Abstract We present a method for enhancing the sampling of configuration space in molecular dynamics and Monte Carlo simulations of systems exhibiting high potential energy barriers. The method builds on a previously described approach, “puddle-skimming”, in which the potential energy function, V( q ) , is modified as follows: wherever V( q ) B , it is set equal to EB, where EB is a selected “puddle height”. V( q ) is unmodified in regions where V( q )>E B . The extended method, “puddle-jumping”, incorporates two or more different puddle heights, with transitions between them. Motion on the higher energy puddles promotes barrier crossings, while motion on the lower puddles more efficiently samples the more important low potential energy regions. We test the method on conformational changes in a model of n-pentane, for which the potential energy parameters have been modified to exhibit a range of barrier heights. The method greatly increases the frequency of barrier crossings while minimizing the time spent sampling unimportant, high-energy regions of configuration space. It is flexible and can be applied to large systems whose features are not known in advance. Finally, it converges rigorously to the correct canonical ensemble distribution; options in its implementation affect only the rate of convergence, not the final converged result.