While the canonical paradigm for protein evolution states that new functions evolve through gene duplication followed by divergence, several observations support the hypothesis that new protein functions evolve through promiscuous intermediates prior to gene duplication. Initial support for this hypothesis derives from observations of promiscuity in individual enzymes and from protein engineering experiments, in which promiscuity can be a vital prerequisite for evolving new functions. Careful analysis of the evolution of function of o-succinylbenzoate synthase (OSBS) has provided some of the strongest evidence supporting this hypothesis. Strikingly, several members of the large and diverse OSBS family exhibit both OSBS and N-succinylamino acid racemase (NSAR) activities. All characterized OSBS/NSARs and many uncharacterized proteins cluster together in the phylogeny, indicating that they have a common origin. Based on operon context, the physiological function of some of these OSBS/NSARs is OSBS, others are required for NSAR activity, and some are bifunctional for both activities. Many other examples of naturally promiscuous proteins are emerging, suggesting that evolving new functions through promiscuous intermediates is a common pathway for enzyme evolution. There are two major implications for these observations. First, the functions of many homologous proteins may only partially overlap; as a result, correctly predicting protein functions in genome projects may require more rigorous methods than merely comparing sequence similarity. Second, if promiscuous proteins represent evolutionary intermediates, using them as starting points for protein engineering may be a fruitful strategy for designing proteins with new and useful properties.