The main contribution of this paper is the use of probability theory to prove that the three-parameter Mittag-Leffler function is the Laplace transform of a distribution and thus completely monotone. Pollard used contour integration to prove the result in the one-parameter case. He also cited personal communication by Feller of a discovery of the result by ''methods of probability theory''. Feller used the two-dimensional Laplace transform of a bivariate distribution to derive the result. We pursue the theme of probability theory to explore complete monotonicity beyond the contribution due to Feller. Our approach involves an interplay between mixtures and convolutions of stable and gamma densities, together with a limit theorem that leads to a novel integral representation of the three-parameter Mittag-Leffler function (also known as the Prabhakar function).