Ova fecundities of Scottish Atlantic salmon Salmo salar, predicted from log(10) regression of ova numbers and female fork length (L(F)), differed widely between upland and lowland stocks within the same river, whereas sea-age, river and year factors had insignificant effects on fecundity once L(F) was accounted for. For upland fish, the relationship between log(10)L(F) and log(10) ova mass (M(O)) was stable between two datasets collected 40 years apart. Although upland and lowland females both produced comparable log(10)M(O) (log(10)L(F))(-1), lowland females partitioned this into 45% more, but smaller ova, whereas upland females produced fewer, but larger, eggs. The possible causes and implications of this are discussed for evolutionary perspectives (lifetime production), population structure (local tributary v. large catchments; environmental effects), population dynamics and stability (density-dependent control mechanisms) and fisheries management (stock-recruitment; short and long-term stock sustainability).