Hi, The silverman's paper introduction offer how to find a mode for one dimensional data based on software http://www.stanford.edu/~kasparr/software/silverman.r, for two dimensional data I use kde2d to smooth it out first, then I get a matrix of densities for all the X(one dimension) cross Y(another dimension). I sort X and Y first before I pass the values to kde2d(x, y, c(hx, hy)), the persp shape changes Does anyone know how to get the modes out of the two dimensional data programmatically. Also if I want to get the minumum of X, Y with modes =2, is the solution unique? Thanks pat