A key problem in experimental protein folding is that of characterizing the conformational ensemble of denatured proteins under folding conditions. We address this problem by studying the conformational propensities of reductively unfolded RNase A under folding conditions, since earlier work has indicated that the equilibrium conformational ensemble of fully reduced RNase A resembles the transient conformational ensemble of a burst-phase folding intermediate of disulfide-intact RNase A. To assess these propensities, the relative disulfide-bond populations of the 1S, 2S, and 3S ensembles of the [C40A,C95A] mutant of RNase A were measured. Thirteen of the fifteen possible disulfide bonds are observed, consistent with earlier results and with the rapid reshuffling and lack of stable tertiary structure in these ensembles. This broad distribution contradicts recent observations by another group, but rigorous cross-checks show unambiguously that our data are self-consistent whereas their data are not. The distributions of disulfide bonds in the wild-type and mutant proteins show a power-law dependence on loop length, with an exponent that is significantly smaller than the exponents of either ideal or excluded-volume polymers. The 65-72 disulfide bond is much more strongly favored than would be predicted by this power law, consistent with earlier peptide studies and the disulfide-bond distributions of the 1S and 2S ensembles in wild-type RNase A. Experimental evidence suggests that this preference results from conformational biases in the backbone, rather than from differing accessibilities or reactivities of the two cysteine residues. In general, the other disulfide species do not deviate significantly from the power-law dependence, indicating that the conformational biases are relatively weak.