The abundance of returning adult Atlantic salmon Salmo salar, in the River Orkla in mid-norway (1 sea-winter, SW, fish) and River Hals in north Norway (1–3 SW fish), was tested against the early marine feeding and the seawater temperature experienced by their corresponding year classes of post-smolts immediately after entry into the Trondheimsfjord (Orkla smolts, 22 years of data) and Altafjord (Hals smolts, 17 years of data). In both river–fjord systems, there was a significant positive correlation between the abundance of returning S. salar and the mean seawater temperature at the time of smolts descending to the sea. The number of 1SW fish reported caught in River Orkla was positively correlated to the proportion of fish larvae in the post-smolt stomachs in Trondheimsfjord. The abundance of returning S.salar was, however, neither correlated to forage ratio (RF) nor other prey groups in post-smolt stomachs in the two fjord systems. In the Altafjord, the post-smolts fed mainly on pelagic fish larva (70–98%) and had a stable RF (0·009–0·023) over the 6 years analysed. In the Trondheimsfjord, however, there was a higher variation in RF (0·003–0·036), and pelagic fish larvae were dominant prey in only two (50 and 91%) of the 8 years analysed. These 2 years also showed the highest return rates of S. salar in River Orkla. These results demonstrate that the thermal conditions experienced by post-smolts during their early sea migration may be crucial for the subsequent return rate of adults after 1–3 years at sea. Pelagic marine fish larvae seem to be the preferred initial prey for S. salar post-smolts. As the annual variation in abundance of fish larvae is related to seawater temperature, it is proposed that seawater temperature at sea entry and the subsequent abundance of returning adult S. salar may be indirectly linked through variation in annual availability of pelagic fish larvae or other suitable food items in the early post-smolt phase.