Advances in spectroscopy, protein engineering, and peptide synthesis have had a dramatic impact on the understanding of the structures and stabilities of transient folding intermediates. The data available from a variety of proteins point to the existence of three common stages of folding. 1. Initially, the unfolded protein collapses to a presumably more compact form containing substantial nonpolar surfaces and secondary structure. This species has little thermodynamic stability and encompasses an ensemble of conformations that are in dynamic equilibrium and may contain non-native elements of structure. This reaction occurs in less than 5 ms and, from a thermodynamic perspective, may be a noncooperative transition. The relatively high content of secondary structure implies that this manifold of states must be far smaller than the manifold for the unfolded protein. 2. The next phase involves the further development of secondary and the beginnings of specific tertiary structure throughout the protein as well as of measurable stability. Nativelike elements of structure appear, possibly in the form of subdomains that are yet to be properly docked. In many instances, the packing is not as tight as is ultimately found in the native conformation, suggesting that the side chains are in general more mobile. Some elements of surface structure, such as loops and the peripheries of sheets and helices, are not yet well defined. This stage, which may consist of more than a single kinetic step, occurs in the 5-1000 ms time range. The ensemble of conformations is much reduced from the first stage; however, it is far from a single, highly populated form. 3. The final stage in folding corresponds to the concerted formation of many noncovalent interactions throughout the protein. The solidlike interior packing is achieved; the final secondary structure forms and the surface structures settle into place. The breadth of these conformational changes reflects the global cooperativity characteristic of protein folding reactions. A pictorial representation of the kinetic and thermodynamic aspects of this process is shown in Figure 1. This folding scheme emphasizes the progressive development of structure and stability through an ever-slowing set of reactions. Because the product of each stage of folding, with the exception of the final step, is an ensemble of related but not identical species, it is an oversimplification to describe the process as a pathway. Perhaps it is better described as a series of transitions between manifolds of structures that are in dynamic equilibrium within any given set. The funneling effect that must occur when the myriad of unfolded conformers is converted into effectively a single native form implies that these ensembles progressively decrease in size as higher-order structure develops. The simple exponential behavior for the interconversions between these ensembles suggests that the passage may involve a common transition state. Thus, the unique aspects of folding mechanisms may not be the transient intermediates themselves but the barriers separating them. This scheme incorporates important elements of several previous proposals to describe the folding process. The initial stage in folding could well involve the rapid formation of isolated elements of secondary structure [the framework model] whose presence could enhance either associations between these elements [the diffusion-collision model] or a hydrophobic collapse [the molten globule model]. The subsequent, intermediary stage in folding leads to higher-order structures, which could be described as critical substructures, subdomains, or midsize autonomous folding units. The final stage in folding involves wide-scale rearrangements in packing and hydrogen bonding even within folded subdomains, consistent with a conformational feedback mechanism. Thus, each of these previous models could play a role in at least one of the stages in the present scheme.