We documented the progression and timing of the annual molt of harbor seals on Tugidak Island, Alaska, from 1997 to 1999. In all years the timing of molting differed among age-sex classes. Yearlings molted first, subadults second, adult females third, and lastly adult males. Timing of molting was nearly identical in 1997–1998, whereas in 1999 molting occurred three to six days later for all age-sex classes except yearlings. Estimated dates when peak proportions of each age-sex class were molting ranged from 2 August (yearlings) to 2 September (adult males). The number of seals hauled out was positively related to the proportion of seals in the molt and negatively related to the proportion of seals in the postmolt. Population trend estimates, based on aerial counts conducted during a narrow window within the molting period, are likely biased toward certain age-sex classes. Statistical models used to estimate trend include covariates to help account for within-year variation in seal numbers, but do not account for compositional changes that occur during molting. Population modeling may elucidate the effects of within-year population structure on trend estimates. Monitoring molting phenology at additional sites is necessary to determine the extent of geographic variation in molting.