This extension of the liquid hydrocarbon model seeks to quantify the thermodynamic contributions to protein stability from the removal of nonpolar and polar surface from water. Thermodynamic data for the transfer of hydrocarbons and organic amides from water to the pure liquid phase are analyzed to obtain contributions to the thermodynamics of folding from the reduction in water-accessible surface area. Although the removal of nonpolar surface makes the dominant contribution to the standard heat capacity change of folding (DELTA-C(fold)o), here we show that inclusion of the contribution from removal of polar surface allows a quantitative prediction of DELTA-C(fold)o within the uncertainty of the calorimetrically determined value. Moreover, analysis of the contribution of polar surface area to the enthalpy of transfer of liquid amides provides a means of estimating the contributions from changes in nonpolar and polar surface area as well as other factors to the enthalpy of folding (DELTA-H(fold)o). In addition to estimates of DELTA-H(fold)o, this extension of the liquid hydrocarbon model provides a thermodynamic explanation for the observation [Privalov, P. L., & Khechinashvili, N. N. (1974) J. Mol. Biol. 86, 665-684] that the specific enthalpy of folding (cal g-1) of a number of globular proteins converges to a common value at approximately 383 K. Because amounts of nonpolar and polar surface area buried by these proteins upon folding are found to be linear functions of molar mass, estimates of both DELTA-C(fold)o and DELTA-H(fold)o may be obtained given only the molar mass of the protein of interest. Use of DELTA-C(fold)o and DELTA-H(fold)o in conjunction with a melting temperature (T(m)) determined under specified solution conditions allows one to estimate the stability (DELTA-G(fold)o) under these conditions at any temperature. Calculated values of DELTA-G(fold)o in the range 293-320 K appear generally to agree with those determined by calorimetric and noncalorimetric methods. This analysis should provide a basis for predicting the stability of proteins whose structural and thermodynamic properties are similar to those in the data set used for this analysis. Deviations from these predictions may be interpretable in terms of deviations from the average properties of the proteins in the data set.