Pyrolysis experiments were conducted on immature petroleum source rocks under various conditions to evaluate the role of water in petroleum formation. At temperatures less than 330°C for 72 h, the thermal decomposition of kerogen to bitumen was not significantly affected by the presence or absence of liquid water in contact with heated gravel-sized source rock. However, at 330 and 350°C for 72 h, the thermal decomposition of generated bitumen was significantly affected by the presence or absence of liquid water. Carbon-carbon bond cross linking resulting in the formation of an insoluble bitumen (i.e., pyrobitumen) is the dominant reaction pathway in the absence of liquid water. Conversely, thermal cracking of carbon-carbon bonds resulting in the generation of saturate-enriched oil, which is similar to natural crude oils, is the dominant reaction pathway in the presence of liquid water. This difference in reaction pathways is explained by the availability of an exogenous source of hydrogen, which reduces the rate of thermal decomposition, promotes thermal cracking, and inhibits carbon-carbon bond cross linking. The distribution of generated n-alkanes is characteristic of a free radical mechanism, with a broad carbon-number distribution (i.e., C₅ to C₃₅) and only minor branched alkanes from known biological precursors (i.e., pristane and phytane). The generation of excess oxygen in the form of CO₂ in hydrous experiments and the high degree of hydrocarbon deuteration in a D₂O experiment indicate that water dissolved in the bitumen is an exogenous source of hydrogen. The lack of an effect on product composition and yield with an increase in H⁺ activity by five orders of magnitude in a hydrous experiment indicates that an ionic mechanism for water interactions with thermally decomposing bitumen is not likely. Several mechanistically simple and thermodynamically favorable reactions that are consistent with the available experimental data are envisaged for the generation of exogenous hydrogen and excess oxygen as CO₂. One reaction series involves water oxidizing existing carbonyl groups to form hydrogen and carbonyl groups, with the latter forming CO₂ by decarbonylation with increasing thermal stress. Another reaction series involves either hydrogen or oxygen in dissolved water molecules directly interacting with unpaired electrons to form a hydrogen-terminated free-radical site or an oxygenated functional group, respectively. The latter is expected to be susceptible to oxidation by other dissolved water molecules to generate additional hydrogen and CO₂. In addition to water acting as an exogenous source of hydrogen, it is also essential to the generation of an expelled saturate-enriched oil that is similar to natural crude oil. This role of water is demonstrated by the lack of an expelled oil in an experiment where a liquid GaIn alloy is substituted for liquid water. Experiments conducted with high salinity water and high water/rock ratios indicate that selective aqueous solubility of hydrocarbons is not responsible for the expelled oil generated in hydrous pyrolysis experiments. Similarly, a hydrous pyrolysis experiment conducted with isolated kerogen indicates that expelled oil in hydrous pyrolysis is not the result of preferential sorption of polar organic components by the mineral matrix of a source rock. It is envisaged that dissolved water in the bitumen network of a source rock causes an immiscible saturate-enriched oil to become immiscible with the thermally decomposing polar-enriched bitumen. The overall geochemical implication of these results is that it is essential to consider the role of water in experimental studies designed to understand natural rates of petroleum generation, expulsion mechanisms of primary migration, thermal stability of crude oil, reaction kinetics of biomarker transformations, and thermal maturity indicators in sedimentary basins.