HX-labeling experiments in the pH-pulse mode show that protein folding can be remarkably fast. A near-native form can be reached within milliseconds. Experimental analysis of the folding process on the millisecond-to-second time scale depends upon the presence of kinetic barriers that avoid apparent two-step folding. A common barrier produces molecular intermediates; disparate barriers produce population heterogeneity that makes analysis more difficult. Results available exhibit an early, native-like two-helix intermediate in cytochrome c, an extensive, native-like, beta-sheet-plus-helix intermediate in RNase A, and a late native-like molten globular intermediate in barnase. These differences appear to reflect chance differences in the placement of the determining kinetic barriers. Requirements for observing kinetic folding intermediates are difficult to satisfy, so most intermediates are not seen, and intermediates that are seen often represent the sum of multiple preceding steps.