The difference between the heat (T(G)) and the cold (T(G)') denaturation temperatures defines the temperature range (T(Range)) over which the native state of a reversible two-state protein is thermodynamically stable. We have performed a correlation analysis for thermodynamic parameters in a selected data set of structurally nonhomologous single-domain reversible two-state proteins. We find that the temperature range is negatively correlated with the protein size and with the heat capacity change (DeltaC(p)) but is positively correlated with the maximal protein stability [DeltaG(T(S))]. The correlation between the temperature range and maximal protein stability becomes highly significant upon normalization of the maximal protein stability with protein size. The melting temperature (T(G)) also shows a negative correlation with protein size. Consistently, T(G) and T(G)' show opposite correlations with DeltaC(p), indicating a dependence of the T(Range) on the curvature of the protein stability curve. Substitution of proteins in our data set with their homologues and arbitrary addition or removal of a protein in the data set do not affect the outcome of our analysis. Simulations of the thermodynamic data further indicate that T(Range) is more sensitive to variations in curvature than to the slope of the protein stability curve. The hydrophobic effect in single domains is the principal reason for these observations. Our results imply that larger proteins may be stable over narrower temperature ranges and that smaller proteins may have higher melting temperatures, suggesting why protein structures often differentiate into multiple substructures with different hydrophobic cores. Our results have interesting implications for protein thermostability.