Hydrogen bonds play an important role in stabilizing (meta-)stable states in protein folding. Hence, they can potentially be used as a way to bias these states in molecular simulation methods. Previously, Wolf et al. showed that applying repulsive and attractive hydrogen bond biasing potentials in an alternating way significantly accelerates the folding process (Wolf, M. G.; de Leeuw, S. W. Biophys. J. 2008, 94, 3742). As the biasing potentials are only active during a fixed time interval, this alternating scheme does not represent a thermodynamic equilibrium. In this work, we present a Hamiltonian replica exchange molecular dynamics (REMD) scheme that aims to shuffle and reorder hydrogen bonds in the protein backbone. We therefore apply adapted hydrogen bond potentials in a Hamiltonian REMD scheme, which we call hydrogen bond switching (HS). To compare the performance of the HS to a standard REMD method, we performed HS and temperature REMD simulations of a beta-heptapeptide in methanol. Both methods sample the conformational space to a similar extent. As the HS simulation required only five replicas, while the REMD simulation required 20 replicas, the HS method is significantly more efficient. We tested the HS method also on a larger system, 16-residue polyalanine in water. Both of the simulations starting from a completely unfolded and a folded conformation resulted in an ensemble with, apart from the starting structure, similar conformational minima. We can conclude that the HS method provides an efficient way to sample the conformational space of a protein, without requiring knowledge of the folded states beforehand. In addition, these simulations revealed that convergence was hampered by replicas having a preference for specific biasing potentials. As this sorting effect is inherent to any Hamiltonian REMD method, finding a solution will result in an additional increase in the efficiency of Hamiltonian REMD methods in general.