were examined. Yield estimates were found to be most sensitive to the estimated recruitment series interacting with the estimate of mortality and the cohort abundances form CMIX, and to a lesser extent the estimated fishing vulnerabilities. A scenario equivalent to that used to recommend yield in 58.5.2 for 2005/06, incorporating the latest estimates of length at age, selectivity and the mixtures estimated from the assignment of RSTS hauls, resulted in an increase in estimated yield.