Experimental values of the Flory-Huggins parameter, X, between polymers and solvents, are frequently used to determine the solubility parameters of the polymers. A method using nonlinear curve fitting of RTchi/V was compared to the linear regression method commonly used. It was found that the formulas for the solubility parameter were the same, but the linear method produced a slightly different entropy term. The nonlinear method gave a lower correlation coefficient and wider confidence intervals and was more effective at distinguishing systems than the linear model. The effect of the deviation of probes in the solubility parameter model is discussed. Using probes with low solubility parameters to measure the polymer solubility parameter gave wider confidence intervals. (C) 2004 Wiley Periodicals, Inc.