The Namibian sardine stock has experienced intense exploitation since 1950, with catches peaking at 1.4 million tonnes in 1968, followed by a collapse in the 1970s and 1980s. Namibia therefore inherited a severely depleted stock at Independence in 1990. The paper describes data available since Independence, which includes research survey data, and length frequencies obtained from port sampling. It also presents age-structured stock assessment models, which consider some of the main sources of uncertainty relating to values for natural mortality, and assumptions about survey bias, density-dependent effects in survey bias and variance, and the form of the stock-recruit curve. These models could be used as "operating models" to underpin the future development of management procedures for Namibian sardine. Despite reasonable fits, 95% confidence intervals for natural mortality are fairly wide and estimates of precision poor for most models considered. However, the model that assumes density-dependence in survey variance shows superior fits and improved estimates compared to the others, despite reservations about the existence of such effects. This indicates the need to explore the possibility of density-dependence in survey variance further if future management procedure performance is negatively affected by it. Keywords: Namibian sardine, Survey bias, Additional variance, Density-dependence, Operating model.