The geological debate about whether, and to what extent, humic coals have sourced oil is likely to continue for some time, despite some important advances in our knowledge of the processes involved. It is clear that not only liptinites, but also perhydrous vitrinites have the potential to generate hydrocarbon liquids in the course of natural coalification. Some liptinites, especially alginite, cutinite, and suberinite, contain a higher proportion of aliphatic moieties in their structure than other liptinites such as sporinite and resinite and are, therefore, more oil-prone. It is of potential value to be able to predict the several environments of deposition in which coals with high liptinite contents or containing perhydrous vitrinites may have been formed. Review of the distribution of oil-prone coals in time and space reveals that most are Jurassic–Tertiary with key examples from Australia, New Zealand, and Indonesia. Methods based both on experimental simulations and the examination of naturally matured samples have been used to determine the order of generation of hydrocarbons from different macerals. Results are not entirely consistent among the different approaches, and there is much overlap in the ranges of degradation, but it seems probable that in the natural environment vitrinites begin to generate early, followed by labile liptinites such as suberinite, then cutinite, sporinite, and, finally, alginite. Petroleum potential may be determined by experimental simulation of natural coalification or inferred through various micro-techniques, especially fluorescence and infrared (IR) spectroscopy, or bulk techniques such as elemental analysis and ¹³C NMR spectroscopy. The latter three techniques enable a measure of the polymethylene component of the coal, which now appears to be one of the best available approaches for determining petroleum potential. No method of experimental simulation of petroleum generation from coals is without criticism, and comparative results are highly variable. However, hydrous pyrolysis, confined pyrolysis, and forms of open-system hydrous pyrolysis approach acceptable simulations. Whether, and to what degree generated liquid hydrocarbons are expelled, has long been the central problem in ‘oil from coal’ studies. The structure of vitrinite was believed until recently to contain an interconnected microporous network in which generated oil would be contained until an expulsion threshold was attained. Recent studies show the pores are not interconnected. Combined with a dynamic model of pore generation, it now seems that expulsion of hydrocarbons is best explained by activated diffusion of molecules to maceral boundaries and ultimately by cleats and fractures to coal seam boundaries. The main reason for poor expulsion is the adsorption of oil on the organic macromolecule, which may be overcome (1) if coals are thin and interbedded with clastic sediments, or (2) if the coals are very hydrogen-rich and generate large quantities of oil. The existence of oil in vitrinite is attested to by solvent extractions, fluorescence properties, and by microscopic observations of oil and bitumen. Experimental simulation of expulsion of oil from coals has only recently been attempted. The relative timing of release of generated CO₂ and CH₄ could have considerable importance in promoting the expulsion of liquid hydrocarbons but the mechanism is unclear. As it is universally agreed that dispersed organic matter (DOM) in some shales readily generates and expels petroleum, it is curious that few consistent geochemical differences have been found between coal macerals and DOM in interbedded shales. Unambiguous evidence of expulsion from coals is limited, and in particular only a few commercial oil discoveries can be confidently correlated to coals. These include Upper Cretaceous Fruitland Formation coals in the USA, from which oil is produced; New Zealand Tertiary coals; and Middle Jurassic coals from the Danish North Sea. It is likely that coals have at least contributed to significant oil discoveries in the Gippsland Basin, Australia; in the Turpan Basin, China; and in the Kutei and Ardjuna basins in Indonesia, but this remains unproven. Early reports that early Jurassic coals in mid-Norway were a major source of the reservoired oils have been shown to be inaccurate. None of the proposed ‘rules of thumb’ for generation or expulsion of petroleum from coals seem particularly robust. Decisions on whether a particular coal is likely to have been an active source for oil should consider all available geological and geochemical information. The assumptions made in computational models should be well understood as it is likely with new understandings of processes involved that some of these assumptions will be difficult to sustain.