Abstract An apparatus capable of directly observing, by an optical method, swelling of coal in CO 2 and other gases is described. The system was used to measure CO 2 -induced swelling in three Australian bituminous coals at temperatures between 25 and 55 °C and pressures up to 15 MPa. All three coals showed similar behaviour, with swelling increasing as a function of pressure up to about 8 to 10 MPa, after which no further increase in swelling was observed. The maximum volumetric swelling was between about 1.7 and 1.9%, even in liquid CO 2 . Significant anisotropy was observed; swelling in the plane perpendicular to the bedding plane was always substantially higher than in the parallel plane. The swelling in these coals was completely reversible with each sample returning to its original dimensions after the gas was removed, even after multiple exposures. Temperature did not directly affect the maximum amount of swelling, however, the swelling tended to occur at lower pressures with decreasing temperature. Expressing the swelling as a function of gas density rather than gas pressure showed that swelling was independent of temperature; in all cases the swelling curves were identical. Swelling could be well represented by a modified Dubinin–Radushkevich model which has been used previously to describe gas adsorption on coal. Volumetric swelling data for two of the coals were combined with gas adsorption results previously measured for these coals. Swelling was roughly proportional to the amount of CO 2 adsorbed up to intermediate pressures, but at high pressures, the relationship was no longer linear; adsorption continued to increase but swelling did not. The swelling data were used to correct the adsorption isotherm measured for these coals, assuming that the volume occupied by the coal was inaccessible to gas and increased in proportion to volumetric swelling. At pressures up to about 5 MPa, the effect of correction for swelling on the adsorption value was negligible but at higher pressures, the volume changes due to the swelling of the coal resulted in an increased measured excess adsorption. At 15 MPa the corrected excess sorption was about 30% greater than the uncorrected value. Correction for swelling had little effect on the calculated sorption capacity of the coal or the heat of sorption (using the modified Dubinin–Radushkevich equation) but had strong effects on the term representing dissolution of carbon dioxide into the coal.