Mitscherlich function (MF), Maskell Resistance function (MRF) and the quadratic function (QF) are mostly the rival fertilizer response functions (FRFs) in practice. A statistical procedure considering bias only is discussed for studying the closeness or divergence of two FRFs. QF approximates MF very closely. The closeness of the approximation is little affected by the number of levels or the value of the parameters in MF. It approximates MRF reasonably well when the soils are not very deficient, the closeness of the approximation is adversely affected when the number of levels is increased though less beyond 8 and marginally after 12. Equi-response providing levels have slight edge over the equi-spaced levels though by an insignificant amount. Thus, the designs which are optimal for estimating QF are expected to be nearly so for estimating MF or MRF. A method of obtaining good starting values throug QF is described for estimating MF and MRF. Yates hypothesis that A′”B′” is the best two-factor interaction component for confounding in 42-design has been disproved and it is shown that A”B” is not only the best for confounding but it also keeps all the other effects completely free from confounding. Some usefull and and unique properties ofthe 4m-design are discussed and plans in the usefull range are presented. The notion of general FRF is introduced alongwlth few such functions with a view to get rid of the present hit and trial practice of selecting an appropriate FRF in a particular situation.
