Two distinct but related approaches for the estimation of von Liebig response functions are presented. The first approach is based upon a two‐phase, ordinary least squares procedure combined with bootstrapping for computing the standard errors of the estimates. The second approach generates maximum likelihood estimates of the parameters and can be implemented according to two different parameterizations of the model. Application of the procedures to a sample of experimental data suggests a satisfactory degree of conformity of the results. A test of normality of the disturbance term based upon the estimated residuals failed to reject the null hypothesis, further supporting the maximum likelihood approach.