The conservation of nucleotides at splicing sites and the characteristics of base composition and base correlation in the adjacent segment sequences have been investigated by use of the method of diversity measure combined with quadratic discriminant analysis. About 4 000 genes in five model genomes have been studied. The splicing sites and the exon/intron boundaries are recognized and predicted. The preliminary calculation shows that, through this simple and unified approach the prediction accuracy on the nucleotide basis is from 92.5% to 97.1% for C.elegans, A.thaliana, D.melanogaster and human. The prediction sensitivity and specificity on the exon basis are 83.7%～94.5% and 87.8%～97.1% respectively for these genomes. Non-canonical splicing has also been analyzed. The prediction capacity of the present method is comparable with GeneSplicer and other current splice site detectors.