The availability of the principal component analysis was tested using two types of artificial community models: one is based on a simllarity matrix and the other is on a matrix of the number of individuals Models were laid out to make species composition in quadrats change along one or two environmental gradients. In the smilarity model，the gradients were represented by the first two principal components，while those of the individualmodel were done by the second and fourth ones. A great 'foiding-in' effect occurred obviously in the two models when the gradient was excessive or when quadrats with peculiar species composition were added. The smilarity model has a merit to ordinate quadrats properly irrespective of the degree of the environmental gradient and a defect to be strongly influenced by quadrats with peculiar composition. On the other hand，the individual model has a merit to discriminate peculiar quadrats easily and a defect that the rang of the gradient to ordilate quadrats exactly is rather limited. The principal component analysis may be available in extracting the environmental gradient，if the quadrat data are so arranged that a great 'folding-in' effect does not take place.