Sexing terns using biometrics: the advantage of within-pair comparisons

Capsule Within-pair comparisons substantially improve the accuracy of sexing from biometrics for two congeneric species of seabird with monomorphic plumage and soft-tissue colouration. Aims To examine the extent to which statistical limitations of sexing birds from biometrics using sample-level analysis could be overcome by sexing Common and Arctic Terns (Sterna hirundo and S. paradisaea) using measurements obtained from breeding pairs. Methods Incubating adults were caught at the nest using walk-in traps and wing, tarsus, head-plus-bill, tail length, tail fork, and body mass measured. Each bird was individually colour-ringed and dyed with picric acid, enabling subsequent sexing by behavioural observations of copulation and courtship feeding. Birds were sexed using biometrics and the proportion of birds sexed correctly this way at the sample level was compared with the accuracy achieved if, within a pair, the larger bird was classified as male. Results Head-plus-bill length was the single most accurate measurement for sexing individuals of both species, and correctly classified 72% of Arctic Terns and 73% of Common Terns. Combinations of measurements derived from discriminant analysis achieved slightly higher accuracy (73% and 78% respectively). Within-pair comparisons were more accurate than sample-level analysis for both single measurements and discriminant functions, and were able to sex 84% of Arctic Terns and 86% of Common Terns correctly. Conclusion Comparing birds within pairs improves accuracy and can eliminate the need to calculate cut points or discriminant functions from a sample of birds of known sex for each particular study. We strongly advocate such comparisons wherever possible to increase accuracy and simplify computational procedures for predicting sex, thus reducing associated sources of error.