Understanding potential sources of bias and error in the biometric sexing of birds

Abstract

ABSTRACT Biometrics, particularly wing length, are commonly used to assign the sex of many birds in species where the plumage is similar for males and females. Virtually all species show some measurement overlap between the sexes, however, and measurement error can add further uncertainty into data sets. This can result in individuals being misclassified as the wrong sex and introducing bias into subsequent analyses, particularly if the life histories of misclassified birds differ from those of others. We used the Marsh Tit Poecile palustris as a case study to examine potential sources of error and bias when assigning sex based on wing length. There was no evidence for a heritability of wing lengths that could result in localised populations of atypical size via a ‘founder effect’, which would otherwise undermine biometric sexing. Additionally, potentially misclassified birds did not differ from others in how long they persisted in the local population, so avoiding any potential bias of misclassification in demographic analyses. Compared to Marsh Tit data collected during intensive studies, the national data set pooled from contributors across Britain showed much greater variation in wing lengths, resulting from wider variation in the accuracy of measurement and recording. This variation in pooled data can have implications for analyses, and we discuss the importance of data quality in ringing schemes.