In response surface analysis (RSA), a surface of predicted, usually second-order response over a range of quantitative and continuous factor levels is estimated. Less information is available on how to apply RSA in the case of multiple response values that do not show similar trends. This paper provides a solution for this situation by applying first a principal component analysis (PCA) and then taking the first principal component in the RSA. This method is applied to select the optimal seed treatment for breaking dormancy in spinach (Spinacia oleracea L.). Fourteen treatment combinations of the two factors gibberellin (GA3) concentration and soaking period, each with five levels were selected and germination percentage, energy of germination and germination index were determined. The three response values were analyzed by PCA and the first principal component was analyzed by RSA. The obtained optimal treatment combination, GA3 at 261mg/L and a soaking time of 22.1h, was able to increase the germination percentage from 56% to 91.3% and the energy of germination from 38% to 68.7%, respectively. The optimal conditions obtained from the individual response values were different from this optimum.