Raphanus sativus (L. Var) is a perennial plant of the Brassicaceae (or Cruciferae) family whose oil has not been investigated in detail for biodiesel production, particularly when ethanol is used as the alcoholysis agent. In this work, response surface methodology (RSM) was used to determine the optimum condition for the ethanolysis of R. sativus crude oil. Three process variables were evaluated at two levels (23 experimental design): the ethanol:oil molar ratio (6:1 and 12:1), the catalyst concentration in relation to oil mass (0.4 and 0.8 wt% NaOH) and the alcoholysis temperature (45 and 65 degrees C). When the experimental results were tentatively adjusted by linear regression, only 58.15% of its total variance was explained. Therefore, a quadratic model was investigated to improve the poor predictability of the linear model. To apply the quadratic model, the 23 experimental design had to be expanded to a circumscribed central composite design. This allowed the development of a response surface that was able to predict 97.75% of the total variance of the system. Validation was obtained by performing one ethanolysis experiment at the conditions predicted by the model (38 degrees C, ethanol:oil molar ratio of 11.7:1 and 0.6 wt% NaOH). The resulting ester yield (104.10 wt% or 99.10% of the theoretical yield of 105.04 wt%) was shown to be the highest among all conditions tested in this study. The second ethanolysis stage of the best RSM product required 50% less ethanol and 90% less catalyst consumption. The amount of ethyl esters obtained after this procedure reached 94.5% of the theoretical yield. The resulting ethyl esters were shown to comply with most of the Brazilian biodiesel specification parameters except for oxidation stability. Addition of 500 ppm of BHT to the esters, however, complied with the specification target of 6 h. The application of 2 wt% Magnesol (R) after the second ethanolysis stage eliminated the need for water washing and helped generate a final product with less unreacted glycerides. (C) 2007 Elsevier Ltd. All rights reserved.