We describe a new method (HOP) for identifying groups of particles inN-body simulations. Having assigned to every particle an estimate of itslocal density, we associate each particle with the densest of the Nhopparticles nearest to it. Repeating this process allows us to trace apath, within the particle set itself, from each particle in thedirection of increasing density. The path ends when it reaches aparticle that is its own densest neighbor; all particles reaching thesame such particle are identified as a group. Combined with an adaptivesmoothing kernel for finding the densities, this method is spatiallyadaptive, coordinate-free, and numerically straightforward. One canproceed to process the output by truncating groups at a particulardensity contour and combining groups that share a (possibly different)density contour. While the resulting algorithm has several user-chosenparameters, we show that the results are insensitive to most of these,the exception being the outer density cutoff of the groups.