Farmer's risk attitudes are modelled using the Cobb-Douglas, transcendental, negative exponential, and conjoint measurement utility functions. The farmers' perception of the riskiness of alternative crops is also measured and a quadratic programming algorithm is used to derive the most efficient expected meanvariance (E-V) frontier of each farmer. The E-V frontiers are then used in conjunction with the utility functions to determine the optimal farm plans. Farm plans that maximise expected profit are also determined. The results reveal that the conjoint measurement utility model predicts actual behaviour better than the other models. The expected profit model, on the other hand, is the worst predictor. This indicates that risk does play a role in decision-making and that the farmers are utility maximizers rather than profit maximizers only. Therefore, more effective programs would be those that tend to reduce risks and uncertainties faced by the farmers.