Although as a species sheep are commonly found in archaeological deposits and occupy a major position in the history of animal husbandry, it remains difficult to discriminate the sexual type (male, female or castrate) of individuals from observation of the remains. However, understanding sheep husbandry through sex ratio analysis, contributes to a better understanding of the history of animal husbandry. Current criteria for determining are limited to visual characters on coxal and skull. This study explores the variability of sheep skeletons according to their sex to submit hypotheses for methods capable of individual sexing. Using 58 complete skeletons and 428 isolated Soay sheep bones from the collection of the Natural History Museum (UK), we initially tested the sexual distinction using the so-called classical methods: the estimates of shoulder heights and the slenderness index. We then evaluated the influence of sex on the shape and size of selected bones with geometric morphometric, using a protocol that we had previously developed and validated. Finally, we classified these data using the K-Nearest Neighbors (KNN) method, quantifying the reliability of sex attribution. We were able to test its application on an archaeological sample, after selecting bones with the best rate of sexual assignment. Our results demonstrate the difficulty of using traditional methods to sex sheep whilst proposing solutions for sex assignment (88% reliable) using geometric morphometric and KNN. In addition, we were able to highlight the influence of sex according to bone type, though this is strongly qualified by various criteria (such as the age of castration). By characterising and quantifying the sex ratio within archaeological assemblages, we may better understand herd management, procurement strategies and feeding habits, and therefore improve our global knowledge of animal husbandry history.