The timing of downstream migration and detection rates of hatchery-reared Atlantic salmon Salmo salar smolts and stream-reared smolts (stocked 2 years earlier as fry) were examined in the Connecticut River (U.S.A.) using passive integrated transponder (PIT) tags implanted into fish and then detected at a downstream fish bypass collection facility at Turners Falls, MA (river length 192 km). In two successive years, hatchery-reared smolts were released in mid-April and early May at two sites: the West River (river length 241 km) or the Passumpsic (river length 450 km). Hatchery-reared smolts released higher in the catchment arrived 7 to 14 days later and had significantly lower detection rates than smolts stocked lower in the catchment. Hatchery-reared smolts released 3 weeks apart at the same location were detected downstream at similar times, indicating that early-release smolts had a lower average speed after release and longer residence time. The size and gill Na⁺/K⁺-ATPase (NKA) activity of smolts at the time of release were significantly greater for detected fish (those that survived and migrated) than for those that were not detected. Stream-reared pre-smolts (>11·5 cm) from four tributaries (length 261–551 km) were tagged in autumn and detected during smolt migration the following spring. Stream-reared smolts higher in the catchment arrived later and had significantly lower detection rates. The results indicate that both hatchery and stream-reared smolts from the upper catchment will arrive at the mouth of the river later and experience higher overall mortality than fish from lower reaches, and that both size and gill NKA activity are related to survival during downstream migration.