Harlequin Ducks (Histrionicus histrionicus) were captured on breeding streams in Alberta, Canada from 1995–2018. Measurements for exposed culmen, tarsus bone, and wing length, and mass, were significantly larger for adult males than those for adult females, while only tarsus was significantly larger for male ducklings than for females. The repeatability of tarsus and culmen measurements across time for the same individuals was significantly different from 0, indicating high consistency. We used logistic regression of culmen and tarsus length of adult ducks to identify the best model to predict the sex of adults, then constructed a classification and regression tree to predict the sex of a test set of juveniles that had been banded and sexed as ducklings but that were sexed by plumage when resighted as adults. The model with the highest accuracy (96.9%) and lowest number of individuals in the zone of uncertainty (thus sexed as unknown) was tarsus < 37.5 mm (classify as females) or ≥ 38.0 mm (classify as males). Individuals in the zone of uncertainty with a penis sheath should be sexed as male, but individuals without a penis sheath that do not match the appropriate tarsus length should be sexed as unknown.