In this study, a comprehensive laboratory investigation was conducted for the recovery of heavy oil from a scaled three-dimensional (3-D) physical model, packed with 18 degrees API gravity crude oil, brine and crushed limestone. A total of 20 experiments were conducted using the scaled 3-D physical model with 30 x 30 x 6 cm(3) dimensions. Basically, four different immiscible CO2-water displacement processes were used for recovering heavy oil: (i) continuous CO2 injection, (ii) waterflooding, (iii) simultaneous injection of CO2 and water, and (iv) water alternating gas (WAG) process. Three groups of well configurations were mainly used: (1) vertical injection and vertical production wells, (2) vertical injection and horizontal production wells, and (3) horizontal injection and horizontal production wells. Base experiments were run with water only and carbon dioxide alone and optimum rates for WAG and simultaneous water-CO2 injection were determined. Tn continuous CO2 injection, highest recovery was obtained by vertical injection-horizontal production (VI-HP), followed by vertical injection-vertical production (VI-VP) and the least by horizontal injection-horizontal production (HI-HP). In VI-HP configuration, the best recovery was obtained as 15.1% OOIP. Higher oil recovery was obtained with a VI-HP wells than with a pair of vertical wells and horizontal wells. The WAG 1:5 ratio yielded a final recovery of 34.5% OOIP with VI-VP well configuration and 17.0% OOIP of additional recovery over waterflooding. In turn, the WAG 1:10 ratio was the best with a final recovery of 20.9% of OOIP with VI-HP well configuration. Oil production from WAG injection is higher than that obtained from the injection of continuous CO2 or waterflooding alone. (C) 2000 Elsevier Science B.V. All rights reserved.