Abstract: This paper uses principal component analysis and the ward’s cluster method of squared euclielian distances of standard data to svaluate fertility grade synthetically for the 12 fertility indexes of 158 paddy soils from medium-low yield regions at Fujian. The results showed that the content of soil organic matter, the status of soil textare, the depth of tillage layer and the content of avalable K2O affected soil fertility more strongly than any other indexes. The scores of four largest principal factors were used as new variables then 158 paddy soils were ferther grouped into seven types of soil fertility on the basis of clustering. Of them there are three types belonging to low fertility, two types belonging to medium and the other two types to high fertility. It is a guide to the classification of medium-low yielding fields.