Vapor-liquid equilibrium measurements for the mixture of isoprene and n-pentane at atmospheric pressure were conducted on a modified Othmer still. The experimental data were carefully examined by the residual analysis and the thermodynamic consistency test. Unfortunately, both methods failed to identify the experimental points with systematic error. Thus an alternative treatment by T - x diagram proposed in this study was used to delete the systematic error points. The remaining data were then used for further data regression. In this study the maxium likelihood principle was chosen as the regression technique to determine the parameters of the van Laar, Wilson, NRTL and UNIQUAC models. The van Laar model has the smallest correlation coefficient of the pair parameters among these. If any experimental point was discarded, we applied the F-test to determine (1) whether the new data set is better for data reduction than the old set, and (2) whether the van Laar model is still appropriate for the new data set, and as good as the other models. From the correlation of experimental data and the correlation coefficients of the pair parameters of the Wilson, NRTL and UNIQUAC models, we may conclude that the isoprene and n-pentane mixture is likely to be a regular solution, which is the basic assumption of the van Laar model.