Rao et al.'s mandibular canine index (MCI) is a simple odontometric method which uses the mandibular canine as the key to sex estimation. This index is defined as the ratio between the right canine mesiodistal dimension and the mandibular canine arch width. The aim of this study was to contribute to sex estimation using dental techniques by analysing the MCI efficiency, and to propose a new approach for its use. Measurements were taken from 120 plaster casts (70 females) in the 16–30 year age group. Although statistically significant sexual dimorphism was observed in both the mesiodistal dimension and the mandibular canine arch width, the MCI showed a low accuracy in sex classification (54.2% correct identifications). This accuracy was improved to 64.2% using receiver operating characteristics curve analysis. Yet, despite the better accuracy, these results reinforce the idea that the MCI may not be particularly useful in sex prediction, since it may not reflect the same degree of sexual dimorphism as its absolute measures.