In this work, we derive the joint distribution of the latent roots of a sample covariance matrix under elliptical models. We then obtain the distributions of the largest and smallest latent roots. In the process of these derivations, we also correct some results present in the literature.