We show how the duration of primary replacement can be calculated from a knowledge of primary loss intervals, primary lengths, and primary growth rates. With this new method, we can now independently evaluate the reliability of regression or maximum likelihood estimators of molt duration in situations where recaptures are not available. Before, molting birds had to be recaptured and scored for molt at least twice in the same episode of molt to check the reliability of duration estimates based on samples of birds scored for molt a single time. This new method assumes only that feathers are lost in sequence. Unlike regression and likelihood estimators of molt duration, it requires neither that the same number of feathers are always replaced nor that molt is confined to a single season and reasonably synchronized across the population. Thus, it can be used to (1) estimate the duration of primary molts in species that molt and breed throughout the year, (2) estimate the time required to replace variable numbers of primaries, and (3) calculate start dates for molts that are initiated at various points in the primaries. We also show how to set confidence limits for these calculated estimates of molt duration. Finally, a variation of the method can be used to calculate where to sample feather material for measuring stable isotopes at evenly spaced time intervals through the primary replacement.