SYNOPTIC ABSTRACT
A family of models involving intersecting straight lines for evaluating response to fertilizer nutrients, for experiments having four to seven levels of a nutrient and exhibiting marked initial response, is examined. A modified quadratic response model is proposed for fitting fertilizer response data which exhibit declining yields at higher rates of application. The bias in estimating optimal rate is studied. A graphical procedure which mitigates the bias problem in multi-nutrient experiments is also examined. Numerical illustrations are given.