Vanadium(III) complex bearing PMePh2 ligand has been synthesized, fully-characterized and investigated for homo- and co-polymerization of ethylene with norbornene, in combination with aluminum compounds, i.e., methylaluminoxane (MAO) and Et2AlCl. The results are carefully compared with those obtained with VCl3(THF)(3) and VCl3 in the presence and absence of ethyltrichloroacetate (ETA), and free phosphine. VCl3(PMePh2)(2) was likely activated through a dissociative mechanism where the dissociation of the labile phosphine ligand was the first and possibly the rate-determining step. The activity of VCl3(THF)3 was slightly higher than that of VCl3(PMePh2)(2) likely due to a different competition of reinsertion rate of dissociated ligand compared to the insertion/coordination of (co)monomers. Copolymers obtained at low norbornene feedstock concentration from VCl3(THF)(3), showed greater non-uniformity in terms of composition distribution with respect to those from VCl3(PMePh2)(2) because of a more pronounced compositional drift in the semi-batch polymerization process. This strongly affects the copolymer's thermal properties, the copolymers from VCl3(PMePh2)(2) exhibiting higher T(g)s. The effect of PMePh2 and THE ligand on reactions responsible for chain growth termination, affecting the copolymers molecular weight, is discussed. Control over the norbornene incorporation and molecular weight of the resultant copolymers proved to be possible by changing the polymerization temperature, and ETA loading. In addition, ad-hoc. experiments, designed to ensure the uniformity of the catalytic copolymerization process even in the presence of the observed relevant (and unavoidable) compositional drift were performed, and copolymers microstructure and catalytic mechanisms were thoroughly investigated. Due to the peculiar features of the catalytic systems, these studies could be performed only after appropriate modifications to well established methods were developed. As a by-product of these investigations, general and original computational methods are proposed, whose applicability goes beyond the cases treated here. (C) 2016 Elsevier B.V. All rights reserved.