Abstract. Nearest neighbor (NN) queries are posed very frequently in spatial applications. Recently a branch-and-bound algorithm based on R-trees has been developed in order to answer efficiently NN queries. In this paper, we combine techniques that were inherently used for the analysis of range and spatial join queries, in order to derive measures regarding the performance of NN queries. We try to estimate the number of disk accesses introduced due to the processing of an NN query. Lower and upper bounds are defined estimating the performance of iNN queries very closely. The theoretical analysis is verified with experimental results, under uniform and non-uniform distributions of queries and data, in the ~dimensional address space.