In establishing fertilizer recommendations for farmers, standard production function analysis has been widely applied to the analysis of fertilizer experimental data in order to derive optimal levels of fertilizer. That is, regression analysis is used to fit a suitable response function and then optimal levels are calculated by setting the marginal productivity of nutrients (given by differentiation of the function) equal to the ratio of the price of nutrients to the price of the crop. Several methodological studies have focused in depth on questions of experimental design and statistics relating to the estimation of the function (e.g. Baum, Heady and Blackmore, Heady and Dillon, Colwell). However, none of these really addresses the practical questions of relating experimental results and prices to those faced by farmers. This note shows that, particularly for small farmers, the naive assumptions on yields and prices usually used by economists in this type of analysis may lead to unrealistically high levels of fertilizer recommendations which the farmer would never find practical (or profitable) to adopt.