The idea of separating the free energy of mixing into two contributions corresponding to the "physical" or "nonspecific" and to the "specific" interactions, respectively) proposed by Painter and Coleman has been extended to ternary polymer/polymer/polymer blends. In this paper experimental phase diagrams of three ternary blends have been reproduced by estimating both contributions for the interaction energy density (B(ij)) which characterizes each polymer pair. In contrast to the Painter-Coleman Miscibility Guide rules, we have found that, in some cases, "specific" contributions positive and unfavorable to the mixing, due to the self-association of one of the components, can describe more adequately the experimental trends.