The first systematic, in-situ, variable-temperature H-1 NMR study, based on the CRAMPS (combined rotation and multiple-pulse spectroscopy) technique, is reported on pyridine-saturated Argonne premium coals. This work focused on how noncovalent associative interactions are affected at the molecular level by thermal treatment and solvent saturation. With improved proton dipolar-dephasing experiments based on CRAMPS detection, we were able to monitor quantitatively changes of molecular mobility of various molecular structural moieties upon pyridine saturation and temperature variation. It was found that pyridine saturation can dramatically promote fast, random molecular motion with correlation times in the range of 1 mus to 30 ns, even at room temperature, while thermal treatment alone is much less effective in promoting molecular motion even at 230 degreesC. For pyridine-saturated coal, the molecular mobility of coal can be enhanced further with just a moderate increase in temperature. Correlations between molecular structure and molecular mobility have been established through comparative variable-temperature 1H CRAMPS studies of coals of different ranks. A mechanism for explaining the enhancement of molecular mobility induced by thermal treatment and solvent saturation is proposed and discussed in terms of these experimental results. Dramatic expansion of void space in coal as a result of disruption of noncovalent associative interactions by pyridine is the key feature of our understanding of these new NMR results, as well as previous work by others. The nature of the molecular phase and macromolecular phase in coal and the relationships between the molecular/macromolecular (M/M) structural model and molecular mobility, as detected experimentally by NMR, have been shown to be much more complicated than previously thought. There is no simple one-to-one correspondence between molecular mobility and M/M phases in coal, but we have developed a coherent view of the molecular structure and dynamics of coal on the basis of the new experimental results. Many discrepancies in the literature are well resolved by this view, and the resolution enhancement due to solvent saturation is satisfactorily explained.