Abstract: In recent years，the research on spatial indexing is arousing more and more interests and attentions.A great number of spatial indexing structures based on secondary memory have been proposed for handling the massive spatial data stored in the spatial database rapidly and efficiently.R-tree，proposed by Guttman in 1984，is the most efficient spatial index structure now.In this paper，a new algorithm is presented for overcoming the diavantages of the old algorithm in distributing node of R-tree.The result of study indicates the new algorithm brings smaller overlap，thereby the probability of multi-way query is efficiently controlled，and the efficiency of spatial query is improved effectively.