Abstract Visual micromodels are a powerful tool for examining the mechanisms of oil recovery from porous media at the pore level. To this end, an algorithm has been developed to create two-dimensional flow network patterns simulating porous media with controlled properties. This has been used to manufacture flow micromodels that have different grain size distributions, permeabilities and heterogeneities. CO2 floods were carried out to determine the effect of several factors, such as injection rate, oil viscosity and the co-injection of water with CO2, on the displacement process and the oil recovery mechanism. Viscous fingering was found to be the dominant displacement mechanism up to solvent breakthrough at all the flooding conditions. Subsequent growth of the fingers was by a much slower dispersion-type process. A mechanism for recovering watershielded oil by a cyclic build-up, thinning-out and then snap-off of the shield has been observed. Introduction Visual micromodels as used in this study can be defined as flow apparatuses that enable visual observation of multiphase flow behaviour in porous media at the pore level. The glass micromodels used in this study of enhanced oil recovery are essentially a flow network etched onto the surface of a glass plate. The first step in building a micromodel was obtaining a high contrast diagram of a flow network pattern. The flow network pattern was then etched onto a glass plate to produce a micromodel. This plate was then sandwiched with another flat plate to seal the channels. This assembly then formed a two-dimensional path through which various flow phenomena were visually observed. A special micromodel holder enabled carrying out floods at reservoir temperature and pressure. Other researchers used glass beads, sand grains, or even thin sections of rock sandwiched between glass plates. Micromodels have proven to be very useful for studying a variety of oil recovery processes such as waterflooding, gels for conformance control, miscible and immiscible displacements, surfactant floods, foam injection, foamy oil flow, microbial EOR and solution gas drive. Micromodels have also been used to study specific aspects relating to flow in porous media such as wettability, capillary pressure, interfacial tension, asphaltene deposition, heterogeneity, mass transfer, scaling, multiple contact miscibility and gravity drainage. A full literature survey has been previously presented(1) and previous research(2) compares glass micromodels with nanotechnology-based polymer ones and presents image analysis techniques for quantitative data interpretation. This paper presents an innovative method for the design of the flow network pattern. This is followed by a discussion of the etching procedure and of the experimental apparatus for carrying out visual micromodel floods at reservoir conditions. Results of several micromodel floods designed to investigate parametrically the influence of several factors on the flooding process are then presented and discussed. Finally, conclusions are drawn. Design of Flow Network Pattern A variety of approaches have been used in the literature to develop pore network patterns(1). These include commercial pattern transfers such as those employed by draughtsmen, computerdrawn regular patterns and thin-section micrographs of reservoir rocks.