Nuclear magnetic resonance (NMR) relaxation times of geologic materials are closely related to pore geometry. In heterogeneous media, however, the details of this relationship are poorly understood because of a phenomenon known as pore coupling, which arises when diffusing protons sample multiple pores before relaxing. Laboratory experiments allow us to explore whether surface geochemistry can influence pore coupling and how this process affects the observed relaxation-time distribution. Measurements of the NMR response for microporous silica gel packs, treated with varying amounts of surface-coating iron, demonstrate that samples with less iron exhibit stronger pore coupling than those with abundant iron. When pore coupling is strong, the relaxation-time distribution grossly misrepresents the underlying bimodal pore-size distribution of micropores and macropores. Specifically, the bimodal relaxation-time distribution becomes merged and the relative amplitude of the peaks fails to reflect the true macropore and micropore volume. A reduction in pore coupling, observed with increasing iron content, is attributed to a decrease in the distance protons are able to diffuse before relaxing. Basic parameters describing the shape of the relaxation-time distributions for this range of samples are well-predicted by a 1D analytical model. Experimental results conclusively demonstrate that surface geochemistry is an important factor determining the degree to which pore coupling occurs and illustrate how this phenomenon can affect the interpretation of NMR relaxation measurements in heterogeneous porous media.