Nuclear magnetic resonance (NMR) has been widely used in petrophysical characterization of sandstones and carbonates, but little attention has been paid in the use of this technique to study petrophysical properties of coals, which is essential for evaluating coalbed methane reservoir. In this study, two sets of NMR experiments were designed to study the pore types, pore structures, porosity and permeability of coals. Results show that NMR transverse relaxation (T(2)) distributions strongly relate to the coal pore structure and coal rank. Three T(2) spectrum peaks identified by the relaxation time at 0.5-2.5 ms, 20-50 ms and >100 ms correspond to pores of <0.1 mu m, >0.1 mu m and cleats, respectively, which is consistent with results from computed tomography scan and mercury intrusion porosimetry. Based on calculated producible and irreducible porosities through a T(2) cutoff time method, we propose a new NMR-based permeability model that better estimates the permeability of coals. In combination with mercury intrusion porosimetry, we also propose a NMR-based pore structure model that efficiently estimates the pore size distribution of coals. The new experiments and modeling prove the applicability of NMR in petrophysical characterization of intact coal samples, which has potential applications for NMR well logging in coalbed methane exploration. (C) 2009 Elsevier Ltd. All rights reserved.