The authors propose two general functional forms, and apply them to the specification of utility functions for predicting farmers' production response. The polynomial utility functions were rejected, based on the result~ of a likelihood-ratio test. The appropriate degree of nonlinearity of the utility function can best be determined by using the general functional forms without a priori specification. Further, farmers' utility functions may exhibit a decreasing absolute risk aversion. The tendency for the Bernoullian utility maximization hypothesis to predict more risky behavior than that actually observed may have been partly due to incorrect specification of the utility function.