The multicanonical (Muca) Monte Carlo method enables simulating a system over a wide range of temperatures and thus has become an efficient tool for studying spin glasses, first-order phase transitions, the helix-coil transition of polypeptides, and protein folding. However, implementation of the method requires calculating the multicanonical weights by an iterative procedure that is not straightforward and is a stumbling block for newcomers. A recursive procedure that takes into account the statistical errors of all previous iterations and thus enables an automatic calculation of the weights without the need for human intervention after each iteration has been proposed. This procedure, which has already been tested successfully for lattice systems, is extended here to continuum models of peptides and proteins. The method is examined in detail and tested for models of the pentapeptide Leu-enkephalin (Tyr-Gly-Gly-Phe-Leu) described by the potential energy function ECEPP. Because of the great interest in the structural mapping of the low-energy region of biomolecules, the energy of structures selected from the Muca trajectory is minimized. The extent of conformational coverage provided by the method is examined and found to be very satisfactory.