All relevant C-13 NMR signals of a series of 19 homogeneous ethylene-propylene copolymers were used to fit the first-order Markov reactivity ratios of the catalyst and the theoretical feeds. The copolymers cover a broad range of comonomer incorporations, from 17.0 to 56.5%, and show both primary (1,2) and secondary (2,1) insertions. As expected, two solutions are found, the normal solution showing a better fit and reliability >> 99.5%. The reactivity ratios, r(12) = 20.0, r(13) = 162.0, r(21) = 0.015, r(23) = 1.3, and r(31) = 0.060, provide direct information about the vanadium-based catalyst, the kinetics, and the chain microstructure. The values also explain the comonomer content-dependent inversion and even predict a 100% secondary insertion PP homopolymer. With these results, the direct peak method shows that the use of all relevant C-13 NMR peaks improves the accuracy of first-order Markov modeling. (c) 2006 Wiley Periodicals, Inc.