Enhanced oil recovery by dry forward in-situ combustion has been studied in a combustion tube. Twelve experiments are reported exploring the effects of three factors: oxygen flow, partial pressure and mole fraction, each factor at two levels. The pressures used went up to 790 kPa, and the oxygen mole fraction to 35%. It was discovered that the oxygen partial pressure had no statistically significant effect. The oil recovery was independent of the factors used. The combustion time was dominated by the oxygen flow, as were the reaction rates, while fuel and oxygen consumption depended mainly on the oxygen mole fraction. Increasing the oxygen mole fraction reduced the consumption figures. The reaction stoichiometry was substantially independent of the three factors. It was also found that the total pressure had no statistically significant effect on oil recovery, combustion time, reaction rates, fuel consumption or stoichiometry. The oil produced by the in-situ combustion process tended to be of lower viscosity and density than the original oil. Oil-water emulsions were produced which could not be broken. The experiments were controlled by a computer, and the PID control algorithms and associated equipment proved succesful. Linked in with the control routines was a model of the process to calculate fluid saturations and flows during the course of the experiment. Measured information was used directly in the mass and energy balances. The resultant fluid saturations supplied a reasonable match with experimental oil saturations from two experiments that were stopped early. The computed liquid production histories also matched up well with the experimental results. The oil saturations from the numerical model were used in developing a robust method for calculating reaction constants from the experimental data. A simplified surface-reaction scheme was used involv~ng low-temperature oxidation and fuel burnoff to explain the effects of flow, pressure and oxygen mole fraction on the process.