Nuclear magnetic resonance (NMR) relaxation time distributions are frequently combined with mercury intrusion capillary pressure (MICP) measurements to allow determination of pore or pore throat size distributions directly from the NMR data. The combination of these two measurements offers an advantage over high-resolution imaging techniques in terms of cost and measurement time, and can provide estimates of pore sizes for pores below imaging resolution. However, the methods that are typically employed to combine NMR and MICP measurements do not necessarily honor the way in which the two different measurements respond to the size distribution and connectivity of the pore system. We present a method for combining NMR and MICP data that is based on percolation theory and the relationship between bond occupation probability and the probability that a bond is part of a percolating cluster. The method yields results that compare very well with pore sizes measured by high-resolution microtomography, and provides particular improvement in media with broad pore size distributions and large percolation thresholds.