Nuclear magnetic resonance T₂ measurements for estimating permeability and drainage capillary pressure, particularly in tight sandstones, are investigated. An assessment of popular permeability models suggested that permeability is hard to predict universally. The best a model may achieve, compared with the measured permeability, is one order of magnitude. As a result of the research the effective surface relaxivity is solved in collaboration with transverse mean relaxation rate using two large siliciclastics datasets. The developed correlation (p = 0.92T₂¯0.935) was used to convert NMR spectrum to capillary pressure. An improved NMR permeability model is developed (K = CaTP30r1+1); whereby the correlation is not influenced by total porosity but uses properties that are directly related to flow. Effective tortuosity correlates with the ratio of bound to free fluid; as a result a new cementation factor equation is derived. A direct prediction of pore volume from NMR is developed, the approch uses the sum of NMR amplitude (PV = 0.017SA — 0.24), and avoids complex computations and corrections. The modelled drainage capillary pressure replicated the measured curvature; it enhances the conventional fixed scaling factor method and allows permeability predictions. However the calibre of the calculated permeabilities is subject to a good match between the predicted and modelled capillary pressure. The developed NMR permeability model assumes a good relationship between the pore throat and pore body and that mean T₂ is correlated to permeability. However, conditions for using NMR to characterise reservoir properties are that its signal is proportional to the pore size and the absence of paramagnetic ions. Nonetheless, this study illustrates clearly that NMR is an important reservoir characterisation tool.