Carbon dioxide (CO2) injection is a well-established method for increasing recovery from oil reservoirs. However, poor sweep efficiency has been reported in many CO2 injection projects due to the high mobility contrast between CO2 and oil and water. Various injection strategies including gravity stable, WAG and SWAG have been suggested and, to some extent, applied in the field to alleviate this problem. An alternative injection strategy is carbonated water injection (CWI). In CWI, CO2 is delivered to a much larger part of the reservoir compared to direct CO2 injection due to a much improved sweep efficiency. In CWI, CO2 is used efficiently and much less CO2 is required compared to conventional CO2 flooding, and hence the process is particularly attractive for reservoirs with limited access to large quantities of CO2 (offshore reservoirs or reservoirs far away from inexpensive natural CO2 resources). This article describes the results of a pore-scale study of the process of CWI by performing high-pressure visualisation flow experiments. The experimental results show that CWI, compared to unadulterated (conventional) water injection, improves oil recovery as both a secondary (before water flooding) and a tertiary (after water flooding) recovery method. The mechanisms of oil recovery by CWI include oil swelling, coalescence of the isolated oil ganglia and flow diversion due to flow restriction in some of the pores as a result of oil swelling and the resultant fluid redistribution. In this article the potential benefit of a subsequent depressurisation period on oil recovery after the CWI period is also investigated.