Experiments with cytochrome c (cyt c) show that an initial folding event, molecular collapse, is not an energetically downhill continuum as commonly presumed but represents a large-scale, time-consuming, cooperative barrier-crossing process. In the absence of later misfold-reorganization barriers, the early collapse barrier limits cyt c folding to a time scale of milliseconds. The collapse process itself appears to be limited by an uphill search for some coarsely determined transition state structure that can nucleate subsequent energetically downhill folding events. An earlier "burst phase" event at strongly native conditions appears to be a non-specific response of the unfolded chain to reduced denaturant concentration. The molecular collapse process may or may not require the co-formation of the amino- and carboxyl-terminal helices, which are present in an initial metastable intermediate directly following the rate-limiting collapse. After the collapse-nucleation event, folding can proceed rapidly in an apparent two-state manner, probably by way of a predetermined sequence of metastable intermediates that leads to the native protein structure (Bai et al., Science 269:192-197, 1995).